|Home | About | Journals | Submit | Contact Us | Français|
Folic acid intake during pregnancy can reduce the risk of neural tube defects (NTDs) and perhaps also oral facial clefts. Maternal autoantibodies to folate receptors can impair folic acid binding. We explored the relationship of these birth defects to inhibition of folic acid binding to folate receptor α (FRα), as well as possible effects of parental demographics or prenatal exposures.
We conducted a nested case–control study within the Norwegian Mother and Child Cohort Study. The study included mothers of children with an NTD (n= 11), cleft lip with or without cleft palate (CL/P, n= 72), or cleft palate only (CPO, n= 27), and randomly selected mothers of controls (n= 221). The inhibition of folic acid binding to FRα was measured in maternal plasma collected around 17 weeks of gestation. On the basis of prior literature, the maternal age, gravidity, education, smoking, periconception folic acid supplement use and milk consumption were considered as potential confounding factors.
There was an increased risk of NTDs with increased binding inhibition [adjusted odds ratio (aOR) = 1.4, 95% confidence interval (CI) 1.0–1.8]. There was no increased risk of oral facial clefts from inhibited folic acid binding to FRα (CL/P aOR = 0.7, 95% CI 0.6–1.0; CPO aOR = 1.1, 95% CI 0.8–1.4). No association was seen between smoking, folate supplementation or other cofactors and inhibition of folic acid binding to FRα.
Inhibition of folic acid binding to FRα in maternal plasma collected during pregnancy was associated with increased risk of NTDs but not oral facial clefts.
Folate deficiency is an established risk factor for neural tube defects (NTDs) and may increase the risk of oral facial clefts as well (Oakley, 1999; Wilcox et al., 2007). Randomized and community-based clinical trials have demonstrated that the prevalence of NTDs can be decreased by providing folic acid to women during the periconception period (MRC, 1991; Czeizel and Dudas, 1992; Berry et al., 1999; Berry and Li, 2002). In addition, folic acid fortification of the food supply has been temporally associated with declines in the prevalence of NTDs in several countries, including the USA (Mersereau et al., 2004; Botto et al., 2006). The mechanism of folic acid protection from NTDs is unknown.
Folate receptor antibodies measured in maternal blood have been previously associated with an increased risk of NTDs and oral clefts (Rothenberg et al., 2004; Bliek et al., 2006; Cabrera et al., 2008; Molloy et al., 2009; Bille et al., 2010). There are no known causes of the development of folate receptor antibodies (with resulting inhibition of receptor binding). One theory is that consumption of cow milk can stimulate antibodies to bovine folate receptor that may also block the human folate receptor, and a milk-free diet has been an effective treatment for cerebral folate deficiency syndrome (Ramaekers et al., 2008). A study of adults in Spain found a correlation between milk intake and folate receptor-blocking autoantibody, but not folate status (Berrocal-Zaragoza et al., 2009b). This could be related to cross-reactivity in the assay, or to a true risk factor for the development of folate receptor autoantibodies. Previous studies of autoantibodies or inhibited binding of folic acid have not considered associations with parental demographics and prenatal exposures. Maternal smoking and folic acid supplement use in the periconception period may also influence folic acid metabolism (Piyathilake et al., 1994).
We evaluated the risk of folate-related birth defects associated with inhibition of folic acid binding to human folate receptor α (FRα) in maternal plasma collected during mid-gestation. Immunoglobulin (Ig)G and IgM antibody titers were also measured, but were ultimately not useable. We hypothesized that each of these assays might be indicative of impaired folate transport into the cell. We had hoped that the Ig assays would indicate whether the autoimmune response was transient (IgM) or long-lasting (IgG) but these titers would not have indicated whether the folic acid-binding activity of FRα was blocked by these antibodies. While the binding assay, we present here is the more function-related assay, we cannot determine what component of maternal plasma is involved in the inhibition. We also considered possible associations of inhibition of folic acid binding with parental demographics and prenatal exposures.
The Norwegian Mother and Child Cohort Study (MoBa), conducted by the Norwegian Institute of Public Health (NIPH, Oslo, Norway), is a cohort consisting of 110 000 pregnancies recruited from 1999 to 2008 (Magnus et al., 2006). Most pregnant women in Norway received an invitation to participate by post prior to the routine 17–18 weeks ultrasound examination offered to all pregnant women in Norway (http://www.fhi.no/eway/default.aspx?pid=238&trg=MainArea_5811&MainArea_5811=5895:0:15,3046:1:0:0:::0:0) (April 28, 2011, date last accessed). Informed consent, initial questionnaires and blood samples were obtained from both parents at the obstetric visit. The study was approved by The Regional Committee for Medical Research Ethics in Norway, the Norwegian Data Inspectorate and the Institutional Review Board of the U.S. National Institute of Environmental Health Sciences.
The current study includes all mothers of cases of NTDs and oral facial clefts included in version three of the quality-assured data files (including 63 182 babies born between 1999 and 2006). All births in Norway are included in the Medical Birth Registry (MBR) and MoBa participants have MBR information, including birth defects, linked to the study. Cases were identified through the ICD10 codes in the MBR.
We identified 20 cases of NTDs enrolled in MoBa, which is far fewer than the expected 63 (1 per 1000) (Petrova and Vaktskjold, 2009). This difference is probably because mothers choose not to enroll in MoBa once a prenatal diagnosis of NTD has been made. Of the mothers who did participate, half did not provide 17-week blood samples. We identified 4 cases of anencephaly (two with available sample), 14 cases of spina bifida (seven available) and 2 cases of spina bifida occulta (both available). These 11 NTD cases were successfully assayed.
We identified 38 cases of cleft palate only (CPO), 21 cases of cleft lip (CLO) and 64 cases of cleft lip and palate (CL&P). We would have expected 50 cases of CPO, 39 cases of CLO and 55 cases of CL&P based on a large study in Norway (Harville et al., 2005). We assessed cleft palate separately from cleft lip with or without cleft palate (CL/P) because the two types of defects have somewhat different risk factors and distinct gestational windows of closure during embryogenesis (Mitchell et al., 2002). We successfully assayed 27 of the 38 cases of CPO, and 72 of the 85 cases of CL/P. We randomly selected 221 control mother samples from 17-week maternal plasma samples on the same 96-well storage plates as the case samples, and all were successfully assayed (Ronningen et al., 2006). These control samples had been processed in the NIPH Biobank concurrently with the case samples. Some case and control babies included in this study had other birth defects and medical conditions recorded in the MBR.
Parents enrolling in MoBa completed their initial questionnaires around 17 weeks gestation, at the time of the first prenatal ultrasound visit. This first questionnaire includes information on reproductive history, vitamin supplement use before and during early pregnancy (including folic acid in single and multivitamin formulations), living habits (such as smoking), occupational and recreational activities. A food-frequency questionnaire was completed by the mothers around 22 weeks gestation regarding their diet since becoming pregnant. All questionnaires are available at: http://www.fhi.no/eway/default.aspx?pid=238&trg=MainArea_5811&MainArea_5811=5903:0:15,3138:1:0:0:::0:0 (April 28, 2011, date last accessed).
Inhibition of folic acid binding to the folate receptor has previously been associated with higher milk intake but not with all dairy products (Berrocal-Zaragoza et al., 2009b). The association did not depend on the fat content of the milk, so we calculated the average daily intake of all types of milk based on seven questions in the food-frequency questionnaire. This questionnaire has demonstrated correlation with a 4-day food diary for milk intake (Spearman r = 0.66) in a validation study within MoBa (Brantsaeter et al., 2009).
Maternal plasma sample plates were located in the NIPH Biobank in Oslo, Norway, and 30 µl of plasma was transferred to a standard 96-well storage plate by two technicians, confirming the placement of each sample. Case and control samples were randomly arranged on the plate, and status was masked from the staff of the Biobank and the laboratory conducting the assays. Plasma samples were frozen and shipped on dry ice to the Institute of Biosciences and Technology at the Texas A&M University Health Science Center in Houston, TX, USA.
While earlier studies used folate receptor from human placental preparations or bovine folate-binding protein, here we used recombinant soluble human FRα protein from a baculovirus expression system engineered by Manohar Ratnam's laboratory, as previously described (Bille et al., 2010). Briefly, FRα cDNA was modified at the N-terminal by replacing the leader peptide with Apis melifica honeybee melitin signal peptide and with a stop codon at position +703 in the C-terminal. The modified cDNA was cloned into baculovirus, producing high-titer viral stocks in SF9 cells for purification of FRα recombinant protein.
Measurement of the inhibition of folic acid binding to FRα was carried out as previously described (Bille et al., 2010). FRα was coupled to 96-well glass slides at a concentration of 10 µg/ml via a monolayer of epoxysilane, then folic acid-depleted subject samples are incubated with the immobilized FRα, followed by incubation with a set amount of folic acid-horseradish peroxidase (FA-HRP) and development of HRP signal. All slides were imaged using an 8-bit UV photography workstation (Kodak, New York, NY, USA). Fluorescence intensities were measured in the area printed with FRα (foreground) and in the unprinted background by Image J and the difference between foreground and background was calculated for each well (NIH, Bethesda, MD, USA).
Competitive inhibition of FA-HRP binding to FRα by unlabeled folic acid in Ig-depleted serum generates a standard curve relating fluorescent signal to nanogram of folic acid blocked from binding FRα per milliliter of serum. The decrease in FA-HRP signal caused by interference with folic acid binding to FRα is converted based on the mean of two standard curves of 16 concentrations from 800 ng/ml to 195 fg/ml (the lower limit of detection for the assay) for each plate. This range covers the physiological range of folate concentrations in human serum (Felkner et al., 2009). The mean standard curve is then used to interpolate the amount of folic acid blocked from binding to FRα in the folic acid-depleted subject serum samples in nanogram per milliliter. The R2 goodness of fit test was >0.95 on all plates (12 df). Previous comparisons in the laboratory found inconsistent differences between serum and plasma samples. For two FRα autoantibody detection assays, glass 96-well slides were prepared and processed (Bille et al., 2010). A 1:20 plasma dilution was applied, followed by a secondary conjugate labeled with alkaline phosphatase specific for detecting human IgG or IgM. Ig assays were run on each plate with four serial-dilution curves of commercially available pooled plasma (7 concentrations from 1 to 0.0025), and four negative controls of antibody-depleted sera. These curves were used to interpolate the level of FRα autoantibody present in case and control samples by constructing a dilution curve from a standard (if not known) amount of FRα autoantibody present in pooled plasma. However, since these curves also resulted in the dilution of numerous other plasma components that could affect the assay, the lowest concentration pooled plasma samples were often not used in the construction of the standard curve. Pooled plasma samples at the same dilution as the case and control samples (1:20, 0.05) were always incorporated into the standard curves. These assays had low intra- and inter-assay coefficients of variation, calculated using 1:20 dilutions of pooled sera (IgG = 1.6 and 2.3%; IgM = 3.2 and 9.6%, respectively). The lower limit of detection for IgG and IgM in pooled sera was at a 1:400 dilution.
All assays results were reported to the National Institute of Environmental Health Sciences prior to unmasking case or control status.
For the inhibition of folic acid binding to FRα assays, the standard dilution curves on each plate were fit to sigmoidal dose–response (variable slope) curves, also referred to as four-parameter logistic equations (Delean et al., 1978). R2 goodness-of-fit tests were used for each plate. The regression curves were used to transform the sample intensity differences to a concentration of folic acid inhibition in nanogram per milliliter. Medians and interquartile range (IQR) were calculated for each of the four case and control groups. Distributions were highly skewed, with high plate-to-plate variation of inhibition patterns; this required statistical adjustment for plate variability in the logistic regression. Adjustment for other factors was considered, although none are thought to be true confounders, and these results are included in the Supplementary data, Tables.
Fluorescence values of the IgG and IgM assays were transformed using the linear portion of the standard curves on each plate. For these assays, the standard curves covered a limited range of the subject sample values (70–95%) and the shape of the curves varied greatly among plates run on different days. Owing to poor correlation with the presumably more functionally relevant binding inhibition assays (and with limited biological samples available for retesting), we did not explore these data further.
Patterns of inhibition of folic acid binding to FRα in the standard curves were similar within the two groups (of three plates each) that were run concurrently. Six of the NTD samples were on the group of plates with higher inhibition of folic acid binding to FRα in the control samples (1.7 versus 0.4 ng/mL), necessitating adjustment for plate group in the analysis. ORs for risk of each defect from inhibition of folic acid binding to FRα were calculated using logistic regression adjusting for plate group. Adjustment for parental demographics and prenatal exposures did not substantially change the OR estimates (data not shown). For rare outcomes such as birth defects, OR can be interpreted as a relative risks. Most parental demographics and prenatal exposures were assessed by Pearson correlations. The level of inhibition of folic acid binding to FRα was compared in the women who did or did not take folic acid supplements using a Satterthwaite t-test for unequal variances. In order to assess a possible association of milk intake with folate receptor-blocking autoantibody production (Berrocal-Zaragoza et al., 2009b), we also calculated a Pearson correlation coefficient between daily milk intake and inhibition of folic acid binding to FRα. Statistical analysis was performed in SAS version 9.1 (Cary, NC, USA). Statistical significance was set at P < 0.05.
The median level of inhibition of folic acid binding to FRα was higher in NTD mothers (1.5 ng/mL, IQR: 0.2–2.2) than control mothers (0.37 ng/mL, IQR < 0.0001 to1.8). Seven of the eleven NTD mothers had folic acid-blocking levels above the control median level. Inhibition of folic acid binding to FRα was associated with an increased risk of NTDs in the offspring [unadjusted odds ratio (uOR) 1.3 (95% confidence interval (CI) 1.0–1.7)]. The association strengthened slightly after adjustment for the group of plates assayed concurrently [adjusted odds ratio (aOR) 1.4 (95% CI 1.0–1.8), Fig. 1]. In contrast, median inhibition levels in both CL/P and CPO were slightly lower than controls (CL/P 0.25 ng/mL, IQR < 0.0001 to 1.2; CPO 0.27 ng/mL, IQR < 0.0001 to 0.61). There was no evidence for increased risk of clefts from inhibition of folic acid binding (CL/P uOR = 0.8 (0.6–1.0), aOR = 0.7 (0.6–1.0); CPO uOR = 0.9 (0.7–1.2), aOR = 1.1 (0.8–1.4)]. Adjustment for parental demographics and prenatal exposures, and removing cases with additional congenital malformations, did not affect the risk estimates (Supplementary data, Table SI). Removing cases with other birth defects also did not change the risk estimates (Supplementary data, Table SII).
IgG and IgM were measured in controls and the three case groups. NTD mothers had higher median levels of IgG compared with control and cleft case mothers (NTD = 0.015, IQR: 0.005–0.0185; CPO = 0.008, IQR: 0.005–0.022; CL/P = 0.008, IQR: 0.004–0.014; control = 0.007, IQR: 0.003–0.013). However, NTD and CPO mothers had lower median levels of IgM compared with control and CL/P mothers (NTD = 0.04, IQR: 0.02–0.07; CPO = 0.04, IQR: 0.02–0.07; CL/P = 0.5, IQR: 0.03–0.1; control = 0.06, IQR: 0.03–0.1). Given the inconsistency of the assay as described in the methods section, small differences between groups, the high variation within groups and the skewed distribution of subject samples that could not be transformed to reasonable normality, we chose not to perform statistical comparison tests for these data.
Case and control mothers were similar in age, gravidity and education (Table I). Fewer mothers of infants with CL/P took folic acid supplements during the first trimester (45 versus 64% in controls), increasing the risk of CL/P from maternal smoking only (OR = 2.2, 95% CI 1.3–3.8). Contrary to expectations, 10 of the 11 NTD mothers took folic acid supplements during this time. A higher proportion of clefts mothers reported smoking at 17 weeks gestation (13% of CL/P, 11% of CPO and 10% of controls) but these differences were not significant. A substantial number of mothers had quit smoking during their pregnancy (18% of CL/P, 7% of CPO and 17% of control mothers).
There may be exposures that affect the likelihood of developing factors that inhibit the binding of folic acid to FRα. Within the control mothers only, we compared parental demographics and prenatal exposures with the inhibition of folic acid binding to FRα. There were no significant correlations between mother's age, father's age, gravidity (Supplementary data, Table SIII), maternal education or smoking and inhibition of folic acid binding to FRα. The strongest correlation was a slight decrease of inhibition with mother's age (Pearson r = −0.1, P= 0.16). Folic acid supplement use was also not associated with folic acid inhibition levels. Of 221 control mothers, 172 (78%) completed the food-frequency questionnaire. There was no correlation between milk intake and inhibited folic acid binding to FRα (Pearson r = 0.014, P= 0.85).
Embryos undergoing rapid cell division after conception require adequate levels of folate to develop properly, and folate insufficiency has been associated with a variety of reproductive problems including NTD and CL&P (Kirke et al., 1993; O'Neill, 1998; George et al., 2002; Wilcox et al., 2007). In preliminary studies, antibodies to the folate receptor have been associated with both of these birth defects (Rothenberg et al., 2004; Bliek et al., 2006; Cabrera et al., 2008). In our study, mothers with plasma that inhibited folic acid from binding to FRα had an increased risk of NTDs in their offspring. Despite the very small number of NTD cases, this association was significant before and after adjusting for assay variation. We saw no similar increase in risk for oral facial clefts.
Previous studies of maternal autoantibodies to folate receptors have found large increases in risks of NTDs [odds ratio (OR) 27, 95% CI 3.8–190] and CL/P (OR 11, 95% CI 1.4–81) despite very small sample sizes (Rothenberg et al., 2004; Bliek et al., 2006). Table II includes a summary of all previous studies of folate receptor antibodies and birth defects, which used blood samples collected during pregnancy as well as at varying lengths of time after pregnancy. The assays performed in these studies also varied in the source of folate receptor (bovine, human placental or human recombinant) and used different assays for antibodies that bind to the receptor or inhibit the binding of folic acid to the receptor.
We assessed several maternal factors for their associations with inhibited folic acid binding to FRα. Smoking and folic acid supplementation are both associated with risk of CL/P and consistent with a previous study in Norway (Wilcox et al., 2007); however, these factors were not associated with inhibited folic acid binding to FRα. In fact the strongest correlation with maternal age was in the opposite direction from a previous non-significant finding of older age in mothers with antibodies (Bliek et al., 2006). While plates of samples processed concurrently had similar distribution of inhibition levels, the large differences between the two groups of plates increased the overall variation. This may have decreased our power to detect modest associations with biologically plausible environmental exposures.
The assays in this study used recombinant human FRα rather than placental or bovine-derived folate receptor. Antibodies to both human placental FR and bovine folate-binding protein (FBP) have been reported to be more common in mothers of NTD cases (Cabrera et al., 2008). No association with oral facial clefts was seen in the only prior study using the recombinant folate receptor (Bille et al., 2010), and we feel that the recombinant source of folate receptor is more consistent than the human placental preparations previously used. High homology between human folate receptor and bovine FBP may lead to antibodies that do not bind specifically to the human form of the receptor, and could potentially lead to assay cross-reactivity with antibodies to bovine FBP from milk exposure. Carriers of FR-blocking autoantibodies have been reported to have higher levels of milk intake (Berrocal-Zaragoza et al., 2009b) but we found no such correlation. Unfortunately there is no crystal structure for human FRα, and the actual epitope is unknown, which complicates the prediction of antibody response.
The majority of NTDs in Norway are detected at the 17-week ultrasound, which is the obstetric visit at which mothers enrolled in MoBa. Most pregnancies which are affected by an NTD in Norway are terminated and would not be included in the MBR, where we identified our cases. If a woman terminated or considered terminating the pregnancy she may be less likely to enroll in MoBa, which is designed to follow the children's development over many years: this presumably explains the much lower than expected number of NTD cases ascertained.
Selection bias may impact our results when considering who enrolled in MoBa and which case mothers had 17-week blood samples available. A study of the MoBa cohort found a self-selection bias for prevalence estimates of 23 exposures and outcomes, but the eight exposure-outcome associations studied were not biased (Nilsen et al., 2009). Of the 85 cases of CL/P we identified in version 3 of the MoBa database, 72 had 17-week samples available (85%). Of the 38 cases of CPO, 27 had samples (71%); while only 11 of 20 NTD cases (55%) had available samples. Control samples were selected randomly from 17-week samples stored on the same plate as the case samples to match time and processing at the Biobank.
It is not known how levels of antibodies that bind to FRα might fluctuate before, during, and after pregnancy. A study of fertility found that levels of inhibited binding to the folate receptor varied (around the assay detection limit) over several measured cycles of attempted conception, with preconception samples from subfertile cases more likely than controls to have positive folate receptor antibody titers (Berrocal-Zaragoza et al., 2009a).
Many previous studies have relied on maternal blood samples which were collected months or years after the delivery of the index pregnancy (Table II). We used blood samples collected at around 17 weeks gestation (~15 weeks post-conception)—a time-point closer to the critical window for closure of the neural tube (Week 4 post-conception) and the lip and palate (Weeks 5–10 post-conception). Our association with NTDs but not oral clefts is consistent with two previous studies that only used samples taken during pregnancy (Cabrera et al., 2008; Bille et al., 2010), but conflicts with two studies that only used samples taken after the pregnancy (Bliek et al., 2006; Molloy et al., 2009). However, these studies also used different assays in different laboratories (Table II).
The direction of causation in this study is not clear. We cannot determine if maternal inhibition of folic acid binding to FRα leads to an increased risk of having a child with an NTD, or if the pregnancy affected by an NTD causes the maternal serum to develop antibodies that inhibit folic acid binding to FRα (for example, by maternal exposure to fetal neuronal tissue). It would be necessary to collect maternal blood before pregnancy to clarify this and to determine the utility of measuring inhibited folic acid binding to folate receptor as a preconception clinical screening tool.
In conclusion, mothers of NTD cases in the Norwegian Mother and Child Cohort Study had more inhibition of folic acid binding to FRα in 17-week gestation plasma samples but no such association was seen for mothers of oral facial clefts. Parental demographics and prenatal exposures were not associated with inhibited folic acid binding to FRα. Given similar findings with respect to high maternal titers of blocking antibodies among women carrying NTD pregnancies in cohorts in both NY, USA and CA, USA, it appears that these blocking antibodies may play a role in the etiologic pathway of NTDs (Rothenberg et al., 2004; Cabrera et al., 2008). Larger sample sizes will be required to determine whether environmental factors, such as smoking, can contribute to the development of these folate receptor-blocking antibodies.
A.L.B, A.J.W., R.T.L. and R.H.F. developed the study design. R.M.C. and R.H.F. developed the assay. J.L.B. and E.B.G. performed the laboratory analysis. A.L.B. and D.R.M. performed the statistical analysis. A.L.B., E.B.G, J.L.B., A.J.W. and R.H.F. wrote the manuscript which was revised and approved by all authors.
This research was supported by the Intramural Research Program of the NIH, National Institute of Environmental Health Sciences; U.S. National Institutes of Health Grant Numbers DE016315 and NS050249; and the Albert and Margaret Alkek Foundation.
The authors would like to thank Jane A. Hoppin, Matthew P. Longnecker, and the staff of the Norwegian Institute of Public Health for their assistance with this project.