A, In the BiHiVE1 discovery cohort, there were significant differences in miR-199a-5p and in miR-374a-5p between the control (n = 21) and hypoxic-ischemic encephalopathy (HIE) (n = 27) groups and differences in miR-374a-5p, miR-376c-3p, and miR-410 between the control and perinatal asphyxia (PA; n = 39) groups. B, In the BiHiVE2 validation cohort, there were significant differences in miR-374a-5p and miR-410 across the control (n = 22) vs HIE (n = 25) groups. In miR-181b-3p, miR-376a-3p, and miR-376c-3p, there were significant differences across the control vs PA groups (n = 26). Finally, in miR-181b-5p, miR-199a-5p, and miR-376a-3p, there were significant differences across the PA vs HIE groups. The horizontal line in the middle of each box indicates the median, whereas the top and bottom borders of the box mark the 75th and 25th percentiles, respectively. The whiskers above and below the box extend to the most extreme point no longer than 1.5 times the interquartile range from the box.
aSignificant at P < .05.
eTable. List of miRNAs included in the BiHiVE qRT-PCR Panel.
eFigure 1. Flowchart of Recruitment, Experimental Procedure, Statistical and in Silico Analysis.
eFigure 2. Functional Enrichment Analysis of GO and KEGG.
Customize your JAMA Network experience by selecting one or more topics from the list below.
O’Sullivan MP, Looney AM, Moloney GM, et al. Validation of Altered Umbilical Cord Blood MicroRNA Expression in Neonatal Hypoxic-Ischemic Encephalopathy. JAMA Neurol. 2019;76(3):333–341. doi:10.1001/jamaneurol.2018.4182
Can umbilical cord blood microRNA levels be used for early detection of perinatal asphyxia and hypoxic-ischemic encephalopathy at birth?
In this validation study including umbilical cord blood of 160 newborns in 2 separate cohort studies of umbilical cord blood from 160 infants, relative quantification identified 2 cord blood–based microRNAs that could distinguish both perinatal asphyxia and hypoxic-ischemic encephalopathy from healthy control infants, respectively. A third microRNA was compared against current markers used at birth to identify eligibility of infants with hypoxic-ischemic encephalopathy for therapeutic hypothermia.
Umbilical cord blood microRNA levels can be measured in a method noninvasive to the infant and may represent an objective measure for early detection of perinatal asphyxia and hypoxic-ischemic encephalopathy.
Neonatal hypoxic-ischemic encephalopathy (HIE) remains a significant cause of neurologic disability. Identifying infants suitable for therapeutic hypothermia (TH) within a narrow therapeutic time is difficult. No single robust biochemical marker is available to clinicians.
To assess the ability of a panel of candidate microRNA (miRNA) to evaluate the development and severity of encephalopathy following perinatal asphyxia (PA).
Design, Setting, and Participants
This validation study included 2 cohorts. For the discovery cohort, full-term infants with PA were enrolled at birth to the Biomarkers in Hypoxic-Ischemic Encephalopathy (BiHiVE1) study (2009-2011) in Cork, Ireland. Encephalopathy grade was defined using early electroencephalogram and Sarnat score (n = 68). The BiHiVE1 cohort also enrolled healthy control infants (n = 22). For the validation cohort, the BiHiVE2 multicenter study (2013-2015), based in Cork, Ireland (7500 live births per annum), and Karolinska Huddinge, Sweden (4400 live births per annum), recruited infants with PA along with healthy control infants to validate findings from BiHiVE1 using identical recruitment criteria (n = 80). The experimental design was formulated prior to recruitment, and analysis was conducted from June 2016 to March 2017.
Main Outcomes and Measures
Alterations in umbilical cord whole-blood miRNA expression.
From 170 neonates, 160 were included in the final analysis. The BiHiVE1 cohort included 87 infants (21 control infants, 39 infants with PA, and 27 infants with HIE), and BiHiVE2 included 73 infants (control [n = 22], PA [n = 26], and HIE [n = 25]). The BiHiVE1 and BiHiVE2 had a median age of 40 weeks (interquartile range [IQR], 39-41 weeks) and 40 weeks (IQR, 39-41 weeks), respectively, and included 56 boys and 31 girls and 45 boys and 28 girls, respectively. In BiHiVE1, 12 candidate miRNAs were identified, and 7 of these miRNAs were chosen for validation in BiHiVE2. The BiHiVE2 cohort showed consistent alteration of 3 miRNAs; miR-374a-5p was decreased in infants diagnosed as having HIE compared with healthy control infants (median relative quantification, 0.38; IQR, 0.17-0.77 vs 0.95; IQR, 0.68-1.19; P = .009), miR-376c-3p was decreased in infants with PA compared with healthy control infants (median, 0.42; IQR, 0.21-0.61 vs 0.90; IQR, 0.70-1.30; P = .004), and mir-181b-5p was decreased in infants eligible for TH (median, 0.27; IQR, 0.14-1.41) vs 1.18; IQR, 0.70-2.05; P = .02).
Conclusions and Relevance
Altered miRNA expression was detected in umbilical cord blood of neonates with PA and HIE. These miRNA could assist diagnostic markers for early detection of HIE and PA at birth.
Neonatal hypoxic-ischemic encephalopathy (HIE) remains a significant cause of long-term neurologic disability and death in newborns and occurs in 1 to 6 per 1000 live full-term births.1 Hypoxic-ischemic encephalopathy is a multifactorial evolving triphasic encephalopathy, involving both the primary injury, a latent window of reoxygenation, and secondary injury.2,3
To date, the only proven treatment for HIE is therapeutic hypothermia (TH), which can improve outcome if introduced within the first 6 postnatal hours.4 However, accurate and timely diagnosis of HIE is difficult, and these infants are currently identified using biochemical and clinical measurements including Apgar scores, initial lactate, and base deficit, which individually have poor positive predictive values.5-8 The Sarnat grading scale of HIE with electroencephalogram grading is accurate when assessed at 24 hours post partum, but this is beyond the limited therapeutic window.9,10 A reliable, validated, and quantifiable biomarker would support clinical decision making and early targeted intervention.
Aberrant microRNA (miRNA) expression has been linked to many disease states11,12 and appears stable in a cell-free state in various biologic fluids.13 Therefore, miRNAs offer potential as reliable and robust blood-based biomarkers in molecular diagnostics.14 Research on miRNA in HIE remains limited. Past studies have focused on cellular ex vivo work, and animal studies focused on regulation of apoptosis.15-18 Our group has previously published what is, to our knowledge, the first description of miRNA profiling in a human neonatal cohort and reported downregulation of cord blood miR-374a-5p in infants with HIE.19
This study aimed to expand on our previous work in the Biomarkers in Hypoxic-Ischemic Encephalopathy (BiHiVE1) study using a separate validation cohort and to explore further additional candidate miRNAs in umbilical cord blood following HI injury. We wished to explore whether miRNAs could reliably assess the development and severity of HIE in infants born with clinical and biochemical signs of perinatal asphyxia (PA).
Infants with a gestation greater than 36 weeks were recruited from May 2009 through June 2011 using 1 or more of the following inclusion criteria: umbilical cord pH less than 7.1, 5-minute Apgar score 6 or less, and need for intubation or cardiopulmonary resuscitation at age 10 minutes. Recruitment procedures have been previously described.19 Infants with suspected or confirmed sepsis or coexisting congenital abnormalities were excluded from analysis. The healthy control population was recruited from the baseline longitudinal birth cohort20; control infants were all healthy full-term infants, with uneventful deliveries, normal neonatal examinations, and without admission to the neonatal intensive care unit. For the study, infants were matched for age, sex, and gestation.
All cases had a continuous multichannel electroencephalogram recorded, as previously described.21 Clinical HIE grade of encephalopathy, if any, was assigned at 24 hours post partum using the modified Sarnat score.9 Standardized neurologic assessment was performed on day 3 and at discharge. Case infants were divided into those with mild, moderate, or severe grade of HIE (HIE group), and those with biochemical or clinical signs of PA but without clinical encephalopathy (PA group).
A second multicenter cohort study, the BiHiVE2 validation cohort, was recruited from March 2013 through June 2015 at Cork University Maternity Hospital, Cork, Ireland, with 7500 deliveries per annum, and Karolinska University Hospital Huddinge, Stockholm, Sweden, with 4400 deliveries per annum. The study was designed to validate and expand on findings from the BiHiVE1 cohort and used identical recruitment criteria.19 The BiHiVE2 cohort also recruited study-specific control infants with uneventful deliveries, normal neonatal examination, and 5-minute Apgar scores greater than 8. The BiHiVE2 study is registered (NCT02019147).
Both studies were approved by the Clinical Research Ethics Committee of the Cork Teaching Hospitals, and the BiHiVE2 cohort was also approved by the regional ethical review board in Stockholm, Sweden. Written informed consent was obtained from parents of all study participants.
Umbilical cord blood of newborns in all cohorts was collected immediately following delivery of placenta and processed within 3 hours of delivery. Three milliliters of cord blood was placed directly into Tempus Blood RNA tubes (Applied Biosystems) and biobanked at −80°C.
Total RNA was isolated from the Tempus system using MagMAX for Stabilized Blood Tubes RNA Isolation Kit (Ambion, Life Technologies) and quantified using a NanoDrop 8000 Spectrophotometer (ThermoScientific NanoDrop). Additionally, RNA quality from the BiHiVE1 study was assessed using an Agilent 2100 Bioanalyzer (Agilent Technologies Inc) to ensure sample quality from a random subgroup, as previously described.19 Isolated RNA was subsequently stored at −80°C before quantitative real-time polymerase chain reaction (qRT-PCR). Five nanograms per microliter of isolated RNA was used for reverse transcription to synthesize complementary DNA using Universal cDNA synthesis kit II (Exiqon), followed by an amplification step using the ExiLENT SYBR Green master mix (Exiqon). The qRT-PCR amplifications were performed in triplicate using LNA PCR Primer sets (Exiqon) on the StepOnePlus (Applied Biosystems). Data were analyzed according to the 2-ΔΔCT method.22
The panel of 12 miRNAs was chosen based on our previously published microarray data19 and previous reports from the literature indicating potential alterations in miR-146a, miR-155, and miR-181b23-25 (eTable in the Supplement). These miRNAs were analyzed using qRT-PCR in the BiHiVE1 cohort, and if the altered expression was confirmed, further validation was performed in the BiHiVE2 cohort. The qRT-PCR was performed using the miRCURY LNA Universal RT microRNA PCR system (Exiqon) using SN4826,27 and miR-223-3p as reference genes. SN48 was tested for stability before qRT-PCR, and miR-223-3p had previously been found to be stable using these methods.19 We also included UniSp6 as a spike-in for complementary DNA synthesis quality control.
Possible downstream targets of the miRNAs altered across both cohorts were identified using the online resource miRWalk 2.0 atlas database. Four prediction databases were included in the retrieval (miRWalk, miRanda, miRDB, and TargetScan), and predicted target genes and their conserved miRNA binding sites were annotated. The criteria used involved a minimal seed length of 7 base pairs and a P value less than .05; for gene ontology, predicted targets needed to be present in 3 or more predictive algorithms. Predicted gene targets of the miRNA were analyzed for fold enrichment in both gene ontology terms and Kyoto Encyclopedia of Genes and Genomes pathways using the DAVID 6.8 bioinformatics database (https://david.ncifcrf.gov/).
Means and standard deviations and medians and interquartile ranges (IQRs) were used for normal and nonnormal distributions respectively. For groupwise comparisons, analysis of variance, Kruskal-Wallis test (n groups), t test, or Mann-Whitney test (2 groups) were used as appropriate. Both cohorts were merged for predictability analysis; this is by dividing the mean relative quantification (RQ) of the control group in the BiHiVE discovery cohort across all groups of the discovery cohort. This step was repeated in the BiHiVE2 validation cohort by dividing the mean RQ of the control group in the validation cohort across all groups of the validation cohort. The predictive ability of the individual markers for HIE was assessed using positive and negative predictive values (NPVs) and area under the receiver operator curves. Correlation of data was performed using the Spearman ρ correlation coefficient. Binomial logistic regression was used to examine combinations of significant markers for associations with current eligibility for therapeutic hypothermia based on a moderate/severe HIE grade. In all cases, a P value of less than .05 was considered statistically significant, and all P values were 2-sided. In silico analyses used an adjusted P value corrected for multiple testing using false discovery rate based on the Benjamini-Hochberg procedure,28 and an adjusted P value was considered significant at less than .10.29 Statistical analysis was carried out using SPSS Statistics, version 22 (IBM) and GraphPad Prism 5.
In the BiHiVE1 discovery cohort, 68 infants had umbilical cord blood collected at delivery, and 66 (39 with PA and 27 with HIE) were included in the final analysis. Healthy control infants (n = 22) were recruited from the BASELINE cohort, and 21 were included in the final analysis.
In the BiHiVE2 validation cohort, 80 infants, including healthy control infants, had umbilical whole blood collected at delivery, and 73 infants (22 control infants, 26 with PA, and 25 with HIE) were included in the final analysis. Ten samples were lost during extraction. The clinical characteristics of both populations are summarized in Table 1.
In the BiHiVE1 discovery cohort, 12 different miRNAs were analyzed. Relative expression levels (2-ΔΔCT) of miRNAs across 3 groups were compared between healthy control infants (n = 21), infants with PA (n = 39), and infants with HIE (n = 27) (workflow in eFigure 1 in the Supplement). Two miRNAs were reduced in the HIE group compared with healthy controls; miR-199a (median RQ, 0.60; IQR, 0.39-0.79 vs median, 0.87; IQR, 0.56-1.40, P = .04) and miR-374a (median, 0.24; IQR, 0.13-0.40 vs median, 0.73; IQR, 0.22-1.46; P = .01). Three miRNAs were reduced in the PA group compared with healthy control individuals: miR-374a (median, 0.21; IQR, 0.07-1.19 vs median, 0.73; IQR, 0.22-1.46; P = .04), miR-376c (median, 0.28; IQR, 0.14-0.55 vs median, 0.55; IQR, 0.41-1.40; P = .01), and miR-410 (median, 0.65; IQR, 0.20-0.78 vs median, 0.99; IQR, 0.43-1.23; P = 0.04) (Figure, A; Table 2).
We next focused within the HIE group (n = 27) and looked at the differences between the mild (n = 16), moderate (n = 5), and severe (n = 6) grade of HIE. miR-181b was significantly reduced in the severe group compared with the mild group (median, 0.08; IQR, 0.04-0.21 vs median, 0.82; IQR, 0.36-1.38; P = .01).
To examine the ability of miRNAs to aid in clinical decision making, we compared infants eligible for TH (11 with moderate or severe encephalopathy) with those ineligible (healthy control infants, those with PA, and those with mild HIE [n = 76]). This analysis revealed reduced expression of 2 miRNAs for infants eligible for TH, miR-199a (median, 0.49; IQR, 0.33-0.60 vs median, 0.75; IQR, 0.43-1.08, P = .049) and miR-181b (median, 0.25; IQR, 0.16-0.32 vs median, 0.61; IQR, 0.26-1.39; P = .03).
From the 12 candidate miRNAs, 5 miRNAs showed consistent significance (miR-374a, miR-376c, miR-410, miR-181b, and miR-199a), and 2 additional miRNA were included: miR-376a, with a strong trend in differentiating PA vs HIE (median, 0.37; IQR, 0.15-0.78 vs median, 0.78; IQR, 0.44-3.06; P = .06), and miR-155 owing to its suggested role in immune response and microglial activation.24 The 5 miRNAs with no significant and consistent changes were not tested further in BiHiVE2.
Of the 7 candidate miRNAs tested, 3 miRNAs showed consistent altered expression levels in the BiHiVE2 cohort. miR-374a was significantly reduced in the HIE group vs control group (median, 0.38; IQR, 0.17-0.77 vs median, 0.95; IQR, 0.68-1.19; P = .009), and miR-376c was significantly reduced in the PA group compared with the control group (median, 0.42; IQR, 0.21-0.61 vs median, 0.90; IQR, 0.70-1.30; P = .004) (Figure, B; Table 2). There was no difference between grades of HIE in the validation cohort.
As in our discovery cohort, we separated infants in the validation cohort into 2 groups, infants eligible for TH (n = 9) and infants ineligible for TH (n = 64), and found miR-181b to be significantly reduced in the group eligible for TH compared with infants not eligible (median, 0.27; IQR, 0.14-1.41 vs median, 1.18; IQR, 0.70-2.05; P = .02) as previously described.30
We merged both cohorts to examine the ability of miR-181b expression to detect moderate/severe HIE. We compared it with current early biochemical and clinical markers, using the standard cutoffs used in previous trials of TH.31 Associations across miRNA and current markers can be seen in Table 3.
miR-181b had an area under the curve of 0.752 (95% CI, 0.610-0.893) for the prediction of moderate-severe HIE. miR-181b performed similar to current biochemical markers and had the highest NPV of 99% (Table 4). Clinical Apgar scores performed better than biochemical markers and miR-181b (Table 4). Logistic regression was performed to assess the ability of a combination of miR-181b and Apgar at 5 minutes to predict moderate-severe HIE. The model was statistically significant at χ2 = 42.576; P < .001. The model explained 60% (Nagelkerke R2) of the variance in the eligibility for TH and correctly classified 93% of cases. However, the combined marker did not make significant improvements to Apgar at 5 minutes alone.
A list of potential miRNA-gene targets were generated for each of the 3 miRNAs significantly altered across both cohorts: miR-376c, miR-374a, and miR-181b. For enrichment of biologic processes, there was an overlap between miR-376c and miR-181b targets for regulation of RNA metabolic process and regulation of RNA biosynthesis processes (eFigure 2 in the Supplement). For enrichment of the cellular component, there was an overlap in miR-374a and miR-376c for postsynaptic specialization (eFigure 2 in the Supplement). Finally, for the enrichment of molecular function, there was an overlap across all 3 miRNAs for regulatory region DNA binding and transcriptional activator activity. For miR-181b and miR-374a, there was an overlap for double-stranded DNA binding, and in miR-374a and miR-376c there was an overlap for sequence-specific DNA binding (eFigure 2 in the Supplement).
We have validated 3 miRNAs in neonatal PA and HIE by using 2 well-defined, independent cohorts. From the 12 candidate miRNAs, 5 miRNAs were differentially expressed in cord blood in the BiHiVE1 discovery cohort. Following this, expression patterns of 3 of 5 microRNAs were validated as consistently altered in the BiHiVE2 multicenter validation cohort. The miRNA biomarkers assessed have shown utility in distinguishing perinatal asphyxia and HIE from healthy control infants and may help identify newborns suitable for TH. In both the BiHiVE1 and BiHiVE2 cohorts, miR-374a, previously identified19 and validated in this study, was found to be expressed at lower levels in the umbilical cord blood of HIE infants compared with the healthy control infants. We also found a consistent downregulation of miR-376c in perinatal asphyxia compared with healthy controls and previously reported downregulation of miR-181b in infants with moderate and severe HIE compared with healthy control infants.30
Research in miRNA is still quite novel, and most studies of miR-374a and miR-376c have been focused on cancer research. MiR-374a was first described in colon cancer32 and miR-376c in ovarian cancer.33 These studies linked them to dysregulation of the Wnt/β-catenin pathway34,35 by targeting associated genes in the pathway via functional assays, including WIF1, PTEN, WNT5A, and LRH-1, respectively. Other pathways explored in miR-374a include PI3K/AKT via CCND136 and for miR-376c include TGF-β signaling via ALK7.37 Our miR-374a in silico analysis found signaling pathways regulating pluripotency of stem cells as a predicted Kyoto Encyclopedia of Genes and Genomes pathway, which encompasses a number of pathways including TGFβ/Smad, PI3K-Akt, and the Wnt/β-catenin signaling pathway. The Wnt/β-catenin signaling described in both miR-374a and miR-376c has also been described in a HIE cell modeling, where high Wnt activity prevents degradation of β-catenin and can inhibit preoligodendrocytes from differentiating into mature oligodendrocytes.38ALK7, a functional target of miR-376c, has also been shown to impair oligodendrocyte differentiation and myelination, where ALK7 is highly expressed in nonrepairing lesions.33
Few studies to date have examined miR-374a in neonatal HIE. A piglet model reported temporal changes in expression after injury39 and a 2017 clinical HIE study confirmed downregulation in umbilical cord blood in neonatal HIE.40
This study also reported the downregulation of miR-181b in infants with moderate and severe HIE and compared the predictive ability of miR-181b with biochemical and clinical markers in HIE (Table 4). The strong NPV of 99% from miR-181b could make this miRNA a useful tool for the clinical setting because it is quantifiable and not subjective. The Apgar score is reliant on subjective observation and can differ in interobserver variability, and past studies have called for more objective and precise measures.41 Additionally, instrumental deliveries were more prevalent in the PA and HIE groups but were not associated with our validated miRNAs.
While miR-199a, miR-376a, and miR-410 were not validated as significantly altered in both cohorts, it is essential to recognize the distribution of Sarnat grades across BiHiVE2 (Table 1). Although recruitment went on for a similar period across 2 different centers, the number of severe-grade HIE infants recruited to the BiHiVE2 validation cohort was significantly lower. Thus, validation with an equal distribution of grades of HIE may highlight further miRNAs altered in more severe grades of HIE in future studies.
Our study has focused on whole blood instead of either serum or plasma; this approach has given a higher yield of miRNA and allowed us to explore both cellular and extracellular miRNA expressed in the blood. This may have masked altered expression seen in extracellular miRNA owing to abundant levels of cellular miRNA in whole blood.42 Previous studies on circulating miRNA have proposed that cellular damage caused by injury may release cellular miRNA into the serum or plasma. This may be comparable to how cellular enzymes are released after apoptosis and necrosis or possibly through a cell-specific transport that releases specific miRNAs.43 We do not know how these decreased circulating expression levels might relate to altered expression within tissues, particularly within the brain. Studies in peripheral tissues would help better understand the mechanism of this altered expression. Future studies will need to explore potential downstream targets of these miRNA to assess whether they are innocent bystanders or play an active role in the pathophysiology of hypoxic injury.
The strength of our study lies in the fact that we have studied the miRNA in 2 completely separate cohorts, recruited using identical and strict recruitment criteria. The BiHiVE2 validation cohort was recruited across 2 different countries, indicating that our findings may be applicable across distinct geographical areas. The altered expression of umbilical cord blood miRNAs immediately after birth implies that changes in miRNA expression may occur during injury. Additional research is required to demonstrate whether these changes persist during the subsequent progression of HI injury.
In conclusion, we have validated the decreased expression of 2 miRNAs, miR-374a and miR-376c, in whole blood from the umbilical cord of newborns with PA and HIE in 2 independent cohorts. We have also expanded on our previous research in miR-181b; its strong NPV may bring this miRNA forward as a quantifiable noninvasive biomarker with a reliable NPV to aid in identifying those infants who will not require intervention.
Corresponding Author: Deirdre M. Murray, MD, PhD, Pediatric Academic Unit, Department of Pediatrics and Child Health, Cork University Hospital, Cork, Ireland (firstname.lastname@example.org).
Accepted for Publication: October 19, 2018.
Published Online: December 28, 2018. doi:10.1001/jamaneurol.2018.4182
Author Contributions: Mr O’Sullivan and Dr Murray had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: O’Sullivan, Looney, Clarke, Boylan, Murray.
Acquisition, analysis, or interpretation of data: O’Sullivan, Looney, Moloney, Finder, Hallberg, Clarke, Murray.
Drafting of the manuscript: O’Sullivan, Finder, Clarke, Murray.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: O’Sullivan, Looney, Clarke.
Obtained funding: Boylan, Murray.
Administrative, technical, or material support: O’Sullivan, Looney, Moloney, Finder, Hallberg, Clarke, Murray.
Supervision: Looney, Clarke, Boylan, Murray.
Other - recruitment of study participants: Finder.
Conflict of Interest Disclosures: None reported.
Funding/Support: The research was funded by the National Children’s Research Centre, Crumlin (B/14/1), the Health Research Board (HRB; CSA/2012/40), and a Science Foundation Research Centre Award (INFANT; 12/RC/2272). Dr Clarke is also funded by Science Foundation Ireland grant 12/RC/2273.
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Meeting Presentation: This paper was presented at the 2nd Congress of Joint European Neonatal Societies; November 2, 2017; Venice, Italy.
Create a personal account or sign in to: