Maternal metabolomic profiling and congenital heart disease risk in offspring: A systematic review of observational studies

Abstract Aetiological understanding and screening methods for congenital heart disease (CHD) are limited. Maternal metabolomic assessment offers the potential to identify risk factors and biomarkers. We performed a systematic review (PROSPERO CRD42022308452) investigating the association between fetal/childhood CHD and endogenous maternal metabolites. Ovid‐MEDLINE, Ovid‐EMBASE and Cochrane Library were searched between inception and 06/09/2022. Case control studies included analysing maternal blood or urine metabolites in pregnancy or postpartum where there was foetal/childhood CHD. Risk of bias assessment utilised the Scottish Intercollegiate Guidelines Network methodology checklist and narrative synthesis was performed. A total of 134 records were screened with eight eligible studies (n = 3242 pregnancies, n = 842 CHD‐affected offspring). Five studies performed metabolomic analysis in pregnancy. Metabolites distinguishing case and control groups spanned lipid, glucose and amino‐acid pathways, with the development of sensitive risk prediction models. No single metabolite consistently distinguished cases and controls across studies. Three studies performed targeted analysis postnatally with altered lipid and amino acid metabolites and raised homocysteine and markers of oxidative stress identified in cases. Included studies reported small sample sizes, analysing different biosamples at variable time points using differing techniques. At present, there is not enough evidence to confidently associate maternal metabolomic profiles with offspring CHD risk. However, several identified pathways warrant further investigation.


| INTRODUCTION
Globally, congenital anomalies affect over 2% of total births, with congenital heart disease (CHD) the most common. CHD prevalence is increasing with similar rates reported across Europe. 1,2 In England, CHD represents one third of congenital anomalies, with severe CHD, generally defined as requiring surgery or treatment in the first year of life, accounting for around half of these. 3 Increasing detection, survival and the need to reduce postoperative mortality in infants with severe and critical CHD emphasise the importance of adequate understanding of CHD pathology on a global level. 4 Cardiac development begins in the second human embryonic week, with the critical period recognised as weeks two to eight. 5,6 Our understanding of CHD aetiology is significantly limited. Around 15% of CHD can be linked to a known genetic cause including aneuploidy, such as trisomy 21 (8%-10%) and single gene defects such as Noonan's syndrome (3%-5%). Approximately 2% of cases are associated with specific environmental factors including maternal medical conditions, for example, diabetes mellitus or phenylketonuria. 7 A meta-analysis and umbrella review of literature has also linked numerous other maternal and environmental factors to foetal CHD including obesity, smoking, infection, air pollution and recreational and prescription drug use. 8,9 Despite such associations, mechanistic understanding is lacking with most CHD causative factors remaining unexplained.
Prenatal screening evaluates individual risk in obstetrics. This is demonstrated by aneuploidy screening, combining foetal nuchal translucency (NT) measurement on ultrasound with maternal serum biomarkers and age. 10 Combined screening tests generally provide more accurate assessment of risk than individual markers alone.
Diagnostic genetic testing by invasive sampling of the placenta or amniotic fluid provides an accurate result but carries a risk of miscarriage of approximately 0.5%. 11 More recently, non-invasive prenatal testing where analysis of maternal serum is undertaken for cell-free foetal DNA has revolutionised prenatal screening for aneuploidy with a significantly higher sensitivity and specificity and no associated risk of miscarriage. 12,13 Whilst these technological developments are being extended to single gene disorders, they are unlikely to provide adequate screening for CHD given its multifactorial aetiology and low single gene causality yield.
The focus of prenatal screening for congenital anomalies primarily rests on ultrasonographic assessment. Increased NT, abnormal ductus venosus flow and tricuspid regurgitation in the first trimester are associated with an increased risk of CHD. 14 The foetal anomaly screening programme recommends a systematic anatomy scan between 18 and 21 weeks, aiming to identify common and major congenital anomalies including serious cardiac anomalies (transposition of the great arteries, atrioventricular septal defects, tetralogy of Fallot and hypoplastic left heart syndrome). 10 Patient uptake is over 98%, illustrating the importance of this screening to pregnant woman. 15 However, targeted detection rates for CHD are only 50%. 10 Recent data in England shows that this detection rate is accurate, with 54.5% of major CHD diagnosed antenatally. 3 Ultrasound has several limitations including intraoperator variability, foetal factors such as position or multiple pregnancy and maternal factors such as obesity. Foetal echocardiography, provided by trained foetal cardiologists, provides greater sensitivity for CHD detection but is generally only performed when an anomaly is suspected on screening ultrasound or in screening high-risk populations. 16,17 Foetal echocardiography can improve detection of CHD following a normal complete foetal anomaly scan; however, performance of 750 scans is required to detect one additional case. 18 Foetal echocardiography is costly and requires significant expertise. 19 As such, there are significant limitations of current antenatal screening strategies for foetal CHD.
Metabolomic analysis utilising chromatography and mass spectrometry or nuclear magnetic resonance spectroscopy detects compounds such as carbohydrates, amino acids, lipids and vitamins in biosamples such as blood or urine. 20 Such profiles reflect global physiology, influenced by genetics, the environment and physiological stressors. [20][21][22] Use of machine learning techniques in large biological datasets has revolutionised utilisation of such information. 23 In CHD, it is hypothesised that changes in foetal perfusion or organ function could be reflected in maternal metabolic alterations. As such, metabolomic analyses provide the potential to derive novel biomarkers to screen for CHD. 21 However, the presence of foetal cardiovascular shunts and the placental interface may limit this. Furthermore, elucidating potential maternal metabolic risk factors could help improve mechanistic understanding and screening for CHD.
A systematic review assessing altered endogenous maternal metabolite levels in maternal screening or risk factor profiling for foetal CHD has not been undertaken previously. We therefore aim to critically appraise available literature to assess the association 648 between altered endogenous maternal metabolite levels and foetal or childhood CHD. Our findings will inform subsequent recommendations for future research and practice.

| Protocol, search strategy and eligibility criteria
The systematic review protocol was registered a priori at PROSPERO (registration number CRD42022308452) and followed the PRISMA 2020 checklist. 24 A systematic literature search of electronic databases MEDLINE (via OVID), EMBASE (via OVID) and the Cochrane library was conducted to identify relevant studies published between the inception of each database and February 3rd 2022 without language restrictions. Searches were re-run prior to final submission (6th September 2022). Reference lists of included studies were manually screened and experts contacted to identify additional emerging and relevant literature. Figure S1 outlines the search strategy for each database. Observational studies were included. It was not anticipated that randomised control or control trials would be available. Case reports and case series were excluded due to high risk of bias.
Case inclusion criteria were mothers of fetuses/children with CHD. Studies with surrogate mothers or pregnancies with egg donation were excluded due to the potential influence of the surrogate or donor environment on the metabolome. The exposure was maternal blood or urine metabolite analysis in pregnancy or postpartum. The primary outcome was foetal/childhood CHD.

| Study selection, data extraction and quality assessment
SM and SR independently screened titles and abstracts to exclude studies that did not meet inclusion criteria. Agreement on potential inconsistencies was reached by consensus. Data extraction from selected studies was performed by SM and checked by SR. Risk of bias assessment was assessed by SM and SR independently utilising the Scottish Intercollegiate Guidelines Network case control study checklist and guidance notes. 25 This facilitates assessment of subject selection, exposure assessment, confounding and statistical analysis.
Studies are assessed overall as high quality (++), acceptable (+) or unacceptable. Unacceptable studies were excluded from analysis due to high risk of bias.

| Data synthesis and analysis
A narrative synthesis was performed including tabulation of study characteristics and results. Quantitative synthesis of data and metaanalysis was precluded due to variation in studies including biomaterial analyses, timing of sampling and metabolic analyses performed. Metabolites distinguishing case and control populations identified through machine learning and confounder adjusted analyses were summarised. For antenatal studies reporting risk prediction, the models developed and included biomarkers are presented.

| Metabolic analysis in pregnancy
Maternal metabolomic analyses was performed in five studies [26][27][28][29] during the index pregnancy. Methods of metabolomic analysis included gas, liquid and direct injection chromatography coupled with mass spectrometry or NMR spectroscopy. Maternal serum was analysed in two studies. 26  Analysis of identified metabolites which significantly differed between case and control groups were assessed by multivariable regression or partial least squares discriminant analysis and synthesised from variable importance in projection plots or selected through volcano plot analysis. Table S1 summarises individual metabolites distinguishing cases and controls with direction of change and method of analysis for each included study. Figure 2A collates metabolites distinguishing case and control groups across included studies. A total of 110 unique metabolites were identified spanning multiple metabolic pathways (excluding 7 unnamed molecules). Differences in lipid metabolism are strongly represented including acylcarnitines, phospholipids, fatty acids, lysophospholipids and sphingolipids, predominantly reduced in cases compared to controls. Metabolites from amino acid, glucose and aerobic metabolism also significantly differ.
Whilst metabolites from similar and related metabolic pathways were identified in multiple included studies, leucine was the only metabolite to significantly differ between cases and controls in two independent studies, with contradictory findings. 28,33

| Mendelian Randomisation
One study utilised three European population cohort studies (n = 33,662, 319 CHD cases) to enable Mendelian Randomisation.
This analysis allowed further validation of potential causal effects of 27 metabolites differentiating cases and controls in pregnancy (n = 44). 33 From this, 11 metabolites were identified with replication and potential evidence of causality, representing amino acid, steroid lipid and succinylcarnitine metabolic pathways.

| Interpretation
Individual metabolites identified to distinguish cases and controls as summarised in Table S1 and Figure 2 physiologically function and interact in several metabolic pathways.

| Lipid metabolism
Lipid metabolites most frequently distinguished maternal metabolic profiles of cases and controls across included studies. Several acylcarnitines and fatty acids, phospholipids including phosphatidylcholine and sphingolipids including sphingomyelin were lower in cases than controls. [26][27][28][29]33 One study further assessing potential causality through Mendelian randomisation suggested higher levels of the acylcarnitine succinylcarnitine in cases than controls. 33 As such, understanding the origin of these metabolic differences is essential for future risk profiling, prediction, and mechanistic understanding of foetal CHD. Metabolic profile changes could represent an inherent maternal risk factor for foetal CHD or may point to a process driven by the fetus or placenta.
Acylcarnitines are formed from the binding of carnitine to long chain fatty acids, facilitating transfer of fatty acids across the mitochondrial membrane for lipid metabolism. 34 Cardiac tissue derives over 95% of its energy from mitochondrial fat metabolism. 35 Maternal changes in lipid metabolites could reflect alterations in foetal fat metabolism within CHD. This is supported by studies demonstrating higher acylcarnitine concentrations in blood samples of children with VSDs compared to CHD not affecting pulmonary blood flow. 36 However, such studies include very small sample sizes (n = 18). Alternatively, intrinsic alteration in placental metabolism could be reflected in maternal blood. The placenta is known to utilise fatty acids in metabolism, with transplacental transfer of free fatty acids from mother to fetus throughout pregnancy. 37,38 Altered placental uptake of lipids and changes in lipid deposition are seen in maternal conditions such as obesity. 39 However, research in this area has not been performed in CHD.
Phosphatidylcholine is the principal phospholipid in human cells.
Phospholipids have essential roles including lipoprotein structure, facilitating lipid storage and transfer to prevent lipotoxicity; cell membrane structure for cellular integrity; and mitochondrial membrane structure and modulation of lipid metabolism. 40  and oxidised glutathione (GSSG) and lower reduced glutathione (GSH). 30,31 Figure 4 illustrates the homocysteine-methionine and transulfuration pathways with published associated outcomes.
Methionine is an essential amino acid, metabolised to homocysteine with SAM and S-adenosylhomocysteine (SAH) as intermediates.
SAH and homocysteine exist in equilibrium, favouring the conversion of homocysteine to SAH. 30,31 SAH is an inhibitor of cellular methyltransferases, influencing gene expression through methylation during organogenesis. 30,45 This is a possible mechanism of homocysteine toxicity. 46 Homocysteine metabolism in the placenta is predominantly folate and B12 dependent conversion to methionine. Raised homocysteine in mothers has been associated with pre-eclampsia, placental abruption, growth restriction and stillbirth illustrating potential toxicity. Homocysteine can cross the placenta, making foetal effects possible. 47 Approximately 50% of homocysteine is irreversibly converted to cystathionine to enter the trans-sulfuration pathway. 31 Several metabolites within this pathway facilitate oxidative stress through formation of reactive oxidative species. 31,46 GSH is an antioxidant that under normal conditions exists in a significantly higher concentration than the oxidised form GSSG. A change in the ratio towards GSSG is characterised by excessive oxidative stress. 48 A shift in this pathway has been seen in animal studies to increase incidence of neural tube defects. 49 Furthermore, lower GSH has been identified in the rat hyperglycaemic model of induced embryonic malformation. 50 This process illustrates a potential mechanism of toxicity through the homocysteine trans-sulfuration pathway.
These findings illustrate potential mechanisms of organotoxicity driven by inherent maternal metabolic alterations through inhibition of cellular methyltransferases and increased oxidative stress. However, mechanistic causative associations have not currently been identified. Furthermore, assessment of maternal profiles was out with the index pregnancy so may not reflect the peri-conceptual period.

| A foetal effect?
True alterations in maternal metabolic profiles associated with foetal CHD could represent underlying maternal risk factors, or derive from a foetal effect driven by CHD or the placenta. 21 Whilst common pathways such as lipid metabolism were identified through included studies, only leucine was replicated in two independent studies with contradictory findings. This makes the foetal effect less likely; however, this cannot be definitively concluded given the small sample sizes and variation of the studies included. At present, established biomarkers utilised in screening for foetal pathologies are often placentally derived, for example, pregnancy associated plasma protein A. 52 To date, no obvious biomarker has emerged in the included studies within maternal blood or urine for screening of CHD.
A potential method to distinguish maternal metabolic risk profile relative to foetal effect is to assess the maternal metabolome out with the index pregnancy in mothers with offspring affected by CHD.
Included studies have assessed the non-pregnant maternal metabolome utilising targeted platforms. [30][31][32] Pilot data suggested persistence of altered maternal lipid profiling in the postnatal period, potentially implicating an underlying maternal risk factor profile.
However, available evidence is limited. Furthermore, untargeted assessment has not been performed. Longitudinal studies of the metabolome in blood and urine over a period of up to 7 years suggests a high degree of conservation. [53][54][55][56] Untargeted metabolomic assessment out with pregnancy would provide further clarification on these hypotheses.

| Strengths and limitations
This is the first study to systematically review the association between

| Conclusions
At present, there is not enough available evidence to utilise maternal metabolomic biomarkers as screening tools for CHD in clinical practice. In addition, further research is required to identify maternal metabolic risk factors and establish biomarker identification and validation across different populations. Delineating maternal risk profiles and a potential foetal effect remains a priority in future research, which can be addressed through metabolomic assessment with index pregnancies. This review identifies several metabolic pathways that illustrate potential CHD pathogenic mechanisms. In addition, the development of combined screening tools using metabolites and ultrasound features is an exciting prospect for future research.