Article pubs.acs.org/jpr
1
H NMR Derived Metabolomic Profile of Neonatal Asphyxia in Umbilical Cord Serum: Implications for Hypoxic Ischemic Encephalopathy
Stacey N. Reinke,†,∥,# Brian H. Walsh,‡,# Geraldine B. Boylan,‡,§ Brian D. Sykes,∥ Louise C. Kenny,§,⊥ Deirdre M. Murray,‡,§ and David I. Broadhurst*,† †
Department of Medicine and ∥Department of Biochemistry, University of Alberta, Edmonton, Alberta, Canada ‡ Neonatal Brain Research Group, Department of Paediatrics and Child Health and §The Irish Centre for Fetal and Neonatal Translational Research (INFANT), University College Cork and Cork University Maternity Hospital, Wilton, Cork, Ireland ⊥ Department of Obstetrics and Gynaecology, University College Cork, Cork, Ireland S Supporting Information *
ABSTRACT: Neonatal hypoxic ischemic encephalopathy (HIE) is a severe consequence of perinatal asphyxia (PA) that can result in life-long neurological disability. Disease mechanisms, including the role and interaction of individual metabolic pathways, remain unclear. As hypoxia is an acute condition, aerobic energy metabolism is central to global metabolic pathways, and these metabolites are detectable using 1H NMR spectroscopy, we hypothesized that characterizing the NMR-derived umbilical cord serum metabolome would offer insight into the consequences of PA that lead to HIE. Fifty-nine at-risk infants were enrolled, together with 1:1 matched healthy controls, and stratified by disease severity (n = 25, HIE; n = 34, non-HIE PA). Eighteen of 37 reproducibly detectable metabolites were significantly altered between study groups. Acetone, 3hydroxybutyrate, succinate, and glycerol were significantly differentially altered in severe HIE. Multivariate data analysis revealed a metabolite profile associated with both asphyxia and HIE. Multiple-linear regression modeling using 4 metabolites (3-hydroxybutyrate, glycerol, O-phosphocholine, and succinate) predicted HIE severity with an adjusted R2 of 0.4. Altered ketones suggest that systemic metabolism may play a critical role in preventing neurological injury, while altered succinate provides a possible explanation for hypoxia-inducible factor 1-α (HIF-1α) stabilization in HI injury. KEYWORDS: asphyxia, hypoxic ischemic encephalopathy, nuclear magnetic resonance, cord serum, metabolomics, succinate, ketone, 3-hydroxybutryate
■
INTRODUCTION
The pathophysiological mechanisms that underlie HIE are multifactorial, complex, and not well understood. Current understanding of the disease suggests that reduced availability of molecular oxygen, the terminal electron acceptor in oxidative phosphorylation (OXPHOS), immediately results in ATP depletion and reactive oxygen species (ROS) production, which lead to mitochondrial membrane depolarization, increased intracellular Ca2+ and apoptosis. Glutamate-mediated excitotoxicity and failure of ATP-dependent ion pumps further add to neurological injury.6 Moreover, as OXPHOS is central to energy metabolites, its impairment can have profound global metabolic effects.7,8 An integral component of HIE that remains to be identified is the metabolic response to hypoxia and its role in reducing or amplifying disease outcome. Low molecular weight biochemicals (metabolites) are the ultimate product of functional biology; their composite status (metabolome) in a given biological system reflects long-term response to gene expression with concurrent immediate
Perinatal asphyxia, defined as neonatal oxygen deprivation around the time of birth, results from any perinatal pathology that interferes with the supply of oxygenated blood to the fetus and may occur in cases of intermittent or acute umbilical cord compression, maternal or fetal hemorrhage, shoulder dystocia, or uterine rupture.1 While most infants completely recover from the hypoxia, others go on to develop hypoxic ischemic encephalopathy (HIE) with lasting neurological sequelae such as cerebral palsy, seizure disorders, cognitive delays, and motor disabilities. HIE remains a global healthy concern, affecting 2 in 1000 live births,2 and is responsible for 1 million full-term neonatal deaths each year.3 Currently the most effective treatment for HIE is therapeutic hypothermia; however, it is only effective for moderate and severe cases of HIE and must be initiated within 6 h of birth.4 Moreover, current methods of diagnosis require a high level of expertise and are unreliable within the time period for treatment initiation.5 Two barriers preventing the development of more effective therapeutic strategies include a reproducible diagnostic test and a complete understanding of disease mechanisms. © 2013 American Chemical Society
Received: June 26, 2013 Published: August 9, 2013 4230
dx.doi.org/10.1021/pr400617m | J. Proteome Res. 2013, 12, 4230−4239
Journal of Proteome Research
Article
response to environmental stimuli.9 Metabolomics, the systematic evaluation of metabolites, provides a temporal snapshot of these biological interactions. As hypoxia inhibits aerobic energy metabolism, which is central to global metabolism, metabolomics offer a powerful systems biology approach to discovering biomarkers and disease mechanisms of HIE. Indeed, several recent studies have employed metabolomics to evaluate perinatal asphyxia in animal models;10−13 however, parallel human studies that integrate metabolic profiles with molecular and pathology knowledge of perinatal asphyxia and HIE are lacking. We recently reported the ability to classify asphyxiated neonates as HIE or non-HIE affected, based on the umbilical cord serum metabolome detected by the Biocrates AbsolutIDQ p180 kit, using direct injection mass spectrometry (DIMS).14 Three metabolite groups were found to be significantly different between healthy and asphyxiated infants: amino acids, acylcarnitines, and phospholipids. However, the scope of this analysis was limited by the 180 metabolites made available for quantification by the Biocrates kit, such that central energy metabolites were not quantified. To develop a more inclusive metabolomic profile of perinatal asphyxiation and investigate its association with neurological injury, 1H NMR spectroscopy was used in the present study. Herein, the differential metabolic responses of asphyxiated neonates with or without HIE are compared. We report both protective and pathological consequences of perinatal asphyxiation and provide insight into mechanisms underlying HIE.
■
A one-to-one matched control population was recruited during the same time period as part of an ongoing birth cohort study (The BASELINE Study, www.baselinestudy.net). Controls and cases were matched to both infant and maternal demographic parameters. These included gestational age, gender, birth weight and centile, method of delivery, maternal ethnicity, maternal age, and maternal body mass index (BMI). Antenatal parental consent was obtained for all control samples collected. Control population infants were clinically healthy, had normal examinations, and had no clinical signs of asphyxia or other medical problems at time of delivery. Sample Collection and Storage
Six milliliters of umbilical cord blood was drawn, using standard operating procedures, and placed in a serum tube without additive (BD Vacutainer no. 366431) within 20 min of delivery of the placenta. The serum was allowed to clot for 30 min at 4 °C and then centrifuged for 10 min at 2400 × g and 4 °C. Serum was removed, placed in a fresh spin tube, and centrifuged for 10 min at 3400 × g and 4 °C. The clarified serum was aliquoted into lithium heparin microtubes and stored at −80 °C until preparation for NMR analysis. The total time between birth and samples being frozen was less than 3 h. Sample Preparation for NMR Spectroscopy
Protein was removed from 200 μL of each serum sample using a BioVision Deproteinizing Sample Preparation Kit (Milpitas, CA, USA). The kit utilizes perchloric acid to precipitate protein from the sample; precipitated protein was removed by centrifugation. A neutralizing solution was added to supernatants to neutralize pH. Prior to NMR analysis, sample pH was adjusted if necessary using hydrochloric acid and sodium hydroxide. During the extraction process, there was a loss of volume; samples were brought to 190 μL with water, and then 10 μL of 5 mM 2,2-dimethyl-2-sila-3,3,4,4,5,5-hexadeuteropentane sulfonic acid (DSS-d6, Chenomx Inc., Edmonton, Alberta, Canada) was added as a concentration reference and chemical shift indicator. Samples were centrifuged to remove any residual particulate matter. The clarified, protein-free serum samples were then transferred to 3 mm NMR tubes.
MATERIALS AND METHODS
Patient Selection
Samples were collected between May 2009 and June 2011 in a single European maternity hospital that delivers approximately 9000 infants per annum. Ethics approval was obtained from the Clinical Research Ethics committee of the Cork Teaching Hospitals. Inclusion criteria included being over 36 weeks of gestation with one or more risk factors for asphyxia and neurological injury.15−18 These risk factors included an arterial cord pH 20% were considered to be below accepted quantification precision and removed from further statistical analyses.26 For each metabolite in turn, data for each matched pair of infants were converted to log2 fold- difference, and then for each arm of the study the null hypothesis that the log2 mean fold difference was zero was tested using Student’s one-sample t test (after testing assumptions for normality using the Lilliefors test). Correction for multiple comparisons was performed using the method described by Benjamini and Hochberg.27 Both pvalues and corrected q-values are reported. The absolute mean fold differences between matched cases and controls were also reported, together with 95% confidence intervals. Comparisons between HIE vs matched controls (ΔHIE), Asphyxia (nonHIE) vs matched controls (ΔAsphyxia), and Grade 3 HIE vs matched controls (ΔGrade 3) are reported. Grade 3 HIE represented the most severe clinical phenotype of hypoxic infants. Additionally, the differences in effect sizes between the arms of the study were tested using 2-sample t test (i.e., ΔHIE vs ΔAsphyxia, and ΔGrade 3 vs ΔAsphyxia), and associated pvalues are reported. In order to compare the univariate results from the two main arms of this study (ΔHIE and ΔAsphyxia) a biplot of mean fold difference for those metabolites significant in either comparison was constructed. Additionally, canonical variate analysis (CVA)28 was performed on this set of differentially changing metabolites in order to visualize the associated multifactorial and correlated discrimination between all 4 clinical groups (asphyxia, HIE, respective matched control groups). Here the matched nature of the data was ignored, and the raw data were log transformed and then autoscaled. Bootstrap resampling/ remodelling was used to determine which metabolites contributed significantly to the CVA following standard protocol for multivariate models.29,30 Finally, to determine if there was a multivariate correlation between metabolome and HIE grade, stepwise multiple-linear regression was performed on the unmatched, log transformed, asphyxia and HIE data only (HIE Grade 0−3). The resulting model was then tested using all available data (cases and controls) and assessed as a disease classifier. Receiver operating characteristic (ROC) curves were constructed from the model predictions, comparing classification of HIE grade against the combined control and asphyxia data sets (i.e., HIE vs non-HIE). Comparison of discriminatory ability was determined using area under ROC curve (AUC).31 All of the statistical analyses were performed using the Matlab scripting language (http://www.mathworks.com).
■
Figure 1. Flow diagram detailing enrollment of study infants.
scoring and early multichannel EEG. Fifty-nine healthy matched control infants were also recruited; infants were matched on clinical variables (gestational age, gender, birth weight and centile, method of delivery, and maternal age, ethnicity, and BMI). Demographic and clinical data are presented in Table 1. Protein-Free Umbilical Cord Serum Analysis 1
H NMR spectroscopic analyses of protein-free cord blood serum yielded reproducible quantification of 37 acid-soluble metabolites across all samples. These metabolites included organic acids central to energy metabolism, amino acids and their catabolic intermediates, urea cycle metabolites, ketones, and phospholipid head groups. The aliphatic region of a representative spectrum, with key metabolites labeled, is shown in Figure 2. Assessment of the QC measurements revealed an average relative standard deviation (QCRSD) of 8% (Supplementary Table S2). Using a critical p-value of 0.05, 18 metabolites were identified as being significantly different between asphyxia versus matched controls (ΔAsphyxia) and 13 were significantly different between HIE versus matched controls (ΔHIE); 12 of these metabolites were common to both groups (Table 2). Figure 3 presents the mean fold differences in metabolite concentration for only those metabolites that were significantly affected in either the ΔHIE or ΔAsphyxia comparisons. Alanine, choline, creatine, glycerol, isoleucine, lactate, leucine, myo-inositol, pyruvate, phenylalanine, succinate, and valine were significantly elevated in both asphyxia and HIE infants. Methionine was the only significantly elevated metabolite unique to the ΔHIE comparison. Acetone, betaine, creatinine, glucose, 3-hydroxybutyrate, and O-phosphocholine were uniquely elevated in the ΔAsphyxia comparison. Succinate increased more than 3-fold in HIE infants, but less than 2-fold in asphyxia infants. Although less pronounced, glycerol also showed a greater increase in the HIE comparison. Conversely, 3-hydroxybutyrate and acetone concentrations increased 2-fold and 1.5-fold, respectively, in the asphyxia comparison but showed no difference in HIE. To determine if these four extreme disease-specific metabolite differences (succinate, glycerol, 3-hydroxybutyrate, and acetone) were correlated with HIE severity, data were stratified by disease grade (Figure 4). 3-Hydroxybutryate and acetone exhibited a 2-
RESULTS
Study Population
One hundred neonatal infants were recruited for this study. Forty-one were excluded (15 had insufficient sample quantity for NMR analysis, 16 had no EEG, 7 had missing clinical data, 3 had alternate diagnoses), leaving 59 infants in the at-risk study population (Figure 1). Twenty-five infants were enrolled with HIE, including 13 mild, 6 moderate, and 6 severe cases, and 34 infants who were asphyxiated but did not present clinical neurological signs, herein referred to as “asphyxia”. The grade of HIE (1-mild, 2-moderate, or 3-severe) was based on Sarnat 4232
dx.doi.org/10.1021/pr400617m | J. Proteome Res. 2013, 12, 4230−4239
Journal of Proteome Research
Article
Table 1. Clinical and Demographic Data of Study Populationa gestational age (wks) gender (M/F) birth weight (g) birth weight centile method of delivery SVD instrumental emergency LSCS elective LSCS HIE grade severe/moderate/mild first pH 1 min Apgar score 5 min Apgar score maternal ethnicity Caucasian African Indian/Pakistani Asian maternal age (yrs) maternal BMI (kg/m2) a
HIE (n = 25)
control HIE (n = 25)
p-value
asphyxia (n = 34)
control asphyxia (n = 34)
p-value
40.6 (40.2,41.3) 18/9 3652 (591) 48.1 (32.5)
40.4 (40.0,41.0) 19/8 3574 (460) 52.7 (32.2)
0.2 1 0.61 0.62
40.1 (39.5,41.1) 22/12 3709 (598) 56.8 (31.1)
40.1 (39.4,41.1) 22/12 3659 (511) 58.8 (31.1)
0.78 1 0.71 0.8
4 (15%) 14 (52%) 9 (33%)
5 (18%) 15 (56%) 7 (26%)
12 (35%) 16 (47%) 5 (15%) 1 (3%)
12 (35%) 16 (47%) 5 (15%) 1 (3%)
6/6/13 7.0 (0.12) 2.5 (1,4.25) 5 (3,7)
--7.30 (0.11) 9 (9,9) 10 (9,10)
--7.03 (0.08) 5.5 (3,7) 8 (6,9)
--7.25 (0.06) 9 (9,9) 10 (9,10)
21 (84%) 2 (8%) 2 (8%)
24 (96%)
32 (94%)
33 (97%)
27.3 (5.0) 24.0 (23.6,28.4)
29.4 (4.1) 23.0 (20.8,24.0)
2 (6%) 30.9 (6.7) 25.1 (22.0,29.0)
1 (3%) 30.1 (5.5) 23.8 (22.0,26.1)