Endocytosis of megalin by visceral endoderm cells requires the Dab2 adaptor protein

The lipoprotein receptors (including the LDL receptor, VLDLR, ApoER2/LRP8, LRP1 and LRP2/megalin) comprise a family of single-pass type I membrane proteins that mediate uptake of various protein cargoes into cells via the endocytic pathway (Krieger and Herz, 1994). Each receptor binds many different cargo proteins (more than 35 for megalin) and continuously recycles to and from the cell surface. The receptors cluster in clathrin-coated pits, are internalized and sorted to early endosomes. Following dissociation of ligands in late endosomes, the receptors are returned to the cell surface. The whole cycle occurs rapidly, just 5 minutes to internalize and another 15 minutes or so to recycle, depending on the cell type and the receptor.

Endocytosis of the prototype lipoprotein receptor, the LDL receptor, requires an FxNPxY signal, with crucial F, N, P and Y residues, in the cytoplasmic tail (Chen et al., 1990). Most lipoprotein receptors contain a similar sequence, although the importance of this signal for endocytosis of some receptors has been questioned (Li et al., 2000). The identity of the protein or proteins that interact with the FxNPxY signal and mediate endocytosis has been unclear until recently.

The first reported interaction, with clathrin, is still of unknown functional significance (Kibbey et al., 1998). However, two protein-interaction–phosphotyrosine-binding (PID-PTB) domain-containing proteins, ARH and Dab2, also bind (Bork and Margolis, 1995; He et al., 2002; Mishra et al., 2002; Morris and Cooper, 2001; Oleinikov et al., 2000), and ARH is required genetically for efficient LDL receptor uptake in liver and lymphocytes (Garcia et al., 2001; Norman et al., 1999). Dab2 is important for megalin transport in the kidney proximal tubule but its role in other tissues is unclear (Morris et al., 2002b; Nagai et al., 2005).

The dab2 gene is alternatively spliced to produce two protein products (Xu et al., 1995), one of which, p96, binds to clathrin and the clathrin adaptor AP2, and localizes to clathrin-coated pits, whereas the other, p67, does not (Mishra et al., 2002; Morris and Cooper, 2001). Significantly, overexpression of a dimer of the PTB domain inhibits internalization of LDL, indicating that the Dab2 PTB domain can displace the protein or proteins that normally mediate LDLR endocytosis (Mishra et al., 2002). However, absence of Dab2 does not affect LDLR endocytosis in a variety of cultured cells (M.E.M. and J.A.C., unpublished results).

Dab2 is essential in the visceral endoderm (VE) for embryonic development (Morris et al., 2002b; Yang et al., 2002). The VE is a polarized epithelial tissue, with a well-defined brush border composed of dense apical microvilli, that surrounds the developing mammalian embryo after implantation. Between embryonic days (E) 5.5 and 7.5, nutrients supplied by the VE support the rapid proliferation of the epiblast (Bielinska et al., 1999; Snow, 1977). The VE also plays an active role in patterning the early embryo (Beddington and Robertson, 1999; Coucouvanis and Martin, 1999; Rossant and Tam, 2004). Conditional knockout of the dab2 gene from most embryonic cells but not the VE allows normal development (Morris et al., 2002b). However, the essential function of dab2 in the VE has previously been unclear.

Here we show that the p96 isoform of Dab2 is essential for normal endocytosis and development. Endocytosis of transferrin (Tf) is decreased and the lipoprotein receptor megalin is mislocalized in the VE of dab2 mutants. The scavenger receptor cubilin lacks a cytoplasmic domain and is dependent on megalin for endocytosis (Christensen and Birn, 2002; Kozyraki et al., 2001). Accordingly, cubilin is also mislocalized in the VE of dab2 mutants. Even though Dab2 p67 is more highly expressed than p96 in the VE, expression of p67 alone led to decreased endocytosis, delayed development and reduced viability compared to control embryos, whereas expression of p96 was sufficient for normal endocytosis and development. These results indicate that Dab2 p96 mediates endocytosis of megalin and cubilin in the VE.

Previously described dab2^-/- mice were used for these studies (Morris et al., 2002b). These mice were maintained on a mixed 129Sv/C57BL/6 genetic background. The phenotype of these mice is similar to that of embryos in which dab2 is not expressed because of the absence of GATA6, a zinc-finger transcription factor that directs expression of dab2 in the VE (Morrisey et al., 2000; Morrisey et al., 1998). Other alleles of dab2 and gata6, both of which were disrupted by β-galactosidase, cause earlier embryonic lethality with a disorganized VE, suggesting that expression of β-galactosidase increases the severity of the phenotype when dab2 is not expressed (Koutsourakis et al., 1999; Yang et al., 2002).

E6.5 and E7.5 embryos were dissected in PBS and allowed to recover for 1 hour in 75% fetal bovine serum (FBS) with 25% Dulbecco's modified Eagle's medium (DMEM) at 37°C in 5% CO₂. Horseradish peroxidase (HRP) and Texas-red-labeled transferrin (TR-Tf) were added at 2 mg/ml and 25 μg/ml, respectively, for 5 minutes at 37°C. Following uptake, embryos incubated with TR-Tf were incubated in ice-cold acid stripping solution (150 mM NaCl, 10 mM NaOAc pH 5.0) for 5 minutes to remove surface-bound Tf. Embryos were fixed in 4% Paraformaldehyde (PFA) for 1 hour at 4°C, followed by three 5-minute washes in PBS. HRP uptake was detected by incubating embryos in diaminobenzidine tetrahydrochloride (DAB) solution plus nickel until a color change was observed (∼10 minutes), followed by a water rinse to stop the reaction. TR-Tf uptake was detected using fluorescence microscopy. To score uptake, a central region of each embryo was used to generate a histogram from which the average pixel intensity was determined. These values were averaged for each genotype, and the standard error was calculated. Significance was determined by using the Mann-Whitney test. Embryos were genotyped by PCR.

E6.5 embryos were fixed in Karnovsky's half-strength fixative for 36 hours and post-fixed in osmium s-collidine for 8 hours. Samples were dehydrated for 1 hour each in 35, 70 and 95% ethanol, followed by two washes with 100% ethanol and propylene oxide. Samples were then infiltrated with 50:50 propylene oxide:Epon 812 and placed in a vacuum oven overnight. Fresh Epon 812 was added and samples were returned to the vacuum oven overnight. Embryos were embedded in fresh Epon 812 and allowed to harden in the oven for 48 hours. Sections (400-600 nm) were placed on 150 mesh grids and stained for 2 hours with 6% saturated uranyl acetate, then with Millonig's lead stain for 4 minutes. Sections were viewed using the JEOL 100SX transmission electron microscope.

Embryos were fixed in 4% PFA, paraffin embedded and sectioned. Immunohistochemistry was performed using the Vectastain Elite ABC Kit (Vector Laboratories, Inc.) as described (Morris et al., 2002b). Briefly, 6-μm paraffin sections were de-waxed in Histoclear, rehydrated in serial dilutions of alcohol and steamed for 20 minutes in 30 mM citrate buffer pH 4.8, for antigen retrieval. Sections were rinsed in PBS and incubated in 3% hydrogen peroxide for 5 minutes to block endogenous peroxidases. Sections were rinsed and blocked for 30 minutes in 5% normal serum with 2% bovine serum albumin (BSA) in PBS. Slides were then incubated overnight at 4°C with a 1:200, 1:200 or 1:400 dilution of mouse anti-Dab2 (p96) (BD Transduction Labs), and goat anti-cubilin (Santa Cruz) or sheep anti-megalin (kind gift from R. Nielsen, University of Aarhus, Aarhus, Denmark) diluted in 5% BSA in PBS. Following three 5-minute washes in PBS plus 0.05% Tween-20 (PBST), sections were incubated for 30 minutes with horse anti-mouse, rabbit anti-goat, or rabbit anti-sheep biotinylated secondary antibodies diluted 1:200 in 2% BSA in PBS. Slides were washed again in PBST and incubated for 30 minutes with Vectastain Elite ABC Reagent. Following the last wash of PBST, sections were incubated in DAB solution plus nickel until a color change was observed (∼5 minutes) and slides were rinsed in water. Sections were counterstained with hematoxylin and mounted using Cytoseal.

Immunofluorescence was performed on whole-mount, paraffin- and cryostat-sectioned E6.5 embryos, and cryostat-sectioned kidneys. Embryos were isolated and fixed in 4% PFA-PBS for 2 hours at 4°C, and kidneys were fixed by perfusion with 4% PFA-PBS. Prior to freezing in optimal cutting temperature compound (OCT, Tissue-Tek), cryostat-sectioned embryos and kidneys were taken through a series of 30% sucrose-PBS:OCT incubations (100%:0%, 50%:50%, 25%:75%, and 0%:100%). 7 μm cryostat sections were rehydrated in PBS. Whole-mount and cryostat-sectioned tissues were permeabilized in 0.1% Triton X-100 in PBS for 20 minutes at 25°C. Tissues were blocked in 5% normal serum with 2% BSA in PBS for 1 hour before incubating with mouse anti-Dab2 (1:200), rabbit anti-ARH (1:20, kind gift from M. Farquhar, University of California, San Diego, CA), goat anti-cubilin (1:200), sheep anti-megalin (1:1000) and/or rabbit anti-EEA1 (1:200) antibody (Affinity BioReagents) overnight at 4°C. Paraffin-embedded embryos were prepared as described above, and immunofluorescence for Dab2 was performed like immunohistochemistry up to the point of secondary antibody addition. Following three 5-minute washes in PBST, whole-mount, cryostat- and paraffin-sectioned tissues were incubated for 1 hour with the appropriate AlexaFluor-labeled secondary antibody (Molecular Probes) diluted 1:1000. Following three 5-minute washes in PBST, 4′,6-diamidino-2-phenylindole (DAPI, 1:1000, Sigma) was added for 10 minutes. Sections were rinsed with water and mounted using the ProLong Antifade Kit (Molecular Probes). For visualization, 0.2 μm serial sections were taken using a Delta Vision microscope (Applied Precision), and the images were deconvolved and analyzed using the softWoRx program (Applied Precision).

Tissues were lysed on ice in lysis buffer (1% Triton X-100, 10 mM HEPES pH 7.4, 150 mM NaCl, 2 mM EDTA, 50 mM NaF, 0.2 M Na₃VO₄, 1% aprotonin, 1 mM PMSF, 10 mg/ml leupeptin) followed by centrifugation at 20,000 g for 10 minutes at 4°C. Samples were boiled for 5 minutes in 2× sample buffer, resolved by SDS-PAGE and transferred to Immobilon P (Millipore). Filters were blocked in 25 mM Tris-HCl, 8 mM Tris-base, 150 mM NaCl, 0.05% Tween-20 and 5 mM NaF with 2% BSA for 1 hour at 25°C. Mouse anti-Dab2 antibody was applied at a dilution of 1:2500 for 2 hours. Goat anti-mouse IgG conjugated to HRP (Bio-Rad) was applied at a dilution of 1:10,000 for 1 hour. Proteins were detected using the Renaissance chemiluminescence reagent (NEN).

Reverse transcriptase (RT)-PCR was used to compare levels of p67 and p96 mRNA in embryos and cells. Tissue was snap-frozen on dry ice and the RNeasy kit (Qiagen) was used to isolate mRNA. cDNA was generated using the Superscript II RT kit (Gibco, Invitrogen) and PCR was performed using oligonucleotides designed to specifically detect either the p67 or p96 cDNA (Fig. 4C). A common 5′ oligonucleotide was used (5′-CCTGATGCTCGAGGAGACAAAATG) in combination with an oligonucleotide that recognized the p96-specific exon (5′-AGAACAGGAGGTGACTCCATTTGTTAAG) or an oligonucleotide that bridged the p67-specific splice junction (5′-CAAGTCGTTTGCTGAAGATGTTGG). PCR conditions were adjusted such that sensitivity of detection was equivalent for both isoforms, and gapdh levels were assessed by PCR as a loading control (5′ primer, 5′-AGTGGAGATTGTTGCCATCAACGACC; 3′ primer, 5′-GGACTGTGGTCATGAGCCCTTCC).

The `plug and socket' targeting strategy was chosen for inserting a cDNA encoding either p96 or p67 into the dab2 locus (Fiering et al., 1999). A genomic clone containing the first two coding exons of Dab2 from a 129Sv mouse library (kind gift from P. Soriano, Fred Hutchinson Cancer Research Center, Seattle, WA) has previously been isolated (Morris et al., 2002b). The socket targeting vector was constructed by subcloning a 1 kb SmaI-EcoRI and a 2.2 kb EcoRI-ApaI genomic fragment as the short and long recombination arms into BSKSII. The diptheria toxin gene was PCR-amplified and inserted into the NotI site upstream of the recombination arms in BSKSII. A fragment containing the 3′ portion of the neomycin-resistance cassette and the hygromycin-resistance cassette flanked by loxP sites was PCR-amplified and inserted into the EcoRI site between the short and long recombination arms. The plug targeting constructs were generated by subcloning a 2 kb BglII-EcoRI genomic fragment into BSKSII upstream of a loxP site followed by the 5′ portion of a neomycin resistance-cassette, which provides the 3′ recombination arm. The SV40 (pA)₃ sequence was added to the 3′ UTR of both the p67 and p96 cDNA before inserting the entire cDNA-(pA)₃ fragment into coding exon 2 of the dab2 genomic sequence within the plug targeting construct using XhoI and BamHI.

AK7 embryonic stem (ES) cells were electroporated with 25 μg of ApaI-linearized socket-targeting vector. Selection (150 μg/ml hygromycin B) was applied 24 hours after electroporation and resistant clones were picked 12 days later, screened by PCR and verified by Southern blotting. Socket clones were electroporated with 25 μg of NaeI-linearized p96 or p67 plug-targeting vector. Selection (225 μg/ml G418) was added 24 hours post-electroporation, and resistant colonies were picked 12 days later. PCR and Southern blotting were used to identify correctly targeted clones that were then electroporated with pPGKCre-bpA (kind gift from P. Soriano) to excise drug-resistance markers. Recombined clones were injected into C57BL/6 host blastocysts and chimeric males were mated to C57BL/6 females to generate mice heterozygous for either the p96 (dab2^p96) or p67 (dab2^p67) allele. Genotyping of embryos and P10 pups was performed by PCR using four primers: one sense primer (5′-GACCACGCTGTCCTTGAACTCAG) was used for all alleles, in combination with an antisense primer for detecting the wild-type allele (5′-GGTGCCCAACATCCTAGTCCCTAG), the p96 allele (5′-AGAACAGGAGGTGACTCCATTTGTTAAG), or the p67 allele (5′-GTCGTTTGCAGAGCTCGTTGG). Mice were maintained on a mixed C57BL/6 × 129Sv genetic background.

To investigate the function of dab2 in the VE, we made use of a previously-described dab2 knockout mouse line (Morris et al., 2002b). Homozygous dab2^-/- embryos arrest development at about embryonic day 6 (E6), with an intact VE. dab2^-/- blastocysts fail to grow when cultured in vitro. Reduced growth and increased apoptosis of dab2^-/- embryos might result from decreased nutrient transport across the VE (Morris et al., 2002b). Consistent with this, EM examination of E6.5 control embryos revealed numerous electron-dense vesicles in the cytoplasm of VE cells (Fig. 1A), whereas in dab2^-/- VE these vesicles were retained along the apical plasma membrane at the base of the brush border microvilli and were absent deeper in the cell (Fig. 1B). Accumulation of these particles at the membrane suggests that movement of materials into the VE from the yolk sac cavity is reduced by dab2 mutation.

To assay for general endocytic activity, we isolated embryos at E7.5 and incubated them with HRP, a marker for fluid-phase pinocytosis. By E7.5, distal VE cells have been displaced by definitive endoderm in wild-type embryos but not in dab2^-/- embryos (Morris et al., 2002b). HRP was actively taken up by VE in the proximal region of control embryos, whereas the definitive endoderm in the distal region was less active (Fig. 1C). By contrast, both proximal and distal VE cells of dab2^-/- embryos exhibited low HRP uptake (Fig. 1C), implying that Dab2 promotes fluid-phase pinocytosis.

To determine whether clathrin-mediated endocytosis was similarly affected by dab2 mutation, we measured the uptake of Tf by dab2^-/- versus control embryos. Live embryos were incubated with TR-Tf, acid-stripped to remove surface-bound Tf and then fixed and visualized by fluorescent microscopy. Afterwards, the embryos were genotyped. TR-Tf uptake by the VE of dab2^-/- embryos was decreased relative to wild-type embryos at both E7.5 (Fig. 1D) and at E6.5, when control and mutant embryos were morphologically similar (Fig. 1E,G). Endosomal labeling was decreased by 66% (P<0.001) in E6.5 dab2^-/- embryos (Fig. 1F,F',H,H'; see Materials and Methods), suggesting that Dab2 regulates receptor-mediated endocytosis in the VE.

Although Tf is taken up by E6.5 embryos, the Tf receptor is not expressed at this stage (Drake and Head, 1990). In polarized epithelia, a complex containing the scavenger receptor cubilin and the lipoprotein receptor megalin can bind and internalize Tf (Kozyraki et al., 2001). Given that both of these receptors are expressed in the VE (Drake et al., 2004) and that Dab2 regulates uptake of megalin ligands in the kidney (Morris et al., 2002b), reduced Tf internalization in dab2^-/- embryos might result from defective endocytosis of megalin and cubilin. Therefore, we examined the subcellular localization of megalin and cubilin in control and mutant E6.5 embryos (Fig. 2).

Immunohistochemistry and immunofluorescence detected Dab2 on the apical brush-border and internal vesicles of wild-type embryos (Fig. 2A-B). The internal vesicles were heterogeneous in size and shape, and could correspond to coated and uncoated vesicles visible by electron microscopy (Fig. 1A). Some of the Dab2 colocalized with EEA1, a marker for early endosomes (Fig. 2B). Dab2 knockout embryos were identified by absence of staining for Dab2 (Fig. 2C-D). EEA1-positive vesicles were reduced in dab2^-/- embryos (Fig. 2D). Megalin and cubilin were detected in the VE of control embryos on the apical brush-border and in apical and basolateral vesicles (Fig. 2E-F and 2I-J). Some megalin colocalized with EEA1 (Fig. 2F). In dab2 knockouts, however, both megalin and cubilin were restricted to the apical surface and very little protein was detected in the cell body (Fig. 2G-H and K-L). Colocalization of megalin and EEA1 was lost (Fig. 2H). A similar redistribution of megalin was recently reported in proximal tubule epithelial cells of conditional dab2^-/- mice (Nagai et al., 2005). Accumulation of these receptors at the apical surface and reduced colocalization with EEA1 suggests that Dab2 is required for their efficient internalization into clathrin coated pits and subsequent movement to endosomes. This supports a model wherein Tf uptake is decreased in dab2^-/- embryos because endocytosis of the cubilin-megalin complex depends on Dab2.

The dramatic reduction of lipoprotein receptor endocytosis in dab2^-/- mutant VE suggests that other adaptor proteins are absent. One such candidate is the adaptor protein ARH, which is responsible for LDLR trafficking in hepatocytes. We examined ARH expression in the pregastrula embryo. Embryos were isolated at E6.5, and ARH and Dab2 protein levels were assessed by immunofluorescence. Dab2 protein was detected in the VE as expected (Fig. 3A). ARH was absent from the VE, but was detected in the parietal endoderm (PE) (Fig. 3B), which overlays the VE. The PE is tightly apposed to the apical surface of the VE, as seen in the overlay of Dab2 and ARH images (Fig. 3C). Absence of ARH from the VE suggests that Dab2 may be the main adaptor for megalin uptake in the VE.

Dab2 is expressed as two alternatively spliced forms, p96 and p67 (Fig. 4A). We examined the isoform-specific expression pattern of Dab2 in the embryo as well as in adult tissues. In E6.5 and E7.5 embryos, p67 was approximately ninefold more abundant, as measured by protein (Fig. 4B, lane 1) and mRNA levels (Fig. 4C). High-resolution deconvolution microscopy showed that Dab2 is present in both membrane-associated and diffuse regions of E6.5 VE cells, as expected if p67 is the major isoform present (Fig. 4D). p67 was also the predominant splice form when ES cells and F9 embryonic teratocarcinoma cells were differentiated into embryoid bodies (EBs) that contained an outer VE-like layer (Cho et al., 1999) (Fig. 4B, lanes 2,3), whereas in kidney and fibroblasts p96 was equally or more abundant than p67 (Fig. 4B,C). The tissue-specific expression pattern of the Dab2 isoforms suggests that they possess non-overlapping cellular activities.

To determine whether the p96-specific exon of Dab2 is important for endocytosis during embryonic development, mice were engineered to express only the p67 or p96 splice form of dab2 (Fig. 5A, see Materials and Methods). Heterozygous p67/+ or p96/+ mice were phenotypically indistinguishable from wild-type littermates and were intercrossed to generate homozygotes. Expression from the targeted allele was assessed by western blotting of protein extracts from tails of postnatal day (P) 11 mice (Fig. 5B). Only the expected isoform was expressed in homozygous knock-in mice. The protein level resulting from the p96 allele was less than 50% of the level of p96 from the wild-type allele (Fig. 5B; lanes 1-4), whereas expression from the p67 allele was near endogenous levels (Fig. 5B; lanes 5-9).

The subcellular distributions of p96 and p67 have previously been compared in tissue culture fibroblasts (Morris and Cooper, 2001). To compare the subcellular localization of the two Dab2 splice forms in vivo in a polarized epithelium, immunofluorescence was performed on kidneys from dab2 wild-type, knockout, p67/p67 and p96/p96 mice. Endogenous Dab2 localized to the apical brush-border of the proximal tubule epithelium (Fig. 5C). Dab2 staining was absent in the kidney of a knockout mouse (Fig. 5D). In p96/p96 kidney, the overall level of Dab2 was reduced, but the subcellular localization was similar to that of endogenous Dab2 (Fig. 5E). In the kidney of a p67/p67 mouse, however, Dab2 was more diffusely localized (Fig. 5F). This confirms that localization of Dab2 to the endocytic apparatus requires the p96-specific exon.

Expression of the p96 allele was sufficient to rescue the phenotype of dab2^-/- mice. p96/p96 pups were recovered at the expected mendelian frequency from p96/+ intercrosses at both developmental and postnatal stages (Fig. 6A). Unexpectedly, we found the viability of p96 homozygotes declined sharply when the dam was concurrently lactating and pregnant (data not shown). During concurrent lactation and gestation, circulating levels of maternal nutrients available for developing embryos is reduced (Johnson et al., 2001). This implies that either the low level of p96 or the absence of p67 in p96 homozygotes (Fig. 5B) results in embryonic lethality under low nutrient conditions.

Tf uptake was examined in E6.5 embryos from p96/+ heterozygous intercrosses (Fig. 7). Internalization of TR-Tf was rescued completely in p96 homozygotes at E6.5. p96/p96 embryos exhibited robust uptake, with values similar to +/+ and p96/+ embryos (Fig. 7A-E). High magnification images showed similar labeling of endosomes in p96 homozygous and control embryos (Fig. 7B',D').

Viability of p67/p67 embryos was compared to dab2^-/- and p96/p96 embryos. At E6.5 and E10.5, p67/p67 embryos were recovered at mendelian frequencies (Fig. 6B). However, although p67 is the predominant isoform expressed in the VE, fewer than 50% of the expected number of homozygous p67/p67 pups were obtained at both P1 and P10 (Fig. 6B). Therefore, p67 rescues the gastrulation defect of dab2^-/- embryos, but approximately half of the p67/p67 embryos were lost later in gestation, between E10.5 and P1. Overall reduced viability of p67/p67 embryos suggests that the p96 isoform, which contains binding sites for major endocytic proteins, fulfills the essential role of Dab2 in the VE better than p67, which lacks these binding sites.

To determine whether p67 can participate in endocytosis in the VE, TR-Tf uptake assays were performed at E6.5. Tf uptake was reduced by 50% in p67/p67 embryos relative to wild-type or p67/+ embryos (P<0.005) (Fig. 7E). Thus the abundant p67 isoform of Dab2 is inefficient at mediating uptake by the VE. Reduced endocytosis at E6.5 by p67/p67 embryos correlates with and might be the cause of later reduced embryonic viability.

Despite impaired Tf uptake by the VE of homozygous E6.5 embryos, the p67 isoform supported development beyond the time when dab2 knockouts die. This suggests either that, p67 has a second function (not evident in the endocytosis assay) that supports post-gastrulation development, or that the partial endocytosis rescue by p67 is sufficient to support development through gastrulation. However, after E10.5, when approximately 50% of p67/p67 embryos were lost, the remaining p67/p67 embryos were smaller and developmentally delayed compared with littermates (Fig. 8A). Reduced size was still apparent postnatally, because p67 homozygotes weighed significantly less than littermates in adulthood (data not shown). Resorbed embryos were found at E16.5, suggesting that nonviable homozygotes arrest by this point of development. These results suggest continuing defects in p67/p67 embryos throughout development.

Growth retardation and lethality of late stage p67 homozygotes might result from defective trafficking of proteins across the visceral yolk sac (VYS). The VYS forms from VE and extra-embryonic mesoderm, and continues to supply the embryo with maternal factors. To test whether megalin and/or cubilin traffic is defective in VYS, we examined the subcellular localization of Dab2 and cubilin in the VYS of control and p67/p67 embryos at E11.5. In control embryos, Dab2 was associated with vesicles within the apical and basolateral region of VYS cells (Fig. 8B). Presumably this is mostly p96, because Dab2 staining was more diffuse in the VYS of p67 homozygotes (Fig. 8C). cubilin staining in wild-type embryos was also vesicle-associated in the apical and basolateral regions of the VYS (Fig. 8D). In the VYS of p67 homozygotes, however, cubilin was restricted to the apical cell surface, and little staining was visible within the cell or associated with vesicles (Fig. 8E). This suggests that p67 is inefficient at mediating the trafficking of cubilin and megalin in the VYS, consistent with the reduced growth and viability of embryos that only express p67.

Our results provide genetic evidence that Dab2 is necessary in cells of the VE and VYS for endocytosis of megalin. This requirement might be evident because ARH is not present (Fig. 3). Deletion of dab2 causes both megalin and its co-receptor cubilin to accumulate at the apical plasma membrane (Fig. 2). Fluid-phase pinocytosis and cubilin-dependent uptake of Tf are decreased in the VE of early dab2^-/- embryos (Fig. 1), at the time when dab2 is essential for development. In addition, coated vesicles are restricted to the apical surface when Dab2 is absent. Both splice forms of Dab2 rescue the gastrulation defect of dab2^-/- embryos (Fig. 6). However, the minor p96 splice form of Dab2, which binds to clathrin and AP2, is more efficient at rescuing embryonic development and endocytosis than the predominant p67 isoform, which does not bind clathrin or AP2 (Mishra et al., 2002; Morris and Cooper, 2001) (Figs 6, 7). Late gestation p67 homozygotes are smaller than littermates throughout development, and fewer than 50% are viable (Figs 6, 8). Cubilin accumulates on the apical surface of the VYS in mid-gestation p67 homozygotes, implying that Dab2 p96 is important for efficient receptor endocytosis in both the VE and VYS (Fig. 8).

These findings suggest a model for Dab2 function in polarized epithelial cells of the VE and VYS (Fig. 8F). Dab2 binds to internalization motifs within the cytoplasmic tails of specific receptors (Morris and Cooper, 2001; Oleinikov et al., 2000). By simultaneously interacting with endocytic proteins, AP2 and clathrin, the p96 isoform recruits receptors into nascent clathrin-coated pits (Fig. 8F, step 1). Although the p67 splice form lacks the AP2 and clathrin binding exon, this isoform still contains endocytic motifs and has partial function in vivo. Both p67 and p96 bind Myosin VI (Inoue et al., 2002; Morris et al., 2002a), a minus end-directed actin motor that is recruited to clathrin-coated vesicles by Dab2 and has been implicated in endocytic trafficking in polarized epithelial cells (Buss et al., 2001; Dance, 2004; Hasson, 2003). The interaction between Dab2 and Myosin VI may facilitate the movement of lipoprotein receptor-containing vesicles towards the cell body, and may help explain the partial function of p67 (step 2). Megalin mediates transcytosis across epithelia and, given the association of Dab2 with vesicles in both the apical and basolateral regions of the VE, Dab2 might facilitate trafficking of the megalin-cubilin protein complex to the basolateral membrane where cargo is released to the embryo (step 3). Molecules trafficked in this way probably include nutrients and morphogens.

The intracellular Dab2- and megalin-containing vesicles could in principle include secretory vesicles. In other systems, post-Golgi transport of lipoprotein receptors might involve adaptor proteins. For example, trafficking of the LDL receptor to the basolateral surface of epithelial cells requires its FxNPxY signal (Matter et al., 1992), and lipoprotein receptor-dependent secretion from C. elegans cells requires the C. elegans Dab2 homolog (Kamikura and Cooper, 2003). However, the observation that megalin and cubilin accumulate on the apical surface implies that their traffic to the apical cell surface does not require Dab2.

Prior to placental formation, transport across the VE is the only route by which maternal proteins are transferred to the developing embryo. Our results suggest that Dab2 is required for endocytosis of megalin and cubilin in the VE. Indeed, dab2 deletion causes patterning defects that might be secondary to reduced levels of various morphogens or growth factors (Morris et al., 2002b). Together, megalin and cubilin mediate the internalization of over 40 different ligands, including lipoproteins, Tf, transthyretin, retinol-binding protein, and morphogens such as sonic hedgehog and BMP4 (Christensen and Birn, 2002; McCarthy et al., 2002; Spoelgen et al., 2005). Some of these ligands carry cofactors that are essential for basic cellular processes, such as cholesterol for membrane biosynthesis. Others carry growth regulators, such as retinol (converted to retinoic acid in the VE) (Bohnsack et al., 2004) and thyroid hormone. Defective transport of these proteins to dab2^-/- embryos might contribute to the death of these embryos.

Like Dab2, the lipoprotein receptors LRP1 and megalin are essential for development (Herz et al., 1992; Herz et al., 1993; Spoelgen et al., 2005; Willnow et al., 1996). However, megalin knockouts progress beyond E6.0 when the dab2^-/- embryos arrest (Willnow et al., 1996), and LRP knockouts arrest at various stages during development (Herz et al., 1992; Herz et al., 1993). It is possible that Dab2 mediates endocytosis of both receptors, accounting for the more severe phenotype. cubilin is also likely to be essential, based on studies of amnionless (Amn). Amn mutants fail to route cubilin to the apical surface of the VE and are lethal by E8.5 (Strope et al., 2004). The early lethality of dab2 knockouts implies that there may be other unidentified receptors that rely on Dab2 for internalization.

Later in development, transport continues across the VYS that is composed of VE and extraembryonic mesoderm. dab2 expression is maintained in the VYS late into development and we have shown that Dab2 p96 is required for cubilin uptake across the VYS. The small size and developmental delay of p67/p67 embryos implies that this later role is also critical. The importance of transport across the VYS is apparent in embryos that lack another protein, Apolipoprotein B (ApoB), involved in lipid transport. ApoB is required in the VYS for repackaging of maternally derived lipids for export to the embryo. In the absence of ApoB, maternal lipids accumulate within the VYS and the majority of embryos die by mid-gestation (Farese et al., 1996; Farese et al., 1995). apoB^-/- embryos are runted and the few that survive into late development display exencephaly or other neural tube defects. A small number of p67 homozygotes exhibited exencephaly, further supporting the idea that Dab2 is essential for the transport of lipoprotein receptors in the VYS.

Our data show that the less predominant isoform of Dab2 expressed in the VE, p96, has an essential role in receptor trafficking and nutrient uptake. The dramatic endocytic defect in the VE of dab2^-/- embryos contrasts with the mild phenotype of mice with conditional deletion of the dab2 gene from most somatic cells (Morris et al., 2002b). These mice have a minor kidney defect that resembles, but is milder than, the one caused by megalin deletion, suggesting a role for Dab2 in megalin endocytosis at the apical membrane in kidney epithelium (Willnow et al., 1996). Apart from the VE and kidney proximal-tubule epithelium, there is no apparent requirement for Dab2 in other tissues. This suggests that other molecules might substitute for Dab2. ARH, which mediates basolateral endocytosis of the LDL receptor in the liver, is one candidate. In the absence of ARH, LDLR is restricted to the sinusoidal (basolateral) surface of hepatocytes, similar to the apical restriction of megalin in dab2^-/- VE (Jones et al., 2003). An intriguing difference between ARH and Dab2 might be the ability to mediate basolateral and apical endocytosis, respectively. In nonpolarized cells, such as fibroblasts, these adaptors might be redundant, whereas in polarized tissues such as VE and liver, specific requirements have now been revealed for both genes.

We thank R. Nielsen, M. Farquhar and P. Soriano for reagents, P. Kronstad-O'Brien for animal husbandry; M. Bender for constructs and advice concerning the plug and socket targeting strategy; and N. Jiang for technical assistance. We also thank A. Davy, D. Kamikura, S. Onrust, C. Moens and S. Parkhurst for critical reading of this manuscript, and E. Robertson, J. Rossant and E. Lacy for advice. This work was supported by National Institute of Health grant GM066257. M.E.M. was supported by a National Science Foundation Predoctoral Fellowship and the Cell and Molecular Biology training grant GM07270.