Differences in Cardiometabolic Proteins in Pregnancy Prioritize Relevant Targets of Preeclampsia

Abstract

Background.

Preeclampsia (PreE) is a hypertensive disorder of pregnancy characterized by widespread vascular inflammation. It occurs frequently in pregnancy, often without known risk factors, and has high rates of maternal and fetal morbidity and mortality. Identification of biomarkers that predict PreE and its cardiovascular sequelae prior to clinical onset, or even prior to pregnancy, is a critical unmet need for prevention of adverse pregnancy outcomes.

Methods:

We explored differences in cardiovascular proteomics (Olink Explore 384) in 256 diverse pregnant persons across two centers (26% Hispanic, 21% Black).

Results:

We identified significant differences in plasma abundance of markers associated with angiogenesis, blood pressure, cell adhesion, inflammation, and metabolism between individuals delivering with PreE and controls, some of which have not been widely described previously and are not represented in the PreE placental transcriptome. While we observed a broadly similar pattern in early (<34 weeks) vs. late (≥34 weeks) PreE, several proteins related to hemodynamic stress, hemostasis and immune response appeared to be more highly dysregulated in early PreE relative to late PreE.

Conclusion:

These results demonstrate the value of performing targeted proteomics using a panel of cardiovascular biomarkers to identify biomarkers relevant to PreE pathophysiology and highlight the need for larger multi-omic studies to define modifiable pathways of surveillance and intervention upstream to PreE diagnosis.

Graphical Abstract:

INTRODUCTION

Preeclampsia (PreE) is characterized by widespread vascular inflammation and increased risks for maternal and fetal morbidity and mortality.^1,2 While PreE occurs more frequently against a backdrop of metabolic dysfunction, many people who experience this condition have no antecedent health conditions.^3,4 People who experience PreE are also at increased risk for development of future cardiovascular disease (CVD) with premature onset in middle adulthood, including heart failure, coronary artery disease, and stroke.^5–9 While landmark studies have focused on the basis for PreE development through placental discovery (e.g., soluble FLT1), a less robust molecular literature exists on shared origins of PreE and long-term CVD complications. While prior studies provide evidence of cardiovascular injury at the time of PreE, they do not provide broader insight into mechanisms of PreE and later development of early-onset CVD. We hypothesized that alterations in the circulating cardiovascular proteome at delivery in patients with PreE identify proteins and pathways relevant to long-term CVD and survival. We performed a secondary analysis of prospectively enrolled cohorts across two major delivery centers. We surveyed a targeted cardiovascular proteome in pregnant people to identify a differential circulating proteome between PreE and non-PreE pregnant persons at delivery, with the goal of identifying relevant pathways of cardiovascular stress and injury to understand potentially causal upstream mechanisms of PreE.

METHODS

The authors declare that all supporting data are available within the article and its online supplementary files. This study was approved by IRBs # 222096 (VUMC), 201911157 (Wash U) and 181917 (UCSD).

Study samples

Pregnancy sample:

Subjects were prospectively enrolled from the inpatient obstetric units at two tertiary-care academic centers between 2015 and 2021. Subjects with PreE were identified from the electronic health record (EHR) based on the American College of Obstetricians and Gynecologists (ACOG) criteria.³ Subjects diagnosed at <34 weeks gestation were categorized as early PreE, and those diagnosed at ≥34 weeks were categorized as late PreE.¹⁰ Plasma samples were obtained within 48 hours prior to delivery. Contemporary normotensive subjects (“control” subjects) were identified by chart review. Individuals with a prior history of hypertension, clinical CVD (coronary artery disease, heart failure, or clinically meaningful arrhythmia), gestational hypertension (as defined by ACOG)³ or pre-term delivery were excluded from the control group. Controls were selected for this analysis based upon gestational age at phlebotomy to match the mean gestational age of individuals with PreE (Table 1). All samples were aliquoted and stored at −80°C for batched analysis.

Table 1.

Characteristics of analytic sample.

Characteristic	Non PreE, N = 91	PreE, N = 165	p-value
Age, years	32 (28, 36)	30 (25, 34)	0.012
Hispanic ethnicity	26 (29%)	40 (24%)	0.5
Race			0.3
White	43 (47%)	70 (42%
Black	13 (14%)	41 (25%)
Native American	2 (2.2%)	7 (4.2%)
Pacific Islander	1 (1.1%)	1 (0.6%)
Asian	9 (9.9%)	10 (6.1%)
Other	23 (25%)	36 (22%)
BMI, kg/m2	30 (28, 36)	34 (29, 40)	0.003
Site			<0.001
UCSD	67 (74%)	84 (51%)
Wash U	24 (26%)	81 (49%)
Gravidity	2.00 (2.00, 3.50)	2.00 (1.00, 3.00)	<0.001
Tobacco Use			0.001
Never	18 (20%)	57 (35%)
Not Current, Prior Unknown	67 (74%)	80 (48%)
Current	2 (2.2%)	11 (6.7%)
Former	4 (4.4%)	17 (10%)
Aspirin Use During Pregnancy			0.001
No	20 (22%)	59 (36%)
Yes	4 (4.4%)	22 (13%)
Unknown	67 (74%)	84 (51%)
Chronic Hypertension	0 (0%)	49 (30%)	<0.001
Pregestational Diabetes	6 (6.6%)	23 (14%)	0.076
Gestational Age at Delivery	38.60 (37.00, 39.30)	36.50 (34.38, 38.10)	<0.001
Gestational Age at Blood Sample Acquisition	36.3 (33.8, 37.5)	36.4 (34.4, 38.0)	0.4
Days between Blood Sample Acquisition and Delivery	−4 (−23, 0)	0 (−1, 0)	<0.001
Delivery Type			0.4
Spontaneous Vaginal	50 (56%)	74 (45%)
Forceps Assisted	0 (0%)	1 (0.6%)
Vacuum Assisted	3 (3.3%)	5 (3.0%)
Cesarean	37 (41%)	84 (51%)
Abruption	0 (0%)	2 (1.2%)	0.5
IUGR	7 (7.7%)	39 (24%)	0.001
Gestational Diabetes	11 (12%)	16 (9.7%)	0.6
Birth Weight (grams)	3,130 (2,762, 3,455)	2,700 (1,970, 3,100)	<0.001

Continuous variables are reported as median (25^th–75^th percentile), categorical variables are reported as n (%). Continuous variables were compared using Wilcoxon rank sum test. Categorical variables were compared using Chi-squared or Fisher’s exact test, as appropriate. Race/ethnicity represented as self-reported. *BMI = Body Mass Index, IUGR = Intrauterine Growth Restriction

Proteomics

Circulating proteomics were performed on an Olink Explore 384 Platform that prioritized proteins putatively involved in cardiovascular and metabolic processes (Cardiometabolic Explore, Olink), using standard manufacturer’s protocols. 33 samples failed during our first run (due to an instrument technical issue), of which 26 were repeated in a subsequent run, with batch correction (median normalization using bridging samples). Three proteins were excluded due to quality control warnings on the Olink panel. Given this was a broad discovery effort, below detection limit values were included as reported (as per current recommendations from Olink), and proteins were not excluded due to technical variability. Our final analytic sample consisted of 366 proteins quantified across 256 participants.

Statistical analysis

We used linear regression to define differences in a cardiovascular proteome at the time of delivery in individuals with and without PreE. Models were specified with protein (in NPX units) as the dependent variable, and PreE status as the independent variable, with nested adjustments first for site (UCSD vs. WashU¹¹) and additionally for age, race and ethnicity (social construct), chronic hypertension, gravidity, gestational age at time of blood draw, smoking status, body mass index (BMI), and gestational diabetes. The regression coefficient for the PreE term in this regression (a binary variable) signifies the adjusted effect on a protein’s normalized proteins expression (NPX) of having PreE (effectively, a fold-difference in individuals with vs. without PreE, given NPX is in log₂ units). Proteins were considered to different between cases and controls if the false discovery rate (FDR; Benjamini-Hochberg method) was <5%.

We visualized these differences across conserved pathways alongside placental expression and its changes in PreE from two sources: placental tissues in Genotype Tissue Expression (GTEx) project¹² (expressed as normalized transcripts per million, nTPM) and the Pregnancy Outcome Prediction Study (Gong et al.¹³; www.obgyn.cam.ac.uk/placentome/, accessed 14 June 2023; nearly ≈300 placental biopsies, including those with and without PreE). Proteins were mapped to their corresponding genes using Ensembl¹⁴, and relevant gene ontology (GO) pathways for each gene were manually curated using Ensembl¹⁴. Given our hypothesis that placental dysregulation of key genes central to cardiovascular dysfunction would be important in PreE, we next prioritized proteins passing a 5% FDR in our cohort (PreE versus control) and examined (1) expression of genes encoding these proteins in GTEx; (2) expression of these genes in the placenta (with minimal fragments per kilobase per million [FPKM] >0) in the Gong et. al. data; (3) log₂-fold-change of these genes in PreE compared to control pregnancies in the Gong et al. data (FDR adjusted P-value<0.1).

Given clinical differences between PreE with and without severe features (“mild” and “severe” PreE, as defined by ACOG³), we performed a sensitivity analysis of these initial model constructions using a three-level variable (mild PreE, severe PreE, control) in adjusted regression models and with unadjusted boxplot visualization (pairwise testing with post-hoc multiple comparisons using Wilcoxon method).

RESULTS

Characteristics of the study samples

The study subjects were recruited from the UCSD (59%) and WashU (41%) (Supplemental Table 1). While there was no significant difference in number of cases enrolled between the two sites (51% vs 49%), significantly more controls were enrolled from UCSD (74% vs 26%). The subjects at UCSD were slightly older (mean age 32 vs 28, P<0.001) and were significantly more likely to self-report Hispanic ethnicity (43% vs 1%, p<0.001) and less likely to self-report Black race (0% vs 51%, p<0.001). Subjects at UCSD had higher rates of pre-gestational hypertension (p = 0.01) and diabetes (p=0.002). Gestational age at plasma sample collection was earlier at UCSD due to the higher number of controls and intentional matching of gestational age at sample acquisition between cases and controls (36.0 vs 37.2 weeks, p<0.001).

Among the 165 cases, 136 (82%) had evidence of severe features. There were no significant differences in race or ethnicity between the case and control groups (Table 1). Case patients were slightly younger (30 vs 32 years, P = 0.01) and had a higher BMI at delivery (34 vs 30, p = 0.003). Consistent with differences in risk factors for and sequelae of PreE, the prevalence of chronic hypertension, diabetes and prophylactic aspirin use were higher in the cases, as well as low birth weight, placental abruption, and cesarean birth. Given that delivery is indicated in the management of PreE with severe features, the mean gestational age at delivery is lower, as expected, in the case group. As the controls were selected to match the cases in relation to approximate gestational age at blood draw, there was no significant difference in gestational age at plasma collection (p = 0.4), but the samples for the controls were drawn at a mean 4 days before delivery compared to the day of delivery for the cases.

Widespread differences in the cardiovascular proteome in PreE

We identified 142 proteins that were present at significantly different levels between PreE cases and controls (≈39% at <5% FDR; Figure 1a and b). Through GO pathway analysis, we identified an array of pathways that were enriched for differentially expressed proteins that are broadly relevant to the pathobiology of CVD, including: (1) extracellular matrix metabolism (MMP7¹⁵, DCN¹⁶, cathepsins¹⁷, COL6A3¹⁸, ADAMTS16¹⁹); (2) angiotensin pathway mediators and vascular function (ACE2, ANPEP²⁰, ACTA2^21,22); (3) innate immune response (CCL15, CXCL8, FCN2²³); (4) hemostasis, angiogenesis, endothelial biology (vWF, PLAT, TIE1²⁴); and (5) direct cardiovascular stress and injury (N-terminal pro-B type natriuretic peptide [NPPB], troponin [TNNI3]^25,26). Most associations were robust to adjustment (for site alone or with a full adjustment, Supplemental Figure 1).

Figure 1.

Global differences in the cardiovascular proteome during PreE. Panel A shows a volcano plot of differences between PreE and control cases. Positive values (right of zero) on the x-axis identify those proteins that are higher in individuals with PreE. Negative values (left of zero) on the x-axis identify those proteins that are higher in control individuals. Blue data points indicate FDR <5%. These results represent models with full adjustment (see Methods). Panel B shows a clustered heatmap of the top 50 proteins across all participants (unadjusted, raw NPX values). While more PreE cases appear to be clustered toward the left of the heatmap (consistent with upregulation of circulating cardiometabolic proteins), we observed significant heterogeneity in these proteins across cases and controls. Panel C shows mapping of the circulating proteome to placental tissues. Significant proteins (PreE versus control <5% FDR) were classified according to manually curated pathways in the exterior layer. The bar plot in the second outer layer denotes the respective coefficients in the association model. The heatmap showed the expression levels of these proteins in the placenta based on GTEx (exterior layer), as well as the protein expression level (middle layer) and relative fold change in subjects with or without PreE (inner layer) from Gong et al. Areas shaded grey in the heatmap indicated proteins not reported in the corresponding dataset.

While several key mediators identified here have been implicated previously in PreE (FCN2²³, DCN²⁷, TIE1²⁸, NID1²⁹, CD59³⁰), many identified gene products have not been extensively reported to be altered in maternal circulation or placental tissue in PreE, despite having mechanisms relevant to CVD pathogenesis¹³ (Figure 1c). We observed several general themes: First, molecules not widely previously reported in PreE were frequently related to extracellular matrix, angiogenesis, or fibrotic pathways, which are pathways that are common between acute vascular stress and chronic vascular remodeling. Second, the directionality of the plasma level in PreE and the putative mechanistic effects on long-term cardiovascular phenotypes were mixed: for example, SOST—an osteocyte product where higher levels have been associated with inhibition of atherosclerosis and vascular stiffness—exhibited higher levels in acute PreE, with a potentially protective effect on long-term atherosclerosis mechanisms³¹. Importantly, many of these proteins did not exhibit differential expression at the mRNA level in placental tissue in PreE cases in a prior landmark study¹³, despite altered plasma protein abundance here, suggesting that they may be expressed by tissues other than the placenta or that their protein levels may not be regulated at the level of RNA abundance (Figure 1c).

Differences in the cardiovascular proteome by early and late PreE

Given potential clinical-biological differences in individuals who are diagnosed with late late (≥34-week gestation) versus early (<34-week gestation) PreE, we fit models including early and late PreE diagnosis (relative to control) to explore the relative strength of association between a given protein and disease subtype. While we observed a broadly similar pattern in early vs. late PreE, several proteins displayed larger differences in early PreE relative to late PreE, including key biomarkers of hemodynamic stress linked to CVD (e.g., natriuretic peptides, NPPB), hemostasis (PLAT), and immune response (IL1RL1, IL6; Supplemental Figure 2).

DISCUSSION

Despite putative origins at the maternal-fetal vascular interface, PreE is a complex, heterogeneous, systemic disorder with broad physiologic impact across multiple systems¹⁰. PreE is consistently recognized as a risk factor for development and progression of CVD, with many considering PreE to be a failed “pregnancy vascular stress test” that unmasks subclinical disease¹⁰. While the study of PreE has often focused on placental biology and longitudinal epidemiology, ^32,33 systemic abundances of cardiometabolic molecular phenotypes at the time of the peak PreE disease state may provide insights into overlapping pathophysiological mechanisms contributing to future CVD.

Through our study of the cardiometabolic proteome in subjects with PreE, we identified significant differences in plasma abundance of key proteins along canonical pathways of CVD. As anticipated, the greatest differences in circulating CV proteins were observed between subjects with PreE and controls. Despite an expected generally consistent pattern across the cardiometabolic proteome between early and late PreE, several proteins (specifically those implicated in cardiovascular stress and inflammation) were higher in individuals with early development of PreE (<34-week gestation). This supports clinical observations that early PreE is associated with greater maternal vascular compromise and downstream CVD.^10,34 Taken together, these data provide support for the shared origins of disease between PreE and CVD which have previously been suggested in epidemiologic studies. They also substantiate the prospect for targeted upstream prevention efforts by subphenotyping patients with PreE and their individual downstream CVD risk.

Consistent with the known systemic nature of PreE, many of the circulating cardiometabolic proteins we identified as differentially expressed in PreE did not localize to the placental transcriptome in a recent landmark study (Figure 1). These findings could result from between-site heterogeneity in study participants, residual confounding from pre-existing disease/risk factors, or of the focus in this study on proteomic targets known to be associated with CVD in non-pregnant populations. It may also support a clinical hypothesis that PreE represents a systemic disorder, with multiple proteins arising as an extra-placental response to a placental disorder.

While early biomarker work at the intersection of PreE and CVD focused on role of angiogenic mediators in maternal vascular phenotypes (sFLT1)^35–38, more recent efforts have identified a potential shared genetic susceptibility between PreE, obesity and hypertension^39,40. Consistent with the findings of this report, both genetic and non-genetic approaches have previously highlighted the role of increased oxidative stress, innate immune activation, and endothelial and myocardial injury in PreE.^41–46 Thus, our work adds to the prior literature by strengthening the evidence for critical pathophysiological overlap between these diseases despite a long time horizon for clinical CVD development after PreE.

The current work is unique in two major aspects: (1) conducting circulating proteomics analysis in a diverse population with PreE at high cardiometabolic risk; and (2) exploring broader implications on CVD through pathway analysis of differentially expressed proteins. Overall, the expressed proteome in human subjects with PreE in this study is consistent with results of animal models of PreE that identify development of myocardial dysfunction and fibrosis, enhanced postpartum vascular response to injury, and persistent alteration in the circulating cardiovascular proteome.^36,47–49 Further, we identified several novel proteins that have not been strongly associated with PreE in the literature, but with putative roles in CVD pathogenesis.

There are several limitations to our analysis. While we adjusted for site in our analyses, there is potential for heterogeneity across enrolled pregnant persons, which may result in residual confounding. Our limited sample size prevents accurate examination of differences across, race, ethnicity, or ancestry and larger studies are needed. We did not have longitudinal proteomic quantification across all study participants prior to, during, and after pregnancy. Serial analysis or circulating proteomics is needed to understand how these findings relate to future CVD development. The concept of examining an acute “challenge” or provoked state to prioritize functional biomarkers in a more chronic disease state is widely used in CVD biology^50,51, though whether PreE reflects preexisting vascular dysfunction or heralds the onset and persistent differential abundances in proteins leading to increased CVD risk requires further study. We recognize that transcripts in the placenta may not necessarily be a source for the circulating cardiovascular proteome. Larger proteomic studies in human placenta across disease state are needed to continue to clarify a role for the placenta in promoting CVD. Other minor limitations include a slightly larger time gap between sample collection and delivery for cases vs. controls and selection of control samples at an earlier gestational age to allow for gestational age matching of the sample acquisition time.

In conclusion, in this study, we identify broad differences in the plasma abundance of a cardiovascular proteome during the peak manifestations of PreE shortly before delivery across major pathways of angiogenesis, vascular function, inflammation, metabolism, and wound healing. Human genetics and curation identified several PreE-associated proteins relevant to long-term CVD phenotypes in this set, suggesting key links between acute responses to vascular stress during PreE and long-term CVD. Further work to dissect potential functional biology of the cardiovascular proteome in broader sample sets and human placental tissue are likely to provide additional critical insight into why and how PreE exacts an early CV toll on certain women but not others.

Highlights.

• Patients with preeclampsia exhibit significantly differences in plasma abundances of relevant cardiovascular proteins at the time of delivery, compared to normotensive patients.

• Differences in circulating proteins in patients with preeclampsia are associated with cardiovascular disease pathways including angiogenesis, blood pressure, cell adhesion, inflammation, and metabolism.

• GO pathway analysis of placental tissue gene expression identified novel gene products relevant to CVD both within and outside of the placenta.

• Despite broad overlap in circulating proteins between patients with early (<34 weeks) vs late preeclampsia, differential expression in biomarkers associated with cardiovascular stress, hemostasis and immune response.

Abbreviation Definition

PreE

Preeclampsia

CVD

Cardiovascular Disease

FLT1

FMS-like Tyrosine Kinase 1

VUMC

Vanderbilt University Medical Center

Wash

Washington University

UCSD

University of California San Diego

HER

Electronic Health Record

ACOG

American College of Obstetricians and Gynecologists

NPX

Normalized Proteins Expression

BMI

Body Mass Index

FDR

False Discovery Rate

GTEx

Genotype Tissue Expression

nTPM

Normalized Transcripts Per Million

GO

Gene Ontology

MMP7

Matrix Metabolism Protein 7

DCN

Decorin

COL6A3

Collagen Type VI Alpha 3 Chain

ADAMTS16

A Disintegrin and metalloproteinase with thrombospondin motifs 16

ACE2

Angiotensin Converting Enzyme 2

ANPEP

Alanyl Aminopeptidase

ACTA2

Actin alpha 2

CCL15

Chemokine Ligand 15

CXCL8

Interleukin 8

FCN2

Ficolin 2

vWF

Von Willebrand Factor

PLAT

Plasminogen Activator, Tissue Type

TIE1

Tyrosine Kinase with Immunoglobulin-like and EGF-like Domains 1

NPPB

Natriuretic Peptide B

TNNI3

Troponin I3

NID1

Nidogen 1

SOST

Sclerostin

IL1RL1

Interleukin 1 Receptor-like 1

IL6

Interleukin 6