Abstract
Abstract. The transmembrane endocytic receptor glycoprotein 330/megalin (hereafter referred to as megalin) is localized to the apical membrane domain of epithelial cells, where it is involved in the uptake of proteins from extracellular sources. The cytoplasmic domain of megalin contains amino acid motifs that have the potential to bind to other proteins, which may influence its localization or function. The yeast two-hybrid system was used to search for proteins that bind to the cytoplasmic tail of megalin, and a protein fragment from a mouse embryonic cDNA library that contained a single PDZ domain was identified. This protein, which was named glycoprotein 330-associated protein (GASP), appears to be a truncated mouse counterpart of the human and rat proteins atrophin-1-interacting protein-1 and synaptic scaffolding molecule, respectively. The interaction of GASP with megalin is mediated by the PDZ domain of GASP binding to the DSDV motif found at the carboxyl-terminus of megalin. A mutant version of megalin that lacks the terminal valine is unable to bind to GASP, illustrating the PDZ domain-dependent interaction between these two proteins. A close homolog of GASP, i.e., membrane-associated guanylate kinase with inverted orientation-1 (MAGI-1), is more ubiquitous in its tissue distribution (including kidney) and is also able to specifically bind to megalin via its fifth PDZ domain. Immunofluorescence studies of adult kidney revealed that MAGI-1 is expressed in the glomerulus of the kidney, in a manner that parallels the expression of the podocyte-specific protein glomerular epithelial protein 1. Western analysis of endogenous MAGI-1 from glomerular preparations suggests that it is associated with the cytoskeleton and seems to be expressed in a different form, compared with cell line-derived endogenous MAGI-1. The association of megalin with MAGI-1 may allow the assembly of a multiprotein complex, in which megalin may serve a nonendocytic function in glomerular podocytes.
The reabsorption of proteins that are filtered in the renal glomeruli takes place primarily in the renal proximal tubules. Because of this uptake, mammalian urine contains only trace amounts of filtered proteins under normal physiologic conditions. Reabsorption in the proximal tubules is mediated by several receptors but is predominantly performed by the transmembrane glycoprotein megalin. Although it was originally identified as one of the major antigens involved in Heymann nephritis, which is a rat model of human membranous glomerulonephritis (1), megalin is also expressed in epithelial cells of a number of other tissues, including brain, lung, epididymis, yolk sac, parathyroid glands, and thyroid glands (2,3). In addition to the proximal tubules, megalin is expressed in the glomerular visceral epithelial cells, or podocytes, of the kidney, although at much lower levels. The cloning of megalin revealed that it is a member of the LDL receptor (LDLR) family (4,5) and, in general, exhibits a restricted pattern of apical domain localization within polarized epithelial cells. Mice that are deficient in megalin expression exhibit abnormalities in epithelial tissues of the lung and kidney, and most die perinatally (6). However, the most distinguishing features of megalin-knockout animals are the numerous abnormalities observed in the brain, which are characteristic of holoprosencephalic syndrome. This syndrome is attributable to impaired proliferation of the neuroepithelium, where megalin is highly expressed on the apical surface (6,7). It is hypothesized that this phenotype results from cellular starvation for cholesterol and fat-soluble vitamins during the time between gastrulation and neural tube closure, when the neuroepithelium may rely on megalin to take up these nutrients from the amniotic fluid. Accordingly, it is important to understand the mechanisms used by epithelial cells that target megalin to the surfaces that expose it to its ligands.
The LDLR family members contain NPXY (where N is arginine, P is proline, X is any amino acid, and Y is tyrosine) motifs in their cytoplasmic domains, which have the potential to bind to a modular protein-protein interaction domain known as the phosphotyrosine binding (PTB) domain (8). In addition, megalin and a Caenorhabditis elegans homolog are the only known LDLR family members to have an SXV (where S is serine, X is any amino acid, and V is valine) stretch of amino acids at their carboxyl-termini. This sequence of amino acids can serve as a binding motif for another protein-protein interaction domain, referred to as the PDZ domain (9). Furthermore, these binding motifs are conserved in all species from which megalin has been isolated to date.
The 80- to 90-amino acid PDZ domain was first identified in the postsynaptic protein PSD-95, the Drosophila dlg tumor suppressor gene, and the tight-junction protein ZO-1. PDZ domains are currently recognized as central organizers of protein complexes at the plasma membrane. These protein complexes mediate the adhesive properties of particular cells, the formation of the paracellular barrier (tight junctions), ion transport, and transmission of signals that regulate growth, differentiation, and homeostasis between adjacent cells. The asymmetrical localization of a number of transmembrane receptors and ion channels has recently been shown to be a consequence of their association with proteins containing PDZ domains (10,11). For example, in C. elegans, the proper localization of the receptor tyrosine kinase Let-23 to the basolateral surface of vulval precursor cells is dependent on a complex of three PDZ proteins, i.e., Lin-2, Lin-7, and Lin-10. The PDZ domain of Lin-7 binds directly to the carboxyl-terminus of Let-23, and mutations in lin-7, as well as in lin-2 or lin-10, result in mislocalization of Let-23 to the apical domain of the plasma membrane (12). The binding of class I PDZ domains is mediated by a (S/T)X(V/L) (where T is threonine and L is leucine) motif found at the carboxyl-termini of their cognate binding partners (9).
To isolate potential PTB or PDZ domain proteins that might bind megalin, we used the yeast two-hybrid system to screen an embryonic mouse brain cDNA library, using most of the cytoplasmic tail of megalin as bait. We report the isolation of a partial cDNA clone that contains a single PDZ domain; we named the protein GASP (glycoprotein 330/megalin-associated protein). GASP cDNA seems to be a truncated version of a larger mouse protein that shares a high degree of identity with human atrophin-1-interacting protein-1 (AIP-1) and rat synaptic scaffolding molecule (S-SCAM). The PDZ domain of GASP is also similar to the fifth PDZ domain of the mouse protein membrane-associated guanylate kinase 1 (MAGI-1) (13). We show, using biochemical techniques, that GASP cDNA and the fifth PDZ domain of MAGI-1 bind directly to the carboxyl-terminus of megalin. In addition, we show that MAGI-1 is expressed in a podocyte-specific manner within kidney glomeruli. We speculate that these associations with megalin are potentially important for the assembly of a multiprotein complex at subcellular locations in glomerular podocytes.
Materials and Methods
Expression Plasmid Construction
The bait construct used for the yeast two-hybrid screening consisted of the carboxyl-terminal 146 amino acids of mouse megalin and was obtained by PCR amplification of an expressed sequence tag (EST) cDNA using mouse megalin-specific primers containing NcoI (5′) and SalI (3′) restriction sites. This segment of megalin contains the three NPXY motifs and the DSDV motif at the carboxyl-terminus. The PCR amplicon was verified by DNA sequencing and subcloned into pAS2-1 (Clontech, Palo Alto, CA), downstream of the GAL4 DNA-binding domain. A mutant cytoplasmic tail for megalin with the carboxyl-terminal valine deleted (megalin ▵V) was generated by PCR, sequenced, and cloned into pAS2-1 in a manner similar to that used for the wild-type (WT) megalin tail.
Glutathione-S-transferase (GST) fusion proteins were constructed in pGSTag (14) using the following inserts: the entire cDNA insert of mbp1b excised from pGAD1318 (see below) (GST-GASP5), a mouse EST cDNA containing the carboxyl-terminal half of the fourth PDZ domain and the entire fifth PDZ domain of MAGI-1 (GST-MAG15), and the WT and mutant ▵V cytoplasmic tails of megalin used in the yeast two-hybrid screening (GST-megalin WT and GST-megalin ▵V, respectively). In addition, complementary sense and antisense oligonucleotides coding for the last 19 amino acids of N-methyl-D-aspartate receptor 2B (NR2B) and the activin type IIA receptor (ActRIIA) were annealed and cloned into pGSTag, yielding GST-NR2B and GST-ActRIIA, respectively. Myc-tagged expression constructs were prepared in pRK5-Myc (15). An EcoRI fragment containing the entire coding sequence of clone mpb1b was cloned into pRK5-Myc to yield Myc-GASP5. The EST cDNA of MAGI-1 (see above) was cloned into pRK5-Myc to yield Myc-MAGI5. MAGI-1 PCR products containing either a variant of the fourth PDZ domain, the first three PDZ domains, or full-length MAGI-1 were cloned into pRK5-Myc to generate Myc-MAGI4b, Myc-MAGI1-3, and Myc-MAGI FL, respectively. The GASP cDNA and EST cDNA of MAGI-1 were subcloned into the plasmid pET28c+ (Novagen, Madison, WI) for the creation of His6-tagged proteins.
Yeast Two-Hybrid Screening
Yeast strain Y190 was transformed with the GAL4-megalin bait construct (see above) and an oligo(dT)-primed embryonic mouse brain cDNA library that was cloned into pGAD1318 (a kind gift from Dr. J. Camonis, Institut Curie, INSERM U-248, Paris, France) downstream of the GAL4 activation domain. A total of 3 × 107 colonies were screened on Leu-Trp-His selection plates. Positive clones that survived additional rounds of selection for histidine and LacZ expression were isolated and grown in liquid selection medium for plasmid DNA isolation. Plasmid DNA isolated from yeast was used to transform DH5α bacteria for subsequent plasmid amplification, purification, and sequencing. The megalin ▵V mutant construct in pAS2-1 was used to confirm the interaction of the positive clones after their nucleotide sequences were established.
Cell Culture and Transfections
Human embryonic kidney 293 (HEK293) cells and MDM cells (a rabbit kidney epithelial cell line provided by Dr. David Kershaw, University of Michigan, Ann Arbor, MI) were grown in Dulbecco's modified Eagle's medium (Life Technologies, Grand Island, NY) supplemented with 10% fetal calf serum, 2 mM L-glutamine, 100 U/ml penicillin G, and 100 μg/ml streptomycin.
HEK293 cells at approximately 50% confluence were transiently transfected with the Myc-tagged expression constructs, using the calcium phosphate precipitation method. Twenty hours after the addition of DNA, the cells were washed once with phosphate-buffered saline (PBS) and fresh medium was added. The cells were incubated for an additional 24 h before collection for lysate preparation.
Tissue and Cell Lysates
Tissue lysates were prepared as follows. Adult mouse tissues were placed in a Dounce homogenizer with Triton lysis buffer [50 mM Hepes, pH 7.5, 150 mM NaCl, 1.5 mM MgCl2, 1 mM ethylene glycol bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid, 10% glycerol, 1% Triton X-100] plus protease inhibitors (4 μg/ml phenylmethylsulfonyl fluoride, 10 μg/ml aprotinin, and 10 μg/ml leupeptin) and were homogenized with 25 to 30 strokes. The lysates were centrifuged at 20,000 × g for 30 min at 4°C. Supernatants were transferred to fresh tubes, and the pellets were resuspended in a volume of lysis buffer equal to the volume of supernatant.
Glomerular extracts were prepared by passing mouse kidneys that had been cut in half through a Nitex mesh with a pore size of 90 μm (Sefar America, Kansas City, MO), using ice-cold PBS. The flow-through fraction was then passed over Nitex mesh with a pore size of 46 μm, and the retained glomeruli were washed extensively with ice-cold PBS. Glomeruli were transferred to a 50-ml tube and centrifuged at 2000 × g for 3 min at 4°C. The pelleted glomeruli were resuspended in lysis buffer, sonicated with three 15-s bursts, and processed as described above.
Transiently transfected HEK293 cells in 60-mm dishes were washed once with PBS, scraped into 200 μl of Triton lysis buffer, transferred to microfuge tubes, briefly vortex-mixed, and incubated on ice for 10 min. The lysates were then processed as described above.
RNA (Northern) Blot Analysis
A 523-bp EcoRI-XbaI fragment of clone mbp1b was labeled with [32P]dCTP (NEN, Boston, MA) using the Rediprime II random prime labeling system (Amersham Pharmacia Biotech, Piscataway, NJ) and was used to probe a mouse multiple-tissue RNA blot (Clontech, Palo Alto, CA). The RNA blot was prehybridized in 10 ml of prehybridization/hybridization solution [5× SSC, 100 μg/ml herring sperm DNA, 50% formamide, 10× Denhardt's solution, 1.5% dextran sulfate, 0.1% sodium dodecyl sulfate (SDS)] for 5 h at 42°C. A total of 1 × 107 cpm of probe was added to the blot and hybridized overnight at 42°C. The blot was washed once with 2× SSC/0.1% SDS at 42°C for 20 min and then twice with 0.1× SSC/0.1% SDS at 42°C for 20 min each. The blot was exposed to x-ray film for 5 d at -70°C.
Reverse Transcription-PCR
Total RNA was isolated from a mouse glomerular preparation and used to synthesize first-strand cDNA with a 3′-RACE kit (Life Technologies, Rockville, MD). Specific primers for mouse MAGI-1 were used to PCR amplify different regions, which were subsequently cloned into pGEM-T Easy vector (Promega, Madison, WI) and sequenced. Individual PCR clones containing the different regions of MAGI-1 were either subcloned into pRK5-Myc or used to construct a full-length cDNA, using conventional DNA cloning techniques.
Antibodies
The polyclonal anti-GASP and anti-MAGI-1 antibodies used for immunoblotting and immunostaining were prepared by injecting rabbits with purified GST fusion proteins, i.e., GST-GASP5 and GST-MAG15, respectively. Anti-GASP antibody UM200 and anti-MAGI-1 antibody UM209 were affinity-purified using His6-tagged protein columns, as described previously (16). Rabbit polyclonal anti-glycoprotein 330 antibody was a gift from Dr. Joachim Herz (University of Texas Southwestern Medical Center, Dallas, TX). The anti-Myc antibody was a mouse monoclonal antibody (clone 9E10) obtained from mouse ascites fluid. The anti-glomerular epithelial cell membrane protein-tyrosine phosphatase 1 (GLEPP1) antibody was a mouse monoclonal against rat GLEPP1 (1B4). Horseradish peroxidase-conjugated sheep anti-mouse secondary antibodies were obtained from Amersham Pharmacia Biotech.
GST Fusion Protein Precipitation
GST fusion proteins were produced in DH5α bacteria as described previously (15). Tissue or cell lysates were combined with 10 to 20 μg of GST fusion protein bound to glutathione-agarose beads and were incubated on a rocker at room temperature for 2 h. The beads were then washed twice with ice-cold PBS. Forty microliters of 1× SDS sample buffer was added to the beads, and the mixture was maintained at 100°C for 5 min. Samples were subjected to SDS-polyacrylamide gel electrophoresis (PAGE), transferred to nitrocellulose membranes, and blotted with the appropriate primary antibodies and horseradish peroxidase-conjugated secondary antibodies. Blot were developed with chemiluminescence reagents (NEN, Boston, MA) and exposed to x-ray film.
Immunofluorescence Microscopy
Cryostat sections (5 μm) of rat kidney were fixed in acetone and blocked with 10% goat serum/10% rat serum in PBS. Sections were double-labeled by incubation with a mixture of affinity-purified polyclonal anti-MAGI-1 (UM209) and monoclonal anti-GLEPP1 primary antibodies for 1 h at room temperature. The sections were washed twice with PBS and incubated for 1 h at room temperature with a mixture of preabsorbed FITC-conjugated goat anti-rabbit IgG (1:50) and rhodamine-conjugated goat anti-mouse IgG (1:100) secondary antibodies. The sections were then washed with PBS, mounted, and photographed with a Nikon 807427 immunofluorescence microscope (Mager Scientific, Inc., Dexter, MI) using the Metamorph imaging program (Version 3.51; Universal Imaging Corp., Downingtown, PA).
Results
Because megalin contains binding motifs for PTB and PDZ domains and seems to be crucial for some aspects of normal development, a search for proteins that are potential binding partners for the cytoplasmic tail of megalin was undertaken. Using the yeast two-hybrid system, we screened a mouse embryonic brain cDNA library with the carboxyl-terminal 146 amino acids of megalin fused to the DNA-binding domain of GAL4 (as bait). A total of 3 × 107 yeast colonies were screened, and three clones survived additional rounds of selection. The three clones, mbp1b, mbp8b, and mbp9, contained inserts of equal sizes (approximately 3200 bp) and were found to be identical when approximately 300 bp of their 5′- and 3′-ends were sequenced. Clone mbp1b was sequenced in its entirety and was found to consist of 978 bp of coding sequence followed by 2182 bp of 3′-untranslated sequence, ending in a poly(A) tail (Figure 1B). The coding sequence of mbp1b contains a single PDZ domain, which could potentially bind to the SDV motif at the carboxyl-terminus of megalin. To verify this interaction, a mutant megalin tail with the carboxyl-terminal valine deleted was constructed in pAS2-1 and used in the yeast two-hybrid assay. Deletion of the valine from the carboxyl-terminus of megalin abolished the interaction with mbp1b (Figure 1A), which supports the idea that the interaction of megalin with mbp1b is mediated via the PDZ domain in the cDNA. We have named this protein GASP.
Yeast two-hybrid assay results and illustrations of the glycoprotein 330-associated protein (GASP) cDNA and protein and GASP amino acid alignment. (A) Yeast strain Y190 was transformed with the GASP cDNA (in pGAD1318) and either the megalin wildtype (WT) tail, a megalin mutant (▵V) tail, or an empty pAS2-1 vector. Double-transformants observed on Leu-Trp (-L-W) selection plates were streaked onto Leu-Trp plates, allowed to grow for 3 d at 30°C, and then replica-plated onto a Leu-Trp-His (-L-W-H) selection plate. The yeast were incubated on the triple-“dropout” plate at 30°C for 5 d, after which they were scored for growth, which is indicative of a positive interaction between the GAL4 fusion proteins of the two plasmids. The megalin WT tail was not transactivating by itself, because no growth was observed on Leu-His plates (data not shown). (B) This diagram of the GASP cDNA shows the PDZ domain (hatched oval) within the coding region, which is illustrated by the thick black line ending with the TGA termination codon. The 3′-untranslated sequence is denoted by the thin black line that ends with a poly(A) tail. The region of the 3′-untranslated sequence that was used as a Northern probe is indicated above the cDNA diagram. (C) The GASP protein was compared with its human counterpart, atrophin-1-interacting protein-1 (AIP-1). GASP is nearly identical to the region of AIP-1 encompassing the fifth PDZ domain and the carboxyl-terminus, except for a small gap in GASP at its amino-terminus and a shorter carboxyl-terminal tail (thick gray line). The guanylate kinase domain (stippled rectangle), WW domains (black squares), and PDZ domains (hatched ovals) are shown in AIP-1. (D) Amino acid alignment of GASP with AIP-1 and synaptic scaffolding molecule (S-SCAM) was performed by using the CLUSTAL W method of MEGALIGN (DNASTAR Inc., Madison, WI). Amino acids in AIP-1 and S-SCAM that are identical to those in GASP are highlighted in black. The asterisk above the tyrosine residue indicates the position at which the sequences diverge in the carboxyl-terminus. This site is homologous to the position at which different carboxyl-terminal tails of membrane-associated guanylate kinase 1 (MAGI-1) begin (13). Gaps in the alignment are represented by dashes. Letters in parentheses next to the names of the proteins indicate the species from which the proteins are derived (m, mouse; h, human; r, rat).
The 5′-end of the cloned GASP cDNA lacks a Kozak consensus sequence for translation initiation, indicating that the cDNA probably represents a truncated version of a presumably larger transcript. This was verified by searching nucleotide and protein databases, because GASP exhibited a high degree of identity with human and rat proteins referred to as AIP-1 and S-SCAM, respectively (17,18). GASP appears to represent the carboxyl-terminal end of a mouse counterpart to these two proteins (Figure 1C). Although the PDZ domain of GASP is identical at the amino acid level to the fifth PDZ domain of AIP-1 and S-SCAM, the sequences outside the PDZ domain seem to diverge somewhat, especially in the sequences downstream from the PDZ domain (Figure 1D). The site at which all three proteins diverge after the PDZ domain is analogous to the site at which alternative splicing results in three different carboxyl-termini in the closely related protein MAGI-1 (13). The PDZ domain of GASP exhibits 76% identity, at the amino acid level, with the fifth PDZ domain of MAGI-1 and brain-specific angiogenesis inhibitor-associated protein 1 (BAP-1), the human counterpart of MAGI-1 (19), and 50% similarity with the last PDZ domain in a C. elegans clone, K01A6.1.
The tissue distribution of GASP was investigated by Northern analysis of a mouse multiple-tissue blot. A 523-bp fragment from the 3′-untranslated region of GASP was used as the probe. GASP was found to be expressed predominantly in brain, with major transcript sizes of approximately 7.4, 6.5, and 3.0 kb (Figure 2). A low level of expression was detected in testis, and no appreciable expression was observed in the other tissues examined. This pattern of expression of GASP is similar to the reported expression of AIP-1 and S-SCAM (17,18).
Northern analysis of mouse tissues for GASP. A 523-bp fragment from the 3′-untranslated region of GASP was radiolabeled with 32P and used as a probe for a mouse multiple-tissue Northern blot. Major transcripts of approximately 7.4, 6.5, and 3.0 kb were detected predominantly in brain. Molecular sizes are indicated on the left, and the sizes of the transcripts are indicated on the right. The different tissues examined are indicated above the lanes. For assessment of the amount of RNA in each lane, the blot was stripped and reprobed with a 32P-labeled mouse β-actin probe, which is shown below the blot probed for GASP.
To further examine the interaction of GASP with megalin, GST fusion protein precipitation assays were performed. A GST fusion protein containing the GASP cDNA was produced in bacteria, bound to glutathione-agarose beads, and added to mouse kidney lysates to precipitate endogenous megalin. The GASP fusion protein was successful in precipitating megalin from kidney lysates, whereas GST alone was not (Figure 3A). In a reciprocal experiment, GST fusion proteins were produced with the WT and ▵V megalin tails, which had been used in the yeast two-hybrid assay, and were added to mouse brain lysates. The megalin WT fusion protein was capable of precipitating endogenous GASP from brain tissue (Figure 3B). In addition, the megalin WT fusion protein was able to precipitate GASP cDNA expressed as a Myc-tagged protein in HEK293 cells (Figure 3C). The megalin ▵V mutant was unable precipitate GASP from either the brain lysate or the lysate of HEK293 cells expressing Myc-tagged GASP cDNA. These data confirm that the PDZ domain within GASP mediates its interaction with megalin and appears to be specific for the fifth PDZ domain in endogenous full-length GASP.
Glutathione-S-transferase (GST) fusion protein precipitation of tissue and cell lysates. (A) Mouse kidney lysates were incubated for 2 h at room temperature with 10 to 20 μg of GST fusion proteins bound to glutathione-agarose beads. The beads were washed twice with phosphate-buffered saline, and proteins bound to the beads were eluted by incubation for 5 min at 100°C in the presence of nonreducing sodium dodecyl sulfate (SDS) sample buffer. Proteins were separated by 5% SDS-polyacrylamide gel electrophoresis (PAGE), transferred to a nitrocellulose membrane, and blotted with an anti-megalin antibody. GST and the GST fusion proteins are indicated above the respective lanes. (B) Mouse brain lysates were incubated with 10 to 20 μg of the indicated GST fusion proteins, as described above, and proteins were resolved by 7% SDS-PAGE. Affinity-purified anti-GASP antibodies were used to observe endogenous GASP on the blot. The last few amino acids of the GST-megalin fusion proteins are indicated below the blot. (C) Lysates from human embryonic kidney 293 (HEK293) cells transiently transfected with Myc-GASP were incubated with 10 to 20 μg of the indicated GST fusion proteins or GST alone. Proteins bound to the fusion proteins were separated by 12% SDS-PAGE, transferred to a nitrocellulose membrane, and blotted with an anti-Myc antibody. (D) Lysates from HEK293 cells transiently transfected with Myc-GASP were incubated with 10 to 20 μg of the indicated GST fusion proteins or GST alone. The amino acids fused to GST in the fusion proteins represent the last 19 amino acids of the indicated proteins. A comparison of the final eight amino acids from these proteins is indicated below the blot, in bold letters [megalin, activin receptor type II (Activin RIIA), and N-methyl-D-aspartate (NMDA) receptor type 2 (R2B)], together with splice variants of activin receptor type II (Activin RIIB) and NMDA receptor type 2 (NMDA R2A). The asterisks at the end of the amino acid sequences denote the termination codon for the respective endogenous proteins. Proteins bound to the fusion proteins were separated by 12% SDS-PAGE, transferred to a nitrocellulose membrane, and blotted with an anti-Myc antibody. In A to D, the lysate lanes represent 1% of the input used for the precipitation, and molecular mass markers (in kD) are indicated to the left of the blots.
The interaction of rat S-SCAM with proteins present at postsynaptic densities within neurons was recently demonstrated (18). One of the proteins found to interact with S-SCAM was NR2B, and the interaction was mediated by the fifth PDZ domain of S-SCAM. In addition, a recently identified mouse protein, which appears to be identical to GASP (except at the extreme carboxyl-terminus), was cloned by means of its interaction with ActRIIA (20). This interaction was reported to occur through the fifth PDZ domain of the mouse protein (referred to as ARIP-1). We tested both NR2B and ActRIIA for their abilities to bind to GASP. GST fusion proteins were constructed with the last 19 amino acids of NR2B, ActRIIA, and megalin and were used in precipitation assays with Myc-tagged GASP. The shorter version of the megalin carboxyl-terminal tail was able to bind to GASP, as was NR2B (Figure 3D). However, ActRIIA was not able to bind to GASP in this assay.
The interaction of megalin with GASP may have a significant function in the brain (see Discussion); however, our interest in the interaction of megalin with other proteins is focused primarily on the kidney and on how these interactions could potentially affect the localization and/or function of megalin. MAGI-1, a close homolog of GASP, is expressed in a more ubiquitous manner, including in kidney. In addition, the domain structure of MAGI-1 is identical to that of the AIP-1/S-SCAM proteins, and the fifth PDZ domain of MAGI-1 displays a high degree of similarity to the fifth PDZ domain in GASP/AIP-1/S-SCAM. Therefore, we wanted to examine a potential interaction between megalin and MAGI-1. To investigate this possibility, an EST cDNA that contained the carboxyl-terminal half of the fourth PDZ domain and the entire fifth PDZ domain of MAGI-1 was fused to GST and used in a precipitation assay with kidney lysates. This GST fusion protein, GST-MAG15, was able to precipitate endogenous megalin from kidney lysates in a manner similar to that of GASP (Figure 3A). This result prompted us to focus on the function of MAGI-1 with respect to its interaction with megalin and to more precisely determine the localization of MAGI-1 within the kidney.
The specificity of the carboxyl-terminus of megalin for the individual PDZ domains of MAGI-1 was investigated by expressing Myc-tagged regions of MAGI-1 in HEK293 cells and using the resulting cell lysates in GST fusion protein precipitation assays. Myc-tagged, PCR-generated regions of MAGI-1 containing either PDZ domains 1 to 3 (Myc-MAGI1-3) or PDZ domain 4 (Myc-MAGI4b) and a Myc-tagged version of the MAGI-1 EST cDNA (Myc-MAG15) (Figure 4A) were combined with GST fusion proteins with megalin WT or mutant ▵V tails. The fusion protein with the megalin WT tail was capable of precipitating Myc-MAG15, whereas the ▵V mutant was not (Figure 4B), illustrating that the interaction of megalin with MAGI-1, like that with GASP, was mediated via a PDZ domain. This interaction is apparently specific for the fifth PDZ domain of MAGI-1, inasmuch as neither Myc-MAGI1-3 nor Myc-MAGI4b was precipitated by the megalin WT tail (Figure 4B).
Specificity of megalin for the PDZ domains of MAGI-1. (A) The PDZ domain (hatched ovals)-containing constructs of MAGI-1 that were tested for their interactions with megalin are shown. The constructs MAGI1-3, MAGI4, and MAGI5 contain PDZ domains 1 to 3, 4b, and 5, respectively. (B) The constructs illustrated in A were transiently expressed as Myc-tagged proteins in HEK293 cells, and the resulting cell lysates were used in precipitation assays with the indicated GST fusion proteins or GST alone. Proteins bound to the fusion proteins were separated by 15% SDS-PAGE, transferred to nitrocellulose membranes, and blotted with an anti-Myc antibody. The lysate lanes represent 1% of the input used in the precipitation assays, and molecular mass markers (in kD) are indicated to the left of the blots.
Although MAGI-1 is apparently expressed in the kidney at the RNA level (13), it has not been further localized. To determine where MAGI-1 is expressed within the kidney, we performed immunofluorescence assays with adult rat kidney sections, using an affinity-purified polyclonal antibody directed against GST-MAGI5. MAGI-1 expression was observed almost exclusively in glomeruli of adult kidneys, where it paralleled the expression pattern of GLEPP1 (Figure 5). GLEPP1 expression in the kidney was previously observed specifically in podocytes (21), suggesting that MAGI-1 is also confined to podocytes.
Immunolocalization of MAGI-1 in rat kidney sections. Rat kidney sections were double-labeled with polyclonal anti-MAGI-1 and monoclonal anti-glomerular epithelial cell membrane protein-tyrosine phosphatase 1 (GLEPP1) antibodies. FITC-conjugated goat anti-rabbit Ig and rhodamine-conjugated goat anti-mouse Ig secondary antibodies were then used to detect MAGI-1 (green) and GLEPP1 (red), respectively. The degree of overlapping expression observed between MAGI-1 and GLEPP1 (yellow in the merge image) suggests that MAGI-1 is confined to the podocytes in its glomerular localization.
The nature of the glomerulus-expressed MAGI-1 was examined by Western analysis of glomerular preparations. Isolated glomeruli obtained from adult mouse kidneys were lysed in one of three different extraction buffers, and samples of the resulting supernatant and pellet fractions were analyzed for the presence of MAGI-1. Endogenous MAGI-1 from glomerular preparations was found to be insoluble in Triton lysis buffer; all of the protein was observed in the pellet fraction (Figure 6A). MAGI-1 was still partially observed in the pellet fraction with RIPA buffer, although most of the protein was soluble. As expected, MAGI-1 was completely soluble in 6 M urea lysis buffer. These results suggest that MAGI-1 is tightly associated in a complex with the cytoskeleton. It was also observed that MAGI-1 existed in three different forms in the glomerular preparations (Figure 6A), with the largest form migrating at approximately 185 to 190 kD. These glomerulus-derived forms of MAGI-1 were larger than the forms observed previously in lysates from MDCK cells (13). We also examined endogenous MAGI-1 in MDM cells (a rabbit kidney epithelial cell line). MDM cells form monolayers and express junctional proteins (e.g., ZO-1) in a manner similar to that of MDCK cells (Patrie KM, Margolis B, unpublished observations). We found that MDM cells contained three forms of MAGI-1, which migrated with sizes comparable to those observed in MDCK cells (Figure 6B). However, the forms of MAGI-1 observed in MDM cells migrated with apparent molecular sizes smaller than those of the forms observed in the mouse glomerular preparations. In addition, MAGI-1 from this cell line was soluble in Triton lysis buffer. It appears, therefore, that the expression of MAGI-1 differs not only in size but also in subcellular association, depending on the source examined. Although we have not precisely defined the subcellular localization of MAGI-1 within podocytes, MAGI-1 has been observed to colocalize with ZO-1 at apparent tight-junctional sites in MDM cells (Patrie KM, Margolis B, unpublished observations) and MDCK cells (Patrie KM, Margolis B, unpublished observations) (22).
Solubility of MAGI-1 from glomerular and cell line lysates. (A) Mouse glomerular preparations were lysed in three different lysis buffers, and equal volumes of the resulting supernatant (supe) and pellet fractions were separated by 7% SDS-PAGE, transferred to a nitrocellulose membrane, and blotted with the affinity-purified anti-MAGI-1 antibody UM209. Arrowheads to the right of the blot indicate the positions of the three forms of endogenous MAGI-1 that were observed. (B) Supernatant and pellet fractions from MDM cell lysates prepared in either Triton or RIPA lysis buffer were separated by 7% SDS-PAGE, transferred to nitrocellulose membranes, and blotted with the affinity-purified anti-MAGI-1 antibody UM209 (left). A blot from an immunoprecipitation of MDM cells using preimmune (PI) UM209 serum or affinity-purified anti-MAGI-1 UM209 antibodies (α-MAGI) is also shown (right). The three forms of MAGI-1 expressed in this cell line are clearly resolved and are indicated by the arrowheads to the right of the blot. In both A and B, the molecular mass markers (in kD) are indicated to the left of the blots.
The appearance of larger forms of MAGI-1 in glomerular preparations suggests that podocytes may express forms of MAGI-1 that are not found in cell lines or other tissues. We therefore wanted to clone a glomerulus-specific MAGI-1 cDNA. Using total RNA isolated from glomerular preparations, single-stranded cDNA was produced with an oligo(dT) primer, and different regions of MAGI-1 were then cloned by PCR. The individual regions of MAGI-1 were ligated together to produce a full-length MAGI-1 cDNA. During the course of this PCR-based cloning, we found that inserts from eight individual PCR clones derived from the carboxyl-terminus of MAGI-1 were the same size; when three of these clones were sequenced, they were found to be identical. The sequences of these clones coded for the variant c tail of MAGI-1 (13). The retrieval of only one of the three known splice variants previously observed for this region of MAGI-1 indicated to us that glomerular podocytes preferentially express this variant, compared with the other two known splice variants. Furthermore, the 3′-untranslated region of this glomerular variant c was not only shorter but also different in nucleotide sequence, compared with the published sequence. The significance of this difference is unknown at this time.
We also identified a number of splice variants in other regions of MAGI-1 that had not been previously reported. Figure 7 illustrates the three different clones obtained with primers in and around the fourth PDZ domain. A plausible explanation for alternative splicing within this PDZ domain may be to create PDZ domains with binding specificities for different target proteins. Other insertions in MAGI-1 were routinely observed and included an 84-bp insertion immediately preceding the third PDZ domain and a 36-bp insertion between the two WW domains (data not shown). The significance of these latter two insertions is not known at this time.
Splicing variants observed in the fourth PDZ domain of MAGI-1 from mouse glomerular cDNA. PCR was performed on mouse glomerular cDNA with primers flanking the fourth and fifth PDZ domains of MAGI-1. Three different-size bands were observed on an agarose gel and were then isolated, subcloned, and sequenced. The upper clone contains sequences that are identical to the published sequence for this region of MAGI-1, and we have designated the fourth PDZ domain in this clone as 4a. The sequences found in the middle and lower clones are identical to those in the upper clone except for the presence of 168-bp (dark gray) and 204-bp (black and dark gray) insertions, respectively. These insertions are in-frame and result in displacement of the carboxyl-terminal half of PDZ4a to a position in which it becomes mostly intervening sequence between the fourth and fifth PDZ domains of the middle and lower clones. The fourth PDZ domains of the middle and lower clones have been designated as 4b and 4c, respectively, and contain novel carboxyl-terminal halves. The insertion in the 4c version would presumably create a longer loop between β-strands βB and βC, compared with 4b. These three splice variants were also observed in PCR assays of adult mouse kidney and mouse embryonic cDNA libraries obtained from Clontech (Palo Alto, CA). Furthermore, the fourth PDZ domain of the human counterpart to MAGI-1, i.e., BAI-associated protein 1 (BAP-1), is nearly identical in amino acid sequence to the 4c splice variant. Therefore, it is likely that these observed insertions in the fourth PDZ domain of MAGI-1 represent legitimate splice variants.
Because the size of MAGI-1 observed in glomerular extracts was larger than those observed in cell lines, the full-length MAGI-1 cDNA that was constructed from the glomerular RNA consisted of clones containing the insertions mentioned above (together with a clone containing the PDZ4c splice variant), as well as the longer c version of the carboxyl-terminus. Lysates from HEK293 cells expressing the full-length MAGI-1 cDNA and glomerular extracts were analyzed by Western blotting. MAGI-1 cloned from glomerular RNA migrated at the same relative size as the largest form of MAGI-1 observed in glomerular extracts (Figure 8A). In addition, the full-length MAGI-1 was precipitated from HEK293 lysates by GST-megalin WT but not by the GST-megalin ΔV mutant (Figure 8B).
Expression of full-length MAGI-1 in HEK293 cells and its interaction with megalin. (A) A full-length cDNA construct of MAGI-1 was expressed as a Myc-tagged protein in HEK293 cells, and its mobility was compared with that of endogenous MAGI-1 from mouse glomerular lysates. Proteins from transfected HEK293 cell lysates and glomerular lysates were separated by 7% SDS-PAGE, transferred to a nitrocellulose membrane, and blotted with affinity-purified anti-MAGI-1 antibody. (B) HEK293 lysates expressing Myctagged full-length MAGI-1 were incubated with the indicated GST fusion proteins. Proteins bound to the fusion proteins were separated by 7% SDS-PAGE, transferred to a nitrocellulose membrane, and blotted with an anti-Myc antibody. The lysate lane represents 1% of the input used in the precipitation assays. Molecular mass markers (in kD) are indicated to the left of the blots in both A and B.
Discussion
Using yeast two-hybrid screening, we cloned a partial cDNA from a mouse embryonic brain library on the basis of its ability to bind to the carboxyl-terminal 146 amino acids of megalin. The interaction of the protein, which we have named GASP, with megalin could be abolished by deletion of the carboxyl-terminal valine from the consensus PDZ binding motif of megalin, indicating that this interaction is mediated by the PDZ domain within GASP. Searches of nucleotide and protein databases revealed that GASP exhibited a high degree of identity with the human AIP-1 and rat S-SCAM proteins and the PDZ domain of GASP was identical, at the amino acid level, to the fifth PDZ domain of AIP-1 and S-SCAM. The tissue distribution of GASP was also identical to that of AIP-1 and S-SCAM; GASP was expressed predominantly in the brain. Therefore, GASP likely represents a truncated mouse ortholog of these proteins. AIP-1 was initially identified as a protein that bound to the dentatorubral and pallidoluysian atrophy gene product atrophin-1 (17). S-SCAM was identified in yeast two-hybrid screening for proteins interacting with a region of SAPAP (18), and it has since been demonstrated to bind to a number of other proteins as well (18,23,24). A scaffolding role has been suggested for S-SCAM, and this function could be extended to GASP. The other protein members of this potential complex may include proteins that reportedly associate with S-SCAM or proteins that have more epithelial restricted expression patterns, including megalin.
GASP also shares a significant degree of homology with the mouse protein MAGI-1 and its human counterpart BAP-1. Together, these two groups of proteins, i.e., GASP/AIP-1/S-SCAM and MAGI-1/BAP-1, comprise a subfamily of membrane-associated guanylate kinase (MAGUK) proteins. The prototypical MAGUK proteins, for example PSD-95, are involved in the clustering of transmembrane receptor proteins and ion channels at precise locations in the plasma membranes of cells. Unlike the restricted pattern of expression exhibited by GASP, however, MAGI-1 is expressed in a more ubiquitous manner (13). Its reported expression in the kidney coincides with our finding that the fifth PDZ domain of MAGI-1 could bind megalin, which is itself abundantly expressed in the kidney. The binding of megalin to MAGI-1 was specifically mediated by the fifth PDZ domain of MAGI-1, inasmuch as the other PDZ domains of MAGI-1 were unable to bind to megalin and a megalin mutant lacking the carboxyl-terminal valine abolished the interaction with the fifth PDZ domain of MAGI-1. By using immunofluorescence assays in mouse kidney sections, we found that the expression of MAGI-1 was predominantly restricted to glomeruli. This pattern of MAGI-1 expression directly overlapped with the expression of the podocyte-specific protein GLEPP1. We are therefore confident that MAGI-1 is also expressed in a podocyte-specific manner. Although megalin is predominantly expressed in kidney proximal tubules, it has also been observed in the sole plates of podocyte foot processes, albeit at a much lower level of expression. The lack of expression of MAGI-1 in proximal tubules may have some significance with respect to megalin function. Megalin is viewed as an endocytic receptor that plays a scavenging role in the proximal tubules, clearing the glomerular filtrate of any remaining proteins that have progressed through the glomerulus. An interaction between the cytoplasmic tail of megalin and MAGI-1 might potentially interfere with the endocytic apparatus responsible for the function of megalin in proximal tubules. In podocytes, however, this interaction would presumably prevent or greatly reduce the endocytosis of megalin. Although endocytosis of megalin in podocytes, either in complex with MAGI-1 or not, cannot be entirely ruled out at this point, a nonendocytic function of megalin can be hypothesized by comparison with other members of the LDLR family that have been demonstrated to interact with potential signaling molecules. Both LDLR and LRP interact with the neuronal adaptor proteins mammalian Disabled (mDab) and FE65 (25). These interactions occur through binding of the PTB domains of mDab and FE65 to the NPXY motifs located in the cytoplasmic tails of LDLR and LRP. Interestingly, one of the three NPXY motifs in megalin is nearly identical (in the core NPXY motif and the flanking amino acids) to the NPXY motif in LDLR and LRP that binds to mDab, which suggests that megalin may also bind to mDab or a closely related protein. Indeed, megalin was recently shown to bind to Dab2, a protein that is highly homologous to mDab, and this binding was mediated through one of the NPXY motifs in the cytoplasmic tail of megalin (26). It is plausible that megalin may also serve in some signaling capacity that could be facilitated by its interaction with MAGI-1. A potential signaling complex could be assembled, with MAGI-1 serving as a docking site for a number of proteins that bind to the other PDZ or WW domains of MAGI-1. Whether MAGI-1 binds to the same proteins as reported for the homologous S-SCAM protein or to other proteins confined to the kidney remains to be investigated.
MAGI-1 was observed in the insoluble fraction of glomerular preparations extracted with Triton X-100. Insolubility of proteins in this detergent is often attributed to tight association with the cytoskeleton. In contrast, MAGI-1 in a rabbit kidney epithelial cell line was found to be Triton-soluble and migrated faster during SDS-PAGE analysis. The bulk of the difference in the sizes of MAGI-1 from these two sources is most likely attributable to the differential expression of splicing variants in the carboxyl-terminus of this protein. We have found that the splicing variant for this region that is preferentially expressed in glomeruli is variant c. This variant is much larger than the other two splice variants found in this region and also contains a number of tracts of basic amino acids. Although these polybasic regions are thought to function as nuclear localization signals (13), small polybasic regions in the MAGUK protein hDLG (a human homologue to the Drosophila dlg protein) are responsible for its interaction with the amino-terminal FERM domain in protein 4.1 and ezrin (27). This interaction would likely result in hDLG associating with actin microfilaments through the F-actin-binding domains of protein 4.1 and ezrin. Whether a similar mechanism occurs with MAGI-1 or whether an interaction with an alternative protein is responsible for the potential cytoskeletal association of MAGI-1 is not known at this time. The fact that MAGI-1 could be bound to an actin-based cytoskeleton fits well with the concept of MAGI-1 anchoring transmembrane proteins at defined cellular locations. Indeed, the morphologic features of the podocyte foot processes are dependent on a microfilament-based contractile apparatus composed of actin, myosin II, α-actinin, talin, and vinculin (28). Recently, other proteins were found to bind to MAGI-1, including β-catenin (29) and the tumor suppressor PTEN (phosphatase and tensin homologue deleted on chromosome 10) (30). Although the binding and localization of these two proteins with MAGI-1 were analyzed in cell lines, it has yet to be determined whether β-catenin and PTEN are expressed in the kidney with patterns similar to that we demonstrated for MAGI-1. Because MAGI-1 has the potential to interact with a number of proteins and form a multiprotein complex within podocytes, it will be of interest to determine whether MAGI-1 plays a role in the overall targeting or anchoring of these proteins and to evaluate the normal function of these complexes. In addition, determination of how these interactions are affected in disease conditions that result in the effacement of podocyte foot processes may lead to a better understanding of the overall changes that occur at the molecular level within podocytes.
Acknowledgments
This study was supported by National Research Service Award 1F32DK09912-01 (to K.M.P.), funds from the Howard Hughes Medical Institute (to B.L.M.) and the Smokler Foundation (to R.C.W.), and National Institutes of Health Grant 5-RO1-DK46073-07 (to R.C.W.).
- © 2001 American Society of Nephrology
References
Jump to section
More in this TOC Section
Cited By...
- Structural Basis of Highly Specific Interaction between Nephrin and MAGI1 in Slit Diaphragm Assembly and Signaling
- Regulation and Involvement in Cancer and Pathological Conditions of MAGI1, a Tight Junction Protein
- MAGI-1a Functions as a Scaffolding Protein for the Distal Renal Tubular Basolateral K+ Channels
- A New Crucial Protein Interaction Element That Targets the Adenovirus E4-ORF1 Oncoprotein to Membrane Vesicles
- Cell junction-associated proteins IQGAP1, MAGI-2, CASK, spectrins, and {alpha}-actinin are components of the nephrin multiprotein complex
- Cultured Podocytes Establish a Size-Selective Barrier Regulated by Specific Signaling Pathways and Demonstrate Synchronized Barrier Assembly in a Calcium Switch Model of Junction Formation
- The Adaptor Protein ARH Escorts Megalin to and through Endosomes
- Role of PDZ Domain-Containing Proteins and ERM Proteins in Regulation of Renal Function and Dysfunction
- JAM4, a Junctional Cell Adhesion Molecule Interacting with a Tight Junction Protein, MAGI-1
- Integration of Endocytosis and Signal Transduction by Lipoprotein Receptors
- Junctional protein MAGI-3 interacts with receptor tyrosine phosphatase {beta} (RPTP{beta}) and tyrosine-phosphorylated proteins
- Megalin and the neurodevelopmental biology of sonic hedgehog and retinol
- Podocyte Biology and Response to Injury
- Interaction of Two Actin-binding Proteins, Synaptopodin and {alpha}-Actinin-4, with the Tight Junction Protein MAGI-1
- Interaction of Synaptic Scaffolding Molecule and {beta}-Catenin
- {beta}1-Adrenergic Receptor Association with the Synaptic Scaffolding Protein Membrane-associated Guanylate Kinase Inverted-2 (MAGI-2): DIFFERENTIAL REGULATION OF RECEPTOR INTERNALIZATION BY MAGI-2 AND PSD-95