An Exploratory Study of White Blood Cell Proportions Across Preeclamptic and Normotensive Pregnancy by Self-Identified Race in Individuals with Overweight or Obesity

Abstract

Objective:

Examine white blood cell (WBC) proportions across preeclamptic (n=28 cases) and normotensive (n=28 controls) pregnancy in individuals with overweight or obesity.

Methods:

WBC proportions were inferred from genome-wide DNA methylation data and compared by case/control status and self-identified race.

Results:

Mean B cell proportions were suggestively lower in cases in Trimester 1 in the overall sample and significantly lower in Trimester 1 in White participants but not in Black participants. More significant WBC proportion changes were observed across normotensive than preeclamptic pregnancy.

Conclusions:

These findings in a small sample demonstrate need for additional studies investigating the relationship between self-identified race and WBCs in pregnancy.

Introduction

Preeclampsia (PE) is a leading cause of maternal and prenatal morbidity, affecting between 2.3% and 3.8% of pregnancies in the United States, and 4.6% of pregnancies globally [1,2]. Individuals who develop PE are 3–4 times more likely to develop hypertension and have double the risk of experiencing a myocardial infarction or stroke later in life [3]. There are presently no therapeutic agents available for individuals who develop PE, and the sole treatment remains birth of the placenta.

The lack of available treatment options for PE patients is a direct consequence of our ongoing, limited understanding of its pathophysiology. PE development highlights the intersection between inflammation and angiogenesis, two directly linked biological processes [4–9]. One potential mechanism of PE proposed in the literature is that inappropriate angiogenesis results in hypoxia, which drives inflammation, fueling a vicious cycle of chronic inflammation, dysregulated angiogenesis, hypoxia or ischemia-reperfusion injury, and systemic endothelial dysfunction [10–12]. A commonality of each of these pathophysiological hallmarks is the involvement of white blood cells (WBCs) of the immune system.

WBCs are known to be elevated in individuals across pregnancy [13–15]. Because the pathophysiology of PE is believed to be driven largely by inflammation, it is possible that individuals who develop PE during pregnancy have WBC values that differ from individuals who experience uncomplicated pregnancy. Characterization of such differences could provide insight into biological underpinnings of PE and potentially be adapted as a non-invasive, clinical tool in pregnancy to screen for risk of PE. Additionally, WBC values have been shown to vary across race in healthy individuals [16]. As such, it is possible that changes in WBCs during pregnancy may also differ with race. Previously conducted studies evaluating differences in specific WBC values in individuals who develop PE compared to normotensive individuals are sparse, cross-sectional, and do not address race, representing a gap in PE literature [17–19].

Typically, quantification of WBCs for biological research is performed by flow cytometry or through patient chart review. Flow cytometry requires immediate sample preparation while chart review can only be done with approved access to clinical laboratory values where a complete blood count with differential was completed. In cases where epigenome-wide DNA methylation exist for clinical or research purposes, an additional under-utilized method is the Houseman approach, which allows for WBC proportions to be inferred from differentially methylated regions that serve as unique signatures for specific cell types [20]. To date, we have not identified any studies that have characterized WBC proportions estimated in this manner across pregnancy. Therefore, the purpose of this study was to characterize WBC proportions estimated from epigenome-wide DNA methylation data across three trimesters of pregnancy in individuals with overweight or obesity who developed PE or remained normotensive throughout their uncomplicated pregnancy. We hypothesized that WBC proportions for individuals who developed PE would differ from those with normotensive pregnancy, reflecting a chronic inflammatory state.

Materials and Methods

Study design, setting, and sample

This was a secondary analysis of an exploratory, case-control study of individuals with preeclamptic or normotensive pregnancy that leveraged biospecimens and phenotype data from a larger prospective cohort study funded by the National Institute of Child Health and Human Development, entitled “Prenatal Exposures & Preeclampsia Prevention Project 3 (PEPP3): Mechanisms of Preeclampsia and the Impact of Obesity” [P01 HD30367]. The purpose of PEPP3, which took place at University of Pittsburgh Medical Center Magee-Womens Hospital from 2008–2014, was to identify factors associated with PE development and obesity. As such, there was an emphasis on the recruitment of participants with overweight or obesity, focusing on those with a body mass index (BMI) ≥25 kg/m². As part of PEPP3, participants were recruited if they were 14 to 40 years of age, had a singleton pregnancy, and had no past history of a medical condition that heighted the risk of PE (e.g., chronic renal disease, diabetes). Participants were followed across pregnancy and detailed demographic, social, and clinical data were collected, and biospecimens were stored.

PE was defined as (1) new onset of gestational hypertension after gestational week 20 in previously normotensive individuals and (2) the presence of proteinuria [21]. Gestational hypertension was defined as new onset of elevated blood pressure (systolic BP ≥140 mmHg and/or diastolic BP ≥90 mmHg) that reverted to baseline by 12 weeks postpartum. The mean of the previous four blood pressures measured upon admission to labor and birth, prior to any therapeutic intervention, was compared to mean blood pressures determined before 20 weeks’ gestation. Proteinuria was defined as (1) ≥300 mg over 24 hours, (2) ≥0.3 protein/creatinine ratio, (3) ≥2+ on a random urine dipstick, or (4) ≥1+ on a catheterized urine specimen. Participants who served as the control phenotype remained normotensive throughout the entirety of pregnancy, did not develop proteinuria, and had an uncomplicated pregnancy outcome. All study protocols were approved by the University of Pittsburgh Institutional Review Board, and we have adhered to all ethical considerations in the protection of human subjects.

From the PEPP3 cohort, 56 participants (28 PE cases and 28 normotensive controls) were selected for inclusion in an ancillary genome-wide association study (R21HD092770) if they (1) had DNA samples available for all three trimesters of pregnancy (to facilitate epigenome-wide DNA methylation data collection) and (2) could be individually (i.e., 1:1) matched on pre-pregnancy BMI, gestational age at sample collection, self-identified race, and self-reported smoking status (to limit confounding).

Biospecimen Sampling, DNA Extraction, and DNA Methylation Data Collection

As part of the PEPP3 study, a peripheral blood sample was collected in EDTA plasma tubes from each participant in the first, second, and third trimester of pregnancy (one sample/trimester). For the subset of participants analyzed here, nuclear DNA was extracted using protein precipitation methods [22] and epigenome-wide DNA methylation data were collected using Infinium® MethylationEPIC Beadchips (Illumina, San Diego, CA, USA) at the Johns Hopkins University Genetic Resources Core Facility SNP Center (data available from dbGAP, accession number: phs001937.v1.p1). As part of our laboratory quality control, 28 technical replicates were included. To mitigate potential batch effects, we attempted to place all samples from each participant on the same chip. To balance potential row and column effects, our sample layouts had an equal number of cases and controls in the columns of each chip and a zigzag pattern was utilized from row to row, alternating between cases and controls.

Our epigenome-wide DNA methylation data cleaning and quality control pipeline was carried out in R using the minfi, ENmix, and funtooNorm packages [23–25]. This pipeline included removal of poorly performing and outlying samples and probes, as well as functional normalization as described [23–27]. Our final data set consisted of 703,200 probes available in 56 participants at up to 3 trimester-specific time points.

Estimation of WBC Proportions

WBC proportions were estimated from the epigenome-wide DNA methylation data using Houseman’s reference-based approach implemented in the ‘estimatecellcounts2’ function from the FlowSorted.Blood.EPIC package in R [20,28]. An external reference data set of epigenome-wide DNA methylation data generated from isolated and pure cell subtypes was used to infer WBC proportions in our sample [28]. The reference data set was derived from blood samples obtained from healthy individuals who had no history of significant or chronic health issues and consisted of samples from 31 male and 6 non-pregnant female donors who had a mean age of 32.6 years and average weight of 86 kg (189.2 pounds) [28].

To estimate WBC proportions, we used 450 CpG probes from the EPIC array that were previously identified as being differentially methylated based on cell type by the Identifying Optimal Libraries (IDOL) algorithm [28]. Of those 450 IDOL probes, DNA methylation data were available for 384 probes in our post-quality control data set. Our approach yielded proportion data for six cell types including neutrophils, monocytes, B cells, natural killer (NK) cells, CD4+ T cells, and CD8+ T cells. Reference data for basophils and eosinophils were not available and therefore estimation of these cell types was not possible. However, proportions and variability of eosinophils and basophils are not typically estimated using this method, presumably as they contribute the least amount of variability in WBC proportions [29].

Statistical Analyses

Statistical analyses were conducted using R version 3.6.0 (R Core Team, Vienna, Austria) and SPSS version 23 (IBM, Chicago, IL). Demographic and clinical characteristics were compared between groups using t-tests or Pearson’s chi square test. WBC proportions were characterized for cases and controls using means and standard deviations (SD) for the entire sample and stratified by self-identified race.

Next, Houseman’s regression calibration approach with a double bootstrap bias estimate was used to formally test differences between WBC proportions in cases versus controls [20]. DNA methylation data for 384 IDOL probes described above were extracted from the external reference data set and our genome-wide DNA methylation data. We then used a linear mixed effects model with a random intercept for batch effects to simultaneously estimate all WBC proportion differences by regressing our WBC proportion matrix on case/control status at trimester-specific time points. This method employs a robust double bootstrap procedure in which both the reference data set and the test data set are sampled with replacement 250 times to compute a standard error estimate used to assess measurement bias, subsequently accounting for variation and noise in WBC proportions estimated from DNA methylation data in small data sets [20]. Within each trimester we computed double bootstrap bias-adjusted regression coefficient estimates, which can be interpreted as the difference in WBC proportions as percentages in cases versus controls, as well as double bootstrap standard errors, 95% confidence intervals (CI), and p-values. Analyses were performed for the entire sample, stratified by self-identified race, and the results were examined graphically using box and whisker plots. To further explore the influence of self-identified race, double bootstrap regressions were also performed while controlling for self-identified race as a covariate. Raw p-values <0.05 were considered suggestive. Correction for multiple testing was made by calculating false discovery rate (FDR) adjusted p-values using the ‘p.adjust’ function from the tidyverse package in R, with adjusted p-values <0.05 being considered significant [30]. We corrected for nine tests per cell type.

Finally, changes in WBC proportions across pregnancy in PE cases and controls were graphically examined using spaghetti plots and box and whisker plots for the entire sample and stratified by self-identified race. We used paired t-tests to formally test differences in WBC proportion changes between Trimesters 1 and 2, Trimesters 2 and 3, and Trimesters 1 and 3. Although these tests were performed post hoc to further characterize variation in cell types between cases and controls, we provide FDR adjusted p-values, calculated as described as above and correcting for all 108 tests (18 tests per cell type for 6 cell types).

Results

Sample Demographics and Clinical Characteristics

Sample demographic and clinical characteristics are presented in Table 1. Our overall sample consisted of 56 participants (28 cases and 28 controls) and the majority of participants self-identified their race as Black (n=42, 75%). By study design, mean (±SD) pre-pregnancy BMI for cases and controls was in the obese range at 33.0 (±7.5) and 33.7 (±7.5), respectively. Between cases and controls, there were no significant differences in maternal age at birth, pre-pregnancy BMI, gestational age at sample collection, proportion of nulliparity, average systolic blood pressure prior to gestational week 20, and proportion of lifetime smokers. Expected significant differences between cases and controls included (1) gestational age at birth (p<0.001), (2) average diastolic blood pressure prior to gestational week 20 (p=0.048), (3) average systolic and diastolic blood pressures in labor (p<0.001), and (4) birthweight of infants (p=0.001).

Table 1.

Demographics and Clinical Characteristics of Sample

Characteristics	Case (PE +)	Control (PE −)	p
n	28	28	-------
Self-Identified Race: Black Participants	21 (75%)	21 (75%)	-------
White Participants	7 (25%)	7 (25%)	-------
Maternal Age at Birth, years	23.94 ± 5.00	23.66 ± 4.26	0.820a
Gestational Age at Birth, weeks	37.38 ± 2.45***	39.52 ± 1.31	<0.001a
Pre-pregnancy BMI, kg/m 2	32.99 ± 7.49	33.68 ± 7.52	0.734a
Gestational Age at Sample Collection, weeks: Trimester 1	8.46 ± 1.72	8.57 ± 1.75	0.817a
Trimester 2	19.74 ± 1.63	19.62 ± 0.79	0.752a
Trimester 3	37.07 ± 2.45	37.41 ± 2.48	0.612a
Nulliparous, n (%): Yes	21 (75%)	24 (85.71%)	0.313b
Average SBP <20 Weeks, mmHg	112.79 ± 7.27	111.54 ± 8.89	0.567a
Average DBP <20 Weeks, mmHg	70.11 ± 5.60*	66.82 ± 6.13	0.048a
Average SBP upon admission to L&D, mmHg	148.29 ± 6.91***	124.54 ± 6.06	<0.001a
Average DBP upon admission to L&D, mmHg	90.54 ± 8.10***	71.01 ± 6.25	<0.001a
Birthweight (grams)	2797 ± 732.79**	3385.61 ± 540.80	0.001a
Lifetime Smoking Status, n (%): No	18 (62.29%)	19 (67.86%)	0.778b

Values represent indicated units ± standard deviation. p= p-value. BMI= body mass index. SBP= systolic blood pressure. DBP= diastolic blood pressure. mmHg= millimeters of mercury. L&D= labor and delivery.

^aTwo sample t-test.

^bPearson Chi-Square test.

^*p < 0.05

^**p < 0.005

^***p < 0.001.

WBC proportions in Cases versus Controls

Mean (±SD) proportions for estimated WBC proportions in cases and controls are presented in Supplemental Table 1. Neutrophils comprised the highest proportion of WBCs at any given time point for both groups, regardless of self-identified race. The remaining five estimated cell types accounted for smaller proportions, with CD4+ T cells representing the second largest proportion, followed by CD8+ T cells, monocytes, B cells, and NK cells.

The results of the double bootstrap regression formally testing WBC proportion differences between cases and controls for the overall sample and stratified by self-identified race are presented in Table 2 and depicted graphically in Supplemental Figure 1. For the overall sample in Trimester 1, the mean proportion of B cells was suggestively lower in cases compared to controls, with an estimate of −1.03% (95% CI=−1.87% to −0.19%, p_raw=0.02, p_FDR=0.09). In the subsample of participants who self-identified their race as White, this difference was significant with an estimate of −2.26% (95% CI=−3.53% to −0.09%, p_raw=0.001, p_FDR=0.009). In participants who self-identified their race as Black, the proportion of B cells in cases was lower than controls, but not statistically significantly. Group differences in B cell proportions in Trimesters 2 and 3 were not significant in the overall sample, or among results stratified by race. No significant differences in the proportions of monocytes, NK cells, CD4+ T cells, or CD8+ T cells were identified.

Table 2.

Regression Coefficient Estimates, Representing Differences in WBC Proportions, in Percentages, in PE Cases Compared to Normotensive Controls

Trimester 1
	Overall Sample, n=50					Black Participants, n=37					White Participants, n=13
	Est	SE	95% CI	praw	pFDR	Est	SE	95% CI	praw	pFDR	Est	SE	95% CI	praw	pFDR
(Intercept, γ0)	0.08	0.28	−0.46 to 0.62	0.77	0.78	0.23	0.34	−0.43 to 0.89	0.46	0.78	−0.46	0.43	−1.30 to 0.39	0.29	0.78
Neutrophils	2.90	2.10	−1.22 to 7.02	0.20	0.60	1.64	2.66	−3.57 to 6.85	0.55	0.80	5.32	2.47	0.47 to 10.17	0.04*	0.18
Monocytes	0.09	0.50	−0.90 to 1.07	0.84	0.95	0.00	0.67	−1.31 to 1.31	0.95	0.95	0.47	0.69	−0.88 to 1.83	0.43	0.65
B cells	−1.03	0.43	−1.87 to −0.19	0.02*	0.09	−0.50	0.49	−1.46 to 0.45	0.34	0.64	−2.26	0.65	−3.53 to −0.99	0.001*	0.009**
NK cells	0.28	0.37	−0.44 to 0.99	0.35	0.69	0.62	0.42	−0.20 to 1.45	0.14	0.68	−0.45	0.60	−1.63 to 0.73	0.43	0.69
CD4+ T cells	−1.05	1.11	−3.24 to 1.13	0.37	0.71	−0.91	1.39	−3.64 to 1.82	0.55	0.71	−1.46	1.50	−4.40 to 1.48	0.36	0.71
CD8+ T cells	−1.16	0.79	−2.70 to 0.38	0.16	0.36	−0.96	0.98	−2.87 to 0.96	0.31	0.56	−1.02	0.53	−2.06 to 0.01	0.05	0.24
Trimester 2
	Overall Sample, n=53					Black Participants, n=40					White Participants, n=13
	Est	SE	95% CI	p raw	p FDR	Est	SE	95% CI	p raw	p FDR	Est	SE	95% CI	p raw	p FDR
(Intercept, γ0)	−0.46	0.29	−1.03 to 0.11	0.12	0.59	−0.27	0.39	−1.04 to 0.49	0.47	0.78	−0.64	0.42	−1.46 to 0.18	0.13	0.59
Neutrophils	0.53	1.57	−2.54 to 3.60	0.80	0.80	1.40	1.83	−2.19 to 5.00	0.35	0.79	−4.86	2.18	−9.14 to −0.58	0.02*	0.18
Monocytes	0.60	0.45	−0.28 to 1.49	0.19	0.47	0.52	0.52	−0.51 to 1.55	0.31	0.56	0.91	0.68	−0.42 o 2.23	0.17	0.47
B cells	−0.07	0.31	−0.69 to 0.54	0.85	0.85	−0.07	0.34	−0.74 to 0.59	0.65	0.81	−0.21	0.33	−0.86 to 0.43	0.51	0.77
NK cells	−0.23	0.32	−0.87 to 0.40	0.46	0.69	−0.31	0.38	−1.06 to 0.44	0.39	0.69	0.55	0.39	−0.22 to 1.32	0.15	0.68
CD4+ T cells	−0.22	0.76	−1.71 to 1.27	0.89	0.95	−0.89	0.94	−2.74 to 0.96	0.27	0.71	2.53	1.32	−0.06 to 5.12	0.06	0.54
CD8+ T cells	−0.09	0.60	−1.26 to 1.08	0.93	0.98	−0.29	0.66	−1.57 to 0.98	0.59	0.76	1.63	0.89	−0.12 to 3.39	0.07	0.24
Trimester 3
	Overall Sample, n=53					Black Participants, n=39					White Participants, n=14
	Est	SE	95% CI	p raw	p FDR	Est	SE	95% CI	p raw	p FDR	Est	SE	95% CI	p raw	p FDR
(Intercept, γ0)	−0.18	0.39	−0.96 to 0.59	0.60	0.78	−0.10	0.37	−0.82 to 0.62	0.69	0.78	0.16	0.50	−0.81 to 1.13	0.78	0.78
Neutrophils	0.45	1.97	−0.34 to 4.32	0.78	0.80	−0.76	2.08	−4.84 to 3.32	0.80	0.80	2.54	3.87	−5.06 to 10.13	0.49	0.80
Monocytes	−0.73	0.56	−1.82 to 0.36	0.17	0.47	−0.88	0.71	−2.26 to 0.51	0.21	0.47	−0.27	0.73	−1.71 to 1.17	0.68	0.87
B cells	0.14	0.36	−0.57 to 0.85	0.72	0.81	0.42	0.41	−0.38 to 1.23	0.36	0.65	−0.63	0.34	−1.29 to 0.03	0.06	0.18
NK cells	−0.14	0.32	−0.77 to 0.48	0.67	0.75	−0.04	0.36	−0.75 to 0.66	0.91	0.91	−0.24	0.45	−1.13 to 0.64	0.57	0.73
CD4+ T cells	0.64	1.02	−1.36 to 2.63	0.51	0.71	0.95	1.00	−1.02 to 2.92	0.40	0.71	−0.12	2.89	−5.79 to 5.55	0.95	0.95
CD8+ T cells	0.02	0.65	−1.26 to 1.30	0.98	0.98	0.54	0.73	−0.90 to 1.97	0.48	0.72	−1.42	0.79	−2.97 to 0.12	0.08	0.24

Est= Double-bootstrap regression coefficient estimate. SE= Double-bootstrap standard error. CI= Confidence Interval. p_raw= raw p-value. p_FDR = False Discovery Rate adjusted p-value. p_FDR reflects multiple testing correction by cell type (9 tests per cell type).

^*suggestive association (p_raw < 0.05)

^**significant association (p_FDR < 0.05).

These data are presented graphically in Supplemental Figure 1.

In Trimester 1, within the subset of participants who self-identified their race as White, the mean proportion of neutrophils was suggestively higher in cases compared to controls, with an effect size estimate of 5.32% (95% CI=0.47% to 10.17%, p_raw=0.04, p_FDR=0.18). Interestingly, in Trimester 2, the mean neutrophil proportions in White participants had an estimate of −4.86% (95% CI=−9.14% to −0.58%, p_raw=0.02, p_FDR=0.18), suggesting a lower proportion in cases compared controls, followed by no significant difference in Trimester 3. No significant differences in the mean proportion of neutrophils between cases and controls were observed in the overall sample or among the subgroup who self-identified their race as Black. The results of the double bootstrap regression formally testing WBC proportion differences between cases and controls while controlling for self-identified race are presented in Supplemental Table 2. When controlling for self-identified race, the mean proportion of B cells was suggestively lower in cases compared to controls, with an estimate of −0.94% (95% CI=−1.74% to −0.10%, p_raw=0.03, p_FDR=0.09).

WBC proportions Across Pregnancy

Longitudinal changes in WBC proportions are displayed as box and spaghetti plots in Figure 1 and Supplemental Figure 2, respectively. Interestingly, comparison of proportions of neutrophils, monocytes, B cells, CD4+ T cells, and CD8+ T cells demonstrated more significant changes across normotensive pregnancy than across preeclamptic pregnancy. The mean proportions of neutrophils increased significantly from Trimester 1 to Trimester 2 in normotensive controls in the overall sample, as well as in Black and White participants in race-stratified results. However, changes in increasing mean neutrophil proportions from Trimester 1 to Trimester 2 were not significant in cases in the overall sample, or in Black and White participant groups separately. Monocytes were the most static cell type across pregnancy, with few significant changes by trimester among normotensive controls and no significant changes among preeclamptic cases.

Figure 1. Estimations of WBC Proportions at the Trimesters of Preeclamptic or Normotensive Pregnancy, Overall and Stratified by Self-identified Race.

Dots represent proportions for individual participants, represented as percentages; lower and upper hinges correspond to the first and third quartiles (25th and 75th percentiles); dashed, horizontal line indicates the mean; solid, horizontal line indicates the median; whiskers extending from the hinge represent the largest and smallest values no further than 1.5 times the interquartile range; data beyond the whiskers represent outliers;; p_FDR reflects multiple testing correction for all 108 tests; paired t-test: p_FDR graphically depicted as: *p_FDR < 0.05, **p_FDR < 0.005, ***p_FDR < 0.0005

B cell proportions significantly decreased from Trimester 1 to Trimester 2 and across pregnancy (from Trimester 1 to Trimester 3) in normotensive pregnancy in the overall sample and in race-stratified results. The same significant changes were observed in preeclamptic pregnancy for the overall sample and for Black participants. However, participants who self-identified as White had no significant changes in B cells across preeclamptic pregnancy.

CD4+ T cell proportions decreased significantly with each trimester and across pregnancy in normotensive pregnancy in the overall sample and in race-stratified results. However, the only significant change in preeclamptic pregnancy was observed in the overall sample, across pregnancy. Similarly, CD8+ T cell proportions decreased significantly from Trimester 1 to Trimester 2 and across normotensive pregnancy in the overall sample and in race-stratified results. However, the only significant difference in preeclamptic pregnancy was seen among self-identified White participants across pregnancy. In contrast, changes in mean NK cell proportions did not appear to demonstrate any striking differences by case/control status across pregnancy.

Discussion

This is the first study to characterize WBC proportions estimated from epigenome-wide DNA methylation data across pregnancy in PE cases and controls. In our sample, neutrophils represented the largest proportion of the estimated cell types, and our exploratory findings demonstrated a steady increase in the mean proportions of neutrophils for both PE cases and controls (Supplemental Table 1, Supplemental Figure 2). This finding is consistent with the literature, which suggests that the increase in WBCs observed in pregnancy is due to an increase in the number of neutrophils [14,31,32]. We observed minimal variations in monocytes in both PE cases and normotensive controls (Supplemental Figure 3). Monocytes have previously been shown to remain relatively static across healthy pregnancy [14], and no studies were found in which monocytes were measured across preeclamptic pregnancy. We did not come across any comparable studies evaluating longitudinal changes in B cells, NK cells, CD4+ T cells, and CD8+ T cell proportions, or values across preeclamptic or normotensive pregnancy for comparison.

Although the sample size for this study was limited, the results highlight potential differences in WBC proportions by self-identified race, in the context of PE. In our analyses comparing PE cases to normotensive controls, in the overall sample, we found a suggestively lower mean proportion of B cells among cases during Trimester 1 (Table 2, Supplemental Figure 1). Interestingly, in our race-stratified results we found this result to be significant among White participants, who represented a smaller proportion of the overall sample. We also found suggestive differences in neutrophil proportions among White participants during Trimesters 1 and 2 (Table 2, Supplemental Figure 1).

In addition to B cells, we estimated proportions for three additional cell types of lymphoid origin including NK cells, CD4+ T cells, and CD8+ T cells, which did not differ between PE cases and controls. Previous studies have demonstrated markedly lower lymphocyte counts in PE patients compared to controls in the first trimester and at term [14,17,18]. A limitation of these studies is that by measuring all “lymphocytes” they cannot discern which particular cell type(s) are responsible for the reduction. Our data suggest that a lower proportion of B cells may be driving the marked reduction reported in lymphocytes in cases in these studies. In pregnancy, B cells have been shown to produce protective antibodies that are directed toward paternal antigens and associated with positive pregnancy outcomes [33]. It is possible that during the first trimester, a reduction in B cells yields insufficient protective antibodies, resulting in a reduction in the immune system’s ‘tolerance’ of the fetus in early pregnancy. These data conflict with the findings reported by Örgül et al. [21] who found no differences in lymphocytes in the first trimester. In the current study, we found B cell proportions to be statistically lower in cases among White participants only, with no statistically significant differences among Black participants. It is possible that these conflicting results can be attributed to genetic differences related to ancestry [34]. Unfortunately, Örgül et al. [21] did not provide racial breakdown for comparison. However, lower B cell proportions among cases compared to controls was not statistically significant when controlling for self-identified race as a covariate (Supplemental Table 2).

Our race-stratified results were suggestive of greater proportions of neutrophils in cases in Trimester 1, as well as a lower proportion of neutrophils in cases in Trimester 2, among White participants only. We did not find statistically differences between cases and controls among Black participants. Also, in the overall sample, no differences were observed (Table 2, Supplemental Figure 1). Previous investigations exploring differences in neutrophil values between PE cases and normotensive controls have reported conflicting results. Lurie et al. [20] found neutrophils to be elevated among both mild (n=30) and severe (n=16) PE cases compared to controls (n=46) at the onset of labor and Örgül et al. [21] reported elevated neutrophils among both early-onset (n=21) and late-onset (n=42) PE cases compared to controls (n=123), during the first trimester. In contrast, Kirbas et al. [19] reported no differences in neutrophils in neither mild (n=288) nor severe (n=326) PE cases compared to controls (n=320) during the first trimester. It should be noted that these studies were performed in samples comprised of individuals with a pre-pregnancy BMI in the normal range, while our sample consisted of mostly individuals with overweight or obesity. Again, a potential explanation for the discrepancies across the literature and our results may be differences in racial demographics of the samples. Unfortunately, the authors of the three papers discussed above did not provide the racial distribution of their samples for comparison.

An additional surprising finding from our study was a lower proportion of neutrophils in cases compared to controls among White participants during Trimester 2, a direct contrast in the direction of effect observed in Trimester 1 (Table 2, Supplemental Figure 1). While this suggestive association may reflect the dynamic nature of inflammation and angiogenesis in PE, it is more likely a reflection of our small sample size of White participants (n=14) and should be considered with caution until examination in a larger sample can examine the result.

Strengths and Limitations

There are multiple strengths to this study, including its rigorous case/control selection, longitudinal approach, and stringent clinical criteria used to precisely define the case and control phenotypes. An additional notable strength of our study was that 75% of our sample self-identified their race as Black. While studies have been conducted to determine differences in complete blood counts in healthy adults by race [16,35], we were unable to identify such studies that have been performed in adults during pregnancy. Samples used for establishing WBC values in preeclamptic pregnancy have either been collected as part of studies of predominantly White participants, or studies that do not report racial demographics. While establishment of WBC parameters for all people is important, it is particularly dire for Black individuals, who have the highest rate of PE, yet continue to be overlooked and under-represented in studies.

Despite these strengths, there are limitations that should be noted. As discussed above, we believe there is great value in focusing on self-identified race as it is a complex social construct which can serve as a proxy for social, environmental, and structural factors, including institutionalized racism [36,37]. Ideally however, holistic examination of both self-identified race and genetic ancestry would assist in better understanding the mechanism responsible for the differences observed here (though we find it prudent to state that we denounce the long-refuted theories of “biological race” which have been used as justification for harm and oppression [38–40]). Unfortunately, we did not have genome-wide genetic data from which to infer genetic ancestry, including information relative to the multi-racial and -ethnic composition of the United States, for comparison [41].

Next, participants were recruited to the parent study between 2008 and 2014, prior to publication of the latest diagnostic criteria to define PE [42]. As such, data related to certain clinical characteristics needed to apply this more recent diagnostic definition were not available for use in defining and selecting PE cases for the present study. Further, our sample consisted primarily of individuals with overweight and obesity. It has been established that obesity is associated with chronic inflammation and increased WBC values, which may have reduced the generalizability of this study [43].

Finally, because of limitations related to our sample size, we were unable to explicitly control for all relevant heterogeneity (e.g., socioeconomic status, diet/nutrition, environmental exposures, family support, PE onset/subtype) [37,44] in our sample to better interpret the racial and cell-type differences observed. Of potential concern is the mix of late-onset (≥ 37 weeks, n=17), pre-term (<37 weeks, n=9), and early-onset (<34 weeks, n=2) PE subtypes in our cases [45]. While we could not rigorously analyze differences between these smaller groups, particularly the two early-onset cases, we did explore WBC proportions across the trimesters of pregnancy by PE sub-group, which did not reveal any striking differences by PE onset (Supplemental Figure 3, Supplemental Table 3). Future work in larger sample and more heterogeneous samples is needed to better understand cellular composition across pregnancy.

Conclusions

We identified potential differences in WBC proportions across preeclamptic and normotensive pregnancy by self-identified race. Researchers have begun to further classify PE pathophysiology into subclasses based on distinct alterations in molecular pathways [46,47]. We speculate that cases of “inflammatory” PE, driven by the immune system, may be more common among White individuals, as evidenced by their changes in WBC proportions. Similarly, it is possible that different subclasses of PE are more prevalent among Black individuals, as we did not see any significant differences among cases in Black participants in our study. Longitudinal studies with larger and more generalized samples representing a population that is racially diverse, includes all BMI categories, and PE subtypes are needed to follow up on our preliminary findings. Establishment of personalized WBC reference values specific to PE and race could provide further insights into PE pathophysiology, including a potential loss of immunomodulation related to B cell insufficiency in early pregnancy.

Data Availability Statement

The data that support the findings of this study are openly available in dbGAP, accession number: phs001937.v1.p1