Regulation of folate receptor 1 gene expression in the visceral endoderm^†

Folate, Birth Defects, and General Health

Folate deficiency has been linked to an increased incidence of congenital malformation (Molloy and Scott, 2001), heightened risk for certain types of cancer (Prinz-Langenohl et al., 2001; Ryan and Weir, 2001; Courtemanche et al., 2004b), reduced immune system performance (Courtemanche et al., 2004a), anemia, and reduced endurance (Lukaski, 2004). Suboptimal folate levels appear to be linked to impaired general health (Singh, 2004) and neurologic symptoms in aging (D'Anci and Rosenberg, 2004; Kim et al., 2008). Furthermore, our own experiments have demonstrated a beneficial effect of folate on the morphogenesis of the skeleton (Kappen et al., 2004). Historically, the finding that neural tube defect (NTD) frequencies may be associated with low folate levels in the mother (Smithells et al., 1976; Yates et al., 1987; Milunsky et al., 1989), and the resulting general hypothesis that nutritional deficiencies could be involved in the etiology of birth defects (Smithells et al., 1976, 1977; Shaw et al., 1995), represented a significant milestone in the understanding of congenital malformations. Importantly, this concept presented a highly feasible therapeutic approach to birth defect prevention simply by supplying vitamin preparations including folate to women who wished to become pregnant (Smithells et al., 1981). In fact, periconceptional supplementation with folate (Locksmith and Duff, 1998; Bailey, 2000; Ladipo, 2000) proved to be highly beneficial to the conceptus, resulting in significantly decreased occurrence of NTDs, craniofacial malformations, and cardiovascular abnormalities among newborns (Wald et al., 1991; Gelineau-van Waes and Finnell, 2001).

Folate Transport and Cellular Uptake

Mammalian cells have developed an elaborate mechanism to harvest extracellular folate (Trippett and Bertino, 1999), involving extracellular, glycolipid-anchored high-affinity folate receptors (Lacey et al., 1989; Wang et al., 1996; Wu et al., 1997), a low-affinity transmembrane carrier (Moscow et al., 1995; Wong et al., 1995), and a proton-coupled folate transporter (Qiu et al., 2006). Five folate receptor genes have been reported for the human genome (Elwood, 1989; Lacey et al., 1989; Ross et al., 1994; Spiegelstein et al., 2000), whereas in the mouse, three folate receptor genes are present. Expression studies on human folate receptors reveal that FOLR1 is mostly expressed in epithelial cells (Lacey et al., 1989; Page et al., 1993; Smith et al., 1999), FOLR3 is specific for the hematopoietic system (Shen et al., 1994), and FOLR2 (Reddy et al., 1999; Ross et al., 1999; Shen et al., 1994) and FOLR4 (Spiegelstein et al., 2000) are found at lower levels in diverse tissues. FOLR4 may play a role for the immune system because of its expression on regulatory T-cells (Walker, 2007; Yamaguchi et al., 2007). The carrier protein encoded by the RFC1 gene and the proton-coupled folate transporter PCFT seem to be widely distributed (Said et al., 1996; Wang et al., 2001; Maddox et al., 2003; Qiu et al., 2007). Mouse embryos lacking the Folr1 (folbp1) gene (Piedrahita et al., 1999) are arrested in their development shortly after gastrulation, fail to close the neural tube, and die in utero at mid-gestation, demonstrating that the mouse Folr1 gene is essential for embryonic development.

Expression of Folate Receptor Genes during Embryonic Development in the Mouse

The essential nature of folate intuitively would suggest that genes involved in folate transport and processing would be ‘housekeeping’ genes, with expression in every cell. In contrast, published observations (Saitsu et al., 2003) and our own in situ hybridization experiments (Kappen et al., 2004) demonstrate that this is clearly not the case: in the mouse embryo, genes for folate receptors are expressed in distinct and specific tissue distributions during development. The expression of Folr1 in the neural tube appears to represent a direct link to NTDs through a cell-autonomous function of the Folr1 gene (Saitsu et al., 2003), and folate supplementation is able to rescue embryos lacking the Folr1 gene (Spiegelstein et al., 2004). However, the literature is not clear about the causes for neurulation defects: both neural tube cells (Copp, 2005) and cells adjacent to the neural tube (Copp et al., 1988; van Straaten et al., 1993) are being implicated in a role for NTDs. Thus, a cell-autonomous model for Folr1 in NTDs may explain only part of the involvement of folate in neural tube closure.

In this context, it was interesting to note that the earliest and strongest expression signal for Folr1 in the developing embryo occurred in cells of the visceral endoderm (Saitsu et al., 2003), and our own data (Fig. 1). Expression was conspicuously absent from the embryo itself, raising the question as to how folate enters the majority of cells in the embryo during crucial periods of morphogenesis. Regarding the visceral endoderm, we found that Folr1 was the only gene of the folate receptor family expressed in this tissue. Consequently, it would appear that the FOLR1 protein represents the gateway for this important micronutrient through the visceral endoderm to the embryo itself. It is therefore possible that the lack of Folr1 in the gene knockout model may not only have a cell-autonomous effect on cells of the neural tube, but an additional indirect, pleiotropic effect on the whole embryo. Such pleiotropy may arise from the absence of Folr1 in the visceral endoderm, a resulting defect in folate transport in the visceral endoderm, failure to supply folate from the visceral endoderm to the embryo, and consequently, a condition of folate deficiency throughout the embryo, with negative consequences for cell proliferation and normal morphogenesis.

Functions of the Visceral Endoderm

The visceral endoderm has important roles in patterning and in nurturing the developing embryo (Brent et al., 1990; Bielinska et al., 1999). Crucial patterning signals for determination of anterior identity arise from the anterior visceral endoderm (Thomas and Beddington, 1996; Tam and Behringer, 1997; Beddington and Robertson, 1998), whereas visceral endoderm adjacent to extraembryonic mesoderm is essential for the induction of blood vessel development (Boucher and Pedersen, 1996; Belaoussoff et al., 1998). In parallel, the visceral endoderm is responsible for nutrient uptake and transport to the embryo (Cross et al., 1994). This function for nutritional support is indispensable during establishment of the chorioallantoic placenta (Brent et al., 1990); once the placenta is functional, the embryo can switch from histiotrophic to hemotrophic nutrition (Burton et al., 2001). However, it is important to consider that crucial developmental processes, such as neural tube closure and patterning of the early heart, occur at a time when the burden of nurturing the embryo lies with the visceral endoderm (Brent et al., 1990). Thus, it stands to reason that birth defects involving the early embryonic patterning processes in relation to a lack of nutrients should be interpreted in the context of visceral endoderm function. Examples are spina bifida and folate deficiency (Smithells et al., 1976), or diabetic embryopathy and the detrimental nutritional milieu brought about by maternal diabetes (Reece et al., 1993; Reece and Eriksson, 1996). In fact, gene expression changes in the visceral yolk sac are thought to contribute to birth defects in diabetic pregnancies (Reece et al., 2006). Therefore, proper regulation of genes involved in nutrient uptake and transport in the visceral endoderm is a crucial prerequisite for successful development of the embryo itself. A targeted mutation of the transcription factor HNF4-alpha underscores that view, as many genes for nutrient transport or metabolism, such as apolipoproteins, glucose transporter 2, transferrin, and cytoplasmic retinoic acid binding proteins are downregulated in the visceral endoderm of HNF4-alpha^−/− embryos (Stoffel and Duncan, 1997). In fact, the failure of HNF4-alpha^−/− embryos to complete gastrulation has been ascribed to “death by starvation” (Copp, 1995; Duncan et al., 1997).

Regulation of Folate Receptor 1 Gene Expression

Gene transcription is typically dependent on regulatory DNA elements such as promoters and enhancers. Like all folate receptor genes, the human FOLR1 gene has multiple promoters that are well characterized. The FOLR1 P1 promoter (so designated as the transcript starts at exon 1) was studied in KB epidermal carcinoma cells and NIH/3T3 fibroblasts (Elwood et al., 1997; Galmozzi et al., 2001). Both the FOLR1 P1 and the FOLR1 P4 promoter (transcript starting at exon 4) contain initiator sequences and respond to the transcription factor SP1 (Sadasivan et al., 1994; Saikawa et al., 1995). The two promoters exhibit differential activity in KB cells, several adult tissues, and ovarian cancer cells (Elwood et al., 1997; Galmozzi et al., 2001). Furthermore, the transcription factor vHNF1 activates the P1 promoter in ovarian carcinoma cells (Tomassetti et al., 2003), and the FOLR1 P4 promoter is modulated by the estrogen receptor in cervical and ovarian carcinoma cells (Kelley et al., 2003). Interestingly, in several cell lines, the FOLR1 gene appears to regulated in response to cellular growth and not in response to folate levels (Doucette and Stevens, 2001), indicating that cellular requirements rather than a simple feedback mechanism control this gene. A recent study reported genetic variation in the FOLR1 promoter region (Nilsson and Borjel, 2004), with potential health implications presumed to be due to altered expression of the gene. As is evident from the literature on folate receptor gene promoters, most experiments have focused on carcinoma cell lines, with emphasis on FOLR1 gene regulation in cancer.

In contrast, regulatory mechanisms for FOLR1 gene expression during embryonic development have not been explored. The expression of FOLR1 in the visceral endoderm is consistent with a role of the visceral endoderm in folate uptake and subsequent release to cells of the embryo itself. Therefore, the mechanism that is responsible for the specific expression of Folr1 in the visceral endoderm is fundamental to ensure folate supply to the embryo. To identify this mechanism, we have undertaken reporter gene experiments to gain further insight into the regulatory events that control Folr1 expression during development, with our focus on the visceral endoderm.

Given the importance of folate for prevention of human birth defects, it is reasonable to assume that deficiencies in folate transport may be cause for susceptibility to congenital malformation in humans. Such deficiencies might arise from genetic variation in the structural part of the FOLR1 gene, but could also be based on mutations in regulatory regions that are required for proper expression of FOLR1. Identifying the human regulatory elements for FOLR1 expression would provide a means to characterize genetic variation in such elements, and investigate the relationship to birth defect susceptibility. To date, this approach was limited to promoter regions of FOLR1 (Barber et al., 2000), because the existence, identity, and location of enhancer sequences were unknown. Therefore, rather than using murine sequences, we chose to attempt identification of regulatory sequences from the human FOLR1 locus, using a transgenic mouse approach to provide an evolutionary conserved in vivo context. In this fashion, any human sequences with regulatory function in vivo may be readily checked for potential genetic variation in human populations in the future. In this study, we report the identification of a sequence from the human FOLR1 locus that can act as a transcriptional enhancer to direct gene expression specifically in the visceral endoderm, and we suggest that the FOLR1 gene, like many other genes for nutrient uptake or transport, is a target for the transcription factor HNF4-alpha.

In Situ Hybridizations

Decidua with mouse embryos at 7.5 days' gestation—with 12 PM on the day of the appearance of a vaginal plug designated as gestation day 0.5—were dissected from the uterus of FVB mice, embedded in O.C.T. compound (Sakura Finetek, Torrance, CA), frozen, and used to generate cryosections of 25 μm thickness. A digoxigenin-labeled antisense riboprobe was generated from a mouse Folr1 full-length cDNA clone, and sections were hybridized as described previously (Salbaum, 1998).

Plasmid Constructs

Reporter constructs for Luciferase assays were generated in pGL3 (Promega, Madison, WI). DNA fragments spanning the P1 and P4 promoter regions, as well as evolutionary conserved sequences flanking the human FOLR1 gene, were generated by PCR from commercially obtained human genomic DNA (Roche, Indianapolis, IN). Genomic coordinates (UCSC genome browser, human genome version hg18, March 2006 assembly) for the amplified fragments were as follows: promoter P1, chr11:71,576,404-71,578,395; promoter P4, chr11:71,578,925-71,580,883. For conserved sequence elements from the FOLR1 gene, we use the abbreviation F1CE (for Folate receptor 1 Conserved Element) followed by a number; genomic coordinates were: F1CE1, chr11:71,560,901-71,561,809; F1CE2, chr11:71,565,324-71,566,907; F1CE3, chr11:71,591,596-71,592,126. The identity of each amplified DNA fragment was confirmed by DNA sequencing. Promoter fragments were generated with flanking MluI and XhoI restriction sites for cloning into pGL3 (Promega); fragments carrying conserved sequences were produced with flanking KpnI and MluI sites for cloning upstream of promoter sequences. Deletions in F1CE2-F1P4-GL3 were generated using existing restriction enzyme sites. A reporter plasmid carrying HcRed as reporter gene was generated by replacing the coding sequence for luciferase in pGL3 with the HcRed coding sequence from pHcRed-N1.1 (Clontech, Mountain View, CA). The plasmid F1CE2-F1P4-GhcR contains the same assembly of conserved sequence and promoter as F1CE2-F1P4-GL3 in the context of the fluorescent reporter, and was used for transgenic mouse experiments.

Transfection Experiments

Conditions to grow F9 mouse embryo carcinoma cells as well as differentiation to either visceral or parietal endoderm phenotype were as described elsewhere (Braunhut et al., 1992; Dong et al., 1990). Cells were seeded in 35-mm dishes at a density of 1 to 2 × 10⁵ cells per dish; transfections using Effectene (Qiagen, Valencia, CA) and 400 to 600 ng of plasmid DNA were performed 24 hours after plating. Twenty-four hours after transfection, the transfection mixture was replaced with media containing differentiation agents; for visceral endoderm, cells were treated with 1 μM all-trans retinoic acid (Sigma, St. Louis, MO), whereas for parietal endoderm, cells were grown on dishes pretreated with 0.1% gelatin and received 1 μM all-trans retinoic acid and 250 μM dibutyryl-cAMP (Sigma). Sixty hours after induction of differentiation, a time at which any retinoic acid-mediated effects of FOLR1 have long ceased, cells were lysed in GloLysis buffer (Promega), and luciferase activity was determined using SteadyGlo substrate (Promega). Luciferase values were normalized to the protein content of the lysate as determined by BCA assay. Each transfection assay was performed in either five or ten replicates. For cotransfections, expression vectors encoding transcription factors (CMV-HNF4-alpha, CMV-TGIF) were obtained from the IMAGE Mammalian Gene Collection of full-length cDNAs (Open Biosystems, Huntsville, AL). The expression vector pMT7-HNF4-alpha (Jiang et al., 1995) was kindly provided by Dr. Francis Sladek (University of California, Irvine, CA). DNA for cotransfection experiments was a mixture of 500 ng of reporter construct DNA as well as 100 ng DNA of the plasmid encoding a transcription factor. An expression vector containing the HcRed fluorescent protein sequence was used as negative control (in place for transcription factor-expressing plasmids) for cotransfection experiments and served for normalization of reporter activity. For all experiments, fold changes were calculated by normalizing all observed values to the average of the respective control experiment. Statistical significance was determined by performing a double-sided t test on control and experimental values normalized to the average of the controls.

Transgenic Mouse Experiments

The construct F1CE2-F1P4-GhcR was used to generate transgenic mouse embryos that were then analyzed for reporter activity by confocal microscopy. Injection DNA free of plasmid backbone sequences was generated by digestion with Asp718 and SalI, followed by agarose gel electrophoresis and purification (Qiagen). DNA was injected into fertilized oocytes of FVB mice as published (Hogan et al., 1996). At gestation day 7.5 (E7.5) as well as 9.5 (E9.5), embryos were dissected from the uterus of CD-1 foster mice, and embryo as well as yolk sac (at E9.5) of each specimen were used for imaging HcRed-specific fluorescence on a Zeiss Confocal microscope (Carl Zeiss Inc., Thornwood, NY); all images were taken at identical intensity settings. Genotyping for transgene presence was performed on DNA extracted from embryos after imaging.

Expression of Folr1 in the Visceral Endoderm

To reveal sites of expression of the Folr1 gene, we performed in situ hybridization experiments on mouse embryos at stages prior to neural tube closure. While our results in general confirm previously published data (Saitsu et al., 2003), we were intrigued by the high level of expression of Folr1 in the visceral endoderm (Fig. 1) at embryonic day 7.5, and the yolk sac at later stages of development. The visceral endoderm is a cell layer thought to play an important role for nutrition of the embryo (Brent et al., 1990). We detected only Folr1 expression; neither Folr2 nor Folr4 expression was found (not shown). It is therefore likely that Folr1 represents the gateway for high-affinity folate transport in the visceral endoderm, and the regulatory mechanisms that direct expression of the Folr1 gene in the visceral endoderm are likely to be of high biologic significance for healthy development of the embryo.

Activity of FOLR1 Gene Promoters in F9 Cells Differentiated toward Visceral Endoderm

We generated reporter constructs comprising 2 kb of sequences of either the P1 or the P4 promoter of the human FOLR1 gene (Fig. 2A) to test their activity in F9 embryo carcinoma cells that were differentiated to visceral endoderm (Dong et al., 1990; Braunhut et al., 1992). Although we observed no activity from the P1 promoter in F9 cells under any circumstance, the construct carrying the P4 promoter showed activity in F9 cells, but only after they had undergone differentiation toward a visceral endoderm phenotype (Fig. 2B). In F9 cells grown without induction of differentiation, the P4 construct did not exhibit any activity higher than a promoterless luciferase control vector. We conclude that the P4 promoter of the human FOLR1 gene has the potential to contribute to expression of the gene in the visceral endoderm.

Activity of FOLR1 Gene Promoters in Transgenic Mice

With the observation of cell type-specific promoter activity from the P4 promoter in the visceral endoderm model, we introduced reporter constructs with a β-galactosidase reporter gene (Fig. 2A) in transgenic mouse embryos to test whether promoter sequences of the human FOLR1, or the mouse Folr1 gene, were sufficient to drive expression of a reporter gene in a pattern resembling the expression of Folr1 in the mouse. Although we were able to generate transgenic specimen at expected frequencies (Table 1), none of those transgenic specimen showed reporter activity that matched expression of Folr1. A few embryos transgenic for the human FOLR1 P4 construct displayed some lacZ activity, but the spatial distributions of these activities were not consistent between individual transgenic samples, and did not match the known Folr1 expression pattern (Saitsu et al., 2003). We did not observe any reporter activity from transgenic samples carrying either the mouse Folr1 P1 or the mouse Folr1 P4 construct. We therefore conclude that the individual P1 or P4 promoter sequences of either the human or the mouse folate receptor 1 gene are not sufficient to drive gene expression in the correct pattern in vivo.

Conserved Sequence Elements at the FOLR1 Gene Locus

Because the gene for folate receptor 1 showed a high degree of sequence conservation between human and mouse, we hypothesized that the regulatory mechanisms controlling the expression of the gene might also be conserved. We used VISTA to generate a sequence conservation landscape around the human FOLR1 gene (Fig. 3). We initially selected the three conserved regions nearest to the FOLR1 gene and generated reporter constructs where each of the conserved sequences were placed in the context of the FOLR1 P4 promoter. An initial transfection survey experiment with constructs containing a fluorescent reporter (HcRed) suggested that the sequence termed F1CE2 conferred transcriptional activation activity upon the P4 promoter. We therefore decided to examine the F1CE2 sequence in further detail.

A Conserved Sequence Upstream of the FOLR1 Gene Shows Enhancer Activity in vitro

Using firefly luciferase as the reporter gene, we compared reporter activity for DNA constructs containing the FOLR1 P4 promoter in the presence or absence of the F1CE2 sequence, or with parts of the F1CE2 sequence deleted from the construct. We tested these constructs in F9 cells differentiated either towards the visceral or the parietal endoderm model. The results are summarized in Figure 4. Compared to the promoter-less vector pGL3, presence of the P4 promoter resulted in an increase in reporter activity at about the same magnitude as observed in the initial experiment; this was observed for both cell models. Activity of the F1P4-GL3 construct was then used to normalize reporter activities and calculate fold change. We found that addition of the conserved sequence F1CE2 to the P4 promoter resulted in an approximately eightfold increase of reporter activity in the visceral endoderm differentiation model, suggesting that the F1CE2 sequence can in fact act as an enhancer. Deletion of approximately two thirds of the F1CE2 sequence between the Asp718 and StuI restriction sites (F1CE2-dAS-F1P4) abolished that enhancement, suggesting that this enhancing activity may reside between these two coordinates. Deletion of the sequence between the two PflI restriction sites had little effect on enhancer activity, as seen for the F1CE2-dP-F1P4 and F1CE2-dPSM-F1P4 constructs. This suggests that the sequence between the second PflI site and the StuI site, which also contains a conserved sequence element, is a major contributor to the observed enhancer effect.

Tissue-Specific Enhancer Activity of a Conserved Sequence Upstream of the FOLR1 Gene

To determine whether the F1CE2 sequence would be able to confer enhancer activity in vivo, we generated transgenic mice with a construct carrying the same F1CE2-F1P4 configuration as described in the in vitro experiment, but using a gene for the red fluorescent protein HcRed as the reporter gene. We performed transient transgenic assays where the analysis for reporter activity was carried out directly on founder embryos. Analysis of embryos at 7.5 days of gestation (Fig. 5) revealed that all eight transgenic embryos exhibited red fluorescence restricted to the region of the visceral endoderm. A second experiment analyzed at 9.5 days of gestation yielded three transgenic specimen; all three showed consistent red fluorescence in the yolk sac. Fluorescence signals in embryos were spurious or not detectable at all, indicating that the reporter activity from this construct was specific for the visceral endoderm at E7.5, and for the yolk sac at E9.5, but not for the embryo proper. This was in excellent agreement with our earlier in situ hybridization results on the expression of the Folr1 gene itself, and suggests that the human F1CE2 region contains an enhancer sequence that is sufficient to drive the reporter gene expression in vivo in a pattern that resembles the expression of the mouse Folr1 gene. Based on the consistency between the in vitro and the in vivo results, we conclude that the F1CE2 sequence can function as an enhancer in the regulation of folate receptor gene expression in early development.

HNF4-Alpha Can Activate the FOLR1 Enhancer

When we examined the PflI-StuI fragment of the F1CE2 region for conservation and for the presence of potential transcription factor binding sites, we noted a sequence 5′-TGGAATTGGACCT-3′ that was identified by rVISTA software (Loots et al., 2002; Loots and Ovcharenko, 2004) as a potential binding site for the transcription factor HNF4-alpha. This suggested the possibility that HNF4-alpha might be involved in the function of the F1CE2 enhancer function and thereby contribute to the regulation of the folate receptor 1 gene. We tested this possibility by performing cotransfection experiments in F9 cells differentiated towards the visceral endoderm. For these experiments, we compared the reporter activity of the F1P4 promoter alone, the F1P4 promoter carrying the full sequence of the F1CE2 region, and the F1P4 promoter with the F1CE2-dPSM deletion. Luciferase reporter plasmids were cotransfected with expression vectors that would express (1) HNF4-alpha from the MT7 promoter (Jiang et al., 1995), (2) HNF4-alpha from the CMV promoter, (3) the transcription factor TGIF from the CMV promoter, or (4) the red fluorescent protein HcRed from the CMV promoter. We used CMV/HcRed to control for the presence of a very strong enhancer/promoter sequence in the transfected cells, and any sequestering of general transcription factors that might occur because of the CMV sequence. To test whether any observed effect would be due to the function of a sequence-specific DNA binding protein and not just due to the increased presence of any DNA-binding protein, we used the transcription factor TGIF, which is not related to the biologic context of the experiment. All reporter activities were normalized to the CMV/HcRed co-transfection to calculate fold-change as a response to presence of HNF4-alpha. In these experiments, we observed that presence of HNF4-alpha lead to a robust and highly significant increase of reporter activity from the F1CE2-F1P4 construct compared to control (Fig. 6). The degree of increase was different for the two HNF4-alpha expression plasmids (60-fold for CMV vs 20-fold for MT7). One reason may be that the two plasmids express different splice variants of HNF4-alpha, which are thought to have slightly different transcriptional activity (Eeckhoute et al., 2003). More likely though, the difference is due to the higher degree of expression of HNF4-alpha from the CMV promoter plasmid compared with the MT7 promoter plasmid. As expected, cotransfection of TGIF as a control for increased DNA binding protein content in the cells did not affect the F1CE2-F1P4 construct in a significant way: as TGIF is a transcriptional co-repressor (Wotton et al., 1999a, 1999b), silencing of the construct might have been expected, but was not observed. Surprisingly, the F1CE2-F1P4 deletion construct carrying the presumed HNF4-alpha site, showed significantly lower activation in response to HNF4-alpha than the construct carrying the intact F1CE2 enhancer element. The construct carrying the F1P4 promoter alone also responded to the presence of HNF4-alpha, although not nearly as strong as the construct with the F1CE2 enhancer. Therefore, it is likely that transcriptional activation via HNF4-alpha involves more than a single binding site on F1CE2. Taken together, these results suggest HNF4-alpha as a part of the regulatory mechanism that controls folate receptor 1 gene expression specifically in the visceral endoderm and the yolk sac.