Genome-wide pathway analysis of folate-responsive genes to unravel the pathogenesis of orofacial clefting in man^†

INTRODUCTION

Clefting of the lip with or without the palate (CLP) is a common congenital malformation that occurs in approximately 14.2 per 10,000 live births in the Netherlands (EUROCAT-Northern-Netherlands, 2006). The etiology of CLP is largely unknown, but is considered multifactorial in origin. As recently reviewed by Krapels et al. (2006) associations between CLP and developmental genes, such as TGFα and MSX1, and linkage disequilibrium of various chromosomal regions, such as 3p21.2, 10p13, and 16p13.3 and CLP emphasize the involvement of genetic components. Of great interest are the findings of the last two decades that environmental factors play a role in CLP etiology as well. Maternal periconception use of folic acid supplements and folate-rich food (Czeizel et al., 1999; van Rooij et al., 2004), medication (van Rooij et al., 2002), and smoking (Shaw et al., 1996; van Rooij et al., 2001) are known to modulate the risk of having a CLP child. We have shown that periconception supplementation with folic acid reduces the CLP birth prevalence rate by approximately 50% (van Rooij et al., 2004), thereby supporting the recommendation of folic acid supplementation in the periconception period, not only to prevent neural tube defects (NTDs) but CLP as well. Other countries have started folic acid fortification programs with beneficial effects on the reduction of NTDs (Mason et al., 2007; Quinlivan and Gregory, 2003). Such programs imply long-term exposure of the total population to folate supplements. It is therefore remarkable that studies investigating the effects of synthetic folic acid and natural folate on biological processes are very scarce (Mason et al., 2007).

After reduction of folic acid to natural folates, folate derivatives serve intracellularly as a one-carbon group donor for the synthesis of purines, pyrimidines, and proteins and the remethylation of homocysteine into methionine. The methionine metabolite, S-adenosylmethionine, is the main methyl donor of the cell and methylates DNA, RNA, proteins, and lipids (Chiang et al., 1996). Folate deficiency induces elevated homocysteine concentrations, uracil accumulation and misincorporation, DNA strand breaks, abnormal DNA and protein methylation patterns, and increased apoptosis (Courtemanche et al., 2004; Duthie and Hawdon, 1998; Kim, 2005; Steegers-Theunissen et al., 2000; Wasson et al., 2006). However, these biological mechanisms cannot explain the protective effects of additional intake of folic acid and food folate on CLP development.

Despite the limited knowledge of the effects of folate on molecular and biological pathways, mothers-to-be and the intrauterine-developing embryo and fetus are supplemented with synthetic folic acid. It is therefore of great interest to elucidate these pathways. In order to explore the possible options for advanced research of specific pathways we performed a pilot study to identify proteomic and genomic changes in response to folate addition. Our previous (unpublished) study on protein changes in response to folate revealed the involvement of glucose metabolism, energy production, nucleocytoplasmic transport, cell cycle regulation, cytoskeleton, protein processing, and DNA transcription and translation. Regarding these findings and the essential role of folate in DNA stability, methylation, and cell death, a clear genomic response is to be expected. To get a first impression of this response, Epstein Barr virus immortalized (EBV) B-lymphoblast cultures were induced with 5-methyltetrahydrofolate. The orientating nature of this approach, and the use of EBV B-lymphoblasts, clearly limits the possible observation of specific developmental functions of folate in facial primordia. However, we expect that there will be mutual consequences of folate supplementation. Therefore the goals of the present study are: (1) to identify folate responsive pathways using gene expression profiling; (2) to identify possible relationships of these differential genes with embryonic pathways involved in palate formation; (3) to compare the folate responsive genes with the earlier identified proteins.

MATERIALS AND METHODS

RESULTS

Folate concentrations in the medium were on average 4.4 nM (standard deviation 0.4) and 25.9 nM (standard deviation 4.0), respectively, before and after the addition of 5-mTHF. The distribution of intensities of the raw microarray data showed increased average intensities for one cell line before folate addition because of high background signal. After normalization, gene expression profiles from samples of the same cell line were highly consistent and therefore grouped together prior to clustering by folate status (Fig. 1). However, the cluster plot also showed sets of genes from which the expressional level seemed to alternate corresponding to the folate status. These potential folate-responsive genes were identified with a SAM analysis. From Figure 1 it is also clear that after normalization the “E before” sample, that is, the sample with the high background signals, showed very low expression levels for almost all genes. For this reason, this sample was left out of further analysis, because it did not contain any usable information. The SAM analysis was performed with the standard false discovery rate of less than 1 per 100 probe sets, corresponding to a delta-value of 4.226. This resulted in the identification of 144 significant up-regulated and 409 down- regulated probe sets after addition of 5-mTHF. The median expression ratio was 1.63 (range: 1.2–3.8) for the significant down-regulated genes and 1.58 (range: 1.2–4.2) for the up-regulated genes.

These potential folate-regulated genes are listed in Table 1 by the pathway or disease in which they are involved. These include cancer development, cell cycle checkpoint regulation, DNA replication, recombination and repair, biosynthesis of amino acids and DNA/RNA nucleotides, protein processing, and cell death. Furthermore, genes that best discriminated between pre- and postfolate-supplemented cell lines were evaluated. Figure 2 shows the 100 best classifying genes arranged by biological function. Of these classifiers, 64% were found to be significantly differentially expressed with the SAM analysis and had functions in similar pathways.

Projection of the data on known canonical pathways reveals involvement of various pathways in cell cycle regulation and phases of the cycle. DNA replication control is represented by regulation of various DNA polymerases and genes from the MCM family and CDC7 involved in the initiation of genome replication. The G2/M DNA damage checkpoint was represented by the CDC25/CDC2/Cyclin B pathway, associated with DNA damage processing via the P53 tumor suppressor.

The regulation of mitosis was represented with, for example, NEK2, a centriole division gene, and AURKA, which formats and stabilizes microtubules at the mitotic spindle pole during chromosome segregation. Furthermore, the CENPA gene, a methylated variant of histone H3 involved in centrosome formation, was found to be significantly up-regulated.

Other pathways linked to cell cycle regulation, such as the nucleotide excision repair pathway, showed diminished activity after folate addition. This was demonstrated with decreased expression of DNA binding proteins ERCC4 and DDB1 and increased expression of single stranded DNA binding protein and various DNA polymerases.

The apoptosis pathway was represented by BCL-2, FHIT, DFFA, and PDCD8. Interestingly, FHIT encompasses the FRA3B fragile site, which is expressed in a folate-deficient environment.

Besides cell cycle regulation, protein processing pathways were represented by various genes of the chaperonin-containing TCP1 complex family, which were down-regulated after folate addition. This complex folds various polypeptides in an ATP-dependent manner into active proteins, including actin and tubulin.

Other known functions of folate, such as the one-carbon group cycle and nucleotide synthesis, were mainly unregulated, although several genes of the purine synthesis pathway were found to be down-regulated.

The results of quantitative real-time RT-PCR on 10 genes belonging to the 100 best classifying probe sets are shown in Table 2. The reference gene β-actin showed no differential expression (ratio after/before folate addition is 1.01). For 8 out of 10 tested genes the direction of regulation found with the RT-PCR experiments was identical to the direction found with the microarrays. Moreover, from the eight directionally correct genes, five reached statistical significance, that is, RBBP6, pTEN, ANKRD11, BATF3, and HSP90AA1. The directions of the HNRPD and the TSC22D3 genes were opposite to the microarray data.

DISCUSSION

In the present study we show the results of a genome-wide expression analysis in B-lymphoblasts derived from CLP patients to identify folate-responsive genes and associated pathways and their relevance in lip and palate development. The forced clustering of the data revealed significant up-regulation of 144 and down-regulation of 409 genes in response to folate. Differential expression was confirmed with quantitative RT-PCR, which showed comparable regulation in 8 out of 10 tested genes, from which five genes reached statistical significance. The regulated genes were not concentrated in specific functions or pathways, but covered several functions at a low level. This indicates a general modifying role of folate in physiological processes, which might be connected with the extensive role of folate as a one-carbon group donor. One-carbon groups are used for the synthesis of purines and pyrimidines, proteins, and the remethylation of homocysteine into methionine, the main methyl group donor of the cell. Interruption of these basal functions leads to various types of cellular and chromosomal damage, such as uracil accumulation and incorporation, abnormal protein and CpG methylation, incorrect imprinting patterns, DNA strand breaks, aneusomy, and cell death (Chiang et al., 1996; Duthie and Hawdon, 1998; Kim, 2005; Steegers-Theunissen et al., 2000; Wang et al., 2004). As a result, altered folate status might influence cell cycle progression. This is supported by our data showing modified expression of a relatively high number of genes involved in cell cycle regulation, especially G2/M checkpoint regulation, S-phase initiation, and regulation of mitosis.

The modest number of genes regulated by folate was confirmed by the unsupervised clustering, which demonstrated that the primary clustering of samples was to the original cell line instead of folate status. The similarities between the samples from the same culture were thus larger than the similarities in folate response. One explanation may be the low number of folate-responsive genes or the low number of samples or even different responses to the folate intervention of the separate cell lines.

In our unpublished study we identified folate-responsive proteins in 30 B-lymphoblast cell lines from CLP and control children using a new proteomic method based on peptide fingerprinting. These proteins involved glucose metabolism, energy production, nucleocytoplasmic transport, cell cycle regulation, cytoskeleton, protein processing, and DNA transcription and translation. The present results confirm the responsiveness of these pathways to folate supplementation. This was especially true with respect to various heat shock proteins and heterogeneous nuclear ribonucleoprotein.

Spiegelstein et al. (2004) studied gene expression in the anterior neural tube of Theiler stage 13/14 FOLBP1 knock-out mice after feeding them a folate-deficient diet and using a 5700 gene array. Biological functions identified as being regulated by folate were comparable with those identified in our study and comprised processes such as proliferation, apoptosis, transcription, and translation. Courtemanche et al. (2004) performed a study using a 695 gene targeted microarray focusing on pathways involved in cellular aging and stress. Interestingly, they also identified cell cycle and DNA damage-related expression as a consequence of folate deficiency. However, their limited array size made it impossible to explore various other pathways.

Developmental genes such as TGFβ and MSX1 and their receptors, which are known to contribute to CLP development (Carinci et al., 2007) and thus might be target genes for folate, were not identified as folate-responsive genes. In the case of TGFβ it was reported earlier that this gene is an important regulator of apoptosis in B-cell precursors (Lanvin et al., 2003) and thus it seems likely that the used B-lymphoblasts were expressing TGFβ. The lack of differential expression of TGFβ in response to folate might therefore indicate that there is no interaction. However, in theory, the possibility remains that interaction is selectively and/or temporarily present in the developing facial structures. Additional testing is needed to clarify these possible gene-specific effects of folate.

An interesting observation is that the expression of genes that code for oncogenes and tumor-suppressors was altered in the present study. Although recent reports on the association of folate and the development of certain cancers are still inconclusive, there are increasing concerns that folate deficiency as well as folate excess might contribute to cancer development (Ames and Wakimoto, 2002; Kim, 2007; Mason et al., 2007). The deregulating effects of folate on normal cell development as shown by our data corroborate with this hypothesis and these results might add new starting points to unravel this apparent gene-nutrient interaction.

This is one of the first studies using human cell lines for genome-wide profiling of the gene expression in response to folate. We observed >500 significant changes in expression of genes involved in a variety of biological pathways. The selection of a homogeneous group of patients all with an identical nonsyndromic cleft contributed to the validity of the results. The measurements of the folate concentration in the medium of the cultures confirmed the actual folate states and thus folate deficiency and the target folate concentration. The B-lymphoblast culture model is a frequently used model for folate studies and was found to be appropriate for the assessment of folate-responsive gene expression and, as such, informative for folate-sensitive congenital malformations such as CLP. However, we realize that other time- and tissue-specific pathways may be active during palatal development and may contribute to CLP development. Secondly, only five cell lines derived from CLP patients could be profiled for this explorative study. Evidently, higher numbers of cell lines will increase the reliability of the data and prevent false-positive identifications. Inclusion of samples derived from healthy control children would also allow case or control specific gene expression relevant for increasing the understanding of CLP development. Future studies are needed to explore the possible role of the present set of genes in the development of CLP and other folate-sensitive congenital malformations. This may eventually lead to the further understanding of gene-environment interactions in the development of congenital malformations such as CLP.