Risk of retinoblastoma is associated with a maternal polymorphism in dihydrofolatereductase (DHFR) and prenatal folic acid intake


  • The second, third, and fourth authors contributed equally to this article.



The incidence of unilateral retinoblastoma varies globally, suggesting possible environmental contributors to disease incidence. Maternal intake of naturally occurring folate from vegetables during pregnancy is associated inversely with the risk of retinoblastoma in offspring.


The authors used a case-control study design to examine the association between retinoblastoma risk and maternal variations in the folate-metabolizing genes methylenetetrahydrofolate reductase (MTHFR) (a cytosine-to-thymine substitution at nucleotide 677 [MTHFR677C→T]; reference single nucleotide polymorphism rs1801133) and dihydrofolate reductase (DHFR) (a 19-base-pair deletion of intron 1a [DHFR19bpdel]; rs70991108). In central Mexico, 103 mothers of children with newly diagnosed unilateral retinoblastoma were enrolled in an institutional review board-approved study along with a control group of 97 mothers who had healthy children. Mothers were interviewed regarding perinatal characteristics, including use of prenatal vitamin supplements, and gave peripheral blood samples, which were used for polymerase chain reaction-based genotyping of rs1801133 and rs70991108.


The risk of having a child with unilateral retinoblastoma was associated with maternal homozygosity for DHFR19bpdel (odds ratio, 3.78; 95% confidence interval, 1.89-7.55; P = .0002), even after controlling for the child's DHFR19bpdel genotype (odds ratio, 2.81; 95% confidence interval, 1.32-5.99; P = .0073). In a subgroup of 167 mothers with data on prenatal intake of supplements containing folic acid (a synthetic form of folate), DHFR19bpdel-associated risk was elevated significantly only among those who reported taking folic acid supplements. Maternal MTHFR genotype was unrelated to the risk of having a child with retinoblastoma.


Maternal homozygosity for a polymorphism in the DHFR gene necessary for converting synthetic folic acid into biologic folate was associated with an increased risk for retinoblastoma. Prenatal ingestion of synthetic folic acid supplements may be associated with increased risk for early childhood carcinogenesis in a genetically susceptible subset of the population. Cancer 2012. © 2012 American Cancer Society.


Retinoblastoma, a malignant primitive neuroectodermal tumor, arises in immature retinal cells. The incidence of retinoblastoma varies geographically. The highest rates of incidence are reported in some populations of Africa, South America, India, and Scandinavia1, 2; whereas, in the United States, comparisons by ethnicity indicate that incidence is highest among Latinos.3

A US-based case-control study in the 1980s suggested that factors associated with perinatal poverty, such as lack of prenatal vitamins and lower maternal education, were associated with having a child with retinoblastoma.4 Global variation in the incidence of retinoblastoma appears to be primarily among the sporadic unilateral form,2 which, in contrast to the bilateral form, involves only 1 eye and is not associated with a family history of the disease. Because the median age at diagnosis for unilateral retinoblastoma is 23 months, the initiating events for tumor development are likely to occur during pregnancy and early infancy. In 85% of cases, the unilateral form is also associated with mutations in the retinoblastoma 1 (RB1) gene that occur in somatic (but not germ) cells. Although the origin of these somatic RB1 mutations is unknown, the majority involve methylated cytosines in CpG islands and result in stop codons and truncated retinoblastoma protein (pRb).5, 6 A recent report has documented imprinting of a maternal RB1 gene.7 In a case-control study in Mexico (before folic acid fortification), we observed that maternal consumption of vegetable-derived folate during pregnancy was protective against having a child with unilateral retinoblastoma.8

Folate provides precursors for nucleotide biosynthesis and methyl groups, through methyl folate, for methylation reactions, including DNA methylation. The potential role of folate in the development of pediatric tumors derived from embryonal layers has not been well studied. Polymorphisms in genes encoding the enzymes involved in metabolism of 1-carbon donors have been studied in relation to the development of birth defects, such as neural tube defects (NTD), as well as cancer incidence in adults.9-11 Because retinoblastoma is a primitive neuroectodermal tumor arising in the retina and, thus, is derived embryologically from the neural tube, we hypothesized that dysregulation of folate metabolism during early retinal development may influence tumorigenesis.

In the current study, we examined the relation between retinoblastoma risk and the prevalence of maternal polymorphisms in 2 genes that encode 2 enzymes of folate metabolism: methylenetetrahydrofolate reductase (MTHFR) (a cytosine-to-thymine substitution at nucleotide 677 [MTHFR677C→T]; reference single nucleotide polymorphism rs1801133) and dihydrofolate reductase (DHFR) (a 19-base-pair [bp] deletion of intron 1a [DHFR19bpdel]; rs70991108). MTHFR catalyzes the synthesis of 5 methyltetrahydrofolate, which is used for homocysteine methylation. The MTHFR 677C→T (rs1801133) polymorphism has been studied extensively regarding increased susceptibility for the development of NTD when folate availability is low9 and for the risk of cancer incidence.11 In addition to its role in thymidylate synthesis, DHFR is responsible for the conversion of orally ingested folic acid (in supplements and fortified foods) into folate (reduced), which can then be incorporated into the body's folate pool. The 19-bp deletion in intron 1a of the human DHFR gene (DHFR19bpdel) is associated with an increased risk for NTD12 as well as an increased risk for breast cancer among women taking folic acid supplements.10 Homozygosity for DHFR19bpdel is associated with decreased efficiency in using folic acid13, 14 and is present in 17% to 19% of the populations that have been studied.10, 13, 14

We conducted a case-control study in central Mexico to examine whether the risk for having a child with retinoblastoma is associated with homozygosity for polymorphisms in MTHFR and DHFR. A second objective was to examine whether prenatal folate intake is associated with the risk of retinoblastoma.


Between January 2000 and December 2009, 108 mothers of children with newly diagnosed, unilateral retinoblastoma at 2 adjacent referral hospitals in Mexico City were invited to participate in this case-control study, which was approved by the institutional review boards of all participating institutions. Mothers of children with a known family history of retinoblastoma were not eligible to participate in the study. The study was designed to recruit 100 cases and 100 controls. Two mothers declined to participate. Case mothers (N = 106) were enrolled during their child's initial visits to the treating hospital. Enrolled mothers were asked to refer a friend (not related by blood to the case mother) who had a child of the same age (±1 year) as their child to serve as a control mother. Among the 97 control mothers, >81% were the first friend the case mother approached, whereas the remaining mothers were the second friend approached. Control mothers were enrolled during home visits in 18 states in central and southern Mexico. At the time of enrollment, all participating mothers provided signed consent. Blood samples were then obtained from all mothers and from most case and control children. Mothers were interviewed regarding supplement intake during the first trimester of pregnancy using a validated questionnaire.15 If women reported taking any supplements, then they were asked about the supplement's brand, dose, and form of administration (tablet, powder, etc). Supplemental intake of folic acid was noted as present or absent after analyzing the folic acid content of the supplements reported using a nutrient content database for supplements available in Mexico.16

Blood was drawn into Becton Dickinson CPT vials (Becton Dickinson, Franklin Lakes, NJ) and stored at 4°C until centrifugation. Buffy coat samples were stored at −80°C until DNA extraction using standard nonorganic methods (Qiagen, Valencia, Calif). Genotyping for the MTHFR 677C→T and DHFR 19bpdel polymorphisms were performed by polymerase chain reaction amplification using restriction fragment-length polymorphism and allele-specific methods, respectively.9, 10, 12 The resulting polymerase chain reaction products were separated onto 3% agarose gels and were observed with 1ethidium bromide. Approximately 20% of samples were selected randomly to run in duplicate with 99% concordance. Laboratory personnel were blinded to sample origin. Batches contained equivalent proportions of case and control samples. Samples and questionnaires were linked through deidentified and bar-coded labels.

Statistical Analysis

Quantitative variables

Maternal age and the number of years of school completed were examined as continuous variables. Genotyping was examined by comparing the homozygous variant genotype with the genotypes that contained at least 1 wild-type allele based on results from analyses done with the Framingham Offspring Cohort demonstrating impaired function with the homozygous 19-bp deletion for DHFR13, 14 and data demonstrating impaired MTHFR function with the homozygous TT genotype.17 Vitamin supplement use was defined as whether or not mothers reported taking vitamins that contained synthetic folic acid during their first trimester.

Statistical methods

We calculated summary statistics to describe sample characteristics and used chi-square tests and t tests to detect group differences in categorical variables and continuous variables, respectively. Variables with skewed distribution were transformed to meet assumptions for a t test. On the basis of the distribution of mothers' genotype, we estimated allele frequency and tested for Hardy-Weinberg equilibrium in the control mothers. A logistic regression model was used to assess the association between maternal genotype and the odds of having a child with disease, with and without adjustment for control variables of the child's genotype. The child's genotype variable had 3 categories with 1 category for “missing genotype” on 33 children. Odds ratios (ORs) and 95% confidence intervals (CIs) were derived from the model parameters, and standard errors were used to aid interpretation.


We recruited 106 mothers who had children with unilateral retinoblastoma at the time of initial diagnosis, of which, 103 were eligible for participation in the study (the child of 1 mother did not have retinoblastoma on histopathologic examination, and it was discovered that the children of 2 mothers had family members with retinoblastoma and, thus, they were excluded). All references to “case mothers” refer to mothers of children with unilateral retinoblastoma. All references to “control mothers” refer to mothers who had healthy children. All control mothers were healthy and met eligibility criteria to participate.

Analyses were performed on a data set that contained measurements from 200 mothers (97 control mothers and 103 case mothers) who had data from interviews regarding perinatal characteristics as well as blood samples and had genotype results for DHFR19bpdel (198 mothers) or MTHFR677C→T (197 mothers). Case mothers did not differ from control mothers in demographic characteristics, such as age at delivery of the index child, the child's birth weight, and maternal education, or in the proportion reporting smoking during pregnancy or folic-acid supplement use in their first trimester of pregnancy (Table 1). Supplement intake was missing for 31 mothers, because they were lost to follow-up before completing their dietary interview. Thus, DHFR and prenatal supplement data were available for 167 mothers (96 controls and 71 cases). Among mothers who were missing data on supplement use, there was a significantly greater proportion who were mothers of case children, were homozygous for DHFR19bpdel, and had lesser education compared with mothers who had data on supplement intake.

Table 1. Comparison of Frequency of Genetic Polymorphisms, Demographic Characteristics, and Vitamin Supplement Intake Between Mothers of Controls and Unilateral Cases
 No. Affected/Total No. Informative (%)
CharacteristicControls, n = 97Unilateral Cases, n = 103
  • Abbreviations: DHFR19bpdel/del, homozygous dihydrofolate reductase (DHFR) 19-base-pair deletion of intron 1a; MTHFR677C→T, methylenetetrahydrofolate reductase (MTHFR) cytosine-to-thymine substitution at nucleotide 677.

  • a

    P = .0002; the difference between cases and controls was detected using a chi-square test.

 DHFR19bpdel/dela14/96 (14.58)40/102 (39.22)
 MTHFR677TT25/96 (26.04)30/101 (29.70)
 Smoked during pregnancy9/96 (9.37)3/98 (3.06)
 Folic acid supplement use during first trimester45/97 (46.39)36/72 (50)
 Median age at delivery [range], yr24.4 [14.4-39.5]25.2 [13.17-38.66]
 Median education [range], yr9 [0-17]9 [0-17]
 Child's median weight at delivery [range], kg3.3 [1.4-4.8]3.3 [2.1-5.0]

There was no significant departure from Hardy-Weinberg equilibrium detected in the 96 control mothers for either DHFR (P = .99) or MTHFR (P = .06). The estimated allele frequency of DHFR19bpdel (rs70991108) was 0.61458. Genotypes for DHFR and MTHFR were not correlated in either control mothers (P = .99) or case mothers (P = .42). The proportion of control mothers who were homozygous for DHFR19bpdel was similar to that reported in US cohorts.10, 14

A higher proportion of case mothers were homozygous for DHFR19bpdel (DHFR19bpdel/del) (39.2%) compared with control mothers (14.5%; OR = 3.78; 95%CI = 1.89, 7.55; P = .0002). After controlling for the child's genotype, the OR for the association between DHFR and having a child with retinoblastoma was 2.81 (95% CI, 1.32-5.99; P = .0075). The distribution of the homozygous genotype for MTHFR677C→ did not differ between case mothers and control mothers (P = 0.68).

Because we were interested in folate intake during initial retinal development, we included questions regarding folic acid supplement intake during the first trimester of pregnancy and during the 3 months before pregnancy. Case children were diagnosed at a median age of 24 months; thus, mothers were interviewed approximately 3 years after their pregnancy began. Other work done by Mejia-Rodriguez et al has demonstrated that reliable information on vitamin intake can be assessed from Mexican mothers 4 to 6 years postpartum.15 Only 4 mothers reported using any supplements during the 3 months before their pregnancy. Therefore, we could not evaluate the impact of supplement use during that period. However, among the 169 mothers who were interviewed about use of vitamin supplements during their first trimester, 46.9% (n = 81) took folic acid supplements in the first trimester. The proportion of folic acid supplement use in the first trimester among the mothers of unilateral cases was 50%, and 46.4% among the control mothers (P = .76) (Table 1), which is similar to the proportion (53%) reported for pregnant women surveyed as part of the 1999 Mexican National Nutrition Survey.16 Use of folic acid supplements in the first trimester was unrelated to DHFR19bpdel or maternal age at delivery; however, those who used supplements had more education than those who did not (P = .006) (Table 2).

Table 2. Comparison of Genotypes and Demographic Characteristics of Mothers by Use of Supplements Containing Synthetic Folic Acid in the First Trimester
 No. Affected/Total No. Informative (%)
CharacteristicUsed Folic Acid-Containing Supplements in First Trimester, n = 81Did Not Use Folic Acid-Containing Supplements in the First Trimester, n = 88
  • Abbreviations: DHFR 19bpdel, dihydrofolate reductase (DHFR) 19-base-pair deletion of intron 1a; MHTFR677C→T, methylenetetrahydrofolate reductase (MTHFR) cytosine-to-thymine substitution at nucleotide 677.

  • a

    P = .006; group mean differences were detected using t tests.

 Homozygous for DHFR19bpdel16/79 (20.25)8/88 (20.45)
 Homozygous for MTHFR677C→T25/79 (31.65)19/87 (21.84)
 Had child with unilateral retinoblastoma36/81 (44.44)36/88 (41.91)
 Median age at delivery [range], yr25.6 [14.4-39.5]24.5 [14.9-39.5]
 Median education [range], yra9 [4-17]9 [0-17]

Among those who reported taking folic acid-containing supplements during the first trimester, women with the DHFR19bpdel/del genotype were more likely to have a child with unilateral retinoblastoma (OR = 3.58; 95%CI = 1.11-11.55; P = .03); whereas the association was insignificant among those who did not take folic acid supplements in the first trimester, even after adjusting for the child's genotype (Table 3). However, there was no significant difference in the odds of a DHFR-retinoblastoma risk association between mothers who did or did not take supplements (P = .18). The odds of having a child with retinoblastoma were not associated with homozygosity of MTHFR 677C→T, regardless of folic acid supplement use during pregnancy (Table 3).

Table 3. Odds Ratios and 95% Confidence Intervals for Gene-Disease Association According to First-Trimester Use of Supplements Containing Synthetic Folic Acid
 OR (95% CI) 
GenotypeAll Cases and ControlsFolic Acid Supplement Intake in First Trimester, n = 79No Folic Acid Supplement Intake in First Trimester, n = 88P for the Difference in OR
  • Abbreviations: CI, confidence interval; DHFR19bpdel, dihydrofolate reductase (DHFR) 19-base-pair deletion of intron 1a; MHTFR677C→T, methylenetetrahydrofolate reductase (MTHFR) cytosine-to-thymine substitution at nucleotide 677; OR, odds ratio.

  • a

    Compared with heterozygotes, or homozygous nonvariant allele.

  • b

    P < .01.

  • c

    P = .03.

  • d

    Adjusted for the child's DHFR genotype.

  • e

    P < .06.

Maternal homozygosity forDHFR19bpdela3.78 (1.89-7.55)b3.58 (1.11-11.55)c1.59 (0.56-4.51).31
Maternal homozygosity for DHFR19bpdela,d2.81 (1.32-5.99)b3.31 (0.95-11.50)e0.98 (0.27-3.53).18
Maternal homozygosity for MTHFR677C→Ta1.20 (0.64-2.24)0.77 (0.30-2.03)1.45 (0.52-4.05).38

Mothers of children who were born after 2001 were more likely to take folic acid supplements than those who were born before 2001, which may be secondary to the greater attention given to promoting folic acid intake coinciding with the implementation of folic acid flour fortification. Adjusting the analysis for before and after full implementation did not affect the strength or direction of our results.

In summary, we have observed that maternal homozygosity for DHFR19bpdel is associated with an increased risk of having a child with unilateral retinoblastoma. The DHFR-associated risk is elevated significantly only among those who reported taking folic acid-containing vitamin supplements during pregnancy.


The results of this case-control study suggest that the risk of developing unilateral retinoblastoma is associated with maternal homozygosity for the 19-bp deletion polymorphism in intron 1a of the DHFR gene.14 When stratified by folic acid supplement use during the first trimester of pregnancy, the DHFR-associated risk of retinoblastoma is elevated only in those mothers who take supplements that contain synthetic folic acid.

Folic acid, the synthetic form of naturally occurring folate, is present in fortified foods and vitamin supplements. Folic acid must be reduced to tetrahydrofolate (THF) by DHFR to enter the cellular metabolic pathway. It has been shown in humans that the capacity of DHFR to convert ingested folic acid into reduced folate is limited.18 Ingestion of 200 μg (mcg) of folic acid in a single dose results in the appearance of circulating unmetabolized folic acid (cUMFA).19

The effect of the DHFR19bpdel polymorphism on the enzyme or its activity has not been characterized. Recent studies of the Framingham Offspring Cohort have shown that homozygotes for DHFR19bpdel have normal concentrations of plasma and erythrocyte folate as well as plasma homocysteine, indicating that this polymorphism does not affect the regeneration of THF from dihydrofolate (DHF).14 However, when daily folic acid intake was <250 μg, homozygosity for the DHFR19bpdel was associated with significantly lower erythrocyte folate compared with those who did not have the polymorphism. In addition, among individuals whose daily folic acid intake exceeded 500 μg, the prevalence of high concentrations of cUMFA was 2-fold greater in DHFR19bpdel homozygotes than in those without the polymorphism.14 Together, these Framingham data strongly suggest that homozygosity for the DHFR19bpdel results in a diminished capacity of the body in converting ingested folic acid into reduced folate. This threshold level of intake is well below both the tolerated upper limit and the Dietary Reference Intake (DRI) for pregnant women (600 μg daily).13 Women who are homozygous for DHFR19bpdel have an increased risk for breast cancer, but this risk is increased only among women taking vitamin supplements, suggesting that the effect of the polymorphism on cancer incidence is apparent only when intake surpasses a threshold.10

The possibility that excessive intake of folic acid or the presence of cUMFA is associated with adverse effects has been documented both in humans and in animal models. The presence of cUMFA is related inversely to natural killer (NK) cell cytotoxicity in older women.20 In pregnant mice, extreme folic acid intake (20 times the DRI) has been associated with embryonic delay and growth retardation21 and with increased incidence of NTD.22 Undoubtedly, the increased intake of folic acid through both food fortification and supplements has been beneficial particularly for preventing NTD; however, potential adverse effects, even involving rarer diseases, need to be identified to inform future policies.

Because cUMFA results from the interaction between DHFR19bpdel and high folic acid intake, it is plausible that the increased retinoblastoma risk may be associated with cUMFA. Mothers who are homozygous for DHFR19bpdel and consume higher amounts of folic acid during pregnancy may have resulting high levels of cUMFA during a critical period of retinal development, potentially increasing tumorigenic risk. Increased levels of cUMFA have been associated with decreased NK cell cytotoxicity in older women.20 Maternal NK cells are particularly active during the first trimester of pregnancy, when they contribute to building decidual tissue and forming new vessels as well as to cytolytic immune surveillance.23, 24 Consequently, the developing fetus may be particularly susceptible to factors that decrease NK function.

The risk of retinoblastoma may be associated with less effective use of ingested folic acid during a key period in the formation of these retinal tumors. There is contradictory evidence suggesting that it acts both as a protectant against and as a promoter of cancer. Both insufficient folate intake and increased folic acid intake have been associated with the development of carcinomas in adults and with the promotion of neoplastic lesions and genetic damage in rodent models.25-28 Vitamin supplementation with folic acid has been associated with an increased risk of breast cancer.10, 23 Prospective studies have demonstrated that the ingestion of folic acid at the tolerated upper limit of 1000 μg may contribute to tumor progression in colonic adenomas and to increased incidence of breast,29 prostate,30 and lung31 cancers.

Flour fortification with folic acid is mandatory in approximately 42 countries, although the levels of fortification vary and remain controversial.32 Fortification has led to higher than predicted ingestion of folic acid, in part because consumption patterns for flour-containing foods have changed and are greater than envisioned when fortification was planned.33, 34 In Mexico, since 2001, wheat flour has been fortified with 2000 μg folic acid per kilogram flour.35, 36 Supplements routinely contain high amounts of folic acid, and even fortified foods often contain more than the required amount.33 In the 1999 prefortification Mexican National Nutrition Survey (prefortification), among nonpregnant women of child-bearing age, 12% were consuming a daily dose above the upper limit.13, 16 Data from the 2006 Mexican National Nutrition Survey demonstrated that poor Mexican women consumed excessive levels of carbohydrates.37 Some of these carbohydrates probably originate from foods that contain fortified flour. Together, these data suggest that consumption at levels present in the diet of some Mexican women plausibly may be above the levels associated with increased cancer risk.29-31 Mothers of children with unilateral retinoblastoma may be consuming elevated quantities of folic acid, potentially through dietary consumption of foods made with folic acid-fortified flour.

For our earlier study in Mexico, we did not have access to parental biospecimens.8 Both that study and the study by Bunin et al in the United states were performed before flour fortification with folic acid.4 All of the data suggesting geographic variations in the incidence of retinoblastoma were compiled before the widespread adoption of fortification.1, 2 Since fortification has begun, no new worldwide incidence data on retinoblastoma have been published, although efforts currently are underway to collect such data, and estimates are expected to be available in 2013.38

An inherent limitation of our study is that it is a retrospective study, a necessity because of the rarity of retinoblastoma. Because of the case-control design, there may be recall bias in the maternal recollection of prenatal vitamin intake; however, this recall bias would not be expected to vary with genotype. Table 1 indicates that our case and control mothers did not differ in the proportion that consumed vitamins containing synthetic folic acid. One additional limitation is that we are unable to fully account for dietary intake, because the nutrient database currently available in Mexico was updated in 2003 and reflects fortification of some industrialized foods, but it does not reflect the synthetic folic acid content of nonindustrialized products containing fortified wheat flour. Because these nonindustrialized foods are consumed frequently by our population, accurate calculation of folic acid intake is not possible without accounting for the content of these foods. Consequently, we are unable to fully examine a threshold effect of synthetic folic acid intake with DHFR. We limited our examination of DHFR to the 19bpdel polymorphism in intron 1a because of its functional significance in the Framingham Offspring Cohort,14 its vitamin-specific carcinogenic effect in breast cancer,10 and its association with NTDs.12 However, if there is linkage disequilibrium between DHFR19bpdel in intron 1a and other DHFR polymorphisms, then this may account in part for our findings. We were unable to fully examine the role of the child DHFR genotype or interactions between child and maternal DHFR genotypes because of our current sample size. Future plans to validate our findings include an expansion of our study population and a determination of dietary folic acid content in nonindustrialized fortified foods, as well as collaboration with other groups to examine our hypotheses in other populations and expansion to include additional MTHFR and DHFR polymorphisms.

In summary, to our knowledge, the current study is the first to suggest that maternal metabolism of folic acid may affect the risk of a child developing retinoblastoma and that maternal genotype for DHFR predicts risk for unilateral retinoblastoma. Our data are consistent with the hypotheses arising from the US and European trials demonstrating increased incidence of cancer among adults who take folic acid supplements.


We gratefully acknowledge invaluable technical assistance from A Hernández, J Romero Rendón, M. F. Lazcano, S. Cardoso Muñoz, and C. Cujar. We thank Drs. Lynette Neufeld, Eduardo Lazcano-Ponce, Nancy Mueller, Fernando López Casillas, Stanislaw Sadowinski, Fernando Cerecedo, Joseph Graziano, Regina Santella, and Javier Torres for critical collaboration and advising.


This work was supported by grants from the National Institutes of Health (R01 CA098180 and P30 ES009089).


The authors made no disclosures.