Construction of a dual chain pseudotetrameric chicken avidin by combining two circularly permuted avidins.

Two distinct circularly permuted forms of chicken avidin were designed with the aim of constructing a fusion avidin containing two biotin-binding sites in one polypeptide. The old N and C termini of wild-type avidin were connected to each other via a glycine/serine-rich linker, and the new termini were introduced into two different loops. This enabled the creation of the desired fusion construct using a short linker peptide between the two different circularly permuted subunits. The circularly permuted avidins (circularly permuted avidin 5 --> 4 and circularly permuted avidin 6 --> 5) and their fusion, pseudotetrameric dual chain avidin, were biologically active, i.e. showed biotin binding, and also displayed structural characteristics similar to those of wild-type avidin. Dual chain avidin facilitates the development of dual affinity avidins by allowing adjustment of the ligand-binding properties in half of the binding sites independent of the other half. In addition, the subunit fusion strategy described in this study can be used, where applicable, to modify oligomeric proteins in general.

Avidin, a glycoprotein found in chicken egg white as well as its distant relative, streptavidin, from Streptomyces bacteria, are nanoscale devices that perform one particular task, the harvesting of free biotin, extremely well. The avidin-biotin interaction is so strong that even the harsh conditions in the digestive tract are insufficient to break the ingested avidinbiotin complexes (1). It is understandable, therefore, that this highly stable interaction has been widely applied in numerous fields of the life sciences to probe, label, affinity separate, and target biomolecules. Collectively these applications are known as the (strept)avidin-biotin technology (2). This technology has been facilitated by the ease with which biotin can be coupled to almost any other molecule without compromising the strong (strept)avidin-biotin interaction or the function of the target molecule.
Both avidin and streptavidin are homotetramers encoded by a single gene (3)(4)(5). This fact, together with the almost perfect 222-point structural symmetry (6,7) and the orientation of the subunits in both proteins, guarantees that all four biotin-bind-ing sites have equally high affinity toward biotin. From the evolutionary point of view this also means that in both proteins all four binding sites co-evolve. The monomers of avidin and streptavidin are simple classical anti-parallel ␤-barrels. They have an identical topology of eight successive ␤-strands and their interconnecting loops. When the three-dimensional structures of these proteins are superimposed, it is evident that the termini and also the biotin-binding pocket are located in topologically analogous regions in both proteins. Two monomer pairs in each (strept)avidin tetramer share a large common interface and, hence, are known as the structural dimers. The complete tetramer is composed of two such structural dimers (6 -9).
During recent years, avidin and streptavidin have been engineered via site-directed mutagenesis by us (10 -16) and by other groups (17)(18)(19)(20)(21)(22)(23)(24)(25). In some studies the focus has been on the adjustment of the physicochemical properties of (strept)avidin, whereas in other studies the target has been the fine-tuning of the biotin binding affinity of the protein. Nevertheless, because these mutant protein subunits are single gene products, the desired changes, produced by mutations, take effect simultaneously in all (strept)avidin subunits.
In several cases, however, it would be of benefit to be able to alter, for example, the binding affinity in only some subunits of the tetramer while conserving the tight binding in the remaining binding sites. Chilkoti et al. (17) developed a partial solution to this problem by producing two separate streptavidin forms, one with natural high affinity biotin-binding capacity and the other with reduced affinity. They denatured and mixed these two forms, after which the mixture was renatured. Nevertheless, the refolding led to a wide variety of alternative forms: some contained four high affinity binding sites, whereas other forms had an ascending series of lower affinity binding sites, finally ending in the form that contained four lower affinity binding sites. It is arguable that genetic fusion of the subunits might be a more straightforward and effective strategy to create (strept)avidin molecules with divergent affinity properties. The N and C termini of the distinct (strept)avidin subunits are, however, located far away from each other in the quaternary structure; therefore, any simple fusion strategy would presumably fail.
A common approach to the study of protein folding and the significance of secondary structure topology is the creation of circularly permuted forms of the proteins in question (26). Usually in this approach the original N and C termini are brought together with a linker peptide, whereas the new termini are typically introduced into a loop region. In most cases proteins withstand these modifications rather well, exhibiting no radical alterations in their structure or function. Chu et al. (23) have described a circularly permuted streptavidin that * 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. The nucleotide sequence(s) reported in this paper has been submitted to the DDBJ/GenBank TM /EBI Data Bank with accession number(s) AJ616762.
§ Both authors contributed equally to this work. ʈ To whom correspondence should be addressed. Tel.: 358-14-2602272; Fax: 358-14-2602221; E-mail: kulomaa@csc.fi. displayed a three-dimensional structure almost identical to that of the native protein. In their study, the circular permutation strategy was used as a tool to delete the loop between ␤-strands 3 and 4 in streptavidin. This loop is functionally important because it undergoes an open-to-closed conformational change upon biotin binding (7,9). Consequently, when the new termini were introduced into this loop, the affinity of the resultant mutant for biotin collapsed by six orders of magnitude in comparison to that of wild-type (wt) 1 streptavidin.
In the present study, our objective was to create an avidin with four binding sites that would consist of only two polypeptide chains. At first, we constructed two circularly permuted forms of avidin in which the new termini were in an ideal position to allow the building of the desired subunit fusion. We describe the construction of this dual chain avidin (dcAvd) through the fusing of the monomers of the structural dimer and show that it formed a wt-like pseudotetrameric quaternary structure and that its four biotin-binding sites exhibited high affinity biotin binding. This construct can be used in the future as a structural scaffold to change the affinity parameters in some subunits while preserving the high biotin binding affinity in the remaining binding sites. This study also sheds light on the factors that should be considered when other multisubunit proteins are engineered by combining separate subunits into genetically continuous units.

EXPERIMENTAL PROCEDURES
Production and Purification of the Mutant Avidins-The circularly permuted avidins and dcAvd were produced in baculovirus-infected insect cells. They were purified from the cell extracts by 2-iminobiotinagarose affinity chromatography essentially as described in detail elsewhere (11,27).
Biotin-Binding Assays-An IAsys biosensor was used to determine the affinity of the proteins toward 2-iminobiotin and the reversibility of biotin binding, essentially as previously described in detail (10). In the reversibility assay, the sample proteins were allowed to bind to a biotinylated cuvette surface in phosphate-buffered saline containing 1 M NaCl. After equilibrium was reached, the cuvette was washed and filled with phosphate-buffered saline, 1 M NaCl containing biotin (423 g/ml), and the dissociation of the proteins was monitored for 1 h. Colorimetric 2-(4Ј-hydroxyazobenzene)benzoic acid assay for the determination of the free biotin-binding sites was performed essentially as described by Green (28) using a Beckman DU640 spectrophotometer.
Gel Filtration Chromatography-Quaternary status of the avidin mutants was determined by high performance liquid chromatography (HPLC) using a Superdex 200 HR 10/30 column (Amersham Biosciences) connected to a Shimadzu HPLC system with a SCL-10A VP system controller, RF-10A XL fluorescence detector, and SPD-M10A VP diode array detector. The data obtained were processed with the Class VP 5.03 program. As a running buffer we used 50 mM sodium phosphate, 650 mM NaCl, pH 7.2. All runs were performed with a flow rate of 0.5 ml/min. The molecular mass markers were bovine serum albumin (68 kDa), ovalbumin (43 kDa), and cytochrome c (12.4 kDa).
StabilityAnalyses-Proteinsampleswereacetylatedandtemperaturedependent dissociation of the subunits after heat treatment for 20 min at temperatures between 25 and 100°C was monitored from Coomassie-stained SDS-PAGE gels essentially as described in detail by Bayer et al. (29). Proteinase K assay was performed essentially as described by Laitinen et al. (11).

RESULTS
Planning of the Constructs-The three-dimensional structure of the avidin tetramer (coordinates kindly given by Dr. Oded Livnah) was examined to identify the ␤-strands of the barrels, along which polypeptide transition from one subunit to the other would create as little structural disturbance as possible. The study revealed that in the quaternary structure ␤-strand 4 in one subunit and ␤-strand 6 in neighboring subunit of the structural dimer are juxtaposed. Moreover, when ␤-strand 4 goes up in one barrel, ␤-strand 6 goes down in the neighboring subunit. Therefore, we decided to use the loops 4 3 5 and 5 3 6 of the neighboring subunits as the monomermonomer transition point in dcAvd (Fig. 1). Determination of the transition point automatically determined the positions of the termini of the two circularly permuted avidins, which were as follows: the N terminus of the first circularly permuted avidin (cpAvd5 3 4) is located before ␤-strand 5 in the wildtype sequence, and the new C terminus follows ␤-strand 4. Likewise, the second circularly permuted avidin (cpAvd6 3 5) started just before ␤-strand 6 and ended after ␤-strand 5 (Fig. 1A).
In cpAvd5 3 4 and cpAvd6 3 5, the original N and C termini were brought together with an artificial hexapeptide loop (GGSGGS). Both constructs started with the normal signal peptide of avidin, followed by the first three N-terminal amino acid residues of wt avidin to ensure appropriate cleavage by signal peptidase. Finally, the desired subunit fusion was constructed by combining the two circularly permuted avidins described above (Fig. 1). A tripeptide linker (SGG) was used to connect the C terminus of cpAvd5 3 4 to the N terminus of cpAvd6 3 5, and the resultant subunit fusion protein (Gen-Bank TM accession number AJ616762) was designated dual chain avidin (dcAvd).
Protein Production and Purification-Production of the proteins was done in baculovirus-infected insect cells essentially as previously reported (27). All constructs were soluble and were efficiently purified in a single step on 2-iminobiotin-agarose column (data not shown).
Biotin Binding Experiments-Reversibility of biotin binding and affinity toward 2-iminobiotin were determined by surface plasmon resonance using an IAsys biosensor (Table I). CpAvd5 3 4 showed binding characteristics similar to those of wt avidin in both assays, whereas cpAvd6 3 5 was more reversible and exhibited reduced affinity when compared with that of wt avidin. Interestingly, dcAvd showed similar reversibility to that of wt avidin, and its affinity toward 2-iminobiotin showed only a negligible decrease. The number of functional biotin-binding sites per dcAvd pseudotetramer ("pseudo" because, in fact, it is a dimer with four binding sites) was determined by a colorimetric 2-(4Ј-hydroxyazobenzene)benzoic acid assay according to Green (28). The results from two independent experiments gave an approximation of 3.3 free biotinbinding sites/molecule. Structure Analyses-According to gel filtration chromatography, the deduced molecular masses (Table II) indicated that all constructs had folded correctly and formed quaternary structures resembling that of wt avidin; circularly permuted forms were tetramers and dcAvd was a dimer (pseudotetramer). In the denaturing SDS-PAGE both circularly permuted avidins had apparent molecular masses corresponding to a monomeric form (Fig. 2), whereas dcAvd had a molecular mass of ϳ32 kDa, which corresponds well to its theoretical molecular mass. Moreover, we also observed that cpAvd6 3 5 was more glycosylated than cpAvd5 3 4.
Stability Analyses-All the avidin proteins were subjected to a heat/SDS-PAGE stability analysis (29). The results (Table II) revealed that the mutants were somewhat less stable than wt avidin. The durability of the avidins in the presence of proteinase K was also tested (Fig. 3), and the mutated avidin forms were found to be more prone to cleavage than wt avidin, both in the absence and presence of biotin. DISCUSSION There are two main reasons for the plethora of studies describing different chemical modifications and genetically engineered forms of avidin and streptavidin. The first is naturally connected to the widespread utilization of these proteins across a multitude of applications. This has led to regulation of their physicochemical properties and biotin-binding properties both to broaden the spectrum of their potential applications and conditions of use and to overcome some of the drawbacks in-herent in the (strept)avidin-biotin system. Secondly, the (strept)avidin-biotin pair serves as an interesting model system of an oligomeric protein that displays extremely high affinity toward a small ligand.
In the present study, we used avidin as a model by engineer- FIG. 1. Description of the avidin mutants. A, representation of secondary structure topology of the two different circularly permuted avidins: cpAvd5 3 4, left, and cpAvd6 3 5, right. The rearranged ␤-strands are numbered according to wild-type avidin. The old termini were connected to each other by the GGSGGS linker, and sequences were opened elsewhere in two different ways. The location of the binding site is indicated, and monomers are positioned roughly as they are in the structural dimer of avidin. Moreover, in dcAvd the C terminus of cpAvd5 3 4 is connected to the N terminus of cpAvd6 3 5 via the short linker (SGG) as indicated. B, the amino acid sequence of dcAvd with the locations of the wt ␤-strands indicated. The underlined ARK denotes the first three amino acids of wt avidin. The hexapeptide GGSGGS linkers and the monomer-monomer transition linker SGG are highlighted with boxes. The part that is derived from cpAvd5 3 4 is underlined with the red bar; the part derived from cpAvd6 3 5 is underlined with the blue bar. C, schematic illustration of the fused monomers of the wt structural dimer in dcAvd. The linker (GGSGGS) that connects the old termini and the intermonomeric linker (SGG) are circled. The left part in red is derived from cpAvd5 3 4. The right part in blue is derived from cpAvd6 3 5. ing the topology of the multisubunit protein that has its N and C termini of the distinct subunits far away from each other. The aim was to modify avidin so that its four binding sites would be composed of only two polypeptide chains, instead of the four found in wt avidin. We accomplished this by joining together the two monomers that form the structural dimer (8) in wt avidin. Understandably, we wanted the resultant protein to retain as far as possible the structurally and functionally important properties of the wt protein. Sanders et al. (30) succeeded in their modification of hemoglobin by first constructing a circularly permuted form of the ␣-chain and then fusing this novel subunit to the wt ␣-subunit. This fused pseudodimeric ␣-unit was able to oligomerize with the wt ␤-subunits, and the resultant chimera was functionally almost indistinguishable from the wt oligomer. Although elegant, this procedure differs somewhat from the dual chain avidin described in the present study: only two subunits of the heterotetrameric complex were joined to a single polypeptide chain and two others entered the complex as single units.
To achieve our goal, we were forced to modulate the natural topology of wt avidin in order to bring the termini of the subunits closer to each other. Therefore, we first constructed two circularly permuted forms. In the first mutant the termini were at one end of the barrel, whereas in the second mutant they were at the opposite end (Fig. 1). Because biotin binds to one end of the avidin barrel, we expected that in cpAvd6 3 5, where the new termini were introduced to that end, it could affect the biotin binding. According to this assumption, in cpAvd6 3 5, where the new termini appeared on the loop between ␤-strands 5 and 6 that contains the biotin-binding residues Trp-70, Phe-72, Ser-73, and Ser-75 (numbering according to wt avidin; Ref. 8), the biotin-binding affinity was slightly reduced. One reason for the lowered affinity might be that the new free termini, and therefore also these biotinbinding residues, have more freedom to move compared with the corresponding loop in wt avidin. However, when compared with the previously mentioned circularly permuted streptavidin (23), the affinity of cpAvd6 3 5 was relatively well pre- The association (k a ) and dissociation (k d ) rate constants and the affinity toward 2-iminobiotin for different avidins were determined with an IAsys optical biosensor. K d (equilibrium) was determined experimentally from the binding curves, whereas the K d (relation) was calculated from the rate constants. The reversibility value corresponds to reversibility from the biotin surface.   served. Furthermore, in cpAvd5 3 4, where the new termini were introduced into the loop between ␤-strands 4 and 5, which is at the non-binding end of the barrel, no major changes were detected in its biotin-binding properties.
Interestingly, dcAvd exhibited binding properties that were somewhere between the two circularly permuted avidins. One reason for this could be that the dcAvd pseudotetramer has two rather well preserved biotin-binding sites exhibiting strong affinity toward biotin originating from cpAvd5 3 4 and two biotin-binding sites with reduced affinity originating from cpAvd6 3 5. On the other hand, in dcAvd two of the new termini are fused with the SGG linker, which provides the transition from one monomer to the other, thereby possibly rescuing part of the structural rigidity of the wt loop (5 3 6) and therefore also retaining part of the binding affinity.
The thermal stability and proteinase K durability of the circularly permuted avidins and dcAvd were somewhat lower than in the case of wt avidin. Proteinase K seemed to digest the new loops faster (not shown) than loop 3-4, which it is able to break in wt avidin (31). This indicates that the artificial hexapeptide loops that connect the old termini were probably not optimal in terms of protein stability. Furthermore, for some unknown reason cpAvd6 3 5 seemed to be more glycosylated than cpAvd5 3 4. However, this may explain why the molecular mass of the former mutant, according to high pressure liquid chromatography, was slightly higher than that of the latter one.
There is an interesting uncertainty concerning the quaternary structure of dcAvd. Because of its structural symmetry, it may have two different quaternary structures. Depending on the outcome of this quaternary structure assembly, the termini of both dcAvd monomers may be orientated to the same or to the opposite face of the pseudotetramer. At present, we do not know what the precise orientation of the dcAvd subunits is. They may form a mixture of two different structural forms; alternatively, some unknown structural phenomenon may favor one over the other. However, it might be possible to introduce a non-symmetrical disulfide bridge between the dcAvd monomers in order to fix their quaternary structure to the desired assembly.
Although in this study we utilized avidin as a model of a multisubunit protein whose topology was extensively transformed, dcAvd also provides a potential structural scaffold for avidins with mixed affinity properties. Such forthcoming dcAvd variants may have enormous value in avidin-biotin technology applications. The rationale for these high expectations derives from the fact that dcAvd is a genetically fused entity, which enables the avidin subunit properties to be engineered separately. It should be possible to maintain the high affinity toward biotin in two of the binding sites while modifying the affinity in the other two sites as desired. One option could be to make two of the binding sites reversible, which would enable bound materials to be mildly detached by free biotin (11,14,32,33). Another procedure might be to modify half of the binding sites with smart polymer conjugates able to respond to changes in pH, light intensity, and temperature (34,35). It might then be possible to alter the binding characteristics of the modified sites by adjusting these physical/chemical parameters. In theory, it is even possible to adjust the binding characteristics so that the resultant protein is able to bind something totally different from biotin with half of the binding sites.