An antibody fragment from a phage display library competes for ligand binding to the low density lipoprotein receptor family and inhibits rhinovirus infection.

Recently antibodies with a wide range of binding specificities have been isolated from large repertoires of antibody fragments displayed on filamentous phage, including those that are difficult to raise by immunization. We have used this approach to isolate an antibody fragment against chicken very low density lipoprotein (VLDL) receptor. It binds to the receptor with good affinity (Kaff = 2 × 108M−1) as measured by plasmon surface resonance, and competes for binding of natural ligands (vitellogenin, VLDL, and receptor-associated protein). The antibody also binds to other members of the low density lipoprotein (LDL) receptor family including rat LDL receptor and human and rat low density lipoprotein receptor-related protein (LRP/α2MR), and it competes for binding of receptor-associated protein to LRP/α2MR. Moreover, the antibody fragment inhibits infection of human fibroblasts deficient in LDL-R but expressing LRP/α2MR by human rhinovirus. Binding of the antibody is abolished upon reduction of the receptors and is strictly Ca2+ dependent. The phage antibody thus recognizes the ligand binding site(s) of several members of the LDL receptor family, in contrast to antibodies produced by hybridoma technology.

The low density lipoprotein (LDL) 1 receptor of mammals is the prototype of a family of related proteins. Members of the LDL receptor family have several structural modules in common; (i) "binding repeats," complement-type domains consisting of ϳ40 residues displaying a triple disulfide bondstabilized negatively charged surface (head-to-tail combinations of these repeats are believed to specify ligand interaction); (ii) epidermal growth factor precursor-type repeats, also containing six cysteines each; (iii) modules of ϳ50 residues with a consensus tetrapeptide, Tyr-Trp-Thr-Asp (YWTD); and (iv), in the cytoplasmic region, signals for receptor internalization via coated pits, containing the consensus tetrapeptide Asn-Pro-Xaa-Tyr (NPXY).
The LDL receptor family includes at least 4 proteins; the LDL receptor (LDL-R), the low density lipoprotein receptor related protein (also termed LRP/␣ 2 MR), gp330 (also termed megalin), and the very low density lipoprotein (VLDL) receptor. The LDL receptor (LDL-R) has a cluster of 7 binding repeats and binds to apolipoprotein B (apoB) and apolipoprotein E (apoE) (1,2). LRP/␣ 2 MR is a giant receptor (4525 amino acids) containing 4 clusters of 2 to 11 binding repeats and has many ligands including apoE (3), ␣ 2 M-proteinase complexes (4,5) among others (4 -12), and a 39-kDa intracellular protein (receptor-associated protein or RAP). RAP binds to LRP/␣ 2 MR with high affinity, co-purifies with LRP/␣ 2 M from liver and placenta (13,14), and competes for binding with all known LRP/␣ 2 MR ligands (4, 7-9, 15, 16). Whereas the majority of the ligands bound by LRP/␣ 2 MR fail to be recognized by LDL-R, human rhinoviruses (HRVs) of the minor receptor group type attach to either of these proteins (8). Recently, it was shown that RAP also binds to LDL-R but with much lower affinity than to LRP/␣ 2 MR (17). Gp330 (megalin) is a membrane glycoprotein (the Heymann nephritis antigen in rats) (18), is closely related to LRP/␣ 2 MR in structure (19), and binds to many of the same ligands (except for ␣ 2 M-proteinase complexes) (20). The VLDL receptor (VLDL-R) is characterized by a cluster of 8 binding repeats, and binds VLDL and other apoE containing lipoproteins (21).
The LDL receptor family is also present in birds; for example, in the laying hen the chicken LDL-R (22) and an LRP/ ␣ 2 MR (23) are expressed in somatic cells, and an LRP/␣ 2 MRlike protein (380 kDa) and a receptor (OVR) for very low density lipoprotein, vitellogenin (VTG) (24), and ␣ 2 -macroglobulin (25) are expressed in oocytes. The binding specificity and sequence of OVR indicate that it is the chicken homologue of VLDL-R (26).
Although polypeptides of the LDL receptor family are highly related, and most of the known receptors bind to apoE, it has proved difficult to map the ligand binding sites with respect to * This work was supported by Austrian Science Foundation Grants P-9999-MOB (to D. B.), P-9508-MOB (to J. N.), and S-7108 (to W. J. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  different ligands. The sites are thought to comprise the cysteine-rich binding repeats, and to involve carboxylate residues on the receptor and lysine and arginine residues on the ligands (2); in apoE most of the positive charges are clustered to one side of the protein (27). Although LDL-R binds to few ligands, LRP/␣ 2 MR (3)(4)(5)(6)(7)(8)(9)(10)(11)(12), gp330 (20), and OVR bind to a wide spectrum of ligands with high affinity (24,25,28). In LRP/␣ 2 MR it appears that the different ligands bind to different clusters of binding repeats (29,30), but OVR and LDL-R contain only one cluster of binding repeats, and more subtle differences must therefore dictate their ligand binding properties.
Mapping of the ligand binding sites has been hampered by the difficulty of raising blocking antibodies by immunization. Recently the display of repertoires of antibody fragments on the surface of filamentous bacteriophage, and the selection of antigen-binding phage (31), has provided a means of making antibodies without immunization (32). Antibody repertoires can be derived from the rearranged V-genes of populations of lymphocytes (32,33) or from V-gene segments rearranged in vitro (34 -36). Antibodies with many different specificities have been isolated from the same repertoire, including some directed against self-antigens (33) and highly conserved proteins (36). Although the binding affinities of the antibody fragments were often moderate, it has been possible to obtain antibodies with good binding affinities (K aff ϭ 10 8 -10 9 M Ϫ1 ) from very large repertoires (34). Here we used a large phage antibody repertoire to isolate antibody fragments against OVR.
LDL-R and LRP/␣ 2 MR both serve as receptors for one group (minor group) of HRVs (8), the main causative agents of the common cold. Due to the large number of different serotypes, vaccination is not possible; therefore, other means of preventing or curing the common cold are being thought of, including inhibition of virus-specific enzymes (37), or blockage of the viral receptors (38,39). In this article we show that viral infection can be blocked with the single chain antibody fragment described.

MATERIALS AND METHODS
Animals and Diets-White Leghorn laying hens were purchased from Heindl (Vienna) and maintained as described (40). Roosters (20 -30-week-old) were treated with 17-ethinylestradiol dissolved in propylene glycol, by injecting 10 mg/kg body weight into the breast muscle. After 72 h, blood was collected from the jugular vein and mixed with the following additives giving the indicated final concentrations (10 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 5 mM leupeptin, 0.1 mM aprotinin).
Preparation of Oocyte Membranes and OVR-Triton X-100 liver membrane extracts from hens and estrogenized rats (41) were prepared according to Ref. 10. Oocyte membranes were prepared from previtellogenic follicles (4 -6 mm diameter) excised from mature laying hens and extracted with 1% Triton X-100 as described (40). For purification of the receptor, membranes were solubilized with 30 mM CHAPS and the extract was subjected to affinity chromatography on VLDL-Sepharose and anti-OVR-Sepharose as described (42). Purified human placental LRP/␣ 2 MR (43) was generously provided by Dr. J. Gliemann, University of Aarhus, Denmark.
OVR was also expressed transiently in COS-7 cells (26). Briefly, COS-7 cells (American Type Culture Collection) were seeded at a density of 1.5 ϫ 10 6 /80-cm 2 dish and incubated overnight in RPMI 1640 medium containing 10% fetal bovine serum, 100 units/ml penicillin, 100 g/ml streptomycin, 2 mM glutamine, and 0.05 mM ␤-mercaptoethanol. The expression vector (pCDMCVR-1) carrying the full-length cDNA for the chicken OVR was transfected into COS-7 cells by electroporation (20 g of DNA/dish) using a Bio-Rad Gene Pulser. Dishes (60-mm diameter) were seeded with 4 ϫ 10 5 cells each in standard medium; after 48 h the cells were washed 3 times with PBS and harvested in PBS containing 0.5 mM phenylmethylsulfonyl fluoride and 2.5 M leupeptin. Cells were pelleted by centrifugation and detergent extracts with Triton X-100 were prepared as described (22).
Preparation of Ligands-VLDL was prepared from plasma from laying hens fed a grower mash by sequential ultracentrifugation according to Ref. 40. Vitellogenin was purified of plasma from estrog-enized roosters by ion exchange chromatography (DEAE cellulose) as described (44). Recombinant RAP was produced as a glutathione Stransferase (GST) fusion protein using a PGEX 2T (Pharmacia Biotech Inc.) derived expression plasmid in DH5 bacteria (15). The fusion protein was purified by affinity chromatography on glutathione-Sepharose (Pharmacia).
Polyclonal Antibodies-Polyclonal antibodies against OVR (26), OVR-and oocyte-specific LRP/␣ 2 MR (45), chicken somatic LRP/␣ 2 MR (23), and mammalian LDL receptor (40) are described in the references indicated. For the various incubations IgG fractions prepared by protein-A affinity chromatography were used. Antiserum against human LRP/␣ 2 MR was kindly provided by Dr. J. Gliemann of the University of Aarhus, Denmark.
Electrophoresis and Transfer to Nitrocellulose-SDS-PAGE under nonreducing conditions was performed according to Laemmli (46) on 4.5-12% gradient slab gels at 180 V for 60 min using the minigels (Bio-Rad). Molecular sizes of proteins were estimated with Bio-Rad markers . Proteins were electrophoretically transferred to nitrocellulose membranes (Hybond-C, Amersham) in 20 mM Tris-HCl, 0.15 M glycine buffer, pH 8.4, for 90 min at 200 mA. After transfer, proteins were visualized by staining the membrane with Ponceau S (2 g/liter in 30% (w/v) trichloroacetic acid) and rinsing with water. Nitrocellulose membranes used for Western blot analysis were blocked with 5% non-fat dry milk in PBS containing 0.1% Tween 20. Bound rabbit antibodies were detected with horseradish peroxidase-conjugated protein-A and an enhanced chemiluminescence system (Renaissance TM system, DuPont). The antibody fragment was detected using anti-myctag antibody 9E10 (Stratagene) and rabbit anti-mouse IgG conjugated with horseradish peroxidase, and the membranes were exposed on Reflectron TM film (DuPont) for the indicated times.
Selection of Antigen Binding Phage by Library Panning-A phage antibody repertoire was prepared as described. 2 Briefly repertoires of human heavy and light chain V-genes were amplified from human lymphocytes from tonsils, and then assembled to encode single chain Fv fragments as described in Ref. 32. The assembled V-genes were cloned into the ampicillin-resistant phagemid pCANTAB6 (Cambridge Antibody Technology, Cambridge, UK) to append a COOH-terminal hexahistidine (His) tag to allow purification of the fragments (47) and a myc-tag (48) to facilitate detection of the fragments. A large repertoire of 1.5 ϫ 10 9 different phage clones was thereby prepared. In the first round of selection 10 13 phage were panned using immunotubes (Maxisorb, Nunc) coated with 10 g/ml purified OVR in 50 mM NaHCO 3 , pH 8.6, overnight at 4°C (33). Four additional rounds were carried out in immunotubes coated with 5 g/ml antigen. Affinity enrichment was carried out as described (34).
Screening and Sequencing of Clones-Phage were isolated from single ampicillin-resistant colonies of infected (suppressor) Escherichia coli TG-1 using helper phage VCS-M13 (Stratagene), and the phage used to infect the (non-suppressor) E. coli HB2151. Single ampicillinresistant colonies were used to inoculate 200 l of culture broth in microtiter plates, and the expression of soluble scFv fragments induced by addition of 1 mM isopropyl-␤-D-thiogalactopyranoside to the cultures (33). Bacteria were pelleted, and the supernatants containing scFv fragments were screened for binding to the antigen by ELISA using the anti-myc-tag antibody 9E10 and anti-mouse IgG conjugated with alkaline phosphatase (Promega) (34). Binding specificity was tested by comparing the signals obtained from plates coated with either 1 g/ml OVR or 10 g/ml bovine serum albumin. Soluble fragments giving an ELISA signal at least 7 times higher for OVR than bovine serum albumin were identified, and using specific primers (36) the encoding variable heavy and light chain regions were amplified via polymerase chain reaction from single colonies. The amplified DNA was subjected to automatic sequencing by the dideoxy method using the TAQ DyeDeoxy Terminator Cycle Sequencing Kit (Applied Biosystems Inc.) and the PCRHLINK primer (Cambridge Antibody Technology, Cambridge, UK) for the heavy chain region and FDSEQ1 (34) for the variable light chains. The sequences of the V-genes were compared with a directory of V-gene segments 3 using SeqEd (Applied Biosystems Inc.) and MacVector 3.5 (IBI Kodak, New Haven, CT).
Preparation of Monoclonal Reagents-For further characterization of the scFv fragments, the infected HB2151 bacteria were grown in 1 liter of 2 ϫ TY medium supplemented with 0.1% glucose and 100 g/ml ampicillin and were induced overnight with 1 mM isopropyl-␤-D-thiogalactopyranoside at 23°C (49). Bacteria were removed using a tangential filtration unit with 45-m filters (Filtron) and the supernatant was concentrated via ultrafiltration through membranes with a cut-off of 10 kDa (Filtron). Samples were dialyzed overnight against PBS and affinity purified by immobilized metal chelate affinity chromatography (47,50) using Ni-NTA (Diagen) as described in Ref. 34. The fractions eluted with 100 mM imidazole were analyzed by SDS-PAGE on 12.5% polyacrylamide gels under nonreducing conditions. The concentration was determined spectrophotometrically assuming A 280 of 1.0 ϭ 0.7 mg/ml (34). Labeling of scFv7 with 125 I was carried out using the chloramine T method with 0.5 mCi of 125 I (Amersham) per 30 g of protein yielding a specific activity of 1.2 ϫ 10 6 cpm/g.
Determination of Binding Kinetics by Surface Plasmon Resonance on a BIAcore Device (Pharmacia Biosensor AB (51)-Monomeric and oligomeric species of purified scFv7 were fractionated on a Superdex 75 FPLC column (Pharmacia) in PBS containing 0.2 mM EDTA and monitored by the absorbance at 280 nm. Peaks corresponding to the monomeric and the dimeric fraction (34) were immediately analyzed on the BIAcore system. Receptors were coupled to a CM5 sensor chip using the Amine Coupling KIT (Pharmacia Biosensor AB (52)) and 45 l of 10 g/ml receptor diluted in 10 mM sodium acetate, pH 4.0, yielding 1411 response units (RU) of immobilized OVR, 2100 RU of LRP, and 550 RU of LDL-R, respectively. All determinations were performed at 25°C in 10 mM HEPES, 150 mM NaCl, 2 mM CaCl 2 , 0.05% BIAcore surfactant P20, pH 7.4 (HBS-M), with a constant flow rate of 5 l/min and an injected sample volume of 45 l for scFv7 and 35 l for GST-RAP, respectively. ScFv7 was injected at concentrations between 1 ϫ 10 Ϫ7 and 2 ϫ 10 Ϫ6 M, GST-RAP was used at concentrations between 2 ϫ 10 Ϫ7 and 2 ϫ 10 Ϫ6 M. Two consecutive 4-l injections of 10 mM HCl were used for regeneration. The calculation of the off-and on-rates was performed with the BIAevaluation Program 2.0 (Pharmacia Biosensor).
Competition Assays-Oocyte membrane extracts were run on 4.5-12% gradient SDS-PAGE gels, electrophoretically transferred onto Hybond-C membranes, and blocked with PBS containing 2% Tween 20 for 1 h. The blots were incubated in PBS, 2 mM CaCl 2 , 0.2% Tween 20 (standard conditions) with 0.66 g/ml (25 nM) of 125 I-scFv in the presence of a 1000-fold molar excess of competitor for 1 h. Blots were dried and exposed overnight on Reflectron TM film (DuPont). Quantitative competition assays were carried out using ELISA plates coated with 1 pmol of OVR or 250 fmol of human LRP/␣ 2 MR per well, respectively, in 50 mM NaHCO 3 , pH 8.6, at 4°C for 16 h. Plates were blocked with 2% milk powder in PBS for 1 h, washed with PBS, and incubated with 80,000 cpm/well (0.066 g) 125 I-scFv7 together with competitors in PBS, 2 mM CaCl 2 for 1 h at room temperature at the concentrations indicated in the figures. The plates were washed 3 times with PBS, 0.2% Tween 20 and 3 times with PBS. Wells were cut, and bound radioactivity was determined using an auto-␥-counter Cobra II (Packard). In order to test the competition of 35 S-HRV2 binding to OVR and LRP by scFv7, 30,000 cpm/well 35 S-HRV2 were used together with competitors in an assay essentially done as described above. Radioactivity bound was determined after addition of scintillation fluid (Beckman) to the cut wells in a Beckman scintillation counter.
Preparation of HRVs and Inhibition of Viral Infection-Human rhinovirus serotype 2 (HRV2) and 14 (HRV14) were grown in HeLa cells (Wisconsin strain) as described (53). 35 S-HRV2 was prepared as described (54). FH cells (NIH repository GM00486A) and normal human fibroblasts were grown in 96-well plates in minimal essential medium (Life Technologies, Inc.) containing 10% heat inactivated fetal calf serum (Life Technologies, Inc.). Preceeding infection, the medium was removed, replaced with infection medium (54), and the plates were placed at 34°C. Infection was done in triplicate with a multiplicity of infection of 10 in the presence of scFv7 at the concentrations as indicated in Fig. 7. In some experiments antibody 9E10 was present at the same concentration as scFv7. After 40 h the supernatants were removed and the progeny virus was titrated by plaque assay.

Isolation of scFvs with Affinity for the OVR from a Phage
Library-The repertoire of scFv fragments displayed on a filamentous phage was selected with OVR. The selected phage were grown and subjected to further rounds of selection and growth, in total 5 rounds. After rounds 3 and 5 the soluble scFv fragments from phage-infected colonies were analyzed for binding to OVR by ELISA. Binding was detected for 15 out of 18 clones from round 5, and for 12 out of 18 clones from round 3. DNA sequencing of the encoded scFv fragments indicated that a single scFv sequence (scFv7) predominated in both rounds 3 and 5 (total 18 out of 36 clones), as shown in Fig. 1. The V H sequence was encoded by the DP7 segment (V H 1 family) (55); and the V L sequence by the L12a segment (V k 1 family) (56). About 5 mg of scFv7 protein was purified from 1 liter of bacterial supernatant by immobilized metal chelate affinity chromatography, and was judged to be about 90% pure, by SDS-PAGE (not shown).
Specificity of Soluble Antibody Fragment scFv7-We analyzed the specificity of binding of scFv7 to OVR and other members of the LDL-R family by Western blotting. As shown in Fig. 2A (lane 1), scFv7 binds specifically to OVR (95 kDa) in crude oocyte membrane extracts. It also reacted with a species of Ͼ400 kDa present in membrane extracts of chicken liver (lane 2, see arrowhead). As seen by overexposure of lane 1, the same protein was also present in oocyte extract, although in a much lower amount (not shown). This appears to be chicken somatic LRP/␣ 2 MR. Lane 4 shows chicken somatic LRP/␣ 2 MR as detected by an anti-peptide antibody directed against the carboxyl terminus (23), and lane 3 shows chicken oocyte LRP as detected with a polyclonal antiserum against OVR and known to cross-react with oocyte LRP (45). This shows that scFv7 binds to chicken somatic LRP, but fails to bind to oocyte LRP on the Western blot. We also analyzed liver membrane extracts from estrogenized rats (41). As shown in lane 5, scFv7 binds rat LRP/␣ 2 MR (as detected in lane 7 using an antibody specific for mammalian LRP/␣ 2 MR). In contrast, scFv7 hardly recognizes the rat LDL-R (as detected in lane 6 with an antibody specific for LDL-R). However, upon longer exposure, we did detect a faint band in lane 5 (not shown) comigrating with rat LDL-R.
In Fig. 2B, we used detergent extracts from COS-7 cells which had been transiently transfected with a plasmid carry- and comigrating with the recombinant OVR probably represents the endogenous simian VLDL-R. There is an additional band migrating slightly slower than OVR in lanes 1 and 2. This protein was not further characterized; it might correspond to the simian LDL-R or to another, still unknown member of the LDL receptor family. The higher band (500 kDa) seen in lanes 1 and 2 is most likely simian LRP/␣ 2 MR which is abundantly expressed in COS cells (57).
Finally, as shown in Fig. 2C, the antibody strongly discriminates between non-reduced (lane 1) and reduced (lane 2) receptors (OVR migrating at 95 kDa and somatic LRP/␣ 2 MR migrating at about 500 kDa, respectively), whereas a control antibody directed against a synthetic peptide derived from the carboxyl terminus of OVR reacted with both forms of OVR equally well, but failed to bind to somatic LRP/␣ 2 MR (lanes 3  and 4). Note that OVR migrates with a much higher apparent molecular weight in its reduced form when compared to the migration of its unreduced form (compare lanes 3 and 4).
Kinetic Analyses-ScFvs can form monomers, dimers, and higher oligomers (34,36). To investigate the kinetics of binding of scFv7 to OVR, the mixture of monomers (30%), dimers (55%), and trimers (15%) was first separated by gel filtration on Superdex 75 (Pharmacia). Kinetic analyses were carried out using a BIAcore surface plasmon resonance device (Pharmacia Biosensor AB); to avoid reassociation, the monomers were used immediately after gel filtration. The chicken OVR was immobilized on a CM5 sensor chip (1411 RU), and scFv7 injected at four concentrations in the range of 1 ϫ 10 Ϫ7 to 2 ϫ 10 Ϫ6 M. Both monomeric and dimeric  4) and electrophoretically transferred to nitrocellulose. Western blotting was carried out using 125 I-labeled scFv7 (0.66 g/ml with a specific activity of 1.2 ϫ 10 6 cpm/g; lanes 1 and 2) and nitrocellulose strips were exposed for 16 h. For lanes 3 and 4, a polyclonal antipeptide antibody specific for the carboxyl terminus of OVR was used at 10 g/ml. Bound IgG was detected as described under A. Exposure time was 2 min. species of scFv7 showed an off-rate of k off ϭ 3 ϫ 10 Ϫ4 s Ϫ1 , which was identical to that of the unfractionated mixture. The on-rate was estimated as k on ϭ 8 ϫ 10 4 M Ϫ1 s Ϫ1 and the association constant K aff ϭ (k on /k off ) ϭ 2 ϫ 10 8 M Ϫ1 .
The analyzed scFv7 was also shown to bind to LRP/␣ 2 M with high affinity and, to a lesser extent, to LDL-R ( Fig. 2A). Affinity constants for human LRP/␣ 2 M and bovine LDL-R were therefore also determined. Immobilization of LRP and LDL-R yielded 2100 and 550 RU, respectively. scFv7 at concentrations of 2.5 ϫ 10 Ϫ7 to 2 ϫ 10 Ϫ6 M was tested for binding to the immobilized proteins. scFv7 bound to LRP with an affinity of K aff ϭ 8 ϫ 10 7 M Ϫ1 and to LDL-R with K aff ϭ 5 ϫ 10 7 M Ϫ1 (Fig. 3).
As it has been recently shown that RAP also binds strongly to the VLDL receptor of mammalian (58) as well as of avian origin (73), we also analyzed the kinetics of binding of RAP to chicken VLDL receptors. Recombinant RAP fused to glutathione S-transferase (GST-RAP) (15) was injected over a Biacore chip coated with OVR at concentrations of 2 ϫ 10 Ϫ7 to 2 ϫ 10 Ϫ6 M, yielding an affinity constant in the same range as scFv7 of 3 ϫ 10 8 M Ϫ1 (data not shown).
Ligand Competition-Competition studies were first carried out using several natural ligands of OVR. Chicken oocyte membrane extract was fractionated on a 4.5-12% gradient SDS-polyacrylamide gel under nonreducing conditions and transferred onto a nitrocellulose membrane. Strips were cut from the membrane and separately incubated with 125 I-scFv7 in the presence of a 1000-fold molar excess of various ligands. VLDL, VTG, GST-RAP, and unlabeled scFv7 strongly competed for binding of 125 I-scFv7 to OVR, whereas excess of the irrelevant scFv D1.3 (31) showed no influence on binding (data not shown). For a more quantitative assay, purified chicken OVR was coated onto ELISA plates and incubated with 125 I-scFv7. Competitors were added at increasing concentrations corresponding to a 2-70-fold molar excess over 125 I-scFv7 (Fig. 4A). Here, a 5-fold molar excess of unlabeled scFv7, VTG, or GST-RAP reduced the binding almost to background. VLDL was much less efficient. Furthermore, rabbit polyclonal antiserum (1000-fold molar excess IgG fraction) to chicken OVR did not compete with binding of the scFv to OVR, suggesting that no (or very low) amount of antibodies binding to the particular epitope recognized by scFv7 were present in the antiserum (not shown). As scFv7 also binds to LRP/␣ 2 MR of chicken and rat ( Fig. 2A), and RAP is known to compete for binding of all known ligands to LRP/␣ 2 MR (4, 7-9, 15, 16, 59, 60), we wondered whether scFv7 and RAP would compete for binding to human LRP/␣ 2 MR. As seen in Fig. 4B, GST-RAP completely eliminated binding of 125 I-scFv7 to LRP/␣ 2 MR at a 7.5-fold molar excess.
Requirement for Ca 2ϩ -Members of the LDL receptor family require Ca 2ϩ ions for ligand binding (9,40,57,61,62). As scFv7 appeared to be binding at the same sites as the ligands, we checked the influence of Ca 2ϩ ions on the binding of 125 I-scFv7 to chicken OVR and to human LRP/␣ 2 MR. Oocyte membrane extract was fractionated on a 4.5-12% gradient SDS-polyacrylamide gel, blotted onto a nitrocellulose membrane, and incubated with 125 I-scFv7 in the presence of 20 mM EDTA and EGTA, respectively. Whereas the blot incubated in the presence of 2 mM Ca 2ϩ showed strong binding of the antibody fragment to the receptor, no binding was evident when EDTA or EGTA was added (data not shown). For a more quantitative assay, we used OVR-and LRP/␣ 2 MR-coated plates, as above, and the binding of 125 I-scFv7 was monitored in the presence of EDTA and EGTA (Fig. 5). Higher concentrations of the chelating agents led to the loss of binding of scFv7 to both OVR (Fig.  5A) and to LRP/␣ 2 MR (Fig. 5B), and this was reversed by addition of Ca 2ϩ in excess (Fig. 5, C-D).
Competition of 35 S-HRV2 Binding to OVR by scFv7-Human rhinoviruses of the minor receptor group use members of the LDL receptor family for entry into their host cells. It was previously shown that RAP inhibits infection of human fibroblasts deficient in LDL-R synthesis, but expressing LRP/␣ 2 MR (8). We have recently shown that HRV2 also binds to OVR. 4 To study whether scFv7 also competes for HRV binding to OVR, oocyte membrane extract was fractionated on a 4.5-12% gradient SDS-polyacrylamide gel, blotted onto a nitrocellulose membrane, and incubated with 100,000 cpm of 35 S-HRV2 in the presence of 10 g/ml scFv7 and RAP, respectively. Both proteins were able to completely abolish binding of 35 S-HRV2 to OVR, while a control scFv did not reveal any influence on binding (data not shown). Quantitative competition assays were carried out using ELISA plates coated with OVR. Plates were then incubated with 35 S-HRV2 together with scFv7 or with RAP at the concentrations indicated in Fig. 6. Binding of FIG. 4. Competition of ligands for binding of 125 I-scFv7 to chicken OVR (A) and to human LRP/␣ 2 MR (B). Purified OVR (1 pmol/well) and LRP/␣ 2 MR (250 fmol/well), respectively, were immobilized on ELISA plates and incubated with 25 nM (1.2 ϫ 10 6 cpm/g) 125 I-scFv7 together with competitors at the molar excess indicated. Radioactivity bound was determined in a ␥-counter and values were plotted against times molar excess of the competitors. Data points given are the mean of triplicate experiments. 35 S-HRV2 to OVR was reduced to about 10% upon addition of 100 g/ml RAP and scFv7, respectively, while the control fragment showed hardly any effect on HRV2 binding.
Inhibition of Viral Infection-Human fibroblasts from a patient with familial hypercholesterolemia (FH cells, deficient in LDL-R) were preincubated with scFv7 at various concentra-tions and infected with either HRV2 or HRV14. Progeny virus was determined by plaque assay and the yields of infectious particles are shown in Fig. 7. From these experiments it becomes clear that scFv7 inhibits viral infection significantly, decreasing the viral yield by more than 3 orders of magnitude at a concentration Ͼ1 g/ml. The inhibition is specific for the minor group virus HRV2, since infection with HRV14 resulted in essentially the same virus yield regardless of the presence of the antibody fragment (data not shown). Addition of the antimyc-tag antibody 9E10 decreased the concentration required for inhibition to about 0.2 g/ml (data not shown). Presence of the irrelevant scFv D1.3 used as a control showed no influence on the viral infection. DISCUSSION The biological functions of the LDL receptor family are important and diverse. For example, the LDL receptor has a key role in cholesterol homeostasis in mammals; mutations disrupting its function leading to severe hypercholesterolemia and premature artheriosclerosis in man (63). The LRP/␣ 2 MR is probably involved in clearing spent proteases and chylomicron remnants from the circulation (64), and may also have a role in development, as mouse embryos with a homozygous knockout for the LRP/␣ 2 MR gene are arrested in various stages of development (59,65). OVR appears to have a role in reproduction mediating growth of oocytes via uptake of the major yolk precursors VLDL and VTG from coated pits in the plasma membrane (66, 67): mutant "restricted ovulator" hens are sterile   (68). As the yolk precursor proteins comprise about 50% of the total weight of the egg yolk the endocytic mechanisms mediated by OVR must be highly efficient.
Antibodies that block the binding of multiple ligands to the LDL receptor family should help in dissecting the roles of these receptors and ligands. Although antisera and monoclonal antibodies have been obtained against several members of the LDL-R family, none have been described that efficiently block the binding of ligands. An exception is IgG-C7, a monoclonal antibody against human LDL-R recognizing the first ligand binding domain of the receptor, which inhibits the binding of LDL or apoE-rich lipoproteins (69). The difficulty in obtaining antibodies with the desired properties by conventional means may be due to the conserved nature of these epitopes of the receptor between different species, or to the presence of ligands in the serum of the animal to be immunized rendering the binding site(s) inaccessible.
We therefore attempted to make blocking antibodies by phage display technology without immunization, and selecting with pure chicken OVR from a large (1.5 ϫ 10 9 clones) repertoire of scFv fragments. We succeeded in isolating an antibody fragment (scFv7) that bound strongly to chicken OVR from membrane extracts from follicles and COS-7 cells transfected with a plasmid encoding OVR.
ScFv7 appears to bind at the same site as several natural ligands as its binding to OVR is competed with VTG and RAP (Fig. 4A).
As shown by BIAcore, the binding affinity to OVR appeared to be good (K aff ϭ 2 ϫ 10 8 M Ϫ1 or K d ϭ 5 nM); the slow off-rate (3 ϫ 10 Ϫ4 s Ϫ1 ) makes the fragment particularly suitable for mapping studies. Indeed the affinity constant is very similar to that of GST-RAP. As well as binding to OVR, scFv7 also binds to chicken, rat, and human LRP/␣ 2 MR ( Figs. 2A and 3), and, to a much lesser extent, to rat and bovine LDL-R (Fig. 3), and its binding to LRP/␣ 2 MR is competed with RAP (Fig. 4B). A 100fold molar excess of recombinant GST failed to inhibit binding of scFv7 to either receptor (not shown).
The binding of scFv7, like the natural ligands, is abolished after reduction of the receptors (Fig. 2C) and also requires Ca 2ϩ ions (Fig. 5). EGTA at about 5 mM led to a significant loss of activity. These observations suggest that scFv7 recognizes a conformational epitope rather than the primary sequence. However, unlike the binding of the natural ligands RAP and apoE which depend on electrostatic interactions, we see no evidence for such interactions with the antibody; the sequence of the antibody reveals no charge clusters in the complementarity determining regions or any sequences resembling those of the natural ligands. Furthermore, the antibody fragment does not bind to heparin, while RAP is strongly interacting with this glycan under identical conditions. 5 Thus the scFv appears to be "seeing" the ligand binding site but in a different manner to the natural ligands; it therefore differs from the anti-integrin antibodies isolated from antibody repertoires with a "built-in" sequence motif from the natural ligand (70).
In addition to the ligands mentioned above, human rhinoviruses have been shown to gain access to the host cell via members of the LDL-R family (8). The large number of different rhinovirus serotypes is divided into two groups dependent on their binding to either the intercellular adhesion molecule 1 (ICAM-1, major group) or to members of the LDL-R family (minor group). Monoclonal antibodies which effectively block infection by major group viruses have been obtained and were used to identify the major group receptor (71,72). No antibodies blocking infection by minor group viruses are available due to the presence of both LDL-R and LRP/␣ 2 MR on the cell surface. These receptors are immunologically distinct but are both used as minor group receptors. ScFv7 cannot block infection of wild type fibroblasts with minor group HRVs but effectively blocks infection of FH cells (human fibroblasts deficient in LDL receptor synthesis).
The protection is even stronger in the presence of the antimyc-antibody 9E10 which renders the scFv7 bivalent by binding two molecules via the myc-sequence tag which is COOH terminally fused to the antibody fragment. Based on the dual specificity it is likely that minor group HRVs recognize a structure or charge pattern equally present in LDL-R and LRP/ ␣ 2 MR which might be detectable using antibodies with a broader specificity and improved affinity toward LDL-R. Experiments to produce such antibodies are presently being carried out in our laboratory.