Identification of Residues in (cid:97) -Macroglobulins Important for Binding to the (cid:97) 2 -Macroglobulin Receptor/Low Density Lipoprotein Receptor-related Protein*

Variants of the receptor binding domain of both hu- man (cid:97) 2 -macroglobulin and the corresponding domain of hen egg white ovomacroglobulin have been expressed in Escherichia coli and refolded in vitro . Competition ex- periments with methylamine-treated (cid:97) 2 -macroglobulin for binding to the multifunctional (cid:97) 2 -macroglobulin re- ceptor identify two Lys residues (residues 1370 and 1374 in human (cid:97) 2 -macroglobulin) spaced by three amino acid residues as crucial for receptor binding. From this re-sult and mutational evidence from other ligands for the (cid:97) 2 -macroglobulin receptor, a tentative sequence motif for receptor binding is proposed. The (cid:97) 2 -macroglobulin receptor/low density lipoprotein recep-tor-related protein ( (cid:97) 2 MR/LRP) is 600-kDa endocytic mem-brane-bound low density which low lipoprotein and the very low

Each member of the ␣-macroglobulin family of proteinase binding proteins is recognized by the ␣ 2 MR/LRP except one, hen ovomacroglobulin (ovoM) (21). The ␣-macroglobulins contain a receptor recognition signal within its COOH-terminal domain (RBD, residues 1314 -1451, human ␣ 2 M), which is exposed upon proteinase complex formation or in some cases by cleavage of the internal thiol esters with methylamine (22)(23)(24)(25). This domain can be expressed in Escherichia coli and refolded, and it binds to ␣ 2 MR/LRP (26,27). A variant (RBDv) of human ␣ 2 M-RBD containing 15 upstream residues in addition to the 138 residues of the domain obtained by proteolysis of ␣ 2 M (23, 24) binds to ␣ 2 MR/LRP with nearly the same affinity as that found for interaction between one RBD in intact ␣ 2 M-proteinase and ␣ 2 MR/LRP (27).
In this study mutant RBDs of both ␣ 2 M and ovoM have been produced in order to study the receptor interaction further. Two Lys residues spaced by three amino acid residues are found to be critical for receptor binding. On the basis of this observation, the mutational evidence provided by other investigations on ␣ 2 M-RAP (45), LPL (43,44), and BPTI (15) and comparison of the sequences of the ligands versus the sequence of ovoM, a common sequence motif of importance for receptor binding is tentatively deduced and presented.

MATERIALS AND METHODS
Preparations of Mutant RBDv-In vitro mutagenesis of ␣ 2 M-RBDv (residues 1299 -1451 of human ␣ 2 M) was performed either on singlestranded template using the T 7 -gen in vitro mutagenesis kit as described by the manufacturer (U. S. Biochemical Corp.) or on doublestranded template using polymerase chain reaction essentially as described (28).
Before expression the mutant insert was ligated into pT 7 H 6 FX a as described (29) and modified by insertion of an oligonucleotide encoding the sequence GSLPQNPFSTSSTLPR (27). The nucleotide sequence of the construct was verified by DNA sequencing using Sequenase version 2.0 (U. S. Biochemical Corp.).
Recombinant protein was produced by expressing the plasmid * This work was supported by grants from the Danish Biotechnology Programme (Biomembrane and Protein Engineering Research Centers). 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 U.S.C. Section 1734 solely to indicate this fact.
Characterization of the Expressed Variants-The variants of RBDv were analyzed by SDS-polyacrylamide gel electrophoresis in order to determine whether its single intrachain disulfide bridge had been formed indicating correct folding (23,24). Gel filtration on a TSK3000SW column in 0.1 M sodium phosphate, pH 7.0, was used to investigate whether the products were monomeric or noncovalenty associated. The amino acid composition of mutant ␣ 2 M-RBDvs was determined after acid hydrolysis with 6 M HCl, 0.1% phenol, and 5% thioglycolic acid for 20 h at 110°C in vacuo using the procedure described (30) modified to permit determination of half-cystine as cysteine (31).
Binding Assay-In vitro binding of 125 I-␣ 2 M-MA to ␣ 2 M/LRP immobilized in microtiter wells and competition for ␣ 2 M binding sites by mutant RBDvs were performed as described (9,27). The concentration of mutant ␣ 2 M-RBDv stock solutions was determined after acid hydrolysis (mean of three analysis). The dissociation constant of each mutant was determined on the basis of at least two independent experiments.

Selection of Target Sites for Mutagenesis-
The selection of residues for mutagenesis studies was based on comparison of the primary structures of the receptor binding domains of nine known receptor binding macroglobulins with the recently determined sequence of the nonbinding hen ovoM (32). As is evident from Fig. 1, the sequences of the RBD part of nine receptor binding ␣-macroglobulins are very conserved. Because previous experiments revealed that blocking Lys residues of human ␣ 2 M-RBD abolished receptor binding (24) and 1425. Some of the residues selected for study were mutated into glutamic acid, glutamine, or asparagine residues to preserve hydrophilic character of residues, others were mutated into alanine. A list of the mutations performed is shown in Fig. 2.
Expression and Characterization of Variant Receptor Binding Domains-In a previous paper (27) we described the refolding of wild type human ␣ 2 M-RBDv (residues 1299 -1451) and demonstrated that we were able to produce and refold the domain and purify it in milligram quantity. All variant domains except the chimeric ovo␣ 2 M-RBD yielded a product with the intramolecular disulfide bridge formed correctly as judged by SDS-polyacrylamide gel electrophoresis analysis. The products were soluble and monomeric under nondenaturing conditions as determined by gel filtration on TSK3000SW, from which the product eluted as an approximately 20-kDa protein.
These results showed that none of the introduced mutations radically changed the folding properties of RBDv, indicating that none of the residues mutated affected interactions critical for the folding or the stability of the final structure. Furthermore, for each variant the amino acid composition was in agreement with the introduced mutation. Final yields of pure ␣ 2 M-RBDv variants were approximately 0.1-0.5 mg/liter culture. The chimeric product resulting from replacement of 20 residues in ovoM with 19 residues from ␣ 2 M was expressed successfully but failed to produce a soluble product, presumably due to disruption of the structural integrity of the domain.
Binding of Variant Domains to ␣ 2 MR/LRP-Apparent dissociation constants of the variant RBDvs were estimated by competition titration against 10 pM 125 I-␣ 2 M-MA on purified immobilized placental ␣ 2 MR/LRP. The titration data are shown in Fig. 2. Mutation of residues Lys-1306, Lys-1333, Lys-1361, Glu-1377, Lys-1425, Asp-1428, and Arg-1384 and mutation of either Lys-1425 or Asp-1428 did not have major impact on binding affinity (apparent K d in the range of 9 -20 nM compared with the 8 nM of wild type RBDv).
OvoM-proteinase complexes are not recognized by the ␣ 2 MR/ LRP (21). From the present experiments it can be concluded that the low affinity of ovoM for ␣ 2 MR/LRP is due to amino acid differences within its domain corresponding to RBDv and not due to permanent masking by other parts of the ovoM molecule.
A clear deviation from ␣ 2 M-RBDv wild type receptor binding affinity was found for the K1370A, K1374E, and K1374E, E1377Q ␣ 2 M-RBDv variants. The apparent dissociation constants of mutant K1374A and double mutant K1374E,E1377Q were both determined to be approximately 100 nM, and the apparent K d of mutant K1370A was found to be greater than 500 nM. Accordingly, these mutants exhibit an approximate 10or 50-fold decrease in affinity, compared with wild type ␣ 2 M-RBDv. Mutation of both Lys-1370 and Lys-1374 into alanine yielded a domain with affinity indistinguishable from the affinity of ovoM-RBDv.
Since Lys-1374 proved important for receptor recognition and is absent in ovoM, whereas the other important Lysresidue, Lys-1370, is present, we then tried to introduce a residue equivalent with Lys-1374 into ovoM by replacing Asp-1358 or Gln-1359 in ovoM, one of which would be expected to be the equivalent of residue 1374 of ␣ 2 M, by a Lys residue. Neither mutation, however, conferred binding capability to ovoM-RBDv. DISCUSSION ␣ 2 MR/LRP binds and mediates endocytosis of multiple ligands. The binding of ␣ 2 M-proteinase and ␣ 2 M-MA has been shown to be mediated by the COOH-terminal approximately 150 residues (24 -29), which is thought to constitute an independent domain (22,23,26,27,41). From chemical modification experiments on RBD, it was found that receptor binding seemed to depend on one or more Lys residues (24). In this work, guided by analysis of the pattern of Lys residues in the RBD part of receptor binding ␣-macroglobulins and the nonbinding ovoM, we have identified, by site-directed mutagenesis, two Lys residues located at positions 1370 and 1374, which are of importance for receptor binding. As seen from Fig. 1, Lys-1374 is replaced by an Asp residue in ovoM, but, curiously, Lys-1370 is conserved in that protein. From sequence comparisons several other sites of potential importance for receptor binding were also identified, but introduction of mutations had no effect on receptor binding. ␣ 2 MR/LRP recognizes a variety of apparently unrelated ligands and itself consists of repeating sets of similar structural modules. It can be speculated whether these ligands essentially bind in similar ways but with minor variations reflecting individual specificity in recognizing different sites on the receptor. If indeed a common structural framework is prerequisite for receptor recognition, it might be reflected in the primary structure of the ligands. It is, therefore, relevant to collate information from mutagenesis and deletion experiments for all known ␣ 2 MR/LRP ligands and inspect sequence information for elements of unity or similarity. Fig. 3 shows our interpretation of such collated information, and the proposed alignment of se- quence segments that are known or suggested to be binding motifs of LPLs, ␣-macroglobulins, ␣ 2 M-RAPs, PAI-1s, tissue plasminogen activators, PA, and BPTI does indeed exhibit some common primary structure features suggestive of an overall similarity in binding motifs.
The most important similarities are as follows. Position 1 preferentially contains an aromatic residue. At position 4 a hydrophobic residue is preferred, most commonly a Leu residue. Position 5 most frequently contains a basic residue. At position 8 a hydrophobic residue is present, most often a Val residue. Position 10 harbors the common Lys residue. Position 14 contains an acidic residue. RAP, however, contains a Gln residue instead. At position 22 a hydrophobic residue is present. Position 24 is an Arg residue, and position 32 is a hydrophobic residue.
Compared with the main group of binding motifs, BPTI contains a deletion at position 15-21. This deletion might well be compensated by other polypeptide segments brought into vicinity by disulfide bridging to cysteine 22. Explicitly, the residues at positions 5, 10, 14, and 24 (Arg-42, Lys-46, Asp-50, and Arg-53) are all exposed on the surface and located less than 10 Å apart in the crystal structure determined for BPTI (42).
Although speculative, this alignment is consistent with the mutational data on ␣ 2 M. Moreover, it is also in agreement with mutational data obtained on LPL locating the binding site to residues 378 -423 of human LPL (43) and providing evidence for the participation of Lys-407 (position 10) but not of Arg-405 (position 7), Lys-413 (position 17), or Lys-414 (position 18) in binding to ␣ 2 MR/LRP (44). Mutation experiments performed on ␣ 2 M-RAP indicating the presence of two binding regions (45), namely an NH 2 -terminal and a COOH-terminal domain, are consistent with the two regions shown in our alignment and, finally, residues that proved important for the binding of BPTI to the ␣ 2 MR/LRP include residues Arg-42 (position 5) and Lys-46 (position 10) of BPTI (15).
In conclusion, we have provided evidence for the participation of Lys-1370 and Lys-1374 in the molecular interaction between ␣ 2 M and its receptor. Several other mutations were performed, but none affected binding significantly. These results proved consistent with data obtained with other ligands, suggesting a common sequence motif for receptor recognition. This motif seems to originate from convergent evolution since none of these ligands appear related otherwise.