Umbilical cord blood metabolomics reveal distinct signatures of dyslipidemia prior to bronchopulmonary dysplasia and pulmonary hypertension

Abstract

Pulmonary hypertension (PH) is a common consequence of bronchopulmonary dysplasia (BPD) and remains a primary contributor to increased morbidity and mortality among preterm infants. Unfortunately, at the present time, there are no reliable early predictive markers for BPD-associated PH. Considering its health consequences, understanding in utero perturbations that lead to the development of BPD and BPD-associated PH and identifying early predictive markers is of utmost importance. As part of the discovery phase, we applied a multiplatform metabolomics approach consisting of untargeted and targeted methodologies to screen for metabolic perturbations in umbilical cord blood (UCB) plasma from preterm infants that did (n = 21; cases) or did not (n = 21; controls) develop subsequent PH. A total of 1,656 features were detected, of which 407 were annotated by metabolite structures. PH-associated metabolic perturbations were characterized by reductions in major choline-containing phospholipids, such as phosphatidylcholines and sphingomyelins, indicating altered lipid metabolism. The reduction in UCB abundances of major choline-containing phospholipids was confirmed in an independent validation cohort consisting of UCB plasmas from 10 cases and 10 controls matched for gestational age and BPD status. Subanalyses in the discovery cohort indicated that elevations in the oxylipins PGE1, PGE2, PGF2a, 9- and 13-HOTE, 9- and 13-HODE, and 9- and 13-KODE were positively associated with BPD presence and severity. This expansive evaluation of cord blood plasma identifies compounds reflecting dyslipidemia and suggests altered metabolite provision associated with metabolic immaturity that differentiate subjects, both by BPD severity and PH development.

INTRODUCTION

Bronchopulmonary dysplasia (BPD) is a chronic lung disease triggered by barotrauma, volutrauma, and oxygen toxicity, resulting in arrest of alveolar development and varying degrees of interstitial fibrosis (17, 51). It is the most common complication of preterm birth, particularly in extremely preterm infants born before 28 wk of gestation, and affects over 10,000 infants per year in the United States (38). Pulmonary hypertension (PH) is a common complication of BPD and is characterized by remodeling of the pulmonary vasculature, increased vascular tone, abnormal pulmonary vasoreactivity, and right ventricular hypertrophy. The development of PH dramatically increases morbidity and mortality among preterm infants (51). The mechanisms mediating pulmonary resistance and altered reactivity are incompletely characterized. While BPD and PH share many risk factors, they have distinct pathophysiologies. It has been postulated that PH is triggered, in part, by fetal exposure to a harmful intrauterine environment (5, 15). This hypothesis is supported by associations between BPD-associated PH and both fetal growth restriction and oligohydramnios (43). Early screening echocardiograms are variable predictors of BPD-associated PH (9), and diagnostic biomarkers of this condition do not yet exist. Umbilical cord blood (UCB) screening presents an opportunity to understand in utero perturbations that lead to the development of PH in preterm infants in the presence of BPD.

Metabolomics is the study of small molecules and biochemical intermediates (metabolites), which are perturbed by or regulators of other regulatory mechanisms (e.g., genome, transcriptome, and proteome) and environmental stimuli. A metabolomics characterization can, therefore, present a detailed organismal phenotype. Successful application of metabolomics to provide novel insights into pathologies and develop prognostic and diagnostic markers of disease are growing and compelling (18, 52, 63). Since the pathophysiology of BPD and BPD-associated PH potentially has strong metabolic components (40), the application of metabolomics represents a promising approach for the identification of diagnostic biomarkers and insights into the metabolic alterations underlying these common diseases of very premature infants.

In the current study, we used a multiplatform metabolomics approach to evaluate alterations in the UCB plasma metabolome of preterm infants who did (n = 21; cases) or did not (n = 21; controls) develop subsequent PH. Metabolites of interest that were reflective of PH development were validated in an independent cohort consisting of UCB plasmas from 10 cases and 10 controls matched for gestational age and BPD status. To our knowledge, this study is the first comprehensive evaluation of metabolic alterations in the context of PH. We hypothesize that fetal circulating biomarkers identified in UCB represent in utero insults that trigger metabolic pathways associated with BPD severity and PH development.

METHODS

Patient sample collection.

Cord blood plasma samples were obtained from an established cohort of premature infants born at Prentice Women’s Hospital in Chicago, IL. This prospective cohort study enrolls all infants born <37 wk (2008 to present), with collection of cord blood at delivery and follow-up of premature infant outcomes, including BPD and PH (62). In the discovery cohort, 21 infants born at <32 wk with PH, determined by routine echocardiogram evaluation at 36 wk corrected-age using a previously published algorithm (15), were included in the study as PH cases. Twenty-one non-PH patient samples matched one to one by gestational age (±1 wk) were included as controls (Table 1). Severity (mild, moderate, severe) of BPD was defined according to the National Institutes of Health consensus workshop definition of BPD (28). In infants with gestational ages <32 wk, moderate BPD is defined by the need for <30% oxygen at 36 wk postmenstrual age or discharge (whichever comes first), whereas severe BPD is defined by the need for >30% oxygen and/or positive pressure at the same time point. In a subsequent sample of preterm infants enrolled at Prentice after the above discovery cohort, we identified a validation cohort (n = 20) in which criteria for cases was restricted to infants with moderate-severe BPD with PH (n = 10) and matched 1:1 with 10 patients of similar BPD severity (moderate-severe only) who did not have associated PH (Table 2). Cord blood plasma samples from these 20 patients were used for validation of the discovery cohort findings. All cord blood was collected at delivery into EDTA tubes and spun at 3,000 revolutions/min for 10 min in a refrigerated table-top centrifuge. Plasma was separated into aliquots and archived at −80°C. Samples were shipped to the West Coast Metabolomics Center at University of California, Davis. The above study was approved by the Institutional Review Board at Northwestern University. Maternal informed consent was obtained before participation of all mothers and their infants.

Table 1.

Discovery cohort patient characteristics

	PH−	PH+	P Value*
n	21	21
Subject characteristics
Birth weight, g	1,079 ± 384	969 ± 486	0.418
Gestation age, wk	27 ± 3	27 ± 3	0.948
Sex/male, n (%)	11 (52)	11 (52)
Ethnicity
African American, n (%)	6 (28.6)	7 (33.3)
Caucasian, n (%)	9 (42.9)	8 (38.0)
Hispanic, n (%)	6 (28.6)	4 (19.0)
Unknown, n (%)	0 (0.0)	2 (10)
BPD characteristics
Yes/no	14/7	20/1
Mild/moderate/severe	7/4/3	2/7/11

Values for weight and age are means ± SD; n = number of participants. PH, pulmonary hypertension; BPD, bronchopulmonary dysplasia.

^*Significance was determined by two-sided t-test.

Table 2.

Validation cohort patient characteristics

	PH−	PH+	P Value*
n	21	21
Subject characteristics
Birth weight, g	1,091 ± 201	878 ± 162	0.02
Gestation age, wk	27 ± 1	27 ± 1	0.13
Sex/male, n (%)	6 (60)	7 (70)
Ethnicity
African American, n (%)	0 (0.0)	5 (50.0)
Caucasian, n (%)	3 (30.0)	3 (30.0)
Hispanic, n (%)	2 (20.0)	0 (0.0)
Unknown, n (%)	5 (50.0)	2 (20)
BPD characteristics
Yes/no	10/0	10/0
Mild/moderate/severe	0/10	0/10

Values for weight and age are means ± SD; n = number of participants. PH, pulmonary hypertension; BPD, bronchopulmonary dysplasia.

^*Significance was determined by two-sided t-test.

Metabolomic, lipidomic, and oxylipin analyses.

The miniX database (48) was used as a laboratory information management system and for sample randomization before all analytical procedures. Samples were analyzed in a single batch on each platform.

For analysis of complex lipids, UCB plasma aliquots (20 µl) were extracted with methyl tert-butyl ether (MTBE) in the presence of analytical surrogates, as previously described (18), and separated with an Agilent 1290A Infinity ultra-high performance liquid chromatograph and detected with an Agilent 6530 accurate-Mass QTOF in both positive and negative mode (18, 12a).

Discovery cohort data were processed using MZmine v. 2.10; validation cohort data were processed in MS-DIAL (59). Lipids were identified by precursor accurate mass and manual MS/MS comparison to LipidBlast mass spectra (30) in addition to confirmation by authentic lipid standards. Data, reported as peak heights for the quantification ion (m/z) at the specific retention time for each annotated and unknown metabolite, were normalized to the class-specific internal standard (annotated) or to the internal standard that had the closest retention time (unknowns). A human plasma laboratory reference material (Bioreclamation plasma, BioIVT, Hicksville, NY) and method blanks were used to assess data quality.

For analysis of primary metabolites, half of the polar (bottom) layer from the MTBE lipid extract was dried under reduced pressure, derivatized by trimethylsilylation/methoximation, and metabolite levels were determined using an Agilent 7890A gas chromatograph coupled to a Leco Pegasus IV time-of-flight mass spectrometer, as previously described (18). Acquired spectra were further processed using the BinBase database (20, 48), including metabolite annotations by retention index and mass spectra matching (18). Data, reported as quantitative ion peak heights, were normalized by the sum intensity of all annotated metabolites across the entire study and used for further statistical analysis.

For analysis of biogenic amines, including arginine, citrulline, and ornithine, the remaining half of the polar (bottom) layer from the MTBE extract was dried under reduced pressure and resuspended in 60 µl of 80:20 acetonitrile/water containing the internal standards 1-cyclohexyl-ureido-3-dodecanoic acid (Sigma-Aldrich), 2 µg/ml l-arginine-¹⁵N₂ (Cambridge Isotope), and Val-Tyr-Val (Sigma-Aldrich). Metabolites were separated by injecting 3-μl sample in hydrophilic interaction chromatography mode using an Agilent 1290A Infinity ultrahigh performance liquid chromatograph pump with a Waters BEH Amide column (2.1 mm × 15 cm, 1.7-μm particles) with solvent A (10 mM ammonium formate + 0.125% formic acid, pH 3) and solvent B (95:5 vol/vol acetonitrile-water with 10 mM ammonium formiate + 0.125% formic acid, pH 3) and the following gradient: 0–2 min 100% (B), 2–7 min 70% (B), 7.7–9 min 40% (B), 9.5–10.25 min 30% (B), 10.25–12.75 min 100% (B), and 16.75 min 100% (B), with a flow rate of 0.4 ml/min. Column temperature was 40°C. Data were acquired on an Agilent 6550 QTOF mass spectrometer at 10,000 resolving power in 4.5 kV positive electrospray ionization mode at 2 Hz with scan range 60–1,200 Da, 3-Da precursor isolation window, and 45-eV collision energy. Agilent MassHunter software was used to quantify peaks according to their external standard curve (ornithine and citrulline) or the internal standard l-arginine-¹⁵N₂ (arginine) and presented as µM concentrations. Pooled Bioreclamation plasma (BioIVT, Hicksville, NY) was included to assess and monitor data quality. Data were processed as given above.

For analysis of oxylipins, analytes were isolated using the Ostro Pass Through Sample Preparation Plate (Waters, Milford, MA). After the addition of plasma (50 μl) to plate wells, plasma was spiked with antioxidants (5 µl of 0.2 mg/ml butylated hydroxytoluene-EDTA in 1:1 MeOH-water) and deuterated analytical surrogates (5 μl of 1,000 nM in MeOH). Acetonitrile (150 μl) with 1% formic acid was forcefully added to the sample and aspirated 3 times to mix. Samples were eluted into glass inserts containing 10 μl 20% glycerol by applying a vacuum at 15 Hg for 10 min. Eluent was dried by speed vacuum. Samples were reconstituted with the internal standards 1-cyclohexyl-ureido-3-dodecanoic acid and 1-phenyl ureido, 3-hexanoic acid at 100 nM (50:50 MeOH-acetonitrile) and filtered at 0.1 µm by centrifugation. Analytes in a 50-μl extract aliquot were separated with a Waters Acquity UPLC (Waters) using modifications of a previously published protocol (24, 53). Separated residues were detected by negative mode electrospray ionization using multiple reaction monitoring on an API 4000 QTrap (AB Sciex, Framingham, MA). Five- to ten-point calibration curves (r² ≥ 0.997) together with internal standard methods were used to quantify analytes. The acquired data were processed with Sciex MultiQuant version 3.0. Standards were either synthesized or purchased from Cayman Chemical (Ann Arbor, MI).

Data analysis.

Statistical analyses were carried out after sex difference and gestation period covariate adjustment of metabolite values. Specifically, a linear model was generated to describe differences in metabolite values because of sex difference and gestational period; the residuals were kept and used for further statistical analyses. Covariate-adjusted data were log₁₀-transformed, and significance was determined using a two-sided t-test or Welch t-test. The significance levels (i.e., P values) were adjusted for multiple hypothesis testing according to Benjamini and Hochberg (4) at a false discovery rate (FDR) of 5% (abbreviated pFDR <0.05). FDR correction was carried out independently on metabolites with known annotations only and on the entire data set. To evaluate the contributions of BPD on detected metabolite levels, a subanalysis was conducted on UCB from subjects that exhibited varying degrees of BPD but did not develop PH (Table 1). Kendall rank correlations were used to assess significant associations (P < 0.05) between UCB metabolite abundances and the presence and severity of BPD. Additionally, to minimize confounding effects because of differences in BPD presence and severity between the PH and non-PH cohort, we performed a secondary subanalysis on only those subjects with moderate to severe BPD who did (n = 18) or did not (n = 7) develop subsequent PH. For subanalyses, significance testing was carried out using a Welch t-test and adjusted for false discovery as described above. For the validation cohort, to test our a prior hypothesis of decreased choline-containing phospholipids in the PH group based on the results of the discovery cohort, significance was determined by one-sided Mann-Whitney U-test.

Network mapping was performed in Cytoscape (49) to encode statistical results through network edge and node attributes. Network edge construction was performed with MetaMapR (25).

All data and detailed information on sample preparation and data acquisition has been archived as project PR000207 can be accessed at https://doi.org/10.21228/M8N30T on the National Institutes of Health Metabolomics Workbench data repository (54).

RESULTS

Comparison of physical and biochemical characteristics.

In the discovery cohort, the mean gestational age was 27 wk for both non-PH and PH cohorts, which showed equivalent representation of boys and girls, gestational age, and birth weight (Table 1). The ethnicity for all study participants was 31.0% African American, 40.5% Caucasian, 23.8% Hispanic, and 4.8% Asian. (Table 1). The prevalence of BPD was higher in preterm infants who developed PH (20/21; 95%) relative to those who did not (14/21; 67%) (Table 1). Preterm infants with moderate to severe BPD were more likely to develop PH (relative risk: 2.9; confidence interval: 1.5–5.7; Fisher exact test P value: 0.001) (Table 1). Validation cohort characteristics included gestational ages of 27.6 and 26.8 wk for the non-PH and PH groups, respectively. Birth weight was significantly lower in the PH group (P < 0.02) (Table 2).

A two-sided t-test or Welch t-test, depending on individual analyte variance, was used to identify 87 significantly altered sex difference- and gestational period-adjusted metabolic features (raw P value < 0.05) between PH and control. No metabolite remained significant after FDR adjustment whether evaluating knowns only or the entire data set. Of the 87 metabolites with a raw P value < 0.05, 43 (49.4%) had known annotations (Table 3).

Table 3.

Significantly differential circulating metabolites and lipids in PH versus non-PH

Compound	Domain	PH−	PH+	Fold Change†	P Value‡
Primary metabolites
Glucose-6-phosphate	Carbohydrate	57.3 ± 40	149 ± 130	2.6	0.007
Threitol	Carbohydrate	174 ± 99	289 ± 220	1.7	0.045
Phosphoenolpyruvate	Organic acid	72.9 ± 64	184 ± 150	2.5	0.007
Asparagine	Amino acid or amide	413 ± 280	598 ± 240	1.4	0.041
Creatinine	Amino acid or amide	669 ± 550	1,010 ± 580	1.5	0.033
Choline	Other primary	5,150 ± 3,600	11,300 ± 8,200	2.2	0.036
Bisphosphoglycerol.NIST	Lipid	148 ± 84	255 ± 180	1.7	0.047
Complex lipids
Eicosenoic acid	Free fatty acid	334 ± 500	464 ± 280	1.4	0.024
α-Linolenic acid	Free fatty acid	506 ± 680	936 ± 700	1.9	0.006
Oleic acid	Free fatty acid	26,500 ± 47,000	36,200 ± 22,000	1.4	0.041
Palmitoleic acid	Free fatty acid	1,200 ± 1,700	2,320 ± 1,700	1.9	0.007
Acylcarnitine (C14:1)	Acylcarnitine	2,000 ± 1,500	3,900 ± 3,300	2	0.042
PC (p40:3) or (o40:4)	Phospholipid	2,400 ± 2,300	970 ± 750	0.4	0.021
PC (p36:2) or (o36:3)	Phospholipid	5,790 ± 3,700	3,330 ± 2,600	0.6	0.043
TG (60:11)	Triacylglyceride	6,840 ± 6,400	14,800 ± 11,000	2.2	0.016
TG (58:9)	Triacylglyceride	28,100 ± 19,000	54,300 ± 32,000	1.9	0.035
TG (58:8)	Triacylglyceride	64,400 ± 46,000	118,000 ± 71,000	1.8	0.035
TG (58:10)	Triacylglyceride	5,690 ± 4,300	11,800 ± 8,700	2.1	0.02
TG (56:9)	Triacylglyceride	4,830 ± 4,100	9,720 ± 5,700	2	0.016
TG (54:8)	Triacylglyceride	2,170 ± 1,800	3,670 ± 2,800	1.7	0.043
TG (53:4)	Triacylglyceride	4,270 ± 2,700	8,410 ± 6,100	2	0.039
Oxylipins
9,10-DiHOME	Diol	1.33 ± 0.89	2.55 ± 1.5	1.9	0.003
9,10-DiHODE	Diol	0.124 ± 0.079	0.187 ± 0.097	1.5	0.023
19,20-DiHDoPE	Diol	2.17 ± 1.4	3.31 ± 1.9	1.5	0.047
15,16-DiHODE	Diol	2.54 ± 3.8	4.91 ± 5.8	1.9	0.006
14,15-DiHETE	Diol	0.795 ± 0.64	1.49 ± 1.6	1.9	0.018
12,13-DiHOME (area ratio%)¥	Diol	1.17 ± 8.2	2.52 ± 1.8	2.2	0.003
9,10-EpOME	Epoxide	0.953 ± 0.82	1.71 ± 1.1	1.8	0.012
15,16-EpODE	Epoxide	2.02 ± 2.5	3.66 ± 3.9	1.8	0.039
12,13-EpOME	Epoxide	2.19 ± 1.4	3.71 ± 2.9	1.7	0.019
12,13-Ep-9-KODE	Epoxide	5.64 ± 2.7	8.92 ± 8.6	1.6	0.048
9-HOTE	Hydroxy acid	1.03 ± 0.6	2.18 ± 1.9	2.1	0.002
9-HODE	Hydroxy acid	19.3 ± 11	34.9 ± 31	1.8	0.008
9-HETE	Hydroxy acid	15.6 ± 10	37.1 ± 37	2.4	0.024
5-HETE	Hydroxy acid	7.98 ± 6.1	16.9 ± 15	2.1	0.02
5-HEPE	Hydroxy acid	0.477 ± 0.25	0.754 ± 0.46	1.6	0.022
17-HDoHE	Hydroxy acid	7.83 ± 7.7	13.9 ± 13	1.8	0.049
13-HOTE	Hydroxy acid	2.17 ± 1.5	4.8 ± 4.5	2.2	0.004
13-HODE	Hydroxy acid	22.4 ± 12	42.6 ± 40	1.9	0.008
9-KODE	Ketone	20.3 ± 9.6	32.1 ± 30	1.6	0.047
13-KODE	Ketone	8.41 ± 4.1	13.8 ± 13	1.6	0.034
PGE1	Prostanoid	0.29 ± 0.19	0.478 ± 0.35	1.6	0.034

Values are means ± SD. Primary metabolites and complex lipids are reported as relative peak heights. Oxylipins are reported as concentrations (nM), with the exception of 12,13-DiHOME, which is expressed as area ratio percent. HODE, 13-hydroxyoctadecadienoic acid; PC, phosphatidylcholine; PH, pulmonary hypertension; TG, triacylglyceride.

^†Fold change of PH⁺/PH⁻ and

^‡significance was determined by two-sided t-test or Welch t-test, depending on individual analyte variance.

Modest perturbations of primary metabolites precede BPD-associated PH.

Analysis of primary metabolites in UCB identified 7 differential (increased/decreased) metabolites between the UCB from subjects who did or did not develop subsequent PH. These included 2.6- and 2.5-fold elevations in the glycolytic intermediate glucose-6-phosphate and phosphoenolpyruvate (Table 3). Similarly, the amino acids asparagine and creatinine were found to be elevated 1.4- and 1.5-fold in neonates who subsequently developed PH, respectively (Table 3). Interestingly, choline was 2.2-fold higher in UCB plasma from the PH group.

Dyslipidemia precedes PH.

UCB plasma from preterm infants who developed subsequent PH showed elevations in circulating free fatty acids (FFAs) eicosanoic acid, α-linolenic acid (ALA), oleic acid, and palmitoleic acid (Table 3). Total FFAs were elevated 1.41-fold in PH relative to non-PH (Table 3). The PH-associated elevation in FFAs was met with a similar twofold elevation in circulating acylcarnitine (C14:1) and increases in numerous triacylglycerides (TG), which was led by a 2.2-fold increase in TG (60:11) (Table 3). Despite the elevation in select acylcarnitines and TGs, no significant differences were observed when looking at global content of the respective lipid classes (Table 4). This may suggest a shift in fatty acid composition, as suggested by the changes in specific fatty acids. Inverse to the elevations in aforementioned lipid species, we observed a 60 and 40% reduction in the select circulating plasmalogen phosphatidylcholine (PC) (plasmanyl-p40:3 or plasmenyl-o40:4) and PC (plasmanyl-p40:3 or plasmenyl-o40:4), respectively (Table 2). Ether linkages of plasmalogens are classified as “p” for alkyl and “o” for alkenyl. Similarly, global abundances of PCs and sphingomyelins (SMs) were reduced 25 and 37% in the PH group relative to control, respectively (Table 4).

Table 4.

PH-associated perturbations in classes of major structural lipids and oxylipins

Lipid Domain*	PH−	PH+	Fold Change§	P Value¶
Acylcarnitines (14)†	1,360,000 ± 610,000	1,720,000 ± 670,000	1.26	0.0915
Ceramides (6)	12,000 ± 6,900	9,090 ± 4,400	0.76	0.1369
Cholesterol esters (7)	215,000 ± 110,000	172,000 ± 76,000	0.8	0.3503
Free fatty acids (19)	79,600 ± 81,000	105,000 ± 45,000	1.32	0.0123
Glucosyl ceramides (2)	17,900 ± 14,000	7,200 ± 5,100	0.4	0.0599
Lysophosphatidyl cholines (12)	410,000 ± 2e + 05	347,000 ± 150,000	0.85	0.3189
Lysophosphatidyl ethanolamines (4)	9,470 ± 4,900	7,100 ± 3,600	0.75	0.2222
Phosphatidyl cholines (34)	11,200,000 ± 3,500,000	8,390,000 ± 3,500,000	0.75	0.017
Phosphatidyl ethanolamines (3)	17,200 ± 7,800	15,800 ± 11,000	0.92	0.3332
Plasmalogen-PCs (24)	384,000 ± 280,000	240,000 ± 150,000	0.63	0.0626
Plasmalogen-PEs (3)	9,270 ± 5,700	9,730 ± 8,700	1.05	0.753
Sphingomyelins (26)	2,050,000 ± 980,000	1,310,000 ± 640,000	0.64	0.0151
Triglycerides (48)	4,810,000 ± 2,200,000	5,790,000 ± 3,200,000	1.2	0.3394
C18_Diols (5)	5.21 ± 4.8	9.33 ± 7.4	1.8	0.0069
C20_Diols (6)	15.1 ± 8.6	20.3 ± 12	1.34	0.1299
C22_Diols (1)	2.17 ± 1.4	3.31 ± 1.9	1.5	0.047
C18_Epoxides (5)	18.3 ± 8	26.4 ± 13	1.4	0.0092
C20_Epoxides (2)	3.69 ± 1.8	7.44 ± 13	2.02	0.1704
C18_Hydroxy acids (4)	44.9 ± 25	84.5 ± 75	1.9	0.0075
C20_Hydroxy acids (10)	263 ± 160	576 ± 1100	2.2	0.1611
C22_Hydroxy acids (2)	80.2 ± 52	183 ± 400	2.3	0.224
C18_Ketones (2)	28.7 ± 14	45.8 ± 43	1.6	0.0441
C20_Prostacyclins or leukotrienes (7)	67.2 ± 67	94.2 ± 100	1.4	0.3806

Values (means ± SD) represent the summed intensities of individual annotated lipids belonging to the respective lipid class. Complex lipids are presented as summed peak intensities; oxylipins are shown as summed concentrations (nM). PH, pulmonary hypertension.

^*Major lipid classes,

^†number of measured molecules,

^§fold change of PH⁺/PH⁻, and

^¶significance was determined by two-sided t-test or Welch t-test, depending on individual analyte variance.

Oxylipins are elevated before the onset of BPD-associated PH.

Of the 43 altered annotated metabolites, 21 (49%) were oxygenated lipids and were uniformly elevated in PH cord blood plasma (Table 3). The largest difference was a 2.4-fold increase in 9-HETE, a nonenzymatic oxidation product of arachidonic acid (AA) (37). Notably, many of the differentially elevated oxylipins were either derived from linoleic acid (LA) (8 of 21) or ALA (5 of 21) (Fig. 1), including nonenzymatic oxidation, lipoxygenase (LOX), cytochrome P450 (CYP450), and soluble epoxide hydrolase (sEH) metabolites (Fig. 1). Ratios of dihydroxy acids and their concordant epoxides, a measure of sEH activity, indicated no significant differences between the two cohorts (data not shown), suggesting an increased production of oxylipins via the P450/sEH pathway rather than a specific increase in sEH activity. When evaluating global abundances of class-specific oxylipins, circulating C18-diols, C22-diols, C18-epoxides, C18-hydroxy acids, and C18-ketones were elevated 1.8-, 1.5-, 1.4- 1.9-, and 1.6-fold in PH compared with control, respectively (Table 4).

Fig. 1.

Biochemical network displaying differences in lipid signaling mediators between pulmonary hypertension (PH) and non-PH. A biochemical network displaying metabolic differences in lipid signaling mediators is shown. Node shape represents the type of lipid mediator. Node color and size reflects significance (gray: not detected or measured; black: not significantly different; red: increased) and fold change in PH relative to control, respectively. Edge color represents relationship between nodes [blue: elongation/desaturation of fatty acid; green: enzymatic; yellow: nonenzymatic (oxidation)]. Edge labels indicate the enzyme or reaction that mediates the transformation. CBR1, carbonyl reductase 1; COX, cyclooxygenase; Cyp(epox), epoxygenase; DGLA, dihomo-γ-linolenic acid; GPx, glutathione peroxidase; LOX, lipoxygenase; LTA4, leukotriene A4; PGDH, 15-hydroxyprostaglandin dehydrogenase; PGS, prostaglandin synthase; sEH, soluble epoxide hydrolase.

In numerous samples, thromboxane B₂ and 12-HETE values were greatly elevated (>10-fold median) in a subset of PH and control samples, suggesting platelet activation/degranulation either during preanalytical plasma handling or in these subjects (39). Therefore, we suspect that the data of these two metabolites may have been compromised and could not be reliably interpreted for this purpose in the study.

Subanalysis of associations between fetal circulating metabolites and BPD.

Given the heterogeneity of BPD status within our examined cohort, we aimed to determine whether putative associations between UCB metabolites and BPD presence and severity exist. For this analysis, only UCB from subjects that did not develop subsequent PH were considered. Using nonparametric correlations, we identified 47 metabolites that indicated significant associations with BPD status (Table 5). Notably, 8 of the 47 significantly associated metabolites were oxygenated lipids (Fig. 2) derived from cyclooxygenase (COX) AA metabolites (PGE1, PGE2, and PGF2α) or LOX and/or nonenzymatic LA products [9-hydroxyoctadecadienoic acid (HODE), 9-KODE, 13-HODE, and 13-KODE] and ALA products (9- and 13-HOTE).

Table 5.

Associations between cord blood plasma metabolites and BPD severity

Compound	Kendal τ Correlation Coefficient	P Value	FDR-Adjusted PH	Direction of Association
Primary metabolites with biogenic amines
Erythritol	0.4652	0.0049	0.2191	Positive
Glucose-6-phosphate	0.4323	0.0079	0.2615	Positive
Aspartic acid	0.4214	0.0081	0.2615	Positive
Conduritol-β-epoxide	0.4542	0.0100	0.2615	Positive
Palmitoleic acid	0.4542	0.0112	0.2639	Positive
Xanthine	0.3886	0.0131	0.2675	Positive
Maltose	0.4214	0.0214	0.3349	Positive
Oleic acid	−0.4104	0.0218	0.3349	Negative
Glycerol-α-phosphate	0.3667	0.0267	0.3647	Positive
Methylgalactose-NIST	−0.3995	0.0267	0.3647	Negative
Proline	0.3557	0.0270	0.3647	Positive
Palmitic acid	−0.3776	0.0280	0.3671	Negative
Piperidinobenzonitrile-NIST	−0.3119	0.0374	0.4350	Negative
Phosphoethanolamine	0.3338	0.0387	0.4350	Positive
Hydroxylamine	−0.3338	0.0472	0.4877	Negative
Niacinamide	0.4761	0.0034	0.1881	Positive
Urocanic acid	−0.4214	0.0159	0.2954	Negative
Methyl histidine	−0.3119	0.0495	0.4877	Negative
Complex lipids
PE (p-36:4) or PE (o-36:5)	0.5527	0.0003	0.0452	Positive
TG (58:8)	−0.5637	0.0003	0.0452	Negative
PE (p-38:4) or PE (o-38:5)	0.4980	0.0010	0.1090	Positive
TG (58:9)	−0.4980	0.0020	0.1399	Negative
TG (60:11)	−0.4980	0.0020	0.1399	Negative
TG (54:4)	−0.4871	0.0022	0.1399	Negative
TG (54:3)	−0.3667	0.0089	0.2615	Negative
TG (56:9)	−0.4323	0.0089	0.2615	Negative
TG (56:8) A	−0.4104	0.0091	0.2615	Negative
Ceramide (d34:1)	0.4214	0.0099	0.2615	Positive
TG (58:10) 1	−0.4104	0.0112	0.2639	Negative
TG (53:4) A	−0.3995	0.0125	0.2675	Negative
TG (56:3)	−0.3776	0.0139	0.2697	Negative
TG (56:4)	−0.3776	0.0181	0.3230	Negative
TG (54:8) B	−0.3776	0.0228	0.3395	Negative
PE (p-38:5) or PE (o-38:6)	0.3338	0.0384	0.4350	Positive
TG (58:10)	−0.3448	0.0407	0.4426	Negative
SM (d34:2)	0.3119	0.0495	0.4877	Positive
Oxylipins
PGE1	0.5746	0.0002	0.0452	Positive
PGE2	0.4652	0.0046	0.2191	Positive
9-HOTE	0.4323	0.0132	0.2675	Positive
PGF2a	0.3886	0.0210	0.3349	Positive
13-HODE	0.3995	0.0214	0.3349	Positive
12,13-Ep-9-KODE	0.3776	0.0303	0.3866	Positive
9-HODE	0.3667	0.0334	0.4142	Positive
9-KODE	0.3667	0.0390	0.4350	Positive
13-KODE	0.3557	0.0468	0.4877	Positive

Associations were determined using Kendall τ correlation test. FDC, false discovery rate; PE, phosphatidylethanolamine; PH, pulmonary hypertension; SM, sphingomyelin; TG, triacylglyceride.

Fig. 2.

Associations between circulating metabolites in cord blood plasma and bronchopulmonary dysplasia (BPD) severity. Box and whisker plots are shown for respective metabolites based on subjects stratified by presence and severity of BPD. Associations were determined using the Kendall τ correlation test. *P < 0.05, **P < 0.01, and ***P < 0.001.

Subanalysis of UCB from subjects with moderate to severe BPD.

To account for the potential effects of BPD status on our initial findings, we conducted an additional subanalysis on UCB samples from subjects with moderate to severe BPD who did (n = 18) or did not (n = 7) develop subsequent PH. Subanalysis identified PH-associated elevations in circulating nonadecanoic acid, whereas lysine, ornithine, phenylalanine, mannitol, phosphate, and niacinamide were decreased (Table 6). Alterations in metabolic products, including an increase in the glycolytic intermediate phosphoenolpyruvate and a decrease in N⁶,N⁶,N⁶,-trimethyl-l-lysine, a compound required for carnitine synthesis, may suggest differences in energy metabolism. None of the seven previously identified primary metabolites (Table 3) retained significance when comparing all subjects.

Table 6.

Significantly differential circulating metabolites and lipids that accompany development of PH based on subanalysis of subjects with moderate to severe BPD

Compound	Domain	Controla	PH+	Fold Changea	P Valueb
Primary metabolites with biogenic amines
Lysine	Amino acid	63,500 ± 27,000	39,900 ± 21,000	0.63	0.0369
N6,N6,N6-Trimethyl-l-lysine	Amino acid	10,400 ± 4,200	5,240 ± 3,300	0.5	0.0322
Ornithine*	Amino acid	57.1 ± 12	38.7 ± 27	0.68	0.0154
Phenylalanine	Amino acid	138,000 ± 32,000	90,300 ± 65,000	0.65	0.0033
Mannitol	Carbohydrate	1,710 ± 870	869 ± 660	0.51	0.0307
Nonadecanoic acid	Lipid	183 ± 72	345 ± 190	1.89	0.0355
Phosphoenolpyruvate	Organic acid	82.2 ± 70	175 ± 94	2.13	0.0164
Hydroxybutyric acid	Organic acid	16,800 ± 5,000	8,240 ± 7,400	0.49	0.0178
Phosphate	Other primary	42,600 ± 17,000	23,200 ± 17,000	0.54	0.049
Niacinamide	Vitamin	63,600 ± 20,000	50,800 ± 39,000	0.8	0.0452
Complex lipids
CE (18:3)	Cholesterol ester	4,990 ± 3,500	2,380 ± 1,500	0.48	0.0286
CE (20:3)	Cholesterol ester	10,900 ± 4,400	6,840 ± 3,000	0.63	0.0245
CE (20:4)	Cholesterol ester	97,700 ± 37,000	51,000 ± 27,000	0.52	0.0236
Ceramide (d34:1)	Ceramide	2,060 ± 730	768 ± 410	0.37	0.0039
Ceramide (d40:1)	Ceramide	2,110 ± 830	978 ± 460	0.46	0.0171
Ceramide (d42:1)	Ceramide	7,000 ± 4,800	3,480 ± 2,100	0.5	0.0422
Ceramide (d42:2)	Ceramide	3,140 ± 780	1,860 ± 930	0.59	0.0466
Gal-Gal-Cer (d18:1/16:0) or lactosylceramide (d18:1/16:0)	Glucosyl ceramide	29,300 ± 15,000	5,910 ± 4,400	0.2	1.00E-04
PC (30:0)	Phospholipid	173,000 ± 99,000	81,300 ± 49,000	0.47	0.044
PC (38:4) A	Phospholipid	1,190,000 ± 460,000	784,000 ± 580,000	0.66	0.0163
PC (40:4)	Phospholipid	26,900 ± 15,000	11,700 ± 7,200	0.43	0.0418
PC (40:5) A	Phospholipid	62,400 ± 41,000	19,900 ± 20,000	0.32	0.0108
PC (o-32:0)	Phospholipid	53,800 ± 33,000	16,700 ± 12,000	0.31	0.0201
PC (p-32:0) or PC (o-32:1)	Phospholipid	28,400 ± 16,000	12,200 ± 9,500	0.43	0.0328
PC (p-34:1) or PC (o-34:2) A	Phospholipid	10,500 ± 5,600	5,110 ± 3,600	0.49	0.043
PC (p-36:2) or PC (o-36:3)	Phospholipid	7,710 ± 3,900	3,400 ± 2,700	0.44	0.0167
PC (p-36:3) or PC (o-36:4)	Phospholipid	214,000 ± 130,000	67,900 ± 47,000	0.32	0.0207
PC (p-36:4) or PC (o-36:5)	Phospholipid	928 ± 370	543 ± 280	0.59	0.0084
PC (p-38:2) or PC (o-38:3)	Phospholipid	3,980 ± 2,800	1,160 ± 900	0.29	0.038
PC (p-38:3) or PC (o-38:4)	Phospholipid	34,600 ± 26,000	9,680 ± 7,400	0.28	0.0216
PC (p-38:4) or PC (o-38:5) A	Phospholipid	708 ± 440	358 ± 220	0.51	0.0294
PC (p-38:4) or PC (o-38:5) A	Phospholipid	92,400 ± 64,000	33,600 ± 25,000	0.36	0.0231
PC (p-40:3) or PC (o-40:4)	Phospholipid	4,030 ± 3,100	825 ± 650	0.2	0.0133
PC (p-40:4) or PC (o-40:5) A	Phospholipid	4,990 ± 3,500	1,210 ± 830	0.24	0.0011
PC (p-40:5) or PC (o-40:6)	Phospholipid	4,420 ± 3,100	1,990 ± 1,800	0.45	0.0187
PC (p-40:7) or PC (o-40:8)	Phospholipid	16,200 ± 9,900	6,240 ± 4,600	0.39	0.0261
PC (p-42:4) or PC (o-42:5)	Phospholipid	5,300 ± 4,400	779 ± 620	0.15	0.0002
PC (p-44:4) or PC (o-44:5)	Phospholipid	7,460 ± 5,900	1,590 ± 1,100	0.21	0.0249
PE (p-36:4) or PE (o-36:5)	Phospholipid	5,740 ± 2,300	3,340 ± 1,600	0.58	0.0164
SM (d34:0)	Sphingomyelin	110,000 ± 51,000	41,700 ± 28,000	0.38	0.0209
SM (d34:1)	Sphingomyelin	788,000 ± 330,000	323,000 ± 180,000	0.41	0.0465
SM (d34:2)	Sphingomyelin	121,000 ± 49,000	46,600 ± 26,000	0.39	0.0207
SM (d36:0)	Sphingomyelin	81,600 ± 36,000	38,300 ± 23,000	0.47	0.0462
SM (d36:1)	Sphingomyelin	225,000 ± 110,000	101,000 ± 69,000	0.45	0.0422
SM (d38:0)	Sphingomyelin	14,500 ± 7,800	6,480 ± 3,700	0.45	0.042
SM (d38:1)	Sphingomyelin	50,500 ± 24,000	24,800 ± 16,000	0.49	0.0383
SM (d39:1)	Sphingomyelin	4,550 ± 2,600	2,370 ± 1,800	0.52	0.0366
SM (d40:1)	Sphingomyelin	115,000 ± 54,000	38,700 ± 25,000	0.34	0.0143
SM (d40:2) A	Sphingomyelin	77,000 ± 56,000	32,800 ± 27,000	0.43	0.0342
SM (d41:1)	Sphingomyelin	12,600 ± 8,000	5,070 ± 4,200	0.4	0.0348
SM (d41:2)	Sphingomyelin	749 ± 290	365 ± 190	0.49	0.0189
SM (d42:2) A	Sphingomyelin	441,000 ± 220,000	150,000 ± 84,000	0.34	0.0317
TG (53:4) A	Triacylglyceride	2,820 ± 3,000	8,670 ± 6,300	3.07	0.0414
Oxylipins
19,20-DiHDoPE	Diol	1.54 ± 1.5	3.28 ± 2.1	2.13	0.0275

Values are means ± SD. Primary metabolites, biogenic amines, and complex lipids are reported as relative peak heights, oxylipins are reported as concentrations (nM). PC, phosphatidylcholine; PE, phosphatidylethanolamine; PH, pulmonary hypertension; SM, sphingomyelin; TG, triacylglyceride.

^aFold change of PH/control,

^bsignificance was determined by two-sided t-test or Welch t-test, depending on individual analyte normality, and

^*nM concentrations.

Alterations in complex lipids remained consistent with our observations from the initial analysis. Significant reductions in several choline-containing phospholipids and elevations in FFAs and a TG (53:4) were characteristic of UCB from preterm infants who developed subsequent PH (Table 6). Cholesterol esters, nonesterified ceramides, and glucosyl-ceramides were also found to be significantly reduced in the PH group relative to controls (Table 6). These findings suggest that the observed reduction in UCB structural lipids, particularly choline-containing phospholipids, and elevation in FFAs and select TGs are more closely related to PH than BPD.

It is interesting to note that many of the identified oxylipins that were significantly different when including all subjects failed to retain significance in our subanalysis. The lone exception was the docosahexaenoic acid metabolite 19,20-DiHDoPE, which was elevated in the PH cohort (Table 6). This would suggest that the elevation in aforementioned oxylipins are more strongly linked to development and progression of BPD rather than PH, a notion that is supported by the observations that many of the oxylipins were also positively associated with BPD status (Fig. 2).

Validation of PH-associated complex lipids in infants with moderate-severe BPD.

To validate our above-mentioned findings, we performed metabolomic analyses on an independent cohort consisting of 10 cases and 10 controls. Importantly, our validation cohort was matched for BPD status (moderate-severe only), thereby removing potential bias because of the presence or severity of BPD. To this end, we focused our analyses on complex lipids on the basis that lipid species, particularly choline-containing lipids, remained statistically different between cases and controls in the discovery cohort, even when taking into account the severity of BPD. Consistent with our initial findings, choline-containing lipid species were significantly reduced in cases relative to controls in the validation cohort (Table 7).

Table 7.

Validation cohort PH-associated perturbations in major structural lipids

Lipid Domain*	PH−	PH+	Fold Change§	P Value¶
Acetylcarnitines (7)†	221,017 ± 52,356	183,407 ± 58,724	0.83	0.07
Ceramides (24)	385,211 ± 125,023	303,119 ± 107,469	0.79	0.1
Cholesterol esters (14)	7,912,370 ± 2,533,645	6,128,807 ± 2,474,167	0.77	0.08
Diacylglycerols (8)	56,515 ± 32,009	54,225 ± 47,990	0.96	0.1
Free fatty acids (14)	889,692 ± 406,594	708,626 ± 275,991	0.80	0.2
Glycosphingolipids (10)	133,211 ± 36,978	88,011 ± 33,673	0.66	0.0011
Lysophatidylcholines (27)	7,318,071 ± 3,117,915	5,376,811 ± 2,473,810	0.73	0.1
Lysophosphatidylethanolamines (7)	70,619 ± 25,357	52,895 ± 21,757	0.75	0.1
Phosphatidylcholines (109)	100,659,668 ± 13,666,756	87,323,531 ± 21,838,007	0.87	0.06
Phosphatidylethanolamines (11)	331,513 ± 138,258	253,044 ± 121,819	0.76	0.032
PlasmalogenLPCs (3)	55,573 ± 15,446	41,833 ± 14,677	0.75	0.022
PlasmalogenPCs (49)	4,175,912 ± 1,377,918	3,018,397 ± 1,455,014	0.72	0.038
PlasmalogenPEs (17)	116,166 ± 36,730	92,079 ± 37,669	0.79	0.045
Sphingomyelins (70)	23,282,302 ± 5,034,895	16,654,834 ± 6,006,540	0.72	0.0093
Triacylglycerols (97)	26,734,745 ± 14,799,333	26,729,596 ± 15,701,640	1.00	0.5

Values (means ± SD) represent the summed intensities of individual annotated lipids belonging to the respective lipid class. Complex lipids are presented as summed peak intensities. PC, phosphatidylcholine; PE, phosphatidylethanolamine; PH, pulmonary hypertension.

^*Major lipid classes,

^†number of measured molecules,

^§fold change of PH⁺/PH⁻, and

^¶significance was determined by one-sided Mann-Whitney U-test.

DISCUSSION

In the current study, a combined metabolomics approach consisting of an untargeted analysis of primary metabolites and complex lipids coupled with a targeted analysis of biogenic amines and lipid signaling mediators identified 1,656 metabolic features, 407 with known structures, in UCB plasma of preterm infants who later developed PH compared with those who did not. To our knowledge, this is the first study to expansively evaluate the systemic alterations in the metabolic profiles of cord blood plasma from preterm infants that precede development of BPD-associated PH. Overall, the observed changes implicate lipid signaling, lipid biosynthesis, and oxidative stress as cofactors associated with the onset of PH in preterm infants.

Although UBC is most commonly acquired for use of its rich stem cell content, it also contains the nutrients and metabolites that may provide insight into the events occurring during fetal growth and help predict the development of disease after birth (31). In relationship to many of our untargeted analysis metabolites investigated, nutrient metabolism and delivery to the fetus are known to be altered by the placenta during pregnancy (11). Furthermore, potentially related to our oxylipin analysis, the placenta expresses phospholipase A₂ and lipoprotein lipase (27) in addition to COX, LOX, and CYP450 (47).

Rodent studies have found associations between BPD and early increases in COX-2 and 5-LOX metabolites (44). Previous metabolomics studies have identified select sugar and amine derivatives in late gestation amniotic fluid and urine collected at or near birth of preterm infants with BPD (19, 41), as well as phospholipids in exhaled breath condensate of adolescent BPD patients (13). Specific long-chain polyunsaturated fatty acid phospholipids in UCB have also been identified as potential predictors of BPD, particularly in infants born at less than 28 wk of gestation (6). A previous study of lipid metabolites in the first 3 days of life in 272 premature infants found higher absolute levels of several HETEs in infants who developed BPD, although these differences became nonsignificant when adjusting for gestational age (45). Serum phospholipids were higher in premature infants than samples of UCB from infants matched for gestational age, predominantly because of increases in LA-containing PC and phosphatidylethanolamine (6). Serum choline levels decrease dramatically from UCB levels within 48 h of birth in preterm infants but not in term infants (8). Combined, these previous observations suggest immaturity in lipid metabolism in preterm infants that is likely exacerbated by the abrupt loss of maternal/placental regulation and the challenges associated with current enteral and parenteral nutritional approaches for extremely preterm infants (high in LA, low in choline).

This study focused on a comprehensive analysis of cord blood as a potential marker of PH risk. When considering all subjects, striking differences in the UCB composition of complex lipids and lipid mediators were observed in neonates who did not develop subsequent PH. Analysis of the lipidome indicated that UCB plasma from PH-positive infants had lower circulating PCs and SMs and elevated choline. These findings were confirmed in an independent cohort of UCB plasma from cases and controls providing confidence in our initial findings. These findings are also consistent with a growth-restriction induced PH rat model, where we recently reported substantial reductions in plasma PC levels in rats that developed PH as compared with controls (34). Moreover, these circulating phospholipids appear to be a primary product of placental metabolism, with fetal lipoprotein metabolism maturing late in the neonatal period of normal pregnancies (29). Phospholipid metabolism is directly linked to the physiological production of pulmonary surfactants that are 70–80% PCs. The late maturation of PC biosynthesis is a well-known risk marker and contributor for BPD development and PH emergence (1), and administration of the PC precursor CDP-choline attenuated hyperoxia-induced lung damage in a neonatal model of BPD (14). In our cohort, circulating levels of choline were higher in the PH positive group. Notably, when evaluating only moderate to severe BPD subjects stratified by PH status, we retained the significant PH-associated reductions in the aforementioned choline-containing phospholipids in addition to reductions in cholesterol esters, nonesterified ceramides, and glucosylceramides. However, choline was not significantly different in the subanalysis; instead, choline tended to be positively associated with BPD presence and severity (Table 4), consistent with previous findings (14).

Despite the potential role for disruption of nitric oxide synthase function (2, 22, 40), we did not observe circulating differences in arginine, citrulline, or ornithine between the PH-positive and -negative groups when including all subjects. However, if only moderate to severe BPD subjects were considered, ornithine concentrations were reduced 1.5-fold in the PH group (data not shown). Moreover, the ratio of citrulline-to-ornithine tended to be higher in the PH group relative to the non-PH group (0.41 ± 0.70 vs. 0.15 ± 0.07; P value: 0.147), suggesting an increase in NO production.

A subanalysis to investigate metabolites associated with BPD severity revealed a strong positive correlation with numerous COX-derived prostaglandins and 5- or 15-LOX-derived LA hydroxy acids and ketones. Interestingly, the positively correlated PGE2 and PGE1 have been shown to be secreted by the placenta to maintain prenatal patency of the ductus arteriosus (23). Persistence of a moderate to large ductus arteriosus following birth is a risk factor for BPD. Furthermore, the previously mentioned hydroxy acids and ketones are associated with placental signaling via activation of peroxisome proliferator-activated receptor (PPAR)γ (10, 35, 61). Specifically, 9- and 13-HODE, and 13-oxo-ODE are putative ligands of PPARγ, the expression of which, in the placenta, has been shown to be vital for proper cardiac tissue development (3) and implicated in the regulation of PH development (55). While these PPARγ ligand oxylipins are typically produced by LOX in a controlled manner, they can be produced nonenzymatically during disease processes when oxidative stress is increased (60). As a result, in these conditions, they are associated with elevated inflammation, apoptosis, and vasoconstriction via GPR132 (G2A) receptor activation by 9-HODE and 9-HpODE and transient receptor potential cation channel subfamily V member 1 and CD36 activation by 13-HODE (33, 36, 60). Furthermore, we observed PH-associated elevations in 9-HETE, a nonenzymatic oxidation product of AA and previously noted marker of oxidative stress and disease (50). Oxidative stress is heavily implicated in the etiology of both BPD and PH, facilitating alterations in the pulmonary vasculature and right ventricle and decoupling endothelial nitric oxide synthase (16).

CYP450-derived fatty acid epoxides and sEH-derived dihydroxy fatty acids have been observed in the placenta previously and are associated with altered uteroplacental remodeling (26). We observed changes in CYP2 epoxygenated and dihydroxy acid linoleate and alpha linolenate-derived oxylipins, including 9,10-EpOME 15,16-EpODE; 9,10-DiHOME; 12,13-DiHOME; 14,15-DiHETE; 15,16-DiODE; and 9,10-DiODE. When evaluating the ratio of the respective dihydroxy acids to their parent epoxides, no significant differences were found between the PH and non-PH group, collectively suggesting this was an increase in both P450 and sEH activity. Regardless, research indicates that linoleate epoxygenated and dihydroxy acid forms may be involved in a variety of biological processes from promotion of neutrophil accumulation, respiratory burst, and pulmonary edema to proliferation of hematopoietic progenitor cells and vascularization (21, 56, 58). Respiratory burst and pulmonary edema are associated with PH and BPD, respectively (12, 46).

An additional subanalysis was conducted in moderate-severe BPD subjects who did or did not develop PH to better differentiate observed alterations that were more associated with BPD versus those that were more associated with PH. Interestingly, all of the lipidomic alterations in terms of reduced PCs and SMs remained statistically significant. Furthermore, decreases in total ceramides and plasmalogen PCs were observed. These findings were recapitulated in our validation cohort of BPD status-matched cases and controls. All of these support the previous discussion regarding their role in PH development.

Our findings collectively demonstrate dyslipidemia in UCB plasma from preterm infants who go on to develop PH. This dyslipidemia is predominately characterized by reductions in circulating choline-containing phospholipids. Consequently, these findings may have value for characterizing cord blood metabolites associated with preterm infants who are at risk for developing PH. We acknowledge that the sample sizes for our cohorts are small; however, it must be emphasized that 1) accessibility to UCB plasma of premature infants diagnosed with BPD that do or do not develop subsequent PH is limited and 2) despite sample size limitations, we were able to confirm that choline-containing phospholipids are reduced in UCB plasma of cases relative to controls in an independent validation cohort, providing confidence in our initial findings.

Larger studies with analysis including both gestational age and BPD severity will be required to fully evaluate the diagnostic value of altered lipid profiles in the context of PH development. While the groups were well matched by gestational age and there were balanced proportions of infant sex difference and low birth status, it must be noted that the data were not adjusted for any of the other numerous prenatal and postnatal factors that influence the development of BPD and PH, including oligohydramnios, maternal preeclampsia, obesity, diabetes, corticosteroid administration, and cesarean section delivery, and neonatal pulmonary abnormalities, among numerous others (32, 42). A larger sample will be needed to take into account these and other important covariates of extremely preterm birth. Due to the possible influence of maternal characteristics on metabolite differences observed in the UCB (7), future studies that concurrently analyze maternal plasma would be optimal.

In conclusion, UCB plasma from preterm infants that did or did not develop BPD-associated PH was characterized by states of dyslipidemia and suggest metabolic immaturity, associated with deficiencies in lipids critical for proper growth and development. Specifically, preterm infants that developed PH displayed reductions in major choline-containing phospholipids and elevations in choline, suggesting alterations in PC biosynthesis. BPD severity was largely characterized by alterations in oxylipin concentrations. UCB reductions in choline-containing phospholipids hold promise as early screening markers for development of BPD-associated PH in preterm infants.

GRANTS

This research is funded by National Institutes of Heath (NIH) Grants U24-DK-097154 (to O. Fiehn) and K23-HL-093302 (to K. Mestan), with instrument support by S10-RR-031630 (to O. Fiehn) and by US Department of Agriculture Intramural Project 2032-51530-022-00D (to J. W. Newman).

DISCLAIMERS

The US Department of Agriculture is an equal opportunity employer and provider.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

K.M., R.H.S., and S.W. conceived and designed research; K.M. and R.H.S. performed experiments; M.R.L.F., J.F.F., and D.G. analyzed data; M.R.L.F., J.F.F., J.W.N., and S.W. interpreted results of experiments; M.R.L.F. and J.F.F. prepared figures; M.R.L.F. and J.F.F. drafted manuscript; M.R.L.F., J.F.F., D.G., T.L.P., J.W.N., O.F., M.A.U., K.M., R.H.S., and S.W. edited and revised manuscript; M.R.L.F., J.F.F., D.G., T.L.P., J.W.N., O.F., M.A.U., K.M., R.H.S., and S.W. approved final version of manuscript.

ENDNOTE

At the request of the authors, readers are herein alerted to the fact that additional materials related to this manuscript may be found at https://doi.org/10.21228/M8N30T on the NIH Metabolomics Workbench data repository (project PR000207). These materials are not a part of this manuscript and have not undergone peer review by the American Physiological Society (APS). APS and the journal editors take no responsibility for these materials, for the Web site address, or for any links to or from it.