Maternal homocysteine in pregnancy and offspring birthweight: epidemiological associations and Mendelian randomization analysis

Abstract

Background: Disturbed one-carbon (1-C) metabolism in the mother is associated with poor fetal growth but causality of this relationship has not been established.

Methods: We studied the association between maternal total homocysteine and offspring birthweight in the Pune Maternal Nutrition Study (PMNS, Pune, India) and Parthenon Cohort Study (Mysore, India). We tested for evidence of causality within a Mendelian randomization framework, using a methylenetetrahydrofolatereductase (MTHFR) gene variant rs1801133 (earlier known as 677C→T) by instrumental variable and triangulation analysis, separately and using meta-analysis.

Results: Median (IQR) homocysteine concentration and mean (SD) birthweight were 8.6 µmol/l (6.7,10.8) and 2642 g (379) in the PMNS and 6.0 µmol/l (5.1,7.1) and 2871 g (443) in the Parthenon study. Offspring birthweight was inversely related to maternal homocysteine concentration—PMNS: –22 g/SD [95% confidence interval (CI): (–50, 5), adjusted for gestational age and offspring gender]; Parthenon: –57 g (–92, –21); meta-analysis: –40 g (–62, –17)]. Maternal risk genotype at rs1801133 predicted higher homocysteine concentration [PMNS: 0.30 SD/allele (0.14, 0.46); Parthenon: 0.21 SD (0.02, 0.40); meta-analysis: 0.26 SD (0.14, 0.39)]; and lower birthweight [PMNS: –46 g (–102, 11, adjusted for gestational age, offspring gender and rs1801133 genotype); Parthenon: –78 g (–170, 15); meta-analysis: –61 g (–111, –10)]. Instrumental variable and triangulation analysis supported a causal association between maternal homocysteine concentration and offspring birthweight.

Conclusions: Our findings suggest a causal role for maternal homocysteine (1-C metabolism) in fetal growth. Reducing maternal homocysteine concentrations may improve fetal growth.

Key Messages.

• Maternal 1-C metabolism is thought to be important for fetal growth.

• Indian mothers have high homocysteine concentrations due to dietary deficiencies (vitamin B12 and protein).

• Higher maternal homocysteine concentrations predict earlier delivery and lower birthweight.

• Mendelian randomization analysis linked variant rs1801133 of the MTHFR gene with neonatal birthweight.

• This supports causality of the association between maternal 1-C metabolism and fetal growth.

• Reducing maternal homocysteine concentrations may improve fetal growth.

Introduction

Small size at birth is associated with poor perinatal outcomes and an increased risk of type 2 diabetes and cardiovascular disease (CVD) in later life.^1–3 Indian babies are among the smallest in the world and this is attributed to the small size and poor nutrition of their mothers.⁴ Studies relating maternal nutrition to newborn size have mostly investigated macronutrients; recent observational and interventional studies have shown a role for maternal micronutrient nutrition.^5–6

The Pune Maternal Nutrition Study (PMNS) investigated the effect of maternal nutrition on fetal growth and future risk of diabetes and CVD. We found that higher frequency of intake of micronutrient-rich foods and higher circulating folate and vitamin C predicted larger newborn size.⁷ In a subset of PMNS participants, we found that mothers of small-for-gestational-age (SGA) babies had higher total homocysteine concentrations compared with those in appropriate-for-gestational-age pregnancies.⁶ Homocysteine is a key metabolite in one-carbon (1-C) metabolism, and homocysteine levels influence several cellular processes including DNA methylation and synthesis of nucleic acids and proteins.⁸ Homocysteine concentration is influenced by a number of genetic polymorphisms that affect the function of different enzymes regulating 1-C metabolism, and also by dietary intake of ‘methyl donors’ (folate, vitamin B₁₂, choline and betaine) and other vitamins (B₆ and B_2, co-factors for enzymes regulating the 1-C metabolism). The association of elevated maternal homocysteine with fetal growth restriction suggests a role of maternal 1-C metabolism in fetal growth and development, and a potential for improvement by dietary manipulations.^9–11 However, observational associations may be confounded by various lifestyle factors and may suffer from reverse causality. Causality is best proved by conducting randomized controlled trials which are time- and labour-intensive, expensive and sometimes unethical if they deviate from local established standards of care. Mendelian randomization is an alternative method that uses genetic variants associated with environmental and lifestyle exposures to overcome these limitations. The genotype is randomized at conception and is unaffected by lifestyle confounders, and there cannot be reverse causality. Therefore the ‘instrumental component’ calculated from genotype can only be causally related. Demonstration of such a relationship, therefore, improves confidence in the causality of associations.¹²

Methylenetetrahydrofolatereductase (encoded by MTHFR) is an important enzyme in 1-C metabolism that helps the conversion of homocysteine to methionine by providing the methyl donor tetrahydrofolate.¹³ Although several homocysteine-raising variants of MTHFR have been identified, the rs1801133 variant (earlier known as 677C→T polymorphism) significantly reduces the enzyme efficiency (by 75%), and individuals carrying the risk genotype ‘TT’ have ∼20% higher circulating homocysteine levels than those with the wild CC genotype.¹⁴ This polymorphism has been used as an instrument for investigating causality of associations of homocysteine with various cardiometabolic phenotypes,^15,16 and in clinical practice to predict high risk of adverse outcomes in pregnancy (neural tube defects and other congenital anomalies).¹⁷

We aimed to investigate the association between maternal homocysteine concentration and offspring birthweight in a large number of mother-offspring pairs by combining data from two Indian birth cohorts. We also used the rs1801133 variant for a Mendelian randomization analysis (instrumental variable and triangulation methods) to test the hypothesis that homocysteine concentrations in mothers have a causal role in determining fetal growth and size at birth.

Methods

Participants

The PMNS was started in 1993 in six villages near Pune to study the relationship between maternal nutrition and fetal growth.⁷ Women were studied twice during pregnancy (18 ± 2 and 28 ± 2 weeks of gestation) for nutritional, biochemical and demographic information. Socioeconomic status (SES) was assessed using a standardized questionnaire for rural India.¹⁸ None of the women were receiving vitamin supplements at enrolment. According to national guidelines, the women were supplemented with iron (60 mg) and folic acid (500 μg) tablets, one daily for 100 days from 18 weeks of gestation onwards.

The Parthenon Cohort Study was set up in 1997 to investigate the relationship between maternal glucose tolerance and fetal growth in the antenatal clinic at the Holdsworth Memorial Hospital (HMH), Mysore.¹⁹ Women were studied at 30 ± 2 weeks of gestation for anthropometric, biochemical and demographic information. SES was assessed using the Kuppuswamy scoring system for urban India.²⁰ The number of deliveries reflects only those who delivered at the HMH. Maternal supplement use was recorded when the women were recruited in mid pregnancy, but not at 30 weeks of gestation. At recruitment, 188 women (36.5%) reported taking vitamin supplements containing both vitamin B₁₂ and folic acid, and 52 (10%) reported taking folic acid alone though the dosage was not recorded.

The institutional ethics committees at both centres approved the study and informed consent was obtained from all women and their families.

Data collection

In the PMNS, neonatal birthweight was measured to the nearest 50 g using a Salter spring balance (Salter Abbey, Suffolk, UK) within 72 h of birth, and in the Parthenon Cohort Study within 24 h, using a digital weighing scale (Seca, Germany) to the nearest 5 g. In the PMNS, an early morning fasting blood sample was collected at home, transported on ice to the local research centre and centrifuged, and plasma aliquots were transferred to the KEM Hospital Research Centre (KEMHRC) and stored at –80°C until further use. In the Parthenon Cohort study, fasting blood was collected in the clinic, plasma separated within 2 h and aliquots stored at –80°C.

Biochemical and genetic analyses

Red cell folate concentrations in PMNS were measured by radioimmunoassay (Becton Dickinson, Oxford, UK),⁷ plasma vitamin B₁₂ by microbial assay^21,22 and homocysteine by the fully automated GC-MS method.²³ For the Parthenon study, plasma folate and plasma vitamin B₁₂ were measured by microbial assays^21,22,24 and homocysteine by fluorescence polarization immune assay at KEMHRC.²⁵ Plasma glucose and triglyceride concentrations were measured by standard enzymatic kits using auto analyzer (Abbott VP Super System, Irving, TX). The inter- and intra-assay coefficient of variation was <5% for all measurements.

Genotyping for both studies was carried out at the CSIR-Centre for Cellular and Molecular Biology, Hyderabad. Genomic DNA was isolated from blood samples using the salting-out method and DNA plated in uniform concentrations. We used a Sequenom-based MassARRAY technology to genotype the rs1801133 variant. The genotyping success rate was >95%. We confirmed the genotypes for ∼10% of the samples by sequencing the purified polymerase chain reaction (PCR) products on an ABI 3730 Genetic Analyzer. Inconsistency of only 0.003% (1/384) suggested high genotyping accuracy.

Definitions

Low vitamin B₁₂ refers to a plasma concentration <150 pmol/l²³ and hyperhomocysteinaemia to a concentration >10 µmol/l.²⁶ Low folate status was defined as an erythrocyte folate concentration <283 nmol/l and plasma folate concentration <7 nmol/l.²⁷ Prematurity refers to delivery at <37 weeks of gestation, low birthweight (LBW) as a birthweight <2500 g and SGA as per Oken et al.²⁸

Statistical methods

Data are presented as mean (SD) for continuous variables, median (25th-75th centiles) for skewed variables and percentages for categorical variables. Vitamin B₁₂, homocysteine and folate concentrations had skewed distributions and were log-transformed for the analyses. They were adjusted for the gestational age at which the sample was taken. Because of differences in methods between the two cohorts, we used study-specific SD scores for folate concentrations and SES scores in the combined analysis. Higher scores refer to better SES. Associations between exposures and birthweight were analysed using multiple linear regression adjusted for gestation and gender. Effect sizes are represented as change in physiological units [95% confidence intervals (CI)] per SD change in the exposure. In the genotype homocysteine association, the effect size is represented as SD change in homocysteine per allele. This aligned the units with those used in the triangulation analysis. Genotype distribution at rs1801133 was tested for compliance with the Hardy-Weinberg equilibrium (HWE). Associations between maternal genotype and continuous variables were examined using an additive model. Since genotype frequency, homocysteine concentrations and birthweight were different between the two centres, we examined associations in each cohort separately and then combined by meta-analysis using a fixed effect inverse variance model after testing for between-study heterogeneity (I²).

We tested the causality of the association between maternal homocysteine concentration and offspring birthweight by instrumental variable and triangulation analysis.²⁹ For the instrumental variable analysis, we used a two-stage least-square process. In the first stage, we calculated the SD change in homocysteine concentration per maternal rs1801133 allele in the PMNS and Parthenon cohorts. In the second stage, this instrument was used to estimate the association between maternal homocysteine and offspring birthweight. The difference between estimates from the observed and instrumental variable analyses was tested using the Durbin-Wu-Hausman test. We also performed a meta-analysis of these results.

In the triangulation method (see Supplementary data available at IJE online), we calculated the expected effect of maternal homocysteine on offspring birthweight by dividing the observed effect between maternal rs1801133 variant and offspring birthweight, and the observed effect between maternal rs1801133 variant and homocysteine concentration, and compared it with the observed effect.

Results

We restricted the analysis to 526 mother-baby pairs in the PMNS and 515 mother-baby pairs in the Parthenon cohort study, whose maternal homocysteine concentrations, maternal rs1801133 genotype and offspring birthweight data were available. These mothers and babies were no different from others in the original cohorts with respect to maternal age, body mass index (BMI), SES and offspring birthweight (data not shown). Mothers in the PMNS were younger and thinner than Parthenon mothers, and more likely to be vegetarian (Table 1). PMNS babies were lighter at birth than Parthenon babies, and had a higher prevalence of SGA births.

Table 1.

Maternal and offspring characteristics in Pune Maternal Nutrition Study (PMNS) and Parthenon Study

Characteristic	PMNS	Parthenon	P-valued
Number	526	515
Maternal
Age (years)a	21.5 (3.5)	23.9 (4.1)	<0.001
Primiparab	160 (30.4)	256 (49.7)	<0.001
SES scorea	27.1 (6.6)	34.0 (6.4)	NA
Vegetarian dietb	164 (31.6)	44 (9.3)	<0.001
Tobacco useb
Smoking	0 (0)	0 (0)	NA
Chewing	68 (13.0)	0 (0)	0.01
BMI (kg/m2)a	20.4 (1.9)	23.7 (3.6)	<0.001
Gestation at measurement of biomarkers (weeks)	29.4 (1.2)	29.1 (1.6)	0.51
Plasma homocysteine (µmol/l)c	8.6 (6.7, 10.8)	6.0 (5.1, 7.1)	<0.001
>10 µmol/lb	170 (32.3)	19 (3.7)	<0.001
Erythrocyte folate (nmol/l)c	958 (734, 1261)	–	NA
<283 nmol/lb	1 (0.2)	–	NA
Plasma folate (nmol/l)c	–	34.4 (16.8, 51.2)	NA
<7 nmol/lb	–	24 (4.7)	NA
Plasma vitamin B12 (pmol/l)c	122 (94, 164)	162 (123, 223)	<0.001
<150 pmol/lb	368 (70.4)	221 (43.0)	<0.001
Offspring
Gender (boys)b	280 (53.2)	247 (48.0)	0.09
Gestation (weeks)a	39.0 (1.7)	39.1 (1.7)	0.93
Preterm (<37 weeks)b	47 (8.9)	46 (8.9)	0.99
Birthweight (g)a	2642 (379)	2871 (443)	<0.001
Low birthweight (<2500 g)b	160 (30.4)	91 (17.7)	<0.001
SGAb	356 (67.7)	217 (42.1)	<0.001
Maternal genotype
rs1081133b
CC	378 (71.9)	415 (80.6)	0.003
CT	132 (25.1)	92 (17.9)
TT	16 (3.0)	8 (1.5)
Offspring genotype
rs1081133b
CC	381 (74.9)	393 (78.9)	0.029
CT	113 (22.2)	101 (20.3)
TT	15 (2.9)	4 (0.8)

SES, socioeconomic status; BMI, body mass index; SGA, small-for-gestational-age.

NA refers to measurements which are not comparable; SES can not be compared as the scales of measurements vary according to rural/urban area.

Values are mean (SD)^a or number (%)^b or median (25th, 75th centile)^c.

^dThe P-value is for the difference between the PMNS and Parthenon cohorts derived using t-tests for continuous and chi square tests for categorical variables.

Maternal homocysteine, folate and vitamin B₁₂ concentrations

Median maternal homocysteine concentration was 8.6 µmol/l in the PMNS and 6.0 µmol/l in the Parthenon cohort (32.3% and 3.7% had hyperhomocystenaemia, respectively) (Table 1). Low folate was rare in both cohorts but 70% of mothers in the PMNS and 43% in the Parthenon cohort had low vitamin B₁₂ concentrations. Both vitamin B₁₂ and folate concentrations were inversely related to homocysteine concentrations (PMNS: r = –0.27 and –0.39, respectively; Parthenon: r = −0.24 and r = –0.28, respectively; P < 0.001 for all). Maternal BMI was inversely associated with vitamin B₁₂ concentrations in the Parthenon cohort (standardized β = −0.4, P = 0.02) but not in the PMNS (standardized β = −0.09, P = 0.3). Maternal folate concentration was directly associated with SES in the Parthenon cohort (r = 0.3, P < 0.001). Maternal B₁₂ was directly associated with SES in the PMNS (r = 0.1, P = 0.04). Non-vegetarians had lower median homocysteine concentrations than vegetarians in the Parthenon cohort (5.9 vs 6.7 µmol/l, P < 0.01).

Maternal homocysteine, folate and vitamin B₁₂ concentrations and offspring birthweight

Folate concentrations were positively associated with birthweight (PMNS: β = 22.8 g/SD, 95% CI: 4.7, 40.9, P = 0.01; Parthenon: β = 8.2 g/SD, 95% CI: –7, 23.8, P = 0.3). Vitamin B₁₂ concentrations were not associated with birthweight in either cohort (PMNS: β = 5.1 g/SD, 95% CI: –14.5, 24.6, P = 0.6; Parthenon: β = 1.6 g/SD, 95% CI: –13.9, 17.1, P = 0.8). Mothers with higher plasma homocysteine concentrations delivered earlier and had lighter babies (Table 2). One SD increase in plasma homocysteine concentration predicted a 0.1-week earlier delivery (P = 0.15) in the PMNS, and a 0.2-week earlier delivery (P = 0.03) in the Parthenon cohort. The effect in both cohorts combined, using fixed effects meta-analysis, was 0.1 week/SD (P = 0.01), P for heterogeneity = 0.5 and I²= 0%. After adjusting for gestational age and gender, one SD increase in plasma homocysteine concentration predicted a 22-g lower birthweight (P = 0.1) in the PMNS, 57 g (P = 0.002) in the Parthenon cohort and 40 g (P  < 0.001) in the meta-analysis, P for heterogeneity = 0.15.

Table 2.

Association between maternal plasma total homocysteine (tHcy) and offspring birthweight and gestational age at birth

Quintiles of tHcy (µmol/l) (median, range)	Birthweight (g) (mean, SD)	Gestation (weeks) (mean, SD)
PMNS (N=526)
1. 5.44 (1.78–6.34)	2713 (348)	39.4 (1.6)
2. 7.14 (6.38–7.78)	2605 (402)	38.9 (1.7)
3. 8.59 (7.79–9.37)	2638 (392)	39.1 (1.8)
4. 10.38 (9.38–11.38)	2655 (389)	39.1 (1.6)
5. 13.33 (11.41–27.59)	2601 (357)	39.1 (1.7)
P1	0.02	0.15
β1 (95% CI)	–39 (–71, –7)	–0.10 (–0.25, 0.04)
P2	0.11	0.16
β2 (95% CI)	–22 (–50, 5)	–0.10 (–0.25, 0.04)
Parthenon (N=515)
1. 4.49 (2.26–4.85)	2921 (431)	39.3 (1.6)
2. 5.31 (4.87–5.68)	2928 (365)	39.4 (1.3)
3. 6.03 (5.70–6.41)	2909 (405)	39.0 (1.6)
4. 6.82 (6.42–7.47)	2821 (445)	38.8 (1.7)
5. 8.51 (7.48–40.73)	2772 (536)	38.8 (2.2)
P1	<0.001	0.03
β1 (95% CI)	–74 (–112, –36)	–0.17 (–0.32, –0.02)
P2	0.002	0.02
β2 (95% CI)	–57 (–92, –21)	–0.17 (–0.32, –0.02)
Meta-analysis
P1	<0.001	0.01
β1 (95% CI)	–56 (–81, –32)	–0.14 (–0.24, –0.03)
P for heterogeneity	0.16	0.5
I2	48.4%	0.0%
P2	<0.001	0.009
β2 (95% CI)	–40 (–62, –17)	–0.14 (–0.24, –0.04)
P for heterogeneity	0.15	0.6
I2	51.0%	0.0%

tHcy concentrations are shown as quintiles for ease of interpretation. However, in regression analysis tHcy is used as a continuous variable in SD units.

P¹: unadjusted; P²: adjusted for gender and gestational age as appropriate.

β1, β2 represent corresponding effect sizes.

Maternal rs1801133 genotype, homocysteine concentrations and offspring birthweight

The genotype frequencies in the mothers and children did not show evidence of deviating from HWE in both the cohorts (P > 0.05). The maternal frequency of the risk allele (‘T’) at rs1801133 was 15% in the PMNS and 11% in the Parthenon cohort; it was similar in mothers and children (Table 1). The TT genotype was uncommon in both cohorts (≤3%), but was more common in the mothers and babies in the PMNS.

The maternal T allele at rs1801133 was associated with higher maternal plasma homocysteine concentrations and TT mothers had the highest plasma homocysteine concentrations (Table 3). Per allele, there was a 0.30-SD (P < 0.001) increase in homocysteine in the PMNS, a 0.21-SD (P = 0.04) in the Parthenon cohort and 0.26-SD (P < 0.001; P for heterogeneity = 0.5) in the meta-analysis. The T allele was not associated with other maternal characteristics which influence birthweight (circulating folate, vitamin B₁₂, fasting glucose, triglycerides, BMI and SES) (data not shown).

Table 3.

Association of maternal MTHFR rs1801133 genotype with homocysteine, offspring birthweight and gestational age at birth

MTHFR rs1801133 genotype	N	tHcy (µmol/l) (median, 25th, 75th centile)	Birthweight (g) (mean, SD)	Gestation (weeks) (mean, SD)
PMNS	526
CC	378	8.4 (6.6, 10.4)	2651  (383)	39.1 (1.8)
CT	132	9.3 (7.3, 11.9)	2620  (349)	39.0 (1.6)
TT	16	9.3 (6.9, 13.8)	2615  (530)	39.8 (1.6)
	P1	<0.001	0 .41	0.77
	β1 (95% CI)	0.30 (0.14, 0.46)	–26  (–88, 36)	0.04 (–0.23, 0.31)
	P2		0 .28	0.77
	β2 (95% CI)		–29  (–81, 23)	0.04 (–0.23, 0.31)
	P3		0 .11	0.72
	β3 (95% CI)		–46  (–102, 11)	–0.05 (–0.34, 0.24)
Parthenon	515
CC	415	6.1 (5.1, 7.1)	2865  (438)	39.0 (1.8)
CT	92	5.8 (4.8, 6.9)	2911  (449)	39.5 (1.5)
TT	8	7.7 (6.5, 10.6)	2679  (610)	39.2 (1.4)
	P1	0.04	0 .56	0.02
	β1 (95% CI)	0.21 (0.02, 0.40)	–26  (–113, 61)	0.39 (0.05, 0.72)
	P2		0 .10	0.02
	β2 (95% CI)		–67  (–148, 13)	0.39 (0.06, 0.72)
	P3		0 .10	0.05
	β3 (95% CI)		–78  (–170, 15)	0.40 (0.01, 0.79)
Meta-analysis
	P1	<0.001	0 .32	0.09
	β1 (95% CI)	0.26 (0.14, 0.39)	–26  (–78, 26)	0.18 (–0.03, 0.39)
	P-heterogeneity, I2	0.49, 0.0%	0 .99, 0.0%	0.10, 60%
	P2		0 .051	0.09
	β2 (95% CI)		–46  (–92, 0)	0.18 (–0.03, 0.39)
	P-heterogeneity, I2		0 .40, 0.0%	0.11, 61%
	P3		0 .02	0.32
	β (95% CI)		–61  (–111, –10)	0.12 (–0.12, 0.35)
	P-heterogenity, I2		0 .44, 0.0%	0.07, 71%

tHcy concentrations are adjusted for gestation at measurement.

P¹: derived by regression using genotype as a continuous variable; P²: additionally adjusted for gender and gestational age as appropriate; P³: additionally adjusted for offspring MTHFR rs1801133 genotype.

β1, β2, β3 represent corresponding effect sizes per allele change. For homocysteine this is in SD units, for others in physiological units

The maternal T allele at rs1801133 was associated with lower offspring birthweight (Table 3). In the meta-analysis, birthweight decreased by 46 g per allele (P = 0.051, adjusted for gestation and gender)and by 61 g (P = 0.019) after further adjustment for fetal genotype. Higher maternal folate protected the fetus against the birthweight-lowering effect of the TT genotype. Thus, the mean birthweight of babies born to TT mothers who had folate concentrations in the highest tertile was 2827 g compared with 2266 g among those in the lowest tertile. (P = 0.047, in the combined data) (Table 4).

Table 4.

Mean offspring birthweight (g) according to maternal MTHFR rs1801133 genotype by tertiles of folate concentrations in combined data

Maternal MTHFR rs1801133 genotype	Tertile of SD score for folatea
	1	2	3
CC	2747	2739	2814
CT	2639	2729	2903
TT	2266	2572	2827

^aWe used SD scores of folate concentrations because of different units of measurements in the two cohorts.

Instrumental variable analysis

We calculated the effect of the rs1801133 genotype on homocysteine concentration, which is the ‘gene effect' and therefore the ‘instrumental variable’ (Table 5). We expressed this as the SD change in homocysteine concentration per allele of the rs1801133 variant. In the second step, we used this variable to predict birthweight. In Pune, birthweight decreased by 84 g (P = 0.3, adjusted for gestation and gender) per SD change in tHcy concentration; in the Parthenon cohort by 309 g (P < 0.001) and in the meta-analysis by 224 g (P < 0.001). The difference between Pune and Parthenon of 231 g [standard error (SE) 130 g] was not statistically significant (P = 0.08). These results were unaffected by additional adjustment for the fetal genotype. There was a difference between the estimates from the observed and instrumental variable analysis (test of ‘endogenecity’). Thus, the instrumental variable effect was larger than the observed effect. These results suggest that the effect of homocysteine on birthweight is causal. Triangulation analysis also provided a similar conclusion (see Supplementary data, available at IJE online).

Table 5.

Observed and instrumental variable analysis of the maternal homocysteine offspring birthweight association

Analysis	Observed change in birthweight (g) per SD change in total homocysteine			Instrumental variable estimate of change in birthweight (g) per SD change in total homocysteine			P1
	Effect size (g)	95% CI	P	Effect size (g)	95% CI	P
Unadjusted
PMNS	–39	–71, –7	0.02	–89	–295, 118	0.40	0.63
Parthenon	–74	–112, –36	<0.001	–268	–462, –75	0.007	0.045
Meta-analysis	–56	–81, –32	<0.001	–198	–338, –57	0.006	0.044
Heterogenity P (I2%)	0.16 (48.4)			0.20 (33.5)
Adjusted for gestational age and gender
PMNS	–22	–50, 5	0.11	–84	–259, 90	0.34	0.48
Parthenon	–57	–92, –21	0.002	–309	–488, –131	0.001	0.005
Meta-analysis	–40	–62, –17	<0.001	–224	–349, –100	<0.001	0.003
Heterogenity P (I2%)	0.15 (51.0)			0.078 (68.0)
Additionally adjusted for fetal genotype
PMNS	–27	–55, 1	0.06	–135	–324, 53	0.16	0.25
Parthenon	–58	–94, –22	0.002	–308	–490, –125	0.001	0.006
Meta-analysis	–43	–65, –20	<0.001	–248	–378, –118	<0.001	0.002
Heterogenity P (I2%)	0.29 (37.9)			0.076 (39.4)

P¹: P-value for the Durbin-Wu-Hausman test for the difference between the estimates from observed and instrumental variable analysis.

Discussion

In this study from two Indian birth cohorts, higher maternal homocysteine concentration predicted lower offspring birthweight. This effect was contributed both by a shorter gestation and by growth restriction. The causality of the association is supported by the instrumental variable and triangulation analyses in the Mendelian randomization framework using maternal genotype at the rs1801133 variant of MTHFR. This showed that the instrumental variable estimate of the maternal homocysteine effect on offspring birthweight was larger than the observed effect of 28-week homocysteine measurements. The instrumental variable is free of confounding and measurement error, is fixed at birth and therefore represents ‘lifetime’ exposure of the conceptus to elevated maternal homocysteine concentration from the periconceptional period to the time of birth. It also eliminates the possibility of reverse causality. Thus it suggests a true underlying causal association between the exposure (homocysteine concentration) and the outcome (birthweight). This association was independent of offspring genotype, suggesting that maternal homocysteine metabolism rather than fetal homocysteine metabolism is responsible. The two Indian cohorts in this study are geographically separated and have demographic and nutritional differences. Moreover, the distribution of the rs1801133 genotype and homocysteine concentrations (exposure) and birthweight (outcome) were different. Such stratification might produce spurious associations. However, the meta-analysis confirms the associations observed in the individual cohorts and there was no evidence of heterogeneity in the results, supporting a true association.

There are only a few human studies which have investigated the role of maternal 1-C metabolism and related nutritional and genetic factors in fetal growth. These are predominantly in well-nourished European populations. The Hordaland study in Norway observed that high maternal plasma homocysteine concentrations (measured many years after the pregnancy) were associated with adverse pregnancy outcomes including low birthweight, and maternal rs1801133 polymorphism increased the risk of intrauterine growth restriction.³⁰ A study in Newcastle, UK³¹ and the Generation R cohort in The Netherlands³² both demonstrated that lower maternal folate concentrations predicted lower offspring birthweight. The variant rs1801133 of MTHFR by itself was unrelated to birthweight, but in combination with low folate predicted lower birthweight. In our study, women with TT genotype gave birth to lighter babies, but higher folate levels provided some protection against this effect. This suggests that both maternal nutrition and her genotype influencing 1-C metabolism have a causal influence on fetal growth and neonatal birthweight.

Mothers in our study are relatively thin, have low intakes of energy, protein and micronutrients and their babies are amongst the smallest in the world. Moreover, mothers in the Indian subcontinent have a different balance between folate and vitamin B₁₂ compared with Europeans: adequate in folate due to vegetarian food and supplements but low in B₁₂ due to a low intake of foods of animal origin.^7,33 Unlike in Europeans, in whom low folate is a major contributor to hyperhomocysteinaemia, vitamin B₁₂ contributed much more to the risk (49%) compared with folate (8%) in our population, similar to the findings in Nepalese pregnant women.³⁴ Low maternal vitamin B₁₂ predicted intrauterine growth restriction in a study in Bangalore, India³⁵ and in babies of migrant Indians in the UK,³⁶ whereas high maternal homocysteine and low folate predicted intrauterine growth restriction in Pakistan.³⁷ These studies support a role for maternal 1-C metabolism in fetal growth and suggest that disturbances in this pathway may lead to fetal programming, so called ‘nutrient-mediated teratogenesis’.³⁸ On this background, it is surprising that a recent study in Norwegian pregnant women failed to show an association between dietary folate intake, supplemental folic acid, maternal plasma folate, homocysteine concentrations in the second trimester and offspring birth size.³⁹

The possible mechanisms of these associations include the role of 1-C (-CH₃) groups in nucleic acid synthesis and methylation of DNA which contributes to regulation of gene expression.³³ In addition, 1-C metabolism provides the essential amino acid methionine for protein synthesis. The importance of dietary methyl donors in fetal programming is well illustrated in a number of animal studies.^40,41 Hyperhomocysteinaemia could also have a direct damaging effect on endothelium by lowering nitric oxide and stimulating pro-inflammatory and oxidative stress pathways, which could impair placental perfusion and function.⁴² Recently a number of studies have demonstrated an association between factors influencing fetal growth (nutrition and smoking), birth size and a number of epigenetic (DNA methylation) markers in the cord blood.^43–50

Our study has several strengths and some limitations. One of the major strengths is the simultaneous measurement of maternal nutritional and genetic markers in relation to fetal growth. Inclusion of two Indian birth cohorts increases the generalizablity of the results within India. The populations studied have a high incidence of micronutrient deficiencies and of low birthweight, increasing the statistical power and relevance of the investigation. Both studies have a comprehensive range of measurements including many well-known confounders and effect modifiers, allowing appropriate statistical adjustments in the analysis. The similarity of the results in the two cohorts, despite many differences in the design and methods, and results of meta-analysis suggests biological consistency. Findings in the instrumental variable and triangulation analysis strengthen the causality of association between maternal homocysteine and offspring birthweight.

One of the limitations was the difference in methods for some of the measurements in the two cohorts. We adjusted for these by using cohort-specific SD scores in the analysis. Another limitation is a relatively small number of mothers with the risk genotype TT because of the low frequency of the risk allele at rs1801133 variant which, however, is consistent with other Indian studies.⁵¹ Given the differences in risk allele frequency and other maternal and neonatal characteristics in the two cohorts, it is difficult to discount an effect of population stratification, but significant results in the meta-analysis argue against stratification. Use of a single genetic instrument, as opposed to multiple instruments that would allow checks for pleiotropy, may also constitute a limitation. However, the rs1801133 variant in MTHFR is a well-established instrument for homocysteine levels and has been used in many different studies for association with many diverse traits.^14,15 Potential pleotropic effects cannot be ruled out. In the presence of vertical pleiotropy or pathway-specific effects, this genetic variant will exert its effect on birthweight through a direct pathway. Whereas this is a potentially informative relationship, it is difficult to rule out alternative intermediate effects through DNA methylation and alteration in function of other genes. This could have horizontal pleiotropic effects, leading potentially to widespread biological effects and more global patterns of secondary effect.⁵² We acknowledge this biological complexity, but would like to emphasize that we have taken the best measures and their instruments to make appropriate use of the data and investigate biological relationships.

Our finding of an important role of maternal 1-C metabolism in fetal growth is likely to promote interest in public health interventions to improve vitamin B₁₂ and folate status in young Indian women. The role of folate in prevention of neural tube defects is well established and it is current standard practice to improve folate intake in the pre- and periconceptional period in developed countries. The role of maternal vitamin B₁₂ is only just emerging; it influences fetal growth, neural tube development and neurocognitive development.^53,54 In a recent vitamin B₁₂ intervention trial in India, started in the first trimester of pregnancy, there was a borderline reduction in incidence of intrauterine growth retardation.⁵⁵ An imbalance with low B₁₂ and high folate in the mother predicts insulin resistance in the child, and high folate predicts higher adiposity in the child.³³ Thus, balanced B₁₂ and folate nutrition in the mothers, taking care to avoid folate excess, may improve fetal growth and reduce risk of later diabetes and CVD, in line with the Developmental Origins of Health and Disease (DOHaD) theory of ‘fetal programming’. Our estimates suggest that if all the mothers had circulating homocysteine concentrations in the lowest quintile (median 4.7 µmol/l) seen in our study, the incidence of low birthweight would fall from 26% to 19%. This would be a substantial public health achievement and should be tested.

Supplementary Data

Supplementary data are available at IJE online.

Funding

This work was supported by funds from the Medical Research Council, UK, the Department for International Development, UK, the Wellcome Trust, UK, Parthenon Trust, Switzerland and Council of Scientific and Industrial Research (CSIR), Ministry of Science and Technology, Government of India, India [XII Five-Year Plan titled ‘CARDIOMED (BSC0122)’]. Some aspects of this work were supported by GEoCoDE, an International Research Staff Exchange Scheme supported by EU which helped travel and interaction to promote analysis and writting of this manuscript.