Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Am J Clin Nutr. Author manuscript; available in PMC 2008 November 25.
Published in final edited form as:
PMCID: PMC2587282

Lymphocyte gene expression in subjects fed a low-choline diet differs between those who develop organ dysfunction and those who do not2



Some humans fed a low-choline diet develop hepatosteatosis, liver and muscle damage, and lymphocyte apoptosis. The risk of developing such organ dysfunction is increased by the presence of single-nucleotide polymorphisms (SNPs) in genes involved in folate and choline metabolism.


We investigated whether these changes that occur in the expression of many genes when humans are fed a low-choline diet differ between subjects who develop organ dysfunction and those who do not. We also investigated whether expression changes were dependent on the presence of the SNPs of interest.


Thirty-three subjects aged 20−67 y were fed for 10 d a baseline diet containing the recommended adequate intake of choline. They then were fed a low-choline diet for up to 42 d or until they developed organ dysfunction. Blood was collected at the end of each phase, and peripheral lymphocytes were isolated and used for genotyping and for gene expression profiling with the use of microarray hybridization.


Feeding a low-choline diet changed the expression of 259 genes, and the profiles of subjects who developed and those who did not develop signs of organ dysfunction differed. Group clustering and gene ontology analyses found that the diet-induced changes in gene expression profiles were significantly influenced by the SNPs of interest and that the gene expression phenotype of the variant gene carriers differed significantly even with the baseline diet.


These findings support our hypothesis that a person's susceptibility to organ dysfunction when fed a low-choline diet is modulated by specific SNPs in genes involved in folate and choline metabolism.

Keywords: Diet, choline, gene expression, lymphocytes, single-nucleotide polymorphism, methylenetetrahydrofolate dehydrogenase, MTHFD, phosphatidylethanolamine methyltransferase, PEMT, choline dehydrogenase, CHDH, choline deficiency


Choline is an essential dietary nutrient involved in a multitude of metabolic roles (1). It is a major source of methyl groups for methionine synthesis and is needed for the structural integrity of cell membranes, the transport of lipids from the liver, and cholinergic neurotransmission (1). Adequate intake (AI) recommendations for choline have been established (2), and, in population studies, diets low in choline were associated with a greater risk of birth defects (3, 4) and with high homocysteine concentrations in blood (5). We previously reported that some humans fed a diet low in choline developed hepatosteatosis, experienced liver and muscle damage, and had greater lymphocyte apoptosis (610). All of these changes were reversed after a period of choline repletion. The susceptibility to these outcomes was not the same in all subjects; it varied with the presence of single-nucleotide polymorphisms (SNPs) within the genes involved in folate and choline metabolism (7, 8). With the notable exception of the carriers of the choline dehydrogenase (CHDH) variant allele (318 A→C; rs9001), who were protected against developing organ dysfunction (8), carriers of the methylenetetrahydrofolate dehydrogenase (MTHFD1) synthase variant (1958 G→A; rs2236225), the CHDH variant (432 G→T; rs12676), and the phosphatidylethanolamine methyltransferase (PEMT) variant (−744 G→C; rs12325817) alleles were at greater risk of developing organ dysfunction when fed a low-choline diet than were the carriers of the corresponding wild-type alleles (7, 8).

In this study, we fed humans a diet containing recommended amounts of choline, then fed the same subjects diets low in choline content, and then assessed gene expression in lymphocytes at the end of each feeding period. We determined whether ingesting a low-choline diet was associated with changes in gene expression, whether those subjects who developed organ dysfunction differed in gene expression from those who did not, and whether changes in gene expression were related to the presence of SNPs.



Healthy men (n = 31) and women (n = 35) were recruited for the study. Inclusion was contingent on an age-typical good state of health as determined by physical examination and standard clinical laboratory tests, such as complete blood count, blood chemistries, and fasting lipids and liver function tests and on the absence of known chronic diseases. Twenty-two subjects admitted to the study had minor elevations in blood lipids that were not deemed of clinical significance by the study physician. Of the originally recruited 66 subjects, 61 completed at least the initial and depletion phases. Of those 61 subjects, 1 was excluded because of a 9-kg weight loss during the study, 3 were excluded because they did not comply with diet restrictions, and 6 were excluded because they did not have baseline measurements; thus, 51 subjects were included in analyses. Of those subjects, 18 were excluded from the gene array analyses (see Microarray data analysis). The remaining 14 men and 19 women ranged in age from 20 to 67 y and had a body mass index (BMI; in kg/m2) between 19 and 31, which they maintained throughout the study. The ethnic distribution of these participants was 61% white, 30 African American, and 9% Asian, which reflects the local population characteristics of the Raleigh-Durham-Chapel Hill area.

Written informed consent was obtained from all participants. The study protocol was approved by the Institutional Review Board at the University of North Carolina at Chapel Hill (UNC-CH).

Study design

The participants were admitted to the UNC-CH General Clinical Research Center, where they remained under the supervision of study staff for the duration of the study. Research diets administered to the subjects, which were composed of 0.8 g high-biologic-value protein/kg body wt, with 30% of energy from fat and 70% of energy from carbohydrate, were prepared in-house to protocol specifications and have been described in detail elsewhere (11). Total food intake was adjusted to be isocaloric and to provide adequate intakes of macronutrients and micronutrients. Initially, all participants received a diet of commonly eaten foods containing 550 mg choline · 70 kg body wt−1 · d−1—the presumed AI (2)—and 400 dietary folate equivalents (DFE)/d. The dietary choline content was confirmed as described previously (11), and the folate content was calculated by using the US Department of Agriculture SR16 database and PRONUTRA software (version; Viocare, Princeton, NJ). After 10 d of this baseline diet, liver fat was measured, and 48 mL blood was collected by venipuncture and processed for peripheral lymphocytes as described below. The subjects were randomly assigned to 2 groups—diet folate only or diet folate supplemented with 400 μg folic acid/d—and then were fed a diet in which the choline content was reduced to <50 mg/d, as confirmed by analysis of duplicate food portions. For the rest of the study, all diets offered to the diet-folate-only group contained 100 DFE/d, whereas the folic acid–supplemented group received an additional 668 DFE/d. Periodic measurements of urinary choline and betaine concentrations (12) were used to confirm compliance with the dietary restrictions. Subjects followed this depletion diet until they developed organ dysfunction associated with choline deficiency or for 42 d if they did not develop organ dysfunction. Blood (48 mL) was again collected at the end of the depletion phase. Humans were deemed to have organ dysfunction associated with choline deficiency if they had a >5-fold increase in serum creatine phosphokinase (CPK) activity (measurements were taken every 3−4 d) or if they had a >28% increase in liver fat content while following the choline-depletion diet (measurements were taken on days 21 and 42 of the choline-depletion diet) and if this increase in CPK activity or liver fat content was resolved when choline was returned to the diet. The change from baseline in liver fat content was estimated by using magnetic resonance imaging (MRI) in a clinical magnetic resonance system (Vision 41.5T; Siemens Medical Solutions, Malvern, PA) with a modified “In and Out of Phase” procedure that was described previously (13). Fat content was derived from measurements across 3−5 liver slices per subject and standardized to similarly measured slices of spleen.

Isolation of lymphocyte RNA and quality assessment

At the end of the baseline and choline-depletion diet phases, peripheral lymphocytes were isolated from blood within 2 h of collection by using Ficoll-Hypaque gradient in evacuated tubes with sodium citrate (Vacuatainer CPT tubes; Becton Dickinson, Franklin Lakes, NJ), and suspended in TRIzol Reagent (Invitrogen, Carlsbad, CA). Total RNA was isolated according to the manufacturer's protocol, purified with the use of an RNeasy kit (Qiagen, Valencia, CA), and diluted to a standard concentration of 100 μg/mL. The subsequent procedures were performed at the UNC-CH Genomics Core Facility. Each sample (0.5 μg), including the reference RNA, was first tested on an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA) to ensure high-quality RNA before being diluted for hybridization.


Genomic DNA was prepared from peripheral blood with a commercial extraction kit (PureGene; Gentra Systems, Minneapolis, MN) according to the manufacturer's instructions and diluted to a standard concentration of 1 μg/mL. The cytoplasmic MTHFD synthase (MTHFD1-G1958A) was amplified by multiplex polymerase chain reaction (PCR), purified, and then analyzed with matrix-assisted laser desorption ionization time-of-flight mass spectrometry (7, 14). For the PEMT and CHDH genes, DNA sequencing was performed on double-stranded DNA templates obtained from genomic DNA by using PCR amplification (8).

Microarray hybridization and data retrieval

We mixed 12 μg of sample and an equal amount of universal human reference RNA (Stratagene, La Jolla, CA) with anchored oligo-dT Primer mixture, and amplified the mixture by using a thermal cycler (70 °C for 5 min to anneal; 42 °C for 1.5 h). The cDNA was then labeled in a reverse transcriptase reaction with dUTP CyDye–labeled nucleotides (Cy3 for reference and Cy5 for experiment) by using a CyScribe First-Strand cDNA Labeling Kit [GE Healthcare (formerly Amersham Biosciences), Piscataway, NJ]. The labeled cDNA was then purified by degrading the mRNA with NaOH at 37 °C for 15 min and washing the samples with Tris-EDTA buffer by using a Microcon PCR filter (Millipore, Bedford, MA). The combined Cy3+Cy5 sample was then applied to a glass slide array, sealed in a hybridization chamber, and incubated in a 65 °C water bath overnight. The human oligo-arrays with 16 000 spots were produced at the Genomics Core Facility in the same batch by using 60-mer oligonucleotides (Compugen USA, Jamesburg, NJ). After hybridization, the array slides were washed and spun dry before being scanned to collect the fluorescent images (GenePix 4000B fluorescent scanner; Axon Instruments, Union City, CA). Images were gridded, and data were collected with the use of GENEPIX PRO microarray acquisition and analysis software (version 5.0; Axon Instruments). Detailed protocols are available at Images obtained were analyzed by superimposing a grid for each array with the use of the GENEPIX PRO software. All spots of poor quality (as determined by visual inspection) were flagged as bad and removed from further analysis.

Microarray data analysis

All of the collected raw data files were further processed by uploading them into the UNC Microarray Database (, and the data were filtered and retrieved according to the following criteria: 1) data were retrieved by the immutable Stanford University Identification (SUID) reference number to average the replicate spots by gene name and present the result as one (ie, to collapse them); 2) spots were selected only if they had both Channel 1 and Channel 2 Lowess-normalized means ≥30% above background; and 3) genes were selected only if they had >70% good data. No cutoffs were selected. Final data were expressed as log(base 2) of lymphocyte RNA-reference RNA Lowess normalized ratio (mean). The final number of arrays (and hence the number of subjects included in further analyses) was influenced by various criteria: not all RNA samples were of good quality, some blood samples did not contain enough RNA, and not all arrays passed the quality test. Therefore, of the 51 subjects initially included in the study, only 33 had arrays that were suitable for analysis. Of these 33, not all had 2 arrays (1 for baseline and 1 for depletion), so the number of arrays available at baseline was 30, and the number of arrays available at depletion was 25 (Table 1).

Table 1
Groups for analyses

Statistical analysis of gene expression

Subjects were grouped for 12 comparisons (Table 1) according to the following criteria: presence or absence of organ dysfunction when fed a low-choline diet (hepatosteatosis and liver and muscle damage); timepoint (end of the baseline phase or end of the depletion phase); and the presence of a specific SNP for PEMT, MTHFD, and CHDH. Significance analysis of microar-rays (SAM) (15) was applied to the final data by using TIGR MEV software [version 3; Dana-Farber Cancer Institute, Boston, MA (16)]. We used the one-class response type and 100 permutations or the maximum allowed permutations (if <100) to select the genes that are significantly changed across each of the groups. An arbitrary false discovery rate (FDR) of maximum 5% was chosen, and the closest threshold value (delta, or Δ) was selected for each group. The FDR is the expected percentage of false predictions; therefore, at 5% FDR, 95% of the observations are reproducible and not due to chance. The software generated a list of significantly overexpressed and underexpressed genes (d score assessment), and the q value (the lowest FDR at which the gene is called significant) was computed for each gene. The data generated by the use of SAM were converted accordingly for subsequent gene ontology (GO) classification: −1 for underexpression, 0 for no change, and 1 for overexpression. The defined groups (Table 1) were clustered according to their changes in gene expression (16). Each group was considered as an experiment, and gene changes were expressed as an average of all the arrays within the group.

Cluster analysis

The TIGR MEV software was used for cluster analysis, in which groups defined in Table 1 were clustered according to changes in gene expression (post-SAM analysis). Hierarchical clustering was assessed by using the average Euclidean distance between groups.

Gene ontology classification

GOMINER software (version 1.22; Georgia Tech University, Atlanta, GA; Internet: was used to construct a GO list of the significantly changed genes (17), by using gene symbols as identifiers. Data from the SAM output file were converted to text files and GOMINER generated a list of genes classified by their various GO classes, according to the default database (com.mysql.jdbc.Driver at jdbc:mysql:// Fisher's exact tests were performed to determine the significance of changes within the total number of genes in each GO class, and significance was separately assessed for the number of genes that were overexpressed, underexpressed, or both (P < 0.05).

Gene expression validation

A small subset of genes was selected to determine the validity of the array-generated data. Real time reverse transcriptase (RT)–PCR was used to assess the expression of 4 of the most overexpressed genes and 4 of the most underexpressed genes. Primers for these 8 genes—FOXA1, PRAME, TERT, CDCA8, CHEK1, IL2RB, TNFAIP3, and NFKBIA—were purchased from SuperArray (Frederick, MD) as was 18S rRNA, which was used as the normalization gene. Equal amounts of RNA were pooled from all subjects at baseline and depletion, respectively. For all genes but one, 100 ng pooled template RNA was used in quintuplicate reactions (one-step RT-PCR) in a QuantiTect SYBR Green RT-PCR kit (Qiagen); for 18S RNA, 10 pg template RNA was used. All reactions were performed on an iQ5 Real-Time PCR Detection System (Bio-Rad Laboratories, Hercules, CA) under the following cycling conditions: reverse transcription for 30 min at 50 °C; initial activation for 15 min at 95 °C; and 45 cycles of denaturation for 15 s at 94 °C, annealing for 30 s at 55 °C, and extension for 30 s at 72 °C, which were followed by data acquisition for the melting curves. The differences and statistical assessment in gene expression were computed by using the comparative threshold (ΔΔCT) method with the REST384 add-in for EXCEL (Internet: (18), and gene expression changes were expressed as 18S-normalized ratios between depletion and baseline. Statistical significance of change was assessed by using both the Student t test and the nonparametric pairwise fixed reallocation randomization test provided by REST (18).

Regulation of gene expression by DNA methylation

We used the DNA Methylation Database (Internet:; last accessed 12/05/2006) to compare genes found to be changed by dietary choline with those reported to have their expression regulated by DNA methylation (19).


Changes in gene expression between baseline and depletion

Of the 12 034 unigenes present within the arrays, 152 were underexpressed, and 107 were overexpressed in subjects after the low-choline diet phase of the study. A list of selected genes (the 20 most overexpressed and the 20 most underexpressed genes) is shown in Table 2. When subjects were further subclassified according to the absence or presence of signs of organ dysfunction (fatty liver, increased plasma CPK, or both) after being fed a low-choline diet, the gene expression profiles of the 2 groups could be differentiated (Table 2). From the ALL SUBJECTS group, we chose a limited number of genes to validate the array data by real-time RT-PCR (Table 3). With the exception of CHEK1, changes in gene expression, given as the relative ratio between depletion and baseline, were confirmed as significant by one or both of the statistical tests (t test and the pairwise fixed reallocation randomization test).

Table 2
Selected genes with expression changes between baseline and depletion1
Table 3
Real-time reverse transcriptase polymerase chain reaction (RT-PCR) validation of the gene array data1

Hierarchical clustering between defined groups

Using the data obtained from SAM analysis, we clustered the groups defined in Table 1 to determine whether specific changes in gene expression profiles were associated with the presence of signs of organ dysfunction when subjects were following a low-choline diet, with the presence of PEMT, CHDH, and MTHFD1 SNPs, or both. In addition, we investigated whether gene expression profiles at baseline were different between these groups (which would suggest a different phenotype at baseline). The tree node of the groups, each one being considered an experiment, is shown in Figure 1. According to the node distribution, 2 main branches were generated for all the experiments. One branch included the PEMT and MTHFD1 groups surrounding the group with all subjects at the end of the depletion phase [ALL(D)]. The other main branch includes both CHDH polymorphisms, and the groups based on the presence (SIGNS) or the absence (NO SIGNS) of signs of organ dysfunction, all centered around the main ALL SUBJECTS group. The ALL SUBJECTS, SIGNS, and NO SIGNS groups were used to compare gene expression at the end of the depletion phase and the end of the baseline phase. ALL(D) compared the SIGNS group with the NO SIGNS group at the end of the depletion period. Four clusters were identified with the smallest distance: 1) subjects with the PEMT SNP fed a low-choline diet [PEMT(D)] associated with women with the PEMT SNP fed a low-choline diet [PEMT(W)]; 2) ALL(D) associated with subjects with the MTHFD1 SNP fed a low-choline diet [MTHFD1(D)]; 3) ALL SUBJECTS associated with subjects who developed organ dysfunction when fed a low-choline diet (SIGNS); and 4) one of the CHDH SNPs [CHDH(G→T)] associated with the FOLATE group.

Cluster analysis. All comparison groups defined in Table 1 were subjected to cluster analysis (see Materials and Methods) to determine the distance of change between groups, as defined by the node structure indicated. ALL SUBJECTS, comparison of all subjects ...

Changes in gene expression across all groups

When the SAM analysis was performed for all groups defined in Table 1, changes for 1054 unigenes were significant. A complete list of these genes is available at On the basis of our GO analysis, we present data for a limited number of genes (Figure 2) that are included in the following GO classes of special interest for our human study: folate-thymidine metabolism (GO 19860, 42083, and 46104), apoptosis (GO 6915, 6916, 6917, 8632, 30693, 43027, 43065, and 43154), DNA damage-repair (GO 77, 724, 725, 31570, 42770, and 51908), cell cycle regulation-proliferation (GO 4861 and 8283), immune-inflammatory response and lymphocyte differentiation-activation (GO 1772, 6952, 6954, 6955, 19731, 19732, 19864, 30098, 42113, 42742, and 50832), epigenetic regulation (GO 8327, 35035, and 45815), and telomere maintenance (GO 3720, 3721, 5697, and 42162).

Gene expression profiles across all groups. The computed change in expression is shown for each gene included in gene ontology (GO) classes pertaining to folate-thymidine metabolism, apoptosis, DNA damage-repair, cell cycle regulation-proliferation, immune-inflammatory ...

Gene ontology analysis

GO analysis was performed for all genes deemed significantly changed within any of the groups defined in Table 1. Fisher's exact test indicated either unidirectional (up or down) or bidirectional (both up and down) changes for 1216 GO classes (a complete list of changes within GO classes is available at Changes across selected GO classes listed above are shown in Table 4, where the vast majority of changes are in one direction, which suggests that, within each GO class, changes in gene expression are functionally convergent.

Table 4
Gene expression changes in selected gene ontology (GO) groups1

Genes regulated by DNA methylation

Previous reports in cell culture and animal model systems show that dietary choline modulates gene expression by altering DNA methylation (20, 21), but no data are yet available in humans fed a low choline diet. Using the only available public database on DNA methylation (see Subjects and Methods), we identified 13 of the 1024 genes changed in our study that were reported to be regulated by DNA methylation in various studies using human tissues or human cells in culture (Table 5).

Table 5
Genes of which the expression is changed by a low-choline diet and that are known to be regulated by DNA methylation1


We found that a low-choline diet induced changes in lymphocyte gene expression in humans. These changes included (but were not limited to) changes in expression of genes functionally involved in folate metabolism, apoptosis, DNA damage-repair, cell cycle regulation, immune response, epigenetic regulation, and telomere maintenance, which suggested that dietary choline deficiency can alter the functionality of many pathways. Subjects who developed organ dysfunction while following a low-choline diet differed from those who did not develop organ dysfunction in their expression of many genes, including some of those related to any or all of apoptosis, the DNA integrity checkpoint, and genes of cell cycle regulation. Subjects with the PEMT (rs12325817) and MTHFD1 (rs2236225) SNPs, previously shown to predispose a person to developing organ dysfunction when fed a low-choline diet (7, 8), differed at baseline from those subjects without the SNP in their expression of apoptosis, the DNA damage checkpoint, and cell proliferation control genes, which suggests that they are phenotypically different even before a low-choline diet is administered.

In different comparison groups, choline deficiency induced different patterns of change (Figure 1). Moreover, many of the results reported within all subjects (baseline compared with depletion, ALL SUBJECTS) could be misleading because these changes were not homogenous when subjects were classified on the basis of signs of organ dysfunction (Table 2 and Figure 2). For instance, 3 of the genes with the largest change in expression (CHEK1, GBE1, and KIF20A) were differently expressed in the NO SIGNS group than in the SIGNS group (Table 2). CHEK1 is involved in DNA repair in human T lymphocytes (22). GBE1 is required for sufficient glycogen accumulation and is normally underexpressed in whole human blood (23). KIF20A regulates the transport of Golgi membranes and associated vesicles along microtubules (24).

By clustering the groups according to genotype (Figure 1), we found that different patterns in gene expression do indeed support this classification. For example, the previously reported (8) protective CHDH (318 A→C) genotype grouped close to the NO SIGNS group—those who had no clinical symptoms while following the low-choline diet—whereas the CHDH (432 G→T) genotype, which was reported to increase susceptibility to choline deficiency (8), grouped with the SIGNS group. Moreover, the NO SIGNS and SIGNS groups were intercalated by the ALL SUBJECTS group, which supports the heterogeneity of the previously reported responses to dietary choline deficiency. Included in the other arm from the clustering analysis were the PEMT and MTHFD1 genotypes, which grouped around the ALL(D) group. This cluster supported our hypothesis that genetic differences account for the presence or absence of organ dysfunction in humans depleted of choline (7, 8). This analysis also found that the presence of the PEMT and MTHFD1 genotypes could confer differences in the phenotypes at baseline, which suggests that different persons may have different susceptibility to dietary choline deficiency and that the risk of choline deficiency is greater in women who are carriers of the PEMT allele [this group was closer to the ALL(D) group than to the PEMT(D) group]. Harder to interpret is the unexpected clustering of the FOLATE group (those receiving folate supplements versus those who did not) close to the CHDH (432 G→T) genotype.

To construct a better picture of the potential functional significance of these gene expression patterns, we used GO analysis to group genes according to their functional roles. The heterogeneity of the response to a low choline diet was also shown at this level of analysis. Genes involved in folate metabolism were most affected in the carriers of the MTHFD1 and CHDH (432 G→T) alleles (Table 4). A different pattern was observed for genes involved in apoptosis (Table 4), in which the MTHFD1 polymorphic allele carriers were the most affected (increases in apoptosis, induction of apoptosis, caspase activity, positive regulation of apoptosis, and decreased expression of genes involved in negative regulation of caspase activity). This result is consistent with our previous data, which showed that humans who developed organ dysfunction when fed a choline-deficient diet were more likely to have the MTHFD1 variant allele (7) and to have increased lymphocyte DNA damage and apoptosis, as measured by caspase activity, than were those who did not develop organ dysfunction (9). It is interesting that the same trends were present in these subjects at baseline, which suggests that their higher susceptibility to choline deficiency may be due to effects present even when subjects are consuming a normal diet. For carriers of the PEMT polymorphic allele, the most affected genes were grouped within GO classes involved in telomere maintenance; notably, these alterations were more extensive in female PEMT allele carriers. Telomeres are the protein-DNA structures that protect chromosome ends from being recognized as double-stranded DNA breaks, and their maintenance is important for cell longevity, normal cell cycling, and prevention of cancer (25).

Our findings suggest that dietary choline deficiency may affect the homeostatic mechanisms responsible for telomere maintenance, perhaps by epigenetic changes in gene expression. The folate status of subjects had little effect on the gene expression changes seen in GO analyses.

Some genes that are regulated by DNA methylation were also identified as being changed by a low-choline diet (Table 5). In cultured human neuroblastoma cells and in rodent models, cho-line deficiency alters both global and gene-specific DNA methylation and the expression of these genes (20, 21). Therefore, we suggest that the observed choline deficiency–induced changes in gene expression occurred because of altered methylation in promoter regions of the genes involved. Among the genes changed in choline deficiency that are known to be regulated by gene methylation, the insulin-like growth factor 2 (IGF2) is an imprinted gene; loss of imprinting is associated with cancer in various experimental models (as reviewed in reference 26). Another gene identified as being changed by dietary choline and known to be regulated by methylation is telomerase reverse transcriptase (TERT), the product of which is the protein component of a ribonucleoprotein polymerase that maintains telomere ends by addition of the telomere repeat TTAGGG, and its deregulation is involved in both cellular senescence (telomere shortening) and carcinogenesis in leukemic cells (27).

In conclusion, dietary choline deficiency induced changes in gene expression profiles in human lymphocytes, and these patterns correlated with the occurrence of organ dysfunction and apoptosis in humans fed a low-choline diet. These outcomes also correlated with polymorphisms in genes that regulate folate and choline metabolism. Further studies are required to determine whether these changes are regulated by epigenetic mechanisms and to identify other populations at risk for dietary choline deficiency.


The authors' responsibilities were as follows—SHZ (principal investigator): designed the human study, oversaw the conduct of the study, and participated in data interpretation and manuscript preparation; LMF: supervised the recruitment of subjects and the conduct of the human diet study; MDN: performed statistical analyses and participated in data interpretation and reverse transcriptase–polymerase chain reaction validation; KD (study coordinator): responsible for sample collection and processing, RNA extraction, quality assurance, and microarray hybridization; and all authors: participated in the writing of the manuscript. SHZ serves on advisory boards for the DuPont Company, the Solae Company, and Hershey Foods; LMF is a consultant with Hannaford Brothers Company. None of the authors had a personal or financial conflict of interest.


2Supported by grant no. DK55865 from the National Institutes of Health (to SHZ) and by grants from the National Institutes of Health to the UNC Clinical Nutrition Research Unit (DK56350), the UNC General Clinical Research Center (RR00046), the Center for Gastrointestinal Biology and Disease (DK34987), and the Center for Environmental Health and Susceptibility (ES10126). The Solae Company donated the lecithin used to formulate the diets.


1. Zeisel SH. Choline: critical role during fetal development and dietary requirements in adults. Annu Rev Nutr. 2006;26:229–50. [PMC free article] [PubMed]
2. Institute of Medicine, National Academy of Sciences . Dietary reference intakes for folate, thiamin, riboflavin, niacin, vitamin B12, panthothenic acid, biotin, and choline. National Academy Press; Washington, DC: 1998. Choline. pp. 390–422.
3. Shaw GM, Carmichael SL, Yang W, Selvin S, Schaffer DM. Periconceptional dietary intake of choline and betaine and neural tube defects in offspring. Am J Epidemiol. 2004;160:102–9. [PubMed]
4. Shaw GM, Carmichael SL, Laurent C, Rasmussen SA. Maternal nutrient intakes and risk of orofacial clefts. Epidemiology. 2006;17:285–91. [PubMed]
5. Cho E, Zeisel SH, Jacques P, et al. Dietary choline and betaine assessed by food-frequency questionnaire in relation to plasma total homocysteine concentration in the Framingham Offspring Study. Am J Clin Nutr. 2006;83:905–11. [PMC free article] [PubMed]
6. da Costa KA, Gaffney CE, Fischer LM, Zeisel SH. Choline deficiency in mice and humans is associated with increased plasma homocysteine concentration after a methionine load. Am J Clin Nutr. 2005;81:440–4. [PMC free article] [PubMed]
7. Kohlmeier M, da Costa KA, Fischer LM, Zeisel SH. Genetic variation of folate-mediated one-carbon transfer pathway predicts susceptibility to choline deficiency in humans. Proc Natl Acad Sci U S A. 2005;102:16025–30. [PubMed]
8. da Costa KA, Kozyreva OG, Song J, Galanko JA, Fischer LM, Zeisel SH. Common genetic polymorphisms affect the human requirement for the nutrient choline. FASEB J. 2006;20:1336–44. [PMC free article] [PubMed]
9. da Costa KA, Niculescu MD, Craciunescu CN, Fischer LM, Zeisel SH. Choline deficiency increases lymphocyte apoptosis and DNA damage in humans. Am J Clin Nutr. 2006;84:88–94. [PMC free article] [PubMed]
10. Fischer LM, daCosta KA, Kwock L, et al. Sex and menopausal status influence human dietary requirements for the nutrient choline. Am J Clin Nutr. 2007;85:1275–85. [PMC free article] [PubMed]
11. Busby MG, Fischer L, Da Costa KA, Thompson D, Mar MH, Zeisel SH. Choline- and betaine-defined diets for use in clinical research and for the management of trimethylaminuria. J Am Diet Assoc. 2004;104:1836–45. [PubMed]
12. Koc H, Mar MH, Ranasinghe A, Swenberg JA, Zeisel SH. Quantitation of choline and its metabolites in tissues and foods by liquid chromatography/electrospray ionization-isotope dilution mass spectrometry. Anal Chem. 2002;74:4734–40. [PubMed]
13. Fishbein MH, Gardner KG, Potter CJ, Schmalbrock P, Smith MA. Introduction of fast MR imaging in the assessment of hepatic steatosis. Magn Reson Imaging. 1997;15:287–93. [PubMed]
14. Meyer K, Fredriksen A, Ueland PM. High-level multiplex genotyping of polymorphisms involved in folate or homocysteine metabolism by matrix-assisted laser desorption/ionization mass spectrometry. Clin Chem. 2004;50:391–402. [PubMed]
15. Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A. 2001;98:5116–21. [PubMed]
16. Saeed AI, Sharov V, White J, et al. TM4: a free, open-source system for microarray data management and analysis. Biotechniques. 2003;34:374–8. [PubMed]
17. Zeeberg BR, Feng W, Wang G, et al. GoMiner: a resource for biological interpretation of genomic and proteomic data. Genome Biol. 2003;4:R28. [PMC free article] [PubMed]
18. Pfaffl MW, Horgan GW, Dempfle L. Relative expression software tool (REST) for group-wise comparison and statistical analysis of relative expression results in real-time PCR. Nucleic Acid Res. 2002;30:e36. [PMC free article] [PubMed]
19. Amoreira C, Hindermann W, Grunau C. An improved version of the DNA Methylation database (MethDB). Nucleic Acid Res. 2003;31:75–7. [PMC free article] [PubMed]
20. Niculescu MD, Craciunescu CN, Zeisel SH. Dietary choline deficiency alters global and gene-specific DNA methylation in the developing hippocampus of mouse fetal brains. FASEB J. 2006;20:43–9. [PMC free article] [PubMed]
21. Niculescu MD, Yamamuro Y, Zeisel SH. Choline availability modulates human neuroblastoma cell proliferation and alters the methylation of the promoter region of the cyclin-dependent kinase inhibitor 3 gene. J Neurochem. 2004;89:1252–9. [PMC free article] [PubMed]
22. Giacomelli L, Nicolini C. Gene expression of human T lymphocytes cell cycle: experimental and bioinformatic analysis. J Cell Biochem. 2006;99:1326–33. [PubMed]
23. Su AI, Cooke MP, Ching KA, et al. Large-scale analysis of the human and mouse transcriptomes. Proc Natl Acad Sci U S A. 2002;99:4465–70. [PubMed]
24. Echard A, Jollivet F, Martinez O, et al. Interaction of a Golgi-associated kinesin-like protein with Rab6. Science. 1998;279:580–5. [PubMed]
25. Aviv A, Valdes AM, Spector TD. Human telomere biology: pitfalls of moving from the laboratory to epidemiology. Int J Epidemiol. 2006;35:1424–9. [PubMed]
26. Feinberg AP. A genetic approach to cancer epigenetics. Cold Spring Harb Symp Quant Biol. 2005;70:335–41. [PubMed]
27. Nowak T, Januszkiewicz D, Zawada M, et al. Amplification of hTERT and hTERC genes in leukemic cells with high expression and activity of telomerase. Oncol Rep. 2006;16:301–5. [PubMed]