Predictors of upstream inflammation and oxidative stress pathways during early pregnancy

Abstract

Background.

Inflammation and oxidative stress are critical to pregnancy, but most human study has focused on downstream, non-causal indicators. Oxylipins are lipid mediators of inflammation and oxidative stress that act through many biological pathways. Our aim was to characterize predictors of circulating oxylipin concentrations based on maternal characteristics.

Methods.

Our study was conducted among 901 singleton pregnancies in the LIFECODES Fetal Growth Study, a nested case-cohort with recruitment from 2007 to 2018. We measured a targeted panel of oxylipins in early pregnancy plasma and urine samples from several biosynthetic pathways, defined by the polyunsaturated fatty acid (PUFA) precursor and enzyme group. We evaluated levels across predictors, including characteristics of participants’ pregnancy, socioeconomic determinants, and obstetric and medical history.

Results.

Current pregnancy and sociodemographic characteristics were the most important predictors of circulating oxylipins concentrations. Plasma oxylipins were lower and urinary oxylipins higher for participants with a later gestational age at sampling (13–23 weeks), higher prepregnancy BMI (obesity class I, II, or III), Black or Hispanic race and ethnicity, and lower socioeconomic status (younger age, lower education, and uninsured). For example, compared to those with normal or underweight prepregnancy BMI, participants with class III prepregnancy obesity had 45–46%% lower plasma epoxy-eicosatrienoic acids, the anti-inflammatory oxylipins produced from arachidonic acid (AA) by cytochrome P450, and had 81% higher urinary 15-series F₂-isoprostanes, an indicator of oxidative stress produced from non-enzymatic AA oxidation. Similarly, in urine, Black participants had 92% higher prostaglandin E₂ metabolite, a pro-inflammatory oxylipin, and 41% higher 5-series F₂-isoprostane, an oxidative stress indicator.

Conclusions.

In this large pregnancy study, we found that circulating levels of oxylipins were different for participants of lower socioeconomic status or a systematically marginalized racial and ethnic groups. Given associations differed along biosynthetic pathways, results provide insight into etiologic links between maternal predictors and inflammation and oxidative stress.

Graphical Abstract:

Introduction

Inflammation and oxidative stress are critical components of a healthy pregnancy [1]. Systemic low-grade inflammation and oxidative stress, however, are associated with adverse pregnancy outcomes, including fetal growth restriction [2, 3], preterm birth [4, 5], and preeclampsia [6, 7]. Along with broader immune function, regulation of inflammation is dynamic and changes across pregnancy to support fetal development [1, 8]. Early pregnancy may be a particularly critical period to establish healthy inflammatory responses for the entirety of gestation [8–10]. However, prior research on inflammation and oxidative stress during pregnancy has been limited by a reliance on nonspecific inflammatory markers such as C-reactive protein (CRP). Although such acute-phase reactants can be relevant indicators of uncontrolled systemic inflammation [11], they are downstream of key regulatory processes and are less likely to play causal roles in the inflammatory milieu during pregnancy [12].

Oxylipins are a large class of bioactive lipids that play a central role in the regulation of inflammation and oxidative stress [13]. The physiological regulation by oxylipins occurs via direct effects on cells and tissues (e.g., vasotension) and indirectly through coordination of immune cell actions [14]. Further, measurements of circulating oxylipins in blood or urine can provide a window into upstream regulatory processes because oxylipins are produced from distinct biosynthetic pathways [15]. These biosynthetic pathways involve oxygenation of polyunsaturated fatty acids (PUFAs) by a variety of enzymatic (cytochrome P450 [CYP], lipoxygenase [LOX], cyclooxygenase [COX]) and non-enzymatic (free radical lipid peroxidation) processes [13]. Measuring oxylipins across these pathways can provide deeper insights into possible biological mechanisms and targets for therapeutic interventions [14, 16]. Our prior work has shown that circulating oxylipins in plasma during pregnancy are associated with newborn size at birth [17], as well as exposure to synthetic chemicals [18]. However, it remains unclear what characteristics of the pregnant person may influence their circulating oxylipin measures. This is a major limitation that prevents understanding how maternal characteristics may influence strategies to modulate maternal inflammation and oxidative stress during pregnancy.

In this study, we measured circulating levels of oxylipins during early pregnancy among participants from the LIFECODES Fetal Growth Study. We measured oxylipins in plasma and urine that are produced from several major biosynthetic pathways, along with PUFAs in plasma. The simultaneous evaluation of key oxylipins in both plasma and urine is novel and provides greater potential to distinguish between inflammation and oxidative damage [15, 19]. Our aim was to examine predictors of circulating concentrations based on participant characteristics, including their current pregnancy, socioeconomic determinants, and obstetric and medical history. Additionally, we evaluated temporal trends in concentrations over the 2007 to 2018 study period, along with other analytical features, that could influence measurements of circulating oxylipin or PUFA levels during pregnancy.

Materials and methods

Study population (and covariates)

Participants were recruited as part of the LIFECODES pregnancy cohort, an ongoing prospective cohort based at Brigham and Women’s Hospital in Boston, MA. Participants are eligible for enrollment if: 1) they are ≥18 years of age; 2) initiate their prenatal care <16 weeks gestation; 3) have a singleton or twin pregnancy; and 4) plan on delivering at Brigham and Women’s Hospital. At the time of enrollment (median 10 weeks gestation), participants provide informed consent and are requested to attend additional study visits that are integrated into their routine prenatal care appointments. Participants provide detailed demographic questionnaires along with blood and urine samples at each study visit.

The LIFECODES Fetal Growth Study (n = 901) is a case-cohort nested within LIFECODES, designed to examine the influences of chemical exposures on fetal growth [20]. Briefly, participants were eligible for sampling into the case-cohort if they had a singleton pregnancy with live birth occur between 2008 and 2018 and had a birthweight in their medical record (N = 3,330 eligible). Sampling into the case-cohort occurred in two phases: 1) a subcohort (n=504) was randomly selected from the entire eligible cohort; 2) two participant enrichment sets were randomly selected from all identified cases of small- (n=199) and large-for-gestational age (n=198) births in the underlying cohort. The focus of the present study was to evaluate oxylipins in early pregnancy, so we utilized plasma and urine specimens and participant data from the first study visit (median 10 weeks of gestation). While all participants (N=901) provided non-fasting plasma or urine at the first study visit, n=882 provided both samples, n=11 only provided urine, and n=9 only provided plasma. This study was approved by the Institutional Review Board at Brigham and Women’s Hospital and was deemed exempt by the National Institute of Environmental Health Sciences.

Covariates for participant characteristics

Participant characteristics were assessed through questionnaires completed at enrollment and medical records abstraction. For the purposes of this study, we created a priori groupings of covariates for participant characteristics of interest, including: 1) current pregnancy; 2) socioeconomic determinants; 3) obstetric history; 4) and medical history. The covariates selected for each category were selected a priori based on demonstrated or plausible inflammatory effects. All covariates were categorized, and missing values were excluded from statistical analyses. Diagnosed conditions that were made before or by first study visit (e.g., history of preeclampsia or gestational diabetes [GDM]) were abstracted after validation by at least two maternal fetal medicine specialists performing medical record review. Abstraction of delivery records was used to determine sex of the fetus.

Current pregnancy covariates included gestational age at sampling, prepregnancy body mass index (BMI), blood pressure, fetal sex, assistive reproductive technology (ART), aspirin use, season of enrollment, and time of day of biospecimen sample collection. Gestational age at sample collection was calculated according to the recommendations by the American College of Obstetricians and Gynecologists [21], and was then categorized as 4–8 weeks, 9–12 weeks, and 13–23 weeks, which approximated the 25^th, 50^th, and 75^th percentiles of the distribution. Prepregnancy BMI was assessed at enrollment and calculated from clinically-measured height and self-reported prepregnancy weight. Weight at enrollment was used if self-reported prepregnancy weight was missing (n=11). Prepregnancy BMI was categorized as underweight (<18.5 kg/m²), normal (18.5–24.9 kg/m²), overweight (25.0–29.9 kg/m²), and by classes of obesity, including class I (30–34.9 kg/m²), class II (35–39.9 kg/m²), and class III (≥40 kg/m²) [22]. Due to limited sample size within the underweight category, participants classified as underweight or normal prepregnancy BMI were combined into one category. Blood pressure was collected at enrollment and categorized according to the American College of Obstetricians and Gynecologists [23], including normal (<120 millimeters of mercury [mmHg] systolic and <80 mmHg diastolic), elevated (120–129 mmHg systolic and <80 mmHg diastolic), and hypertensive subtypes, including stage I (130–139 mmHg systolic and 80–89 mmHg diastolic) and stage II (≥140 mmHg systolic and ≥90 mmHg diastolic) hypertension. Aspirin use was self-reported at the time of the study visit and verified using electronic records. Use was defined as either standard or baby aspirin use, as no participants reported using both. Data on aspirin use was unavailable for participants recruited between 2007–2008. Season of enrollment was determined based on the enrollment visit occurring during Summer (June to August), Fall (September to November), Winter (December to February), or Spring (March to May). Finally, the time of day of biospecimen (urine and plasma) sample collection was defined as either AM (before 12:00 pm) or PM (after 12:00 pm). Information on sampling time was unavailable for participants who delivered prior to 2010.

Covariates of socioeconomic determinants included maternal race and ethnicity, age, educational attainment, insurance provider (private or uninsured/Medicaid [uninsured]), smoking during pregnancy, and alcohol use during pregnancy. While these factors are often strong predictors of socioeconomic status at the population level, we recognize they can be limited at the level of an individual participant [24]. We summarized race and ethnicity as non-Hispanic White (White), non-Hispanic Black (Black), Hispanic, Asian, and Multiracial/Other. Participants were categorized as Hispanic when they reported that ethnicity, regardless of their selected racial identity. Categories of insurance provider were defined as 1) Private or health maintenance organization and 2) self-payment, Medicaid, or uninsured; which were labeled as private and uninsured, respectively. Covariates of obstetric history included parity, gravidity, prior preeclampsia, and prior GDM. Medical history was determined from medical records abstraction and included chronic hypertension, cardiac disorder, diabetes, thyroid disorder, and asthma.

Oxylipin quantification

Plasma and urine samples were aliquoted and stored at −80°C shortly after collection. Samples were shipped on dry ice to the Eicosanoid Core Laboratory at Vanderbilt University Medical Center, where they were analyzed for panels of oxylipins and fatty acids (PUFAs). Liquid chromatography with tandem mass spectrometry (LC-MS/MS) was used to quantify the non-esterified (free) portion of oxylipins and fatty acid levels in maternal circulation. Although the free portion of oxylipins in plasma is much less abundant than the esterified portion [25], we targeted the free portion because it is generally considered to be more bioactive component [14]. We provide a detailed description of the analytical processes of our LC-MS/MS approach in plasma and urine in Appendix A. For purposes of interpreting results, we grouped oxylipins by their primary biosynthetic pathways of endogenous production, which groups them by precursor fatty acid and enzyme group [15, 17]. This included the fatty acid precursors linoleic acid (LA), arachidonic acid (AA), docosahexaenoic acid (DHA), and eicosapentaenoic acid (EPA), as well as the enzyme groups of CYP, LOX, and COX. Based on these criteria, we targeted a total of 9 fatty acid-enzyme groupings across plasma and urine samples: LA-CYP, LA-LOX, AA-CYP, AA-LOX, AA-COX, DHA-CYP, DHA-LOX, AA-COX (urine), and AA-non-enzymatic (urine). Additionally, we targeted six PUFAs in plasma, including LA, AA, dihomo-gamma-linolenic acid (DGLA), alpha-linoleic acid (ALA), DHA, and EPA. We grouped PUFAs as omega-6 (LA, AA, DGLA) or omega-3 (ALA, DHA, EPA) because each class has distinct sources and health impacts [14]. We restricted our statistical analysis to oxylipins and fatty acids that were detected in ≥50% of samples. Values below the limit of detection (LOD) were replaced by the LOD divided by the square root of 2.

Along with our use of individual oxylipin and PUFA analytes, we created ratio measures. Among oxylipins, we calculated 4 ratios to get indirect measures of soluble epoxide hydrolase (sEH) activity, an enzyme family that hydrolyzes oxylipin epoxy-residues to diols. The structural change between epoxide- to dihydroxy-forms of these oxylipins can cause meaningful changes to inflammatory potential and risks for adverse health outcomes [26–28]. We calculated two ratios for LA-CYP oxylipins, including 9,10-epoxy:dihydroxy-octadecenoic acid (9,10-EpOME:DiHOME) and 12,13-EpOME:DiHOME, along with two ratios for AA-CYP oxylipins, including 11,12-epoxy:dihydroxy-eicosatrienoic acid (11,12-EET:DHET) and 14,15-EET:DHET. Lower ratios of EpOME:DiHOME may indicate higher sEH activity. This could increase inflammatory potential due to the pro-inflammatory effects of DiHOMEs [26, 28]. Similarly, lower ratios of EET:DHET may indicate higher sEH activity, but with opposing effects. Given EETs have anti-inflammatory effects, higher sEH activity may increase inflammatory potential by reducing EET levels [14, 27]. Among the PUFAs, we calculated a single ratio of omega-6 to omega-3, which is an objective measure of dietary PUFA intake [29]. Along with providing a marker of nutritional intake and downstream oxylipin metabolite profiles, the ratio of circulating PUFAs may independently influence inflammatory potential [30, 31]. While higher omega-6 levels can promote inflammation, higher omega-3 promotes anti-inflammatory effects [14]. Additionally, there is competitive inhibition between omega-6 and omega-3 PUFAs for enzymatic reactions [14]. Thus, a higher omega-6:omega-3 ratio may indicate greater inflammatory potential. We calculated the omega-6:omega-3 ratio by dividing the sum of omega-6 analytes (LA, AA, and DGLA) by the sum of omega-3 analytes (ALA, DHA, and EPA).

In Figure 1, we overview the biosynthetic pathways of targeted oxylipins that met our inclusion criteria. The biological matrix in which an oxylipin was measured (plasma versus urine) was based on the best matrix for biological relevance and stability [32, 33]. We measured two of the major urinary metabolites of the pro-inflammatory prostaglandin E₂ (PGE₂), including PGE-M and tetranor-PGE₁ (TN-E) [34, 35], which have yet to be thoroughly investigated during pregnancy. We also measured two metabolites of the pro-inflammatory oxylipin thromboxane (TX) A₂, including thromboxane B₂ (TXB₂) in plasma and 11-dehydro-thromboxane B₂ (11dTxB₂) in urine. While TXB₂ is best measured in plasma, 11dTxB₂ is preferred in urine based on its stability and relevant metric of circulating TXA₂ production [36].

Figure 1.

Overview of biosynthetic pathways for circulating oxylipins targeted and detected in the LIFECODES Fetal Growth Study.

Along with the urinary oxylipin prostaglandin F_2α (PGF_2α) that is produced enzymatically via AA-COX, we also measured four PGF_2α -related F₂-Isoprostanes (F₂-IsoP) in urine that are predominantly formed by non-enzymatic lipid peroxidation. This included the two most abundant urinary F₂-IsoPs, 15-series and 5-series F₂-IsoP, which have rarely been measured simultaneously in human studies [37]. While 15-F₂-IsoP, also known as 8-isoprostane, is mainly generated by non-enzymatic oxidative stress, it can also be produced by COX metabolism [38]. However, 5-series F₂-IsoP (5-F₂-IsoP) is only produced non-enzymatically [19]. In this study, along with urinary 15-F₂-IsoP, we also measured its two major metabolites, 2,3-dinor-15-F₂-IsoP and 2,3-dinor-5,6-dihydro-15-F₂-IsoP (2,3-dinor-5,6–15-F₂-IsoP). The 2,3-dinor-5,6–15-F₂-IsoP is considered to be the primary metabolite because it is produced in higher abundance than 2,3-dinor-15-F₂-IsoP and is the more terminal metabolite of 15-F₂-IsoP [37].

Statistical approach

Urinary oxylipins were corrected for urine dilution using specific gravity (SG), which was measured using a digital handheld refractometer (AtagoCo., Ltd., Tokyo, Japan). We corrected concentrations of urinary oxylipins using the Boeniger method [39, 40]: $M_{SG}=M_o\frac{(S{G}_{\mathit{\text{median}}}-1)}{(S{G}_o-1)}$, where M_SG is the SG-corrected oxylipin measurement, M_O is the observed oxylipin measurement, SG_median is the median SG of the study population in early pregnancy (1.017), and SG_O is the observed SG. To correct for potential batch-effects and standardize concentrations, we performed a two-stage residualization procedure commonly used in molecular epidemiology [41, 42]. In the first stage, we regressed the observed analyte concentration on a categorical batch covariate. In the second stage, we normalized the model residuals using the rank-based inverse normal transformation [43]. This procedure produced normally distributed z-scores for each analyte. Ratios, including the sEH and PUFA measures, were calculated prior to performing the same two-stage residualization procedure.

We first examined temporal variability in standardized analyte concentrations over the study period. Given our 11-year study period, we were interested in determining temporal variability in concentrations that could be attributable to changes in study procedures or long-term storage effects [44, 45]. Additionally, we were interested in potential changes in dietary intake of PUFAs over time [29]. Calendar time was based on year of conception, which we calculated using the date and gestational age at sampling. We calculated the mean and 95% confidence interval (95% CI) of z-scores for each analyte based on year of conception, categorized as 2-year increments from 2007–2018. We tested for overall differences across categories of years for each analyte using multivariate Wald tests. To facilitate interpretability, we converted z-scores to percentiles. We then visualized percent differences in mean percentile scores across categories using heat maps. Heat maps included numerical values of all percent differences, but only statistically significant differences (Wald tests p-value<0.05) were highlighted with color and bolded text.

Next, we contrasted analyte levels across our four groups of participant characteristics. Given several analytes showed changes in concentrations over calendar time, as well as evidence that some characteristics of our participants changed over time [46], we calculated the population marginal mean and 95% CI using a generalized linear model that adjusted for year of conception, which was modeled continuously in 2-year increments. We again converted marginal mean (95% CI) z-scores to percentiles and visualized percent differences in means using heat maps. We grouped results by covariate categories, biosynthetic pathway of oxylipin production, and biological matrix (plasma or urine). Given we had no missing values for calendar year (year of conception), the adjusted marginal means maintained the same sample size as those for the unadjusted means.

We performed all statistical analyses using inverse probability of sampling weights to account for participant selection into the case-cohort using the survey package (version 4.1.1) [47]. We generated sampling weights by case status using guidelines for secondary analyses of case-cohort data [46, 48]. Statistical analyses were performed in R (version 4.2.1) [49].

Results

Participant characteristics

The study was composed of 901 participants with oxylipins measured in at least one plasma or urine sample and characteristics shown in Table 1. Nearly 80% of participants enrolled in the study prior to 12 weeks of gestation. Most participants had normal or underweight (53%) to overweight (25%) prepregnancy BMI, normal blood pressure (70%), did not use ART to conceive (88%), and conceived prior to 2013 (72%). Regarding socioeconomic determinants, most participants were White (58%), enrolled after age 30 (69%), had completed a college education (66%), and had access to private health insurance (73%).

Table 1.

Distribution of maternal characteristics in participants from the LIFECODES Fetal Growth Study (N=901).

Variable	Category	n (%)
Current pregnancy
Gestational age at sampling	4–8 weeks	246 (27)
	9–12 weeks	472 (52)
	13–23 weeks	183 (20)
Prepregnancy BMI	normal or underweight	476 (53)
	overweight	223 (25)
	class I obesity	120 (13)
	class II obesity	51 (6)
	class III obesity	31 (3)
Blood pressure	normal	624 (70)
	elevated	125 (14)
	stage I hypertension	121 (14)
	stage II hypertension	20 (2)
Fetal sex	female	433 (48)
Assistive reproductive technology	yes	105 (12)
Aspirin use	standard	29 (4)
	baby	8 (1)
Season of enrollment	Summer	276 (31)
	Fall	273 (30)
	Winter	189 (21)
	Spring	163 (18)
Year of conception	07–08	229 (25)
	09–10	149 (17)
	11–12	269 (30)
	13–14	134 (15)
	15–16	71 (8)
	17–18	49 (5)
Time of biospecimen collection	AM	241 (39)
	PM	375 (61)
Socioeconomic determinants
Race and ethnicity	White	526 (58)
	Black	127 (14)
	Hispanic	141 (16)
	Asian	56 (6)
	Multiracial/Other	51 (6)
Age	>35	302 (34)
	30 – 34.9	315 (35)
	25 – 29.9	188 (21)
	<25	96 (11)
Education	college or greater	587 (66)
	some college or technical school	190 (21)
	high school or less	112 (13)
Insurance provider	private	648 (73)
	uninsured or Medicaid	241 (27)
Current smoker	yes	66 (7)
Alcohol use	yes	61 (7)
Obstetric History
Parity	0	345 (38)
	1	357 (40)
	2	141 (16)
	3+	58 (6)
Gravidity	0	214 (24)
	1	248 (28)
	2–3	273 (30)
	4+	165 (18)
Prior preeclampsia	yes	54 (6)
Prior preterm birth	yes	139 (15)
Prior GDM	yes	21 (2)
Medical History
Chronic hypertension	yes	40 (4)
Cardiac disorder	yes	66 (7)
Diabetes	yes	35 (4)
Thyroid disorder	yes	94 (10)
Asthma	yes	160 (18)

¹Sampling weights were used to generate sample size counts and proportions. Participants had missing values for: Blood pressure (n=13), Fetal sex (n=1), Aspirin use (n=217), Time of biospecimen collection (n=284), Education (n=14), Insurance provider (n=10), Alcohol use (n=8). Data on aspirin use was unavailable for participants who conceived between 2007–2008.

Abbreviations: BMI, body mass index; GDM, gestational diabetes; HMO, health maintenance organization

Oxylipin detection and temporal patterns

Of the oxylipins from 7 biosynthetic pathways of production we targeted in plasma and urine, oxylipins from 5 pathways had sufficient detection (i.e., >50% samples above LOD) to be included for statistical analysis (Table 2). All 6 PUFAs were highly detected, including the 3 omega-6 (LA, AA, and dihomo-gamma-linolenic acid [DGLA]) and 3 omega-3 (alpha-LA [ALA], DHA, and EPA) PUFAs. A total of 6 plasma oxylipins produced from 4 biosynthetic pathways, including pro-resolving metabolites of DHA, were excluded due to low detection (Table S1 in Appendix B). Correlation patterns of oxylipins supported our a priori grouping by biosynthetic pathways and based on plasma or urine measurement (Figure 2). Analytes displayed temporal differences in concentrations (Figure S1 and Table S2 in Appendix B). We observed significant differences in mean levels across years for at least one oxylipin from each biosynthetic pathway, as well as among PUFAs. However, for the oxylipins with significant differences, there were no consistent temporal trends in concentrations (e.g., linear decrease in levels over time).The exception to this was for PUFAs, as participants from more recent years had lower levels of the omega-6 LA and higher of the omega-3 ALA. For example, participants that conceived in 2017–2018 had 33% lower LA and 93% higher ALA compared to participants who conceived in 2007–2008. Importantly, this finding was confirmed by the observation that omega-6:omega-3 ratios decreased over the study period, including 60%−71% lower ratios for participants in 2015–2018 compared to those in 2007–2008. Based on these observations, we adjusted all subsequent analyses of marginal means for year of delivery.

Table 2.

Circulating oxylipins and polyunsaturated fatty acids (PUFAs) with ≥50% of samples above limit of detection (LOD) (N=901).

Pathway grouping1		Full name	Abbreviation	Matrix	Median (IQR)2	LOD (ng/ml)	Above LOD (n [%])3	Total (n)3
Oxylipins
Precursor	Enzyme
LA	CYP	9,10-epoxy-octadecenoic acid	9,10-EpOME	Plasma	8.49 (5.35, 14.5)	0.075	887 (1.00)	891
		9,10-dihydroxy-octadecenoic acid	9,10-DiHOME	Plasma	1.64 (1, 2.5)	0.075	872 (0.98)	891
		9,10-epoxy:dihydroxy-octadecenoic acid4	9,10-EpOME:DiHOME	Plasma	5.37 (3.47, 7.99)	-	-	890
		12,13-epoxy-octadecenoic acid	12,13-EpOME	Plasma	10.43 (6.8, 16.69)	0.075	886 (0.99)	891
		12,13-dihydroxy-octadecenoic acid	12,13-DiHOME	Plasma	4.23 (2.66, 6.4)	0.075	879 (0.99)	891
		12,13-epoxy:dihydroxy-octadecenoic acid4	12,13-EpOME:DiHOME	Plasma	2.5 (1.84, 3.5)	-	-	890
	LOX	13-hydroxy-octadecadienoic acid	13-HODE	Plasma	7.98 (5.06, 12.82)	0.075	866 (0.97)	891
AA	CYP	5,6-dihydroxy-eicosatrienoic acid	5,6-DHET	Plasma	1.66 (0.04, 2.41)	0.050	621 (0.70)	862
		11,12-epoxy-eicosatrienoic acid	11,12-EET	Plasma	5.22 (3.29, 7.78)	0.075	857 (0.96)	891
		11,12-dihydroxy-eicosatrienoic acid	11,12-DHET	Plasma	0.43 (0.3, 0.58)	0.050	825 (0.93)	891
		11,12-epoxy:dihydroxy-eicosatrienoic acid4	11,12-EET: DHET	Plasma	12.64 (7.34, 21.38)	-	-	890
		14,15-epoxy-eicosatrienoic acid	14,15-EET	Plasma	3.46 (2.4, 5.36)	0.075	875 (0.98)	891
		14,15-dihydroxy-eicosatrienoic acid	14,15-DHET	Plasma	0.52 (0.39, 0.67)	0.050	842 (0.94)	891
		14,15-epoxy:dihydroxy-eicosatrienoic acid4	14,15-EET: DHET	Plasma	7.04 (4.55, 12.21)	-	-	890
		11-hydroxy-eicosatetraenoic acid	11-HETE	Plasma	0.9 (0.05, 1.64)	0.075	541 (0.61)	891
		20-hydroxy-eicosatetraenoic acid	20-HETE	Plasma	231.4 (167.67, 307.55)	0.100	891 (1.00)	891
	LOX	5-hydroxy-eicosatetraenoic acid	5-HETE	Plasma	1.47 (0.49, 2.87)	0.075	685 (0.77)	891
		12-hydroxy-eicosatetraenoic acid	12-HETE	Plasma	0.96 (0.54, 1.67)	0.075	828 (0.93)	891
		15-hydroxy-eicosatetraenoic acid	15-HETE	Plasma	2.42 (1.9, 3.17)	0.075	886 (0.99)	891
	COX	Thromboxane B2	TXB2	Plasma	0.62 (0.04, 1.26)	0.050	493 (0.55)	891
AA	COX	Prostaglandin E2 metabolite 1 (metabolite of PGE2)	PGE-M	Urine	4.8 (2.28, 8.99)	0.005	887 (0.99)	888
		Tetranor prostaglandin E2 metabolite (metabolite of PGE2)	TN-E	Urine	1.97 (1.21, 3.22)	0.040	876 (0.98)	890
		11-dehydro-thromboxane B2	11dTxB2	Urine	0.46 (0.28, 0.71)	0.025	729 (0.82)	888
		Prostaglandin F2-alpha	PGF2α	Urine	2.82 (1.78, 4.54)	0.020	884 (0.99)	890
	Nonenzymatic	15-F2-isoprostane	15-F2-IsoP	Urine	1.05 (0.74, 1.52)	0.020	872 (0.98)	890
		2,3-dinor-5,6-dihydro-15-F2-IsoP (primary metabolite of 15-F2-IsoP)	2,3-dinor-5,6–15-F2-IsoP	Urine	7.53 (5.34, 10.76)	0.500	883 (0.99)	890
		2,3-dinor-15-F2-IsoP (secondary metabolite of 15-F2-IsoP)	2,3-dinor-15-F2-IsoP	Urine	3.08 (1.56, 5.68)	0.020	767 (0.86)	892
		5-series F2-isoprostane	5-series F2-IsoP	Urine	2.12 (1.23, 3.47)	0.020	879 (0.98)	890
	PUFAs
	Omega-6	Linoleic acid	LA	Plasma	27.99 (23.06, 33.16)	1.000	860 (0.96)	890
		Arachidonic acid	AA	Plasma	7.53 (6.01, 9.04)	1.000	885 (0.99)	890
		Dihomo-γ-linolenic acid	DGLA	Plasma	15.88 (13.08, 23.22)	1.000	890 (1.00)	890
	Omega-3	α-linoleic acid	ALA	Plasma	174.89 (153.03, 205.62)	1.000	890 (1.00)	890
		Docosahexaenoic acid	DHA	Plasma	4.68 (3.42, 6.41)	1.000	731 (0.82)	890
		Eicosapentaenoic acid	EPA	Plasma	2.09 (1.31, 3.28)	1.000	855 (0.96)	890
		Omega-6:Omega-34	N6:N3	Plasma	0.28 (0.24, 0.34)	-	-	888

¹Enzyme abbreviations: COX, cyclooxygenase; CYP, cytochrome P450; IQR, interquartile range; LOX, lipoxygenase; PUFA, polyunsaturated fatty acid

²Median (IQR) concentrations of PUFAs are presented as μg/ml and concentrations of oxylipins are presented as ng/ml. Ratio measures do not have units.

³Count and proportion of total samples is based on non-missing sample values. Missing values were due to no collection of urine (n=9) or plasma (n=11) samples, or due to exclusion from insufficient sample volume.

⁴Ratio measures do not have units or LOD values. Oxylipin ratios were based on individual analytes, while the omega-6:omega-3 ratio was based on the sum of all omega-6 analytes to the sum of all omega-3 analytes.

Figure 2.

Spearman correlations among oxylipins grouped by biosynthetic pathway and sampling matrix.

Pregnancy predictors of oxylipins

The main pregnancy predictors of circulating oxylipin concentrations were gestational age at sample collection and prepregnancy BMI, with generally opposing trends observed between plasma and urinary oxylipins (Figure 3, Table S3 in Appendix B). Gestational age at sample collection was associated with oxylipins across nearly all biosynthetic pathways. In plasma, higher gestational age was associated with lower oxylipin levels, but the inverse was displayed in urine. For example, compared to participants sampled at 4–8 weeks, those sampled at 9–12 weeks and 13–23 weeks had 5%−11% and 16%−27% lower plasma oxylipin levels, respectively. In urine, there was a positive, monotonic association between gestational age at sampling and oxylipin levels, but solely among the oxylipins produced enzymatically. Compared to those sampled at 4–8 weeks, participants samples at 9–12 weeks and 13–23 weeks had 11%−35% and 34%−54% higher urinary levels of AA-COX oxylipins.

Figure 3.

Heatmap of percent differences in marginal mean levels of oxylipins by characteristics of the current pregnancy.

Prepregnancy BMI was positively associated with urinary oxylipins produced both enzymatically and nonenzymatically. For example, compared to those with normal or underweight prepregnancy BMI, those with class I, II, or III obesity had 57%, 68%, and 81% higher 15-F₂-IsoP, respectively. Similar magnitudes of differences were observed for the other nonenzymatically produced oxylipins, including the two 15-F₂-IsoP metabolites, 2,3-dinor-5,6-dihydro- and 2,3-dinor-15-F₂-IsoP, and 5-F₂-IsoP. Higher prepregnancy BMI was also associated with lower levels of plasma oxylipins produced from LA-CYP and AA-CYP pathways. For example, the anti-inflammatory epoxy-eicosatrienoic acids (11,12- and 14,15-EETs) produced by AA-CYP were 45–46% lower among participants with the more severe class III obesity compared to those with normal or underweight BMI.

Some other associations between current pregnancy characteristics and oxylipins were noted. Blood pressure categories only displayed significant differences in plasma thromboxane B₂ (TXB₂), a product of AA-COX and stable indicator of the pro-inflammatory, vasoconstrictor TXBA₂. The marginal mean of TXB₂ was 24% higher among those with stage II hypertension (mean rank-normalized percentile [percentile]: 60, 95% CI: 42, 77) compared to those with normal blood pressure (percentile: 49, 95% CI: 45, 52). Interestingly, blood pressure was one of the only pregnancy characteristics associated with the oxylipin ratio measures of sEH activity (Figure S2 and Table S4 in Appendix B). Specifically, higher 12,13-EpOME:DiHOME ratio of the LA-CYP pathway were associated with higher blood pressure, including 61% higher 12,13-EpOME:DiHOME among those with stage II hypertension compared to those with normal blood pressure.

Participants with fetuses of male sex had higher levels of some LA-derived oxylipins in plasma, including EpOMEs from CYP and 13-HODE from LOX. Overall, null associations were observed between oxylipins and other current pregnancy covariates, including aspirin use, ART, season of enrollment, and time of day of sample collection. However, we observed that levels of AA-COX oxylipins in urine, particularly PGE-M and 11dTxB₂, were lower among participants who took any form of aspirin. For example, those who took standard (adult) aspirin had 62% lower 11dTxB₂ compared to non-users (percentile: 19 [95% CI: 11, 31] versus 51 [95% CI: 46, 55], respectively).

Sociodemographic predictors of oxylipins

Sociodemographic determinants displayed diverse and numerous associations with circulating oxylipins (Figure 4, Table S3 in Appendix B). In plasma, socioeconomic determinants were predominantly associated with LA-CYP oxylipins. Concentrations of the pro-inflammatory epoxy-octadecenoic acids (EpOMEs) and their dihydroxy-eicosatrienoic acids (DiHOMEs) metabolites are produced from the LA-CYP pathway and were lower among participants who were of Black or Hispanic race/ethnicity, younger age, lower education, and uninsured. For example, plasma DiHOME levels were 23%−42% lower among those without compared to those with a college degree, 22%−32% lower among those without compared to those with private health insurance, and 17%−32% lower among those of Black compared to White race/ethnicity. These patterns were similar for oxylipins from other pathways in plasma, but with less consistent statistical significance. Alcohol use, however, had null associations with all plasma oxylipins. Contrary to individual oxylipins, we observed no significant associations between socioeconomic determinants with plasma oxylipin ratios for sEH activity (Figure S2 and Table S4 in Appendix B).

Figure 4.

Heatmap of percent differences in marginal mean levels of oxylipins by socioeconomic determinants.

In urine, every sociodemographic determinant, except alcohol, was associated with circulating oxylipin levels. Further, these relationships paralleled those observed among plasma oxylipins, but in the opposite direction. Participants who were of Black or Hispanic race/ethnicity, younger age, lower education, uninsured, and those who smoked during pregnancy had higher marginal means of urinary oxylipins. For example, Black participants had 92% higher urinary PGE-M compared to White participants (percentile: 72 [95% CI: 66, 78] versus 38 [95% CI: 34, 41], respectively), while those without private health insurance had 76% higher PGE-M compared to those with it (percentile: 70 [95% CI: 65, 75] versus 40 [95% CI: 36, 43], respectively).

Obstetric and medical history predictors of oxylipins

Overall, obstetric and medical history displayed fewer associations with circulating oxylipins (Figure 5, Table S3 in Appendix B). However, there were several exceptions. Participants with prior pregnancy complications (preterm birth, GDM, or preeclampsia) had lower plasma levels of 15-HETE, an AA-LOX oxylipin, while those with a cardiac disorder or asthma had lower plasma levels of LA-CYP oxylipins. In urine, participants with prior preeclampsia and higher parity or gravidity were more likely to have higher circulating oxylipins, particularly for nonenzymatic metabolites. For example, those with prior preeclampsia had 28%−36% higher nonenzymatic oxylipins. On the other hand, participants with chronic hypertension, diabetes, or thyroid disorder generally had lower urinary oxylipins. For example, marginal means of the nonenzymatically-produced 5-F₂-IsoP were 28% lower among those with chronic hypertension compared to those without (percentile: 36 [95% CI: 24, 48] versus 50 [95% CI: 46, 53], respectively)

Figure 5.

Heatmap of percent differences in marginal mean levels of oxylipins by participant obstetric and medical history.

Predictors of circulating PUFAs

Overall, compared to the oxylipins, plasma PUFAs displayed fewer associations with participant characteristics (Figure S3 and Table S5 in Appendix B). Higher gestational age at sampling was associated with higher levels of the omega-6 PUFAs AA and DGLA, and lower levels of the omega-3 EPA. However, most other covariates of current pregnancy had null associations with PUFAs, including prepregnancy BMI, season of enrollment, or time of day of biospecimen collection. Regarding socioeconomic determinants, lower plasma EPA and DHA was associated with Black or Hispanic race and ethnicity, younger age, lower education, and uninsured. Participants who consumed alcohol during pregnancy generally had higher PUFA levels, particularly for the omega-6 analytes. Finally, we observed predominantly null associations between circulating PUFAs and obstetric and medical history characteristics.

Discussion

In a well-characterized pregnancy cohort with the largest sample size to date, we present associations between a comprehensive set of plasma and urinary oxylipins and key pregnancy, sociodemographic, and obstetric and medical history characteristics. We observed consistent evidence that characteristics of the current pregnancy and socioeconomic determinants were associated with oxylipins, with results that were replicated within biosynthetic pathways, but few associations with obstetric and medical history. Although we did not have longitudinal data within participants, our results showed that during early pregnancy plasma oxylipins increased and urinary oxylipins decreased. In addition, based pregnancy or demographic characteristics, we observed that participants with traits that could make them more vulnerable or historically marginalized had concentrations of oxylipins that were lower in plasma and higher in urine. Specifically, these differences were among those with prepregnancy obesity, a non-White race or ethnicity, younger age, lower education, uninsured or on Medicaid, or who smoked during pregnancy. Notably, these groups did not display clear differences in circulating PUFAs, indicating associations may have been driven more by inflammation or oxidative stress than diet. Overall, our study provides novel evidence that these characteristics are associated with distinct and upstream biological pathways of inflammation and oxidative stress.

Our results are consistent with prior studies showing that populations with the same vulnerable characteristics (e.g., higher prepregnancy BMI, non-White race/ethnicity, uninsured) are more likely to have elevated inflammation and oxidative stress [50–53]. Previous studies have reported higher levels of CRP among vulnerable populations [51, 53]. However, CRP is a downstream inflammatory marker, and provides minimal insight to causal biological processes [12]. We observed that plasma levels of oxylipins, particularly those derived from LA-CYP and AA-CYP pathways, were lower among these groups, which suggests a key role of CYP activity. Specifically, we observed that the epoxy (EpOMEs or EETs) and their dihydroxy (DiHOMEs or DHETs) metabolites were lower among these groups. Experimental evidence demonstrates that expression of CYP isoforms can change in response to inflammation [54], even instigating a signaling cascade that reduces CYP activity and levels of downstream oxylipin metabolites like EETs [55]. This is potentially problematic and reflective of ongoing inflammation, as oxylipins such as EETs promote the resolution of inflammation [56]. Thus, the lower plasma concentrations we observed may be indicative of a systemic, pro-inflammatory signaling cascade involving CYP-produced oxylipins.

At the same time, we showed novel evidence that the participants with potentially greater social and health vulnerabilities had higher levels of urinary oxylipins produced enzymatically by COX. The urinary oxylipins PGE-M, TN-E, 11dTxB₂, and PGF_2α are produced along the AA-COX pathway and are considered to be indicators of inflammation [34–36], but have not been thoroughly investigated in the context of pregnancy [15, 57]. This finding is consistent with evidence from oxylipins produced nonenzymatically, which have been evaluated during pregnancy. A prior pooled analysis of four US cohorts and over 2,000 pregnant people also evaluated 15-F₂-IsoP and its primary metabolite, 2,3-dinor-5,6–15-F₂-IsoP [50]. The study also found significantly higher urinary levels of these 15-series F₂-IsoP among participants with prepregnancy obesity or of non-Hispanic Black race/ethnicity [50]. These results are consistent with other studies showing that non-Hispanic Black women are at particularly higher risk of elevated oxidative stress [50, 52, 53], which is hypothesized to partially stem from experiencing stressful life events associated with racial discrimination [58]. Our results build upon these lines of evidence because we observed consistent results with the 5-F₂-IsoP, which have not been previously examined in this context. This is important because the 5-series version of isoprostane may be a more pure oxidative stress marker than the 15-series since it cannot be created by COX [19].

Our study also provides new evidence that increasing severity of obesity, particularly class II or III, are associated with oxylipins during pregnancy, which corresponds to prior evidence using fewer analytes [50]. Experimental and observational evidence shows a clear relationship between higher maternal BMI and altered inflammation during pregnancy [59–61], possibly acting via dysregulated activation of placental inflammation [62, 63]. While the prior studies have typically found pro-inflammatory markers were higher among those with higher prepregnancy BMI [60, 61], we only observed this positive relationship for urinary oxylipins and the inverse for plasma oxylipins. This may point towards the upstream and complex nature of oxylipin metabolism, as well as effects, as causal inflammatory mediators compared to these other indicators of broad systemic inflammation. However, the directionality of this relationship is unclear and may be cyclical since adipocyte accumulation can promote pro-inflammatory conditions which in turn promote adipogenesis [64–66]. While we cannot determine the directionality, we believe our results demonstrate that distinct biosynthetic pathways of oxylipin production may play an important role in this cardiometabolic endpoint.

Finally, we provide evidence that the gestational age at sample collection is an important factor for circulating oxylipin levels. This evidence corresponds to our prior work, in which we observed that concentrations of many plasma oxylipins decline between early to mid-pregnancy [17]. Similar to our findings, a prior study using serum inflammatory markers like IL-8 and CRP found concentrations during the early phases of pregnancy were lower than later phases [61]. This decline may be attributed to changes in maternal inflammation and placental transfer of fatty acids during this gestational period to meet the changing nutritional needs of the fetus [67, 68], as well as in response to placentation [69]. At any rate, accounting for gestational age at sample collection is necessary when evaluating circulating oxylipins. However, it is important to note that other analytical factors, including the season or time of day of sample collection, were not associated with oxylipins or PUFAs, and are thus unlikely to bias pregnancy studies of circulating levels.

We observed few significant associations between circulating oxylipins and characteristics of maternal obstetric or medical history. However, we did see certain associations that would be expected based on maternal clinical factors, such as current aspirin use. We observed significantly lower levels of prostaglandin metabolites, particularly urinary 11dTxB₂, among those reporting aspirin use. This closely matches clinical evidence that aspirin use can decrease 11dTxB₂ production [36, 70], and provides confidence that aspirin use was reported accurately by our participants. Although aspirin-related inhibition of COX would also reduce TXB₂ production [36], we did not observe significant differences by aspirin use for plasma TXB₂. The difference in associations may be due to the higher stability of 11dTxB₂ in urine than TXB₂ in plasma or urine [71]. This highlights the importance of accounting for aspirin use when evaluating urinary AA-COX oxylipins, though it may be less important when examining oxylipins from this pathway measured in plasma.

Our study showed long-term temporal trends in circulating oxylipins as well as PUFAs. Trends in oxylipins may be due to analytical features such as autooxidation or sampling conditions. Even at −80°C, there is evidence that levels could change over the course of years during storage [44]. However, it is also possible that subtle changes in sample processing procedures (e.g., time stored on ice before freezing) could also artificially impact oxylipins measurements [45]. Our results show the importance of adjusting or standardizing circulating oxylipin concentrations by year of collection when participant recruitment and sampling has been conducted over a long period of time. For measures of PUFAs, omega-6 levels decreased across the study period and levels of omega-3 PUFAs increased. For example, measured omega-3 ALA was nearly twice as high in participants who conceived in 2017–2018 compared those who conceived in 2007–2008. Given LA and ALA are essential fatty acids that cannot be synthesized de novo in humans and must be attained via diet, we expect that these trends are indicative of changes in dietary patterns among pregnant participants. This notion is supported by our results using the omega-6:omega-3 ratio, an objective dietary PUFA indicator [29], which showed a decreased average ratio over the study period. While the trend of increasing circulating omega-3 levels matches prior observations in the general US population [72], decreasing levels of omega-6 PUFAs opposes general dietary consumption patterns [73, 74]. However, it is conceivable that dietary PUFA patterns would change differentially among pregnant people compared to the general US population. Pregnancy often increases the likelihood that an individual will change their dietary habits and follow physician recommendations [75]. Further, several clinical guidelines for dietary PUFA intake during pregnancy were updated during our study period (2007–2018), including recommendations to increase the intake of foods rich in omega-3 such as seafood and certain vegetables [76, 77].

Strengths and limitations

Our study had key strengths. The LIFECODES Fetal Growth Study provided a large sample size in a demographically and clinically diverse pregnant population. We were able to evaluate many participant characteristics due to the rich clinical and questionnaire data available in the cohort. Additionally, we were able to examine clinically-relevant subcategories of covariates that are rarely assessed in perinatal epidemiology, including subclasses of obesity and blood pressure. Additionally, we simultaneously measured of a large set of oxylipins in both plasma and urine, which is rare among pregnancy studies [15]. This provided us the ability to simultaneously evaluate oxylipins from several key biosynthetic pathways of production, all of which have recognized biological and causal relevance to processes of inflammation and oxidative stress. Further, we utilized ratios of several pairs of oxylipins to gain insight into other potential enzymatic processes, namely epoxide hydrolase activity that could influence inflammatory potential. Although we observed few significant associations with the sEH indicators, this approach provides additional ways to maximize insight into biological processes using observational oxylipin data.

We also had several limitations. First, although circulating oxylipins are useful for portraying issues with systemic inflammation, we miss information on tissue-specific inflammatory conditions [13]. This could be a reason why we saw mostly null associations between clinical history and oxylipins. Second, oxylipin analytic methods have not been standardized, nor do they have accepted clinical thresholds of significance [15]. In the event that individual markers are firmly linked to disease status or progression, this may be important future work. Third, we chose to measure oxylipins from samples collected during a single timepoint in pregnancy. Although we expect early pregnancy to play a critical role in determining maternal and fetal health throughout gestation, our prior work shows that longitudinal changes in oxylipins may also be important [17]. Fourth, we measured the free portion of circulating oxylipins, which may limit the ability to compare our results to other studies that exclusively assessed the esterified portion. Although the esterified portion is much more abundant [25, 78], we measured oxylipins in the free (non-esterified) portion of plasma and urine as this compartment is typically considered to play a more active role in mediating ongoing inflammation and oxidative stress [14]. One approach to improve this limitation would be to separately measure oxylipins in both the free and esterified portions, but few human studies to date have performed such combined analyses due to the complexity of additional hydrolysis steps and greater sample volume required [78]. . Fifth, although we sought to measure oxylipins produced from biosynthetic pathways other than LA and AA, those analytes did not meet our detection threshold to be included for statistical analysis. Several of these oxylipins, particularly those produced from EPA or DHA, are known to play important physiological roles in inflammation and methods for measuring them accurately at low levels deserve attention in future work [79]. Finally, although plasma concentrations of free fatty acids may approximate concentrations in adipose tissue when a fasting blood sample is collected [3], samples in our study were collected from non-fasting participants. Thus, we are unable to state the degree to which free PUFA levels were the result of a recent meal. However, we do not expect there were systematic differences in recent dietary intake across most participant covariate categories because we did not observe significant associations between plasma PUFA concentrations and most participant characteristics. We did observer, however, similar trends in the omega-6:omega-3 ratio by calendar year were observed across participant characteristics. Future studies could investigate the impact of recent dietary intake more thoroughly by pairing dietary data (e.g., food frequency questionnaires) with biospecimen measurements of oxylipins and PUFAs.

Conclusions

In a large study of pregnant participants, we observed novel evidence that characteristics of the current pregnancy and determinants of maternal socioeconomic status were the most important predictors of circulating oxylipins in early pregnancy. We observed hallmarks of dysregulated inflammation and elevated oxidative stress for participants with higher prepregnancy BMI, non-White race and ethnicity, and indicators of lower socioeconomic status. Given oxylipins are key biological mediators of inflammation and oxidative stress, these results may inform efforts to mitigate the adverse effects of systemic inflammation or oxidative stress in pregnancy.

Highlights.

• Largest study of early pregnancy predictors of upstream inflammatory pathways to date

• Novel assessment of circulating oxylipins in both plasma and urine biospecimens

• Sociodemographic and pregnancy characteristics associate with oxylipin dysregulation

• Vulnerable groups may experience greatest propensity for oxylipin dysregulation

• Results provide insight on biological mechanisms and targets of pregnancy conditions

Abbreviations

AA

arachidonic acid

ALA

alpha-linoleic acid

COX

cyclooxygenase

CYP

cytochrome P450

DGLA

dihomo-gamma-linolenic acid

DHA

docosahexaenoic acid

DHET

dihydroxy-eicosatrienoic acid

DiHOME

dihydroxy-octadecenoic acid

EET

epoxy-eicosatrienoic acid

EPA

eicosapentaenoic acid

EpOME

epoxy-octadecenoic acid

GDM

gestational diabetes

HETE

hydroxy-eicosatetraenoic acid

IsoP

isoprostane

LA

linoleic acid

LOD

limit of detection

LOX

lipoxygenase

PG

prostaglandin

PGE-M

prostaglandin E₂ metabolite

PUFA

polyunsaturated fatty acid

TN-E

tetranor-PGE₁

TX

thromboxane