Untargeted Maternal Plasma Metabolomics in Hirschsprung Disease: A Pilot Study

ABSTRACT

Background and Aims:

Hirschsprung disease (HSCR) is a congenital disorder of unknown etiology affecting the enteric nervous system (ENS). Since the early gestational development of the ENS is dependent on the prenatal maternal metabolic environment, the objective of this pilot study was to explore the role of specific maternal plasma metabolites in the etiology of HSCR.

Methods:

In this cross-sectional study, postnatal (as a surrogate for prenatal) plasma samples were obtained from mothers of children diagnosed with HSCR (n = 7) and age-matched mothers of normal children (n = 6). The plasma metabolome was analyzed by ultra-high-pressure liquid chromatography and mass spectrometry. Metabolites were identified by mzCloud using Compound Discoverer software. Using an untargeted metabolomics workflow, metabolites with case versus control group differences were identified.

Results:

A total of 268 unique plasma metabolites were identified and annotated in maternal plasma. Of these, 57 were significantly different between case and control groups (P < 0.05, t-test). Using a false discovery rate corrected cutoff of 10% to adjust for multiple comparisons, 19 metabolites were significantly different in HSCR cases, including carnitines, medium-chain fatty acids, and glutamic acid. Pathways affected were for amino acid and lipid metabolism.

Conclusion:

Disordered prenatal metabolic pathways may be involved in the etiopathogenesis of HSCR in the developing fetus. This is the first study to assess maternal plasma metabolomics in HSCR.

INTRODUCTION

Hirschsprung disease (HSCR) is a congenital disorder affecting the enteric nervous system (ENS) of a child and is characterized by aganglionosis in the distal bowel. Affected children may present with intestinal obstruction, chronic constipation, failure to thrive, or enterocolitis, and if left untreated, HSCR can be life-threatening. The etiology of the disease is unknown but is thought to be due to a combination of genetic and nongenetic factors.[1] Dozens of genes, including RET proto-oncogene, have been identified to be associated with HSCR; however, the expression and penetrance of mutations are variable and depend on gene–environmental interactions.[2] Environmental factors that have been identified in experimental models include drugs such as mycophenolate, statins, artemisinin, high-dose ibuprofen, and Vitamin A deficiency.[3]

The prenatal or intrauterine environment of the mother has long been suspected to influence the health or disease of a child;[4] however, the role of maternal metabolites in the ontogeny of the gut is unknown. The enteric neurons of the gut are derived from neural crest cells which enter the fetal foregut, proliferate, differentiate, and then colonize the bowel craniocaudally between the 4^th and 12^th weeks of gestation.[5,6] A perturbation of the intrauterine environment during any of these key events of bowel innervation within can lead to HSCR. Indeed, maternal prenatal Vitamin A deficiency has already been proven to be a risk factor in experimental mice models.[7]

Metabolomics has emerged as a cutting-edge research technique to understand the underlying mechanistic etiology of a disease, by providing insights into the associated disordered biochemistry. Serum metabolomics were used to explore the biochemical profile of HSCR patients in a recently published study, which found that tryptophan metabolism was particularly disordered.[8] We hypothesized that circulating maternal metabolites could influence the disease risk of HSCR in the fetus and sought to compare the plasma metabolomic profile of mothers of children with HSCR and mothers of normal children. However, it is not feasible to evaluate the maternal plasma metabolome during the first trimester when HSCR develops as no prenatal diagnosis exists for HSCR. The earliest that HSCR can be diagnosed is in the neonatal period, making postnatal measurements shortly after HSCR diagnosis the nearest option for evaluating risk. Thus, while postnatal measurements may not be the ideal surrogate of first-trimester nutritional or metabolic status, they are the closest approximation one can get. This study, therefore, aimed to compare the plasma metabolomic profile of mothers of infants with HSCR with mothers of normal children, to evaluate their disease risk.

METHODS

Study population

This observational cross-sectional study was conducted at St. John’s Medical College Hospital, Bengaluru, India, over 8 months, after approval from the Institutional Ethics Board (IEC 89/2019). The mothers of infants (≤1 year of age) who were admitted to the pediatric surgery ward for suspected HSCR were screened. Mothers with excess gestational weight gain (>20 kg) or weight loss (>5 kg), or those on regular medications such as statins, antidepressants, and antipsychotics, or on supplements other than iron and calcium, were excluded. Mothers were recruited as soon as the diagnosis of HSCR was made in the child, by histopathological biopsy of the colon or rectum, with or without radiological evidence by contrast enema (n = 7). Consenting mothers (n = 6, controls) of children with no developmental or biochemical anomalies, matched for their age and the age of their child, were also recruited. History was obtained using structured questionnaires for sociodemographic data, breastfeeding practice, as well as antenatal and perinatal events. Medications or supplements taken during the periconceptual period, pregnancy, and till the time of recruitment were specifically recorded. Dietary recall of foods consumed in the periconceptual period was collected by a food frequency questionnaire.[9] The weight of the subjects was recorded in minimal clothing on a digital scale with a precision of 0.1 kg, and their height was recorded to the nearest 0.1 cm.

The subjects were instructed to report for blood sample collection at 7 a.m. after an overnight fast of 10 h. Ten milliliters of blood was collected in EDTA tubes (BD Vacutainer, Becton, Dickinson and Company, Franklin Lakes, NJ) between 07:30 and 08:00 a.m. by antecubital vein puncture. A free flow of blood without a tourniquet was ensured before sampling in all subjects. Samples were stored immediately in an ice box and centrifuged within 1 h of collection. A cooling centrifuge (REMI C-23 BL, Mumbai, India) was used for the separation of plasma at 3500 rpm for 15 min at 4°C, which was then aliquoted into cryovials and stored at −80°C until further analysis.

Untargeted plasma metabolomics

Plasma samples (100 μL) were spiked with a mixture of isotopically labeled and unlabeled internal standard (IS, 20 μL of a U-²H labeled amino acid mixture, 98%, Cambridge Isotope Laboratories, MA, USA); 12-[(cyclohexylcarbamoyl)amino]dodecanoic acid (CUDA, ≥95%, Cayman Chemical Co, MI, USA); Val-Tyr-Val, ≥98%; and creatine-(methyl-D₃) monohydrate, 98% (Sigma-Aldrich, Taufkirchen, Germany) and analyzed on a high-resolution accurate mass platform consisting of an ultra-high pressure liquid chromatography (Thermo Scientific, Vanquish Flex Binary, MA, USA) coupled to an orbitrap mass spectrometer (Q Exactive, Thermo Scientific, San Jose, USA) as described previously.[10] Quality control samples, which included four technical replicates of a pool of aliquots from plasma samples, were analyzed along with the biological samples daily. Metabolites were identified by mzCloud using Compound Discoverer software (version 3.1.0.305, ThermoFisher Scientific, Waltham, MA, USA).[11] For untargeted metabolomic data, groups area ratios, fold change (log2 scale), group-wise coefficient of variance, trend charts, principal component analysis (PCA), P value calculated by t-test, and adjusted P values (using the Benjamini–Hochberg correction for the false discovery rate) were analyzed within the Compound Discoverer™ 3.1 software. Metabolites with a false discovery rate (FDR) of <0.10 were considered statistically significant.

RESULTS

A total of seven cases and six controls were recruited. The average age of the mothers in the case group was 25.1 ± 2.5 years and in the control group was 28.3 ± 5.9 years (P = 0.22). The body mass index between the case and control groups was similar (23.1 ± 3.8 vs. 25.6 ± 7.1, P = 0.43). The median age of recruitment was within a month of delivery (range 0–8 months), and at the time of recruitment, the mean hemoglobin measured was 10.5 ± 0.7 g%. All 13 subjects were breastfeeding their children, and none were reported to be on regular medicines or supplements during the periconceptual period. All participants were initiated on folate in the first trimester and received iron and calcium supplements during the second and third trimesters. The total energy intake as well as macronutrient intake calculated from dietary recall was similar in both groups.

A total of 268 unique plasma metabolites were identified in maternal plasma and annotated. Of these, 57 were found to be significantly different between case and control groups (P < 0.05 by t-test). Using an FDR-corrected cutoff of 10% to adjust for multiple comparisons, 19 metabolites were significantly associated with HSCR. Decreased concentrations were observed for amino acids such as glutamic acid and ethionine (ethyl analog of methionine), hydroxy fatty acids such as hydroxyoctanoic acid and oxopentanoic acid, oxysterols like 7-ketocholesterol, and a few drugs (vigabatrin and piperidine). On the other hand, dimethyllysine, dihydrothymine, methyl tocotrienol, 3-hydroxyoctanoylcarnitine, and 3-hydroxydecanoyl carnitine concentrations were increased in cases compared to controls [Table 1]. Box plots representing the normalized areas of three amino acids are presented in Figure 1. The normalized areas of three metabolites related to lipid metabolism are represented as box plots in Figure 2. Although not significant when based on the adjusted P value cutoff of < 0.1, the normalized areas of tryptophan and kynurenine were higher in mothers with HSCR children [unadjusted P < 0.03, Table 1].

Table 1.

Plasma metabolites with significantly different abundances between case and control groups

Metabolites	Log2 fold change: (Case)/(control)	P value: (Case)/(control)	Adjusted P value: (Case)/(control)
DL-Glutamic acid	−0.69	<0.001	0.005
Diosgenin	0.82	<0.001	0.005
8-Methyltocotrienol	0.87	<0.001	0.005
Vigabatrin	−0.39	<0.001	0.009
3-[5-(4-Methoxyphenyl)-2-oxazolyl] pyridine	0.44	<0.001	0.010
3-hydroxyoctanoyl carnitine	0.68	<0.001	0.014
7,8-Didehydro-4,5-epoxymorphinan-3,6-diol	−4.28	<0.001	0.014
3-hydroxydecanoyl carnitine	0.87	<0.001	0.015
(2S)-2-Amino-8-hydroxyoctanoic acid	−1.34	<0.001	0.030
Fenoxycarb	−2.71	<0.001	0.030
L-Ethionine	−0.50	0.002	0.060
Piperine	−1.53	0.003	0.070
(1S,4S)-4-hydroxy-3-oxocyclohexane-1-carboxylate	−0.43	0.004	0.081
Terbinafine	1.22	0.004	0.082
N (6), N (6)-Dimethyllysine	0.27	0.005	0.091
Dihydrothymine	0.55	0.005	0.091
Natriuretic peptide B	0.34	0.005	0.091
2-oxopentanoic Acid	−0.40	0.005	0.091
7-Keto cholesterol	−0.90	0.006	0.094
Oleic acid	0.5	0.007	0.102
Propionylcarnitine	−0.4	0.007	0.106
Guvacine	−0.84	0.008	0.110
Maleamic acid	0.3	0.008	0.110
1-Methylhistamine	0.33	0.008	0.110
D-Pantothenic acid	0.64	0.008	0.114
Nalorphine	−1.42	0.010	0.136
2,3,5-Trimethylphenol	0.35	0.012	0.147
3-Methyladenine	0.21	0.013	0.151
Estradiol undecylate	−0.85	0.013	0.151
Octadecyl (E)-p-coumarate	−0.88	0.014	0.162
Atagabalin	−0.35	0.015	0.166
Hydroxyhexanoycarnitine	0.94	0.015	0.167
Ametryn	0.28	0.015	0.166
LysoPC (20:5 (5Z,8Z,11Z,14Z,17Z))	−0.44	0.016	0.173
N-Methylpyrrolidone	−1.27	0.017	0.177
trans-2-Dodecenoylcarnitine	0.98	0.018	0.180
L-(-)-Asparagine	0.37	0.018	0.183
Indole	0.28	0.018	0.184
Creatinine	0.3	0.020	0.187
7-Hydroxy-Pyrazolo[4,3-D] Pyrimidine	0.38	0.020	0.187
Indoleacrylic Acid	0.2	0.020	0.190
DL-Tryptophan	0.21	0.020	0.190
Isoprene	0.15	0.020	0.190
Trimethylamine N-oxide	−0.78	0.021	0.191
Calystegin A3	−1.68	0.021	0.195
Leu-Val	0.86	0.023	0.201
N-Boc-4-piperidone	0.31	0.025	0.209
Kynurenic acid	0.44	0.027	0.215
(2E)-3-(3,4-Dihydroxyphenyl)-N-[2-(3,4-dihydroxyphenyl) ethyl] acrylamide	−1.5	0.030	0.230
Acetyl Arginine	0.67	0.031	0.224
Nadolol	0.85	0.033	0.227
3-[(2,6-Dimethylheptanoyl) oxy]-4-(trimethylammonio) butanoate	−0.26	0.034	0.230
Butenylcarnitine	0.7	0.038	0.250
Octadecanedioic acid	−0.63	0.039	0.251
24-Epibrassinolide	−0.49	0.041	0.257
Asp-Leu	−0.18	0.045	0.274
Sorbic acid	0.12	0.046	0.277

Fifty-seven plasma metabolites with significantly different abundances between case and control groups (P<0.05 by t-test). The top 19 metabolites are significant with a false discovery rate of <10%, P<0.1

Figure 1.

Box plots of significant metabolites in amino acid metabolism depicting levels of (a) glutamate, (b) tryptophan, and (c) kynurenic acid in cases (light blue) and controls (orange). “P” denotes the unadjusted P value and “q” denotes the adjusted P value after multiple testing corrections. The bold dark line of the box-whisker plot indicates the median coefficient of variation, the top and bottom ends of the rectangle are indicative of the 25th and 75th percentiles, and the whiskers measure the 95% confidence interval. Orange dots denote outliers in the control group

Figure 2.

Box plots of significant metabolites in lipid metabolism depicting levels of (a) 3-hydroxyoctanoyl carnitine, (b) 3-hydroxydecanoyl carnitine, and (c) 2-Amino-8-hydroxyoctanoic acid in cases (light blue) and controls (orange). “p” denotes the unadjusted P value and “q” denotes the adjusted P value after multiple testing corrections. The bold dark line of the box-whisker plot indicates the median coefficient of variation, the top and bottom ends of the rectangle are indicative of the 25th and 75th percentiles, and the whiskers measure the 95% confidence interval. Blue and orange dots denote outliers in the case and control groups, respectively

Metabolic profiles were modeled according to group (case and control). Based on PCA, the two subject groups could be separated along principal component 1 (PC1). PC1 and PC2 explained 17.5% and 12.5%, respectively, of the variance in the metabolomic data. Examination of the plot showed that the plasma metabolome of the cases was generally separated from the plasma metabolome obtained from the controls, with the exception of a single outlier [Figure 3].

Figure 3.

Principal component analysis of maternal plasma untargeted metabolomic profiles. Plot of principal component PC1 and PC2 of the three plasma samples from cases (dark blue circles) and controls (orange circles) showing separation between groups

DISCUSSION

This study found that 19 metabolites in the maternal postnatal plasma metabolome were significantly associated with the diagnosis of HSCR in infants, including carnitines, medium-chain fatty acids, and glutamic acid. Amino acid, lipid metabolism, and tryptophan–kynurenine pathways were the main metabolic pathways that were found to be affected in mothers who had children with HSCR.

Carnitines transport long-chain fatty acids into the mitochondria for energy production, and in our study, in particular, 3-hydroxyoctanoyl carnitine and 3-hydroxydecanoyl carnitine were significantly increased in the mothers of HSCR children. A corresponding significant decrease in the fatty acid hydroxyoctanoic acid is in line with the previously reported increase in liver mitochondrial fatty acid oxidation in male rats maintained on a Vitamin A-deficient diet.[12] Low vitamin A has been identified as a maternal risk factor for HSCR in studies using retinol-binding protein-deficient (Rbp4^−/−) pregnant mice,[13] and a deficient state is thus corroborated by changes in the maternal plasma metabolome. However, a direct link to retinol metabolism was not observed with either plasma retinol or other carotenoids in our study. Enhanced fatty acid oxidation was also measured in the metabolome of HSCR colonic tissue in a study conducted on 75 HSCR patients and 75 controls.[14] The fatty acids involved in the colonic tissue study were long-chain fatty acids that were involved in the synthesis of PGE2. Our study, however, did not find any alteration in prostaglandin levels in the maternal blood or any changes in long-chain fatty acids. This may be due to localized production and action of prostaglandins in the colonic tissue of the child with no corresponding changes in the metabolome of the mother.

A recent study of the serum metabolome of HSCR patients[8] shows disordered tryptophan metabolism which is also reflected in the maternal metabolic profile in our study. The mothers with HSCR children had a change in the tryptophan–kynurenine pathway (unadjusted P < 0.03) which may point toward inflammation which can independently affect ENS structure and function. A few drugs were found significant in the plasma metabolome in spite of all individuals not being on any medication as per history. This finding may be explained, however, by the possibility of medications given during the time of delivery as 61.5% of the subjects were within the 1^st month of the postnatal period.

Plasma glutamic acid concentration was reduced in mothers with HSCR children, which is a finding consistently seen in other biological samples of HSCR. In a study of 50 preoperative children with short-segment HSCR, serum glutamic acid concentration was found to be much lower than in 50 normal children (0.57 nmol/μL vs. 0.93 nmol/μL, P < 0.005).[15] The western blot analysis of the colonic tissue of HSCR children from this study (n = 90) showed that the abundance of glutamate (normalized to beta-actin abundance and expressed as relative gray values) was higher in the ganglionic compared to the aganglionic segments of the colon (0.198 vs. 0.04, respectively). In 38 postoperative HSCR patients, fecal metabolome revealed that glutamic acid was significantly reduced, in comparison to 21 healthy siblings (P = 0.030).[16] Furthermore, strong expression of glutamate synthetase has been reported in the normal human gut and ganglionic segments of HSCR colon; however, this is absent in the aganglionic segments of HSCR colonic tissue.[17] Our study reports similar lower concentrations of glutamic acid in postnatal plasma of mothers of HSCR children. The significance of this in the etiology of HSCR is unknown.

This is the first study to assess maternal plasma metabolome in HSCR patients. It was ensured that the mothers were recruited as early as possible, and in six out of the seven cases, HSCR was diagnosed within the first 3 months of life. Care was also taken to match the control mothers for the postpartum month of diagnosis of the HSCR cases. In addition, all participants were administered a food frequency questionnaire to ensure that there were no drastic changes in dietary intake between the groups that could affect the metabolome. Precautions were taken to ensure a standardized sample processing and storage protocol to reduce changes in the biochemical profile due to technical factors. Besides the inability of obtaining a prenatal plasma sample, limitations of the study include the small sample size which precludes definitive conclusions being made on HSCR risk.

Larger-scale maternal plasma metabolomic analyses can provide valuable insights into putative biological etiologies of HSCR and may also be useful in biomarker discovery for early detection. This has clinical implications for prevention in high-risk mothers, timely diagnosis of HSCR, early surgical intervention, and improved outcomes in the child.

CONCLUSION

This pilot study points to changes in amino acid and lipid metabolism in the postnatal state in mothers of children with HSCR, which may reflect events and nutritional status in the first trimester. Further studies with a larger sample size are required to assess the role of maternal glutamate, tryptophan metabolism, and fatty acid oxidation pathways in relation to HSCR risk in the child.

Financial support and sponsorship

This study was supported by the Clinical Research Training Program (CRTP) grant from the Wellcome Trust/Department of Biotechnology (DBT) India Alliance to AVK (IA/CRC/19/1/610006). SGH is a Clinical Research Fellow of the CRTP grant. AM (IA/CPHI/19/1/504593) is independently supported by Wellcome Trust/DBT India Alliance Fellowships.

Conflicts of interest

There are no conflicts of interest.