LRP2 mediates folate uptake in the developing neural tube

The low-density lipoprotein (LDL) receptor-related protein 2 (LRP2), also known as megalin (Saito et al., 1994), is a multifunctional cell-surface receptor that is structurally related to the LDL receptor (Nykjaer and Willnow, 2002). Endocytosis through LRP2 is important for the efficient uptake of several vitamins and hormones bound to their carrier proteins. Examples are the re-uptake of vitamin D and retinol from the primary urine (Christensen et al., 1999; Nykjaer et al., 1999), as well as the uptake of androgens and estrogens into steroid-responsive cells (Hammes et al., 2005).

LRP2 plays a crucial role in forebrain development. The receptor is highly expressed in the neuroepithelium and loss of receptor activity in the developing central nervous system (CNS) in gene-targeted mice causes holoprosencephaly (HPE), a failure of the forebrain hemispheres to separate along the midline (Spoelgen et al., 2005; Willnow et al., 1996). Patients with autosomal recessive LRP2 gene defects suffer from Donnai-Barrow syndrome, a disorder associated with forebrain anomalies (Kantarci et al., 2007). We previously clarified the molecular mechanism underlying the HPE phenotype by identifying LRP2 as a novel component of the sonic hedgehog (SHH) signaling machinery in the ventral forebrain neuroepithelium (Christ et al., 2012). LRP2 deficiency in mice leads to failure of the neuroepithelium to respond to SHH and, consequently, to improper specification of structures of the ventral forebrain midline.

Besides the HPE phenotype, we have also noted additional cranial neural tube defects (NTDs) in LRP2-deficient mice that cannot be explained by loss of SHH signaling in the developing forebrain. The cause of these NTDs in Lrp2^−/− embryos is so far unknown. NTDs are a group of congenital malformations that occur when the neural tube fails to close during embryonic development. In human pregnancies, NTDs are the second most frequent malformations after congenital heart defects (Wallingford et al., 2013). Among the candidate genes associated with risk for human NTDs are genes important in folate metabolism.

A functional link between LRP2 activity and endocytic folate uptake in vitro in kidney sections has been suggested previously (Birn et al., 2005). The binding of soluble folate receptor 1 (sFOLR1; also known as FBP1 or FBP, for folate binding protein) to immobilized LRP2 was shown by surface plasmon resonance analysis, as well as by binding of sFOLR1 to sections of kidney cortex and uptake of sFOLR1 by BN-16 cells. In this study, we addressed the question of whether LRP2 expressed in the developing neural tube is required for the delivery of folate into neuroepithelial cells during neurulation.

We systematically examined the neural tube closure phenotype of Lrp2^−/− mice and Shh-null mutant mice. In agreement with previous reports (Chiang et al., 1996; Murdoch and Copp, 2010), SHH deficiency does not prevent neural tube closure (Fig. 1). Despite impaired specification of the ventral neural tube, 95.5% of all somite-stage 17–26 SHH-deficient embryos showed normal neural tube closure, a number not significantly different from that of wild-type embryos, where rostral neural tube closure is completed at the 15-somite stage [embryonic day (E) 9.0; Fisher's exact test, P>0.2] (Fig. 1). By contrast, a significant proportion (38%; Fisher's exact test, P<0.0001) of LRP2-deficient embryos exhibited an open rostral neural tube at somite-stages 17–26 (Fig. 1). In all of these cases, the posterior neural tube had closed and, in most of them, the most anterior portion of the forebrain was also closed. This pattern indicates a defect in neural tube closure at the level of Closure 2 (Copp et al., 2003), which is initiated at the forebrain-midbrain boundary, in Lrp2^−/− embryos.

Other LDL receptor family members might be affected in our Lrp2-deletion model and might therefore contribute to the NTDs. Loss- and gain-of-function mouse models for LRP6 suffer from compromised neural tube development (Gray et al., 2013; Gray et al., 2010). We thus measured the expression levels of Lrp1, Lrp1b, Lrp4, Lrp5 and Lrp6 in the developing brain. As shown in supplementary material Fig. S1, no differences in expression levels of these receptors were detected by comparing mutant with control samples. These results support the idea that the neural tube phenotype is caused by LRP2 deficiency and not by altered expression of other LDL receptor family members.

One established risk factor for NTDs is impaired uptake of folate (vitamin B₉) bound to folate-binding proteins and disturbed metabolism of the vitamin (Gelineau-van Waes et al., 2008a; Piedrahita et al., 1999; Spiegelstein et al., 2004). There are different routes by which cells take up folate. Relevant for embryonic development of the nervous system are the bidirectional solute carrier family 19 (folate transporter) member 1 (SLC19a1; also known as reduced folate carrier 1, RFC1) and the folate receptor 1 (FOLR1; also known as folate binding protein, FBP1), which exists both as a glycosylphosphatidylinositol (GPI)-anchored and as a soluble isoform, sFOLR1 (Lacey et al., 1989; Spiegelstein et al., 2000). FOLR1 regulates folate uptake through endocytic mechanisms and is important for proper neural tube closure (Piedrahita et al., 1999; Tang and Finnell, 2003). Because none of the FOLR1 receptor isoforms contains a transmembrane segment, high efficiency uptake relies on their association with an additional as-yet-unidentified transmembrane receptor.

We asked whether LRP2 might mediate the uptake of FOLR1 in vivo. To test this hypothesis, we injected adult control and LRP2-deficient mice with Alexa-Fluor-488 (A488)-labeled sFOLR1. After 2 hours, the labeled protein was detectable in intracellular compartments of kidney proximal tubules of control mice (Fig. 2A). By contrast, no uptake was seen in LRP2-deficient animals (Fig. 2A), indicating that LRP2 is required for efficient uptake of sFOLR1 in vivo.

We next tested whether LRP2 is required for cellular uptake of sFOLR1 in neuroepithelial cells during neurulation stages. To this end, we cultured control and LRP2-deficient E8.5 embryos before neural tube closure in medium containing A488-labeled sFOLR1. After 2 hours of incubation, sFOLR1 could be detected bound to the surface and within the apical compartment of control neuroepithelial cells (Fig. 2B). However, no sFOLR1 was detectable in the LRP2-deficient neuroepithelium (Fig. 2B). Furthermore, adding inhibitory anti-LRP2 antibody blocked sFOLR1 uptake in wild-type neuroepithelial cells (Fig. 2C). These results indicate that LRP2, expressed in the neuroepithelium, is required for the cellular uptake of sFOLR1 during neurulation. To exclude the possibility that LRP2 facilitates endocytosis of the A488 tag we incubated wild-type embryos with sFOLR1–Alexa-Fluor-647 (A647) and GST–A488. No uptake of GST–A488 was seen in these experiments, whereas uptake was seen for sFOLR1–A647 (supplementary material Fig. S2).

We next examined whether LRP2-mediated uptake of FOLR1 results in the simultaneous uptake of folic acid bound to this carrier. To do so, whole-embryo cultures were incubated with Cy3-labeled folic acid complexed with sFOLR1–A488. Uptake experiments showed intracellular signals both for sFOLR1–A488 and Cy3-labeled folic acid in the neuroepithelial tissue of E8.5 wild-type mouse embryos. However, no sFOLR1–A488 uptake, and severely reduced internalization of folic-acid–Cy3, was observed in the rostral neural folds of Lrp2^−/− littermates (Fig. 2D). Thus, LRP2 appears to present the main uptake pathway for this vitamin in the neural tube. To explore sFOLR1-independent uptake of folic acid, we incubated embryo cultures with folic-acid–Cy3 without adding soluble FOLR1, under serum-free conditions. In this scenario, the GPI-anchored FOLR1 should be able to mediate vitamin uptake both in wild-type and mutant embryos. However, also in this case, LRP2-deficient neural folds showed reduced uptake of folic acid compared with wild-type tissue (Fig. 2E), indicating that internalization of the complex formed by folic acid and GPI−FOLR1 is also dependent on LRP2 (see model, Fig. 5).

Taken together, these experiments demonstrate that efficient cellular uptake of folic acid and its binding proteins sFOLR1 and GPI–FOLR1 is mediated in a LRP2-dependent manner at the apical surface of the developing neuroepithelium. Importantly, our results suggest that endocytosis of a vitamin-binding protein, in this case FOLR1, by LRP2 directly influences crucial events during embryonic development, as cellular uptake of folate is required for the normal progression of neural tube closure.

To investigate whether the impaired uptake of FOLR1 caused by LRP2 deficiency in the neuroepithelium influences folate concentrations in the embryonic neuroepithelium, we determined tissue folate levels in E9.5 Lrp2^−/− embryos and somite-matched control littermates. Folate concentrations were significantly lower in tissue samples isolated from LRP2-deficient embryonic anterior neural tube compared with samples derived from control embryos (Fig. 3A). This finding is consistent with an impaired uptake of folic acid and FOLR1 in the neuroepithelium of LRP2-deficient embryos.

To provide further evidence for a functional link between FOLR1 and LRP2, we carefully compared the expression pattern and protein localization in the developing brain of E9.5 and E10.5 embryos (supplementary material Fig. S3). The most prominent overlap for LRP2 and FOLR1 was seen in the dorsal neuroepithelial midline, which corresponds to the site where Closure 2 is initiated, and in the ventral midline of the midbrain neuroepithelium.

We next asked whether the impaired uptake of folate and decreased tissue folate concentrations are primary consequences of LRP2 deficiency rather than of defects in the expression of folate receptors and carriers. We thus examined the expression of FOLR1 (protein and mRNA) and Slc19a1 (mRNA) in Lrp2 mutants and in control littermates. Using in situ hybridization and immunohistochemistry, we showed that the expression pattern of FOLR1 in the neural tube of LRP2-deficient embryos was comparable to that of wild-type controls (Fig. 3B). The expression levels of Folr1 and Slc19a1 were significantly increased in Lrp2^−/− embryonic heads compared with their expression in control samples (Fig. 3C,D), a fact that could be explained by a compensatory upregulation of folate receptors and folate carriers in response to LRP2 deficiency.

In the kidney, where LRP2 mediates the uptake of sFOLR1 into proximal tubule cells (Fig. 2A), the expression pattern of Folr1 mRNA and the immunohistological detection of FOLR1 protein was not different between Lrp2^−/− and control kidneys (Fig. 3E,F). Thus, a loss of expression of the established folate receptors or carriers does not account for the defects in folate metabolism seen in LRP2-deficient organisms.

Little is known about the mechanisms underlying the rescue of neural tube closure defects by folate. Recently, it was shown that expression of the gene Alx3 (aristaless-family homeobox transcription factor 3) is specifically dependent on folate, and that ALX3 is important for neural tube closure (Kessaris et al., 2006; Lakhwani et al., 2010).

In situ hybridization analysis and quantitative RT-PCR showed that Alx3 mRNA expression was reduced in LRP2-deficient embryos compared with controls (Fig. 4A,C). Interestingly, expression of the Alx3-related folate-independent transcription factor Alx4 (Lakhwani et al., 2010) was unchanged in LRP2-deficient embryos (Fig. 4B,C). Importantly, impaired Alx3 expression in LRP2-deficient embryos is unlikely to be a downstream effect of aberrant SHH signaling, because we detected normal expression patterns and mRNA levels of Alx3 in Shh-mutant embryos (Fig. 4A,C). These findings suggest that in the absence of LRP2, the endocytosis of folate is impaired, leading to alterations in the expression of folate-dependent genes including Alx3, ultimately contributing to the pathogenesis of NTDs.

Over the past years it became well accepted that members of the LDL receptor family, including LRP2, play an important role in signaling during embryonic development. Here, we show that LRP2 is essential for the uptake of folic acid and folate receptors in the developing neural tube. Our results also demonstrate that LRP2 deficiency leads to impaired cellular uptake of folate during neurulation and impaired expression of the transcription factor Alx3, which likely contributes to rostral neural tube closure defects in Lrp2^−/− embryos.

LRP2 is a multifunctional endocytic receptor expressed early in embryonic development on the apical surface of the neuroepithelium. In our recent work, we established a role for LRP2 in the ventral midline of the rostral diencephalon, where the receptor mediates SHH signal transduction in the early embryo, acting as a co-receptor for patched1 (Christ et al., 2012). However, the function of LRP2 in other neural tube domains during embryonic development remained unclear. Here, we postulate that LRP2 plays a role in the process of neural tube closure, because 38% of LRP2-deficient embryos showed impaired rostral neural tube closure, which is unlikely to be linked to loss of SHH signals (Fig. 1) (Chiang et al., 1996; Murdoch and Copp, 2010).

Folic acid (vitamin B₉) is known to reduce the risk of human NTDs, and mouse models with loss-of-function mutations affecting the Folr1 or Slc19a1 genes show NTDs and embryonic lethality (Blom et al., 2006; Cabrera et al., 2004; Gelineau-van Waes et al., 2008a; Piedrahita et al., 1999; Zhao et al., 2001). LRP2 has been shown to mediate the uptake of many hormones and vitamins bound to their respective carrier proteins, including vitamin D-binding protein (DBP), retinol-binding protein (RBP) and sex hormone-binding globulins (SHBG). Impaired uptake of these complexes in LRP2-deficient mice affects adult vitamin homeostasis (DBP and RBP) as well as late embryonic and early postnatal development (SHBG) (Christensen et al., 1999; Hammes et al., 2005; Nykjaer et al., 1999).

A functional link between LRP2 activity in the kidney and endocytic uptake of folic acid and the folate receptor has been suggested (Birn et al., 2005). Additional evidence for the role of LRP2 in folate uptake in embryonic tissue comes from a study by Gelineau-van-Waes and colleagues, showing that components of the LRP2–cubilin receptor complex are upregulated in SLC19a1-deficient embryos (Gelineau-van Waes et al., 2008b). Thus, increased expression of Lrp2 could compensate for the lack of folate uptake through SLC19a1 by increasing the folate uptake through FOLR1 and LRP2. Our results demonstrate that LRP2 plays an essential role in the uptake of folic acid and its binding protein FOLR1 in the developing neural tube (Fig. 2). LRP2 deficiency leads to impaired cellular uptake of folate during neurulation, which might consequently contribute to NTDs in Lrp2^−/− embryos.

The weak uptake of folic acid seen in Lrp2^−/− tissue (as shown in Fig. 2D,E) might be explained by the internalization of folic acid through LRP2-independent routes. For example, folic acid can enter the cell by the folate carrier SLC19a1. Expression levels for this folate carrier are upregulated in LRP2-deficient embryos (Fig. 3C,D), indicating a compensatory effect as a consequence of impaired LRP2- and FOLR1-mediated uptake of folate. However, considering the weak uptake of folic acid in LRP2-deficient tissue, the SLC19a1-mediated folic acid uptake and alternative routes for the uptake of sFOLR1 or GPI–FOLR1 (depicted in Fig. 5) seem to play a minor role in the efficient uptake of folic acid in the neural folds. This further supports the hypothesis that efficient uptake of folate is dependent on co-receptor-mediated endocytosis of sFOLR1 and GPI–FOLR1.

A putative co-receptor for FOLR1 had not yet been identified. Our data suggest that the endocytosis of soluble FOLR1 and membrane-anchored GPI–FOLR1 requires the interaction of these proteins with the single-spanning transmembrane endocytic receptor LRP2 (Fig. 5). This ‘dual-receptor complex’ hypothesis for the uptake of folate is analogous to the uptake of renal and intestinal vitamin B₁₂ bound to intrinsic factor through cubilin, a peripherally attached glycoprotein. Internalization of the cubilin–ligand complex strongly relies on the interaction with LRP2 (Denz-Penhey and Murdoch, 2009; Horbinski et al., 2009).

Gene defects and environmental factors, or a combination thereof, can cause NTDs. Studies on animal models have identified a large number of candidate genes involved in the etiology of NTDs (Harris and Juriloff, 2010). In humans, however, little is known about the interaction of most of these risk genes and the maternal factors. Therefore, further experiments will be required to elucidate the mechanisms underlying the protective effects of maternal folic acid supplementation for the embryo (Blom et al., 2006; Finnell et al., 2010; Wallingford et al., 2013). Moreover, to better understand the molecular mechanisms of LRP2-dependent NTDs, it remains to be investigated whether other factors, such as altered morphogen pathways, cilia function and apico-basal polarity of neuroepithelial cells (Eom et al., 2011), function in addition to the impaired folate uptake to modulate the NTD phenotype in LRP2-deficient mice.

The generation of mice with targeted disruption of the Lrp2 gene has been described previously (Willnow et al., 1996). Analyses of the embryonic neural tube defects were carried out in LRP2-deficient and in somite-matched wild-type and heterozygous littermates on a C57BL/6N background. Wild-type and heterozygous littermates show no differences in any of the analyses, and both genotypes are therefore referred to as controls in the experiments. sFOLR1-uptake studies in the kidney were performed on LRP2-deficient mice on an inbred FVB/N genetic background, because surviving adult LRP2-deficient mice can be obtained on this genetic background. Shh^{tm1(EGFP/cre)} mutant mice (Harfe et al., 2004), kindly provided by Clifford Tabin (Harvard Medical School, Boston, MA) were used to analyze the neural tube phenotype in SHH-deficient embryos and wild-type somite-matched littermates. All experiments involving animals were performed according to institutional guidelines following approval by local authorities.

sFOLR1–A488, sFOLR1–A647 and GST–A488 were generated by labeling folate-binding protein from bovine milk (F0504, Sigma) and GST using the Alexa Fluor 488 or Alexa Fluor 647 protein labeling kit (A10235, A20173, Life Technologies). Cy3–PEG-folic acid was obtained from Nanocs (PG2-FAS3-3k). E8.0 and E8.5 embryos were incubated in their yolk sacs in DMEM containing 5 µg/ml sFOLR1–A488, sFOLR1–A647, GST–A488 or folic-acid–Cy3 at 37°C under 5% CO₂ and 95% humidity. Folic-acid–Cy3 was incubated with sFOLR1–A488 for 30 minutes at room temperature before adding both reagents to the embryo culture. The yolk sac and amnion were opened by an incision to enable the diffusion of proteins. After 2 hours of incubation, the yolk sac and amnion were removed, and embryos were fixed in 4% paraformaldehyde (PFA) for 15 minutes. Embryos were cryosectioned at 10 µm. LRP2 receptor activity was blocked by incubation with goat anti-LRP2 antiserum (1∶100); controls received goat non-immune serum (1∶100).

To test sFOLR1 uptake by kidney proximal tubule cells, sFOLR1–A488 was injected into the tail veins of adult mice (300 ng/µl in PBS; 180 µl total volume). After 90 minutes, animals were sacrificed and kidneys were isolated and fixed in 4% PFA (at room temperature for 90 minutes). Kidneys were cryosectioned at 12 µm.

Folate to total protein ratios (nmole/g) were determined for head tissue samples from LRP2-deficient embryos (n = 13; 23–26 somites) and somite-matched heterozygous (n = 21) and wild-type (n = 17) controls. The assay procedure used to identify the concentration of folate in tissue samples was a modification of the folate-inhibition assay (Cabrera et al., 2008). Briefly, 50 µg/ml bovine folate-binding protein (Sigma Aldrich, St Louis, MO), diluted in 100 mM NaHCO3 pH 8.3, was printed onto the plate surface in 2 µl volumes at 4°C overnight. Sample tissues were suspended in 150 µl of lysis buffer [1× PBS, 0.1% Tween (v/v), 1% ascorbate (w/v)]. Samples were homogenized mechanically and placed in a boiling water bath for 10 minutes. Samples were then spun at 14,000 g for 7 minutes, and the supernatant was collected. A volume of 2 µl of 5% NaOH (w/v) was added for neutralization of the solution. Unlabeled folic acid was spiked by serial dilution (250–0.244 ng/ml) in PBS-Tween buffer in order to generate a standard regression curve for the determination of relative folate concentrations in samples. Standard and sample solutions were mixed 1∶5 with horseradish peroxidase (HRP)-labeled folic acid solution (FA–HRP; Ortho-Clinical Diagnostics, Raritan, NJ) and incubated in sample wells for 2 hours. Plates were washed six times with 1× PBS-Tween, and the HRP signal was detected with SuperSignal ELISA Femto Substrate (Pierce). The 96-well plates were imaged and analyzed on a Q-view chemiluminescent imager (Quansys Biosciences). Data were analyzed in GraphPad Prism (GraphPad Software) using a Student's t-test.

Standard immunohistochemical analysis was performed by incubation of tissue sections with the following primary antibodies at the indicated dilutions; sheep anti-LRP2 antiserum (1∶5000; kindly provided by Renata Kozyraki, Institute De La Vision, Paris) and rabbit anti-FOLR1 (1∶500; Abcam). Bound primary antibodies were visualized using secondary antisera conjugated with Alexa Fluor 488, 555 or 647 (1∶500; Invitrogen). Nuclei were counterstained with DAPI (1∶8000; Roche). Alternatively, bound primary antibodies were visualized with secondary antisera conjugated with biotin (1∶300), avidin–biotin complex (ABC; 1∶300; Vectastain) or diaminobenzidine (DAB; Sigma Aldrich). Analyses were performed using a Leica SPE confocal microscope and Leica DMI6000B.

Whole-mount in situ hybridization (WISH) was performed as described previously (Hammes et al., 2001). In situ hybridization (ISH) on sections was performed as described previously (Travers and Haas, 2004). Plasmids were purchased from Source Bioscience: FolR1, IRAKp961N1711Q; Alx3, D230004L08; Alx4, IRAVp968G0971D.

Total RNA from E9.5 embryo heads was extracted using an RNeasy Micro kit (Qiagen). cDNA was synthesized by using the High Capacity RNA-to-cDNA kit (Life technologies), and quantitative PCR was performed using an ABI7900. The expression of Alx3, FolR1 and Slc19a1 was normalized to that of Gapdh (Mm99999915_g1; Life Technologies). The following gene expression assays were used: Alx3, Mm01204737_m1; Alx4, Mm00431780_m1; Folr1, Mm00433355_m1; Slc19a1, Mm00446220_m1; Lrp1, Mm00464608_m1; Lrp1b, Mm00466712_m1; Lrp4, Mm00554326_m1; Lrp5, Mm01227476_m1; and Lrp6, Mm00999795_m1 (Life Technologies). Data were analyzed in GraphPad Prism (GraphPad Software) using an unpaired Student's t-test.

We are indebted to Cliff Tabin (Harvard Medical School, Boston, MA) for providing the Shh^{tm1(EGFP/cre)} mouse line, to Renata Kozyraki (Institute de la Vision, Paris, France) for providing the sheep anti-LRP2 antibody, to Kristin Kampf and Maria Schmeisser (MDC, Berlin, Germany) for excellent technical assistance. We thank Vanessa Schmidt (MDC, Berlin, Germany), Katja Herzog-Reynolds (MDC, Berlin, Germany) and Gary Lewin (MDC, Berlin, Germany) for constructive discussions.