Molecular Dissection of the Intrinsic Factor-Vitamin B12 Receptor, Cubilin, Discloses Regions Important for Membrane Association and Ligand Binding*

Cubilin, the receptor for intrinsic factor-vitamin B12, is a novel type of high molecular weight receptor consisting of a 27 CUB (complement components C1r/C1s, Uegf, and bone morphogenic protein-1) domain cluster preceded by 8 epidermal growth factor repeats and a short N-terminal sequence. In addition to binding the vitamin B12-carrier complex, cubilin also binds receptor-associated protein. To delineate the structures for membrane association and ligand binding we established a panel of stable transfected Chinese hamster ovary cells expressing overlapping segments of rat cubilin. Analysis of conditioned media and cell extracts of transfected cells revealed that the N-terminal cubilin region conveys membrane association. Helical plotting of this region demonstrated a conserved amphipathic helix pattern (Lys74–Glu109) as a candidate site for hydrophobic interactions. Ligand affinity chromatography and surface plasmon resonance analysis of the secreted cubilin fragments showed ligand binding in the CUB domain region. Further dissection of binding-active fragments localized the binding site for intrinsic factor-vitamin B12 to CUB domains 5–8 and a receptor-associated protein-binding site to CUB domains 13–14. In conclusion, the N-terminal cubilin region seems crucial for membrane association, whereas the CUB domain cluster harbors distinct sites for ligand binding.

Cubilin, the receptor for intrinsic factor-vitamin B 12 , is a novel type of high molecular weight receptor consisting of a 27 CUB (complement components C1r/C1s, Uegf, and bone morphogenic protein-1) domain cluster preceded by 8 epidermal growth factor repeats and a short N-terminal sequence. In addition to binding the vitamin B 12 -carrier complex, cubilin also binds receptor-associated protein. To delineate the structures for membrane association and ligand binding we established a panel of stable transfected Chinese hamster ovary cells expressing overlapping segments of rat cubilin. Analysis of conditioned media and cell extracts of transfected cells revealed that the N-terminal cubilin region conveys membrane association. Helical plotting of this region demonstrated a conserved amphipathic helix pattern (Lys 74 -Glu 109 ) as a candidate site for hydrophobic interactions. Ligand affinity chromatography and surface plasmon resonance analysis of the secreted cubilin fragments showed ligand binding in the CUB domain region. Further dissection of binding-active fragments localized the binding site for intrinsic factor-vitamin B 12 to CUB domains 5-8 and a receptorassociated protein-binding site to CUB domains 13-14. In conclusion, the N-terminal cubilin region seems crucial for membrane association, whereas the CUB domain cluster harbors distinct sites for ligand binding.
Uptake of dietary vitamin B 12 /cyanocobalamin (B 12 ) 1 depends on gastric intrinsic factor (IF), which undergoes a specific and strong complex formation with the vitamin in the intestinal lumen (1). The vitamin-carrier complex is subsequently internalized after being recognized by the high affinity membrane receptor, cubilin (2,3). The binding of IF-B 12 is suggested to occur via the N-terminal region of IF (4) and only when B 12 is bound to the carrier protein (1)(2)(3)5). The physiological importance of the specific B 12 carrier and its receptor is evident from the B 12 deficiency, characterized by the megalo-blastic anemia and neurological symptoms related to defective functional expression of IF or cubilin (6 -11). In addition to binding IF-B 12 , cubilin also binds receptor-associated protein (RAP) (5), and growing evidence suggests that the receptor has other biological functions. This is in particular indicated by the fact that kidney and yolk sac epithelia exhibit a high cubilin expression (2,5,12), albeit IF is synthesized in the gastric epithelium and virtually only is present in the gastrointestinal lumen.
The recent determination of the primary structure of the rat and human cubilin molecule (13,14) revealed an ϳ3600-aa protein with a short N-terminal segment followed by 8 epidermal growth factor (EGF) repeats and a large cluster of 27 CUB 2 domains. No classical transmembrane segment has been predicted, and the overall structure is distinct from known endocytic receptors, including the RAP-binding members of the low density lipoprotein (LDL) receptor family (15). Two mutations, one causing a Pro/Leu substitution in CUB domain 8 and another suggested to activate an intronic cryptic splice site leading to truncation of the receptor in CUB domain 6, have recently been identified in the cubilin gene of patients with inherited B 12 malabsorption (11).
To link the structural and functional features of cubilin, we established a comprehensive panel of transfected Chinese hamster ovary (CHO) cells expressing various segments of the receptor. Biochemical analyses of the cells and expression products now delineate the regions important for membrane association and binding of IF-B 12 and RAP.

EXPERIMENTAL PROCEDURES
Ligands and Receptors-Porcine and human IF-B 12 were purified from gastric mucosa extract as described (16). Human RAP was produced as a recombinant protein in Escherichia coli. Rabbit cubilin was purified from renal cortex by IF-B 12 and RAP affinity chromatography, as described previously (5).

Construction of Plasmids for Expression of Recombinant Rat Cubilin
Fragments-Cubilin cDNA fragments extended with enzyme restriction sites were amplified by polymerase chain reaction with the Expand TM High Fidelity PCR System (Roche Molecular Biochemicals) and purified with the QIAEX II gel extraction kit (Qiagen, CA). The polymerase chain reaction products were subcloned into the expression vector (pSecTag2B or pcDNA3.1/Zeo(Ϫ) from Invitrogen, Groningen, The Netherlands) by use of the appropriate restriction enzymes (New England BioLabs, Beverly, MA; Amersham Pharmacia Biotech) and the T4 DNA ligase (Amersham Pharmacia Biotech). Plasmids were transformed using XL1-Blue competent cells (Stratagene, LaJolla, CA), and plasmid DNA was isolated by the Qiagen Maxiprep method (Qiagen) and sequenced before transfection as described previously (13). The following 14 constructs were subcloned and expressed 3 : N-EGF1-8 The rat cubilin constructs were expressed by using the pSecTag2B vector except for the two constructs, N-EGF1-8 (variant b) and N-EGF1-8ϩCUB1-3, which were expressed by using the pcDNA3.1/Zeo(Ϫ) vector. The pSecTag2B vector encodes the murine Ig chain leader sequence, which was used as leader peptide for protein secretion in these vector constructs, whereas the leader peptide of cubilin was used as leader sequence in the pcDNA3.1/Zeo(Ϫ) vector constructs.
Expression and Purification of Recombinant Rat Cubilin Fragments-CHO K-1 cells were transfected by using DOSPER liposomal transfection reagent (Roche Molecular Biochemicals), and stable transfected CHO clones were established by limited dilution and selection with 500 g/ml Zeocin (Invitrogen). Clones were grown in serum-free medium for CHO cells (HyQ-CCM ® 5 from HyClone ® , Logan, Utah) with 300 g/ml Zeocin. Cells were lysed in a solution of 10 mM NaH 2 PO 4 , 150 mM NaCl, 0.6 mM CaCl 2 , 1% Triton X-100 (Merck), 1 mM phenylmethylsulfonyl fluoride (Sigma), pH 7.4. Secretion of the recombinant rat cubilin fragments was analyzed by Western blotting of growth medium and cell lysate using a rabbit polyclonal antibody against rat cubilin (5) and a monoclonal antibody specific for the c-myc epitope coded for by the pSecTag2B vector (Invitrogen). The expressed CUB5-8 and CUB7-14 constructs were purified from the conditioned medium by IF-B 12 and RAP affinity chromatography, respectively, as described previously for cubilin purification (5).
Immunofluorescence Microscopy-Immunofluorescence analysis of CHO cells expressing the EGF cluster with and without the preceding N-terminal region (N-EGF1-8 and EGF1-8, respectively) was performed on cells fixed in formaldehyde, washed with an isotonic Tris-NaCl solution with 0.05% Triton X-100, and incubated with a polyclonal antibody against rat cubilin. Antigen-antibody complexes were visual-ized by use of fluorescein isothiocyanate-anti-rabbit Ig (DAKO A/S, Copenhagen, Denmark).
Ligand Affinity Precipitation-The conditioned medium of the transfected cells secreting cubilin fragments into the medium were incubated with CNBr-activated Sepharose 4B beads (Amersham Pharmacia Biotech) coupled with porcine IF-B 12 or human recombinant RAP (1-5 mg/ml gel). Beads were washed with a solution of 10 mM NaH 2 PO 4 , 150 mM NaCl, 0.6 mM CaCl 2 , pH 7.4, before and after incubation with medium containing the fragments. Supernatants were removed upon centrifugation at 5000 ϫ g for 2 min. Larger scale purification of the CUB1-8 and CUB5-8 constructs from conditioned media was carried out by IF-B 12 affinity chromatography (5), whereas the CUB7-14 and CUB13-20 constructs were purified by RAP affinity chromatography (5).
Surface Plasmon Resonance (SPR)-SPR analysis of the binding of IF-B 12 (human and porcine) and human RAP on a BIAcore 2000 instrument (Amersham Pharmacia Biotech) was carried out essentially as described (5). The BIAcore sensor chips (type CM5, Amersham Pharmacia Biotech) were activated with a 1:1 mixture of 0.2 M N-ethyl-NЈ-(3-dimethylaminopropyl) carbodiimide and 0.05 M N-hydroxysuccimide in water. Proteins were immobilized at a concentration of up to 40 g/ml in 10 mM sodium acetate, pH 4.0, and the remaining binding sites were blocked with 1 M ethanolamine, pH 8.5. The surface plasmon resonance signal generated from immobilized cubilin and cubilin fragments corresponded to 30 -80 fmol of receptor/mm 2 . The flow cells were regenerated with 1.6 M glycine-HCl, pH 3. The binding data were analyzed using the BIAevaluation program version 3.0. Fig. 1. shows the structural elements of cubilin and the initial strategy for mammalian expression of cubilin fragments in stable transfected CHO cells. The EGF repeat and CUB domain products were designed to overlap to ensure that the investigated binding sites that might encompass several domains were present in at least one expression product.

Identification of the Membrane Binding Region in Cubilin-
Immunoblotting of the conditioned media from transfected cells showed an effective secretion of all expression products except for the products containing the N-terminal region (Fig.  2). For example, the eight-EGF repeat region (EGF1-8 product) was effectively secreted, whereas the same region added the N-terminal cubilin region (N-EGF1-8 product) was not. Immunoblotting of solubilized transfected cell clones showed that the expression products containing the N-terminal segment rat cubilin aa sequence, the cubilin cDNA sequence in the forward and reverse primers, and the flanking enzyme sites used for subcloning. were retained in the cells, whereas the expression products without this segment were virtually absent in the cells, thus indicating an immediate secretion upon synthesis. Immunofluorescense microscopy of the cells (not shown) expressing the N-terminal region indicated that the expression product was localized to the membrane and intracellular vesicles as previously reported for native cubilin in cultured yolk sac cells (13). Using either the cubilin signal peptide or the murine Ig kappa chain signal peptide as the leader sequence in the construct encoding the N-terminal product (N-EGF1-8) led to cellular retention of the expressed protein. This indicates that the membrane association of the N-terminal region of cubilin is not because of membrane binding of a noncleaved leader peptide.
No hydrophobic stretch similar to classical transmembrane segments has been identified in the N-terminal region of cubilin (13). However, helical plotting of this region of rat and human cubilin (Fig. 3) predicted a conserved amphipathic helix structure encompassing aa Lys 74 -Glu 109 and Gly 73 -Asp 108 , respectively. A nearly identical cluster of 10 and 11 hydrophobic residues are located in the same side of the putative amphipathic helices of rat and human cubilin, respectively. The structure has some similarity to the A-type amphipathic helices seen in lipid-embedded regions of apolipoproteins (17).
Identification of the Regions for Binding of IF-B 12 and RAP-Ligand affinity precipitation of the expression products displayed in Fig. 2 with IF-B 12 -Sepharose beads (Fig. 4) revealed IF-B 12 binding to the CUB1-8 product, whereas none of the other regions exhibited any binding. Because no binding was seen to the EGF1-8ϩCUB1-3 product, shorter segments of the CUB domains 4 -8 region were expressed. The CUB5-8 product was effectively precipitated by IF-B 12 beads, whereas CUB4 -6 was not. These data combined with the absent IF-B 12 binding to CUB7-14 indicate that the binding site for IF-B 12 is localized within CUB domains 5-8 and is dependent on the structure of several domains within this region. Similar expression of human cubilin CUB domains 5-8 also revealed a binding site for IF-B12 (data not shown).
Precipitation of the expression products using RAP-Sepharose beads (Fig. 5) revealed that the CUB7-14 and CUB13-20 products bind RAP. Binding of RAP to the two large CUB domain constructs may reflect RAP binding to the overlapping CUB domains 13 and 14. Expression of smaller fragments further indicated this because the CUB11-14 and CUB13-16 products bound RAP, whereas no binding was seen to CUB9 -12.
The expressed recombinant CUB domain regions containing the distinct sites for IF-B 12 (CUB1-8 and CUB5-8) and RAP (CUB7-14 and CUB13-20) were readily purified to homogeneity by IF-B 12 and RAP affinity chromatography. Figs. 6 and 7 show the purification of CUB5-8 and CUB7-14 and the ligandfragment interactions as recorded by SPR analysis. These data further confirmed that the sites for IF-B 12 and RAP are distinct. The K d for human and porcine IF-B 12 binding to CUB5-8 was ϳ35 and ϳ45 nM, respectively, versus ϳ5 nM for the binding to native rabbit and human cubilin (Refs. 13 and 14 and data not shown). The K d for binding of RAP was ϳ70 nM versus ϳ170 nM for the binding to native rabbit cubilin (13). DISCUSSION In the present study we used a mammalian expression approach to identify the cubilin regions important for membrane association and the binding of IF-B 12 and RAP. Fig. 8 is a schematic presentation of the present mapping of functional regions in cubilin.
Membrane Association of Cubilin-Cubilin is a membrane protein although more loosely attached to the membrane compared with other membrane proteins (13). The present expression data now indicate that the N-terminal cubilin region is crucial for membrane association. The region has no classical transmembrane segment, but helical plotting demonstrated a conserved amphipathic helix motif similar to the lipid-embedded amphipathic helix structures described in apolipoproteins (17). Membrane association by means of this structure may be a hairpin-like anchoring with the extreme N-terminal end facing the noncytoplasmic environment. This will allow the N-terminal trimming by furin, the Golgi proteinase suggested to recognize cubilin and induce cleavage after the furin recognition sequence preceding Arg 12 in rat cubilin (14). Alternatively, the amphipathic region may be involved in the assembly of cubilin. A very recent electron microscopic study of purified bovine cubilin has shown the formation of trimers in solution without detergent and lipid (18). A region of four heptad repeats, actually a part of the putative amphipathic helix structure, has been proposed to account for the assembly. Future structural studies of membrane-associated cubilin may further elucidate how the receptor is physically attached to the membrane and how it is organized there. Maybe cubilin is functionally a monomer, but when shed from the membrane, the short hydrophobic region causes self-assembly into homodimer/ homotrimers.
The IF-B 12 Binding and RAP Binding Sites in Cubilin-The identification of the binding site for IF-B 12 in the region encompassing CUB domains 5-8 establishes the CUB domain structure as a ligand binding domain. Although this domain has long been known as a structural entity in other membrane proteins, its biological function in these proteins is largely unknown. In view of the present mapping data, it is tempting to suggest that some other identified receptor-like membrane proteins with CUB domains, e.g. SRP-ductin (19) are orphan receptors with ligands remaining to be identified.
The recently determined three-dimensional CUB domain structure (20) agrees well with a ligand binding function. The CUB domain is a barrel-like structure containing two layers of five-stranded ␤-sheets with the ␤-turns in a surface-exposed position as in the antigen binding regions of immunoglobulins. Furthermore, structural analysis of CUB domain dimers (21) suggests that the dimers may associate via the ␤-sheet layers. If the CUB domains of cubilin are arranged in this way, on top of each other and with the ␤-sheets facing, it will leave the less conserved ␤-turns of the CUB domains available for interactions. In view of the existence of 27 CUB domains in cubilin, a high number of putative sites for various protein interactions may then be predicted.
Another interesting perspective arising from the recognition of the binding site for IF-B 12 in cubilin relates to the recent identification of mutations in cubilin of patients with autosomal recessive B 12 malabsorption (Imerslund-Grä sbeck syndrome) (9, 10). The two mutations identified in the human cubilin gene (11) both relate to the critical region now known to harbor the IF-B 12 binding region. The mutation, suggested to activate a cryptic intronic splice site leading to truncation of the receptor in CUB domain 6, may lead to an inactive translation product, whereas the P/L substitution in CUB domain 8 might cause a structural change specifically impairing IF-B 12 binding. Further expression and mutational studies are now in progress to identify the crucial residues of the IF-B 12 binding site and the structural-functional implications of this mutation in human cubilin.
A site for binding of RAP was localized within CUB domains 13-14. The physiological importance of this interaction remains to be determined, but it is tempting to speculate that RAP may assist the processing/folding of cubilin in line with its suggested chaperone function and importance for normal processing of LDL receptor-related protein and megalin (22). Interestingly, the RAP binding to cubilin differs from the RAP binding to other receptors in the sense that RAP is not a general inhibitor of ligand binding in cubilin.
The estimated K d values for binding of IF-B 12 was 6 -9-fold higher than the estimated affinities of IF-B 12 to native rabbit and human cubilin (12,13), whereas the estimated K d values for binding of RAP to recombinant and native cubilin only differed about 2-fold. The lower affinity of the IF-B 12 binding to CUB5-8 may be explained by the small size of the fragment making it more vulnerable to immobilization compared with full-length cubilin. Alternatively, other parts of the receptor are required for complete structural integrity of the binding site. However, the existence of other independent sites for IF-B 12 binding seems unlikely in view of the absent IF-B 12 binding to other expressed regions of cubilin.
No specific function has yet been dedicated the cluster of EGF repeats in cubilin. Its position between the membraneassociated segment and the CUB domain region might indicate a structural role as a spacing segment for positioning the ligand binding region in an appropriate distance from the membrane. Interestingly, a similar cluster of EGF repeats is positioned between the transmembrane segment and the ligand binding regions of LDL receptor-related protein (23). Neither has a function of these EGF repeats, constituting the extracellular part of the LDL receptor-related protein ␤-subunit, been defined apart from binding the LDL receptor-related protein ␣-subunit.
In conclusion, the present molecular dissection of cubilin has identified the regions important for membrane association and binding of IF-B 12 and RAP. The high number of cubilin CUB domains suggests a high number of CUB domain interactions. To address this question we are currently investigating the existence of other cubilin ligands binding to the CUB domain region.