Engineering Soluble Monomeric Streptavidin with Reversible Biotin Binding Capability*

Monomeric streptavidin with reversible biotin binding capability has many potential applications. Because a complete biotin binding site in each streptavidin subunit requires the contribution of tryptophan 120 from a neighboring subunit, monomerization of the natural tetrameric streptavidin can generate streptavidin with reduced biotin binding affinity. Three residues, valine 55, threonine 76, and valine 125, were changed to either arginine or threonine to create electrostatic repulsion and steric hindrance at the interfaces. The double mutation (T76R,V125R) was highly effective to monomerize streptavidin. Because interfacial hydrophobic residues are exposed to solvent once tetrameric streptavidin is converted to the monomeric state, a quadruple mutein (T76R,V125R,V55T,L109T) was developed. The first two mutations are for monomerization, whereas the last two mutations aim to improve hydrophilicity at the interface to minimize aggregation. Monomerization was confirmed by four different approaches including gel filtration, dynamic light scattering, sensitivity to proteinase K, and chemical cross-linking. The quadruple mutein remained in the monomeric state at a concentration greater than 2 mg/ml. Its kinetic parameters for interaction with biotin suggest excellent reversible biotin binding capability, which enables the mutein to be easily purified on the biotin-agarose matrix. Another mutein (D61A,W120K) was developed based on two mutations that have been shown to be effective in monomerizing avidin. This streptavidin mutein was oligomeric in nature. This illustrates the importance in selecting the appropriate residues and approaches for effective monomerization of streptavidin.

(Strept)avidin with reversible biotin binding capability can extend the applications of the biotin-(strept)avidin technology. These molecules can be applied for affinity purification of biotinylated biomolecules, screening of ultratight binders binding to biotinylated biomolecules displayed on the phage display system, and development of reusable biosensor chips, protein/ antibody microarrays, and enzyme bioreactors (1,2). (Strept)avidin is a homotetrameric molecule with a biotin binding site in each subunit (3). The three-dimensional structure of (strept)avidin (4,5) suggests that a complete biotin binding pocket in each subunit requires the contribution of a tryptophan residue from an adjacent subunit. Site-directed mutagenesis studies also demonstrate the importance of this residue for tight biotin binding and subunit communications (6 -8). Therefore, development of monomeric (strept)avidin can be an attractive approach to engineer (strept)avidin with reversible biotin binding capability.
The engineering of (strept)avidin to its monomeric form is technically challenging. In the case of avidin, the first generation of engineered monomeric avidin can exist in the monomeric state only in the absence of biotin (9). This problem has been solved by the recent development of the second generation of monomeric avidin (10), which carries two mutations (N54A,W110K). Structural alignment of avidin and streptavidin indicates that these two residues correspond to Asp-61 and Trp-120 in streptavidin (11). However, a streptavidin mutein designated AK, 1 which carries the corresponding double mutations (D61A,W120K), does not become monomeric as demonstrated in the present study. This illustrates the need to identify a new set of critical residues in combination with effective approaches to generate monomeric streptavidin with minimal mutational changes. Development of monomeric streptavidin has been reported previously through mutation of three residues to alanine (12). However, the low affinity of this mutein toward biotin (K d ϭ 1.7 ϫ 10 Ϫ6 M) makes it less than ideal for many applications.
To develop better versions of monomeric streptavidin, three residues (Val-55, Thr-76, and Val-125) were selected for sitedirected mutagenesis. In combination with L109T mutation, a series of single, double, and quadruple streptavidin muteins were created and produced from Bacillus subtilis via secretion. As they were produced in the soluble form without the requirement of refolding (13,14), their oligomeric state can be rapidly analyzed by SDS-PAGE using culture supernatants from the streptavidin mutein production strains. The muteins were purified and further characterized by different approaches to confirm their monomeric state. Their kinetic parameters for biotin binding were determined by surface plasmon resonancebased biosensor studies. based oligonucleotide-directed mutagenesis. Five mutants (V125R, V125T, V55R, V55T, and T76R), each bearing a single mutation that results in the change of an amino acid residue as the name suggests, were constructed using pSSAV-Tcry as the template and the primers listed in Table I. The amplified products were digested with the pair of enzymes listed in Table I and cloned into pSSAV-Tcry. Five plasmids (pV125R, pV125T, pV55R, pV55T, and pT76R) resulted.

Construction of Streptavidin
Two double mutants (M2 and AK) were also constructed. For M2 (T76R,V125R), a ScaI/NheI-digested fragment of pV125R was used to replace the corresponding fragment in pT76R. For AK (D61A,W120K), the fragment bearing the two mutations was amplified by PCR using the primers SAVD61AF and SAVW120KB (Table I) and the template pSSAV-Tcry. The amplified fragment was digested by XbaI/ScaI and used to replace the corresponding fragment in pSSAV-Tcry.
Production and Purification of Streptavidin-Wild-type streptavidin was produced by B. subtilis WB800(pSSAV-Tcry) cultured in a defined medium (14). The secreted protein was purified to homogeneity using cation exchange followed by iminobiotin affinity chromatography (12). Production and purification of streptavidin muteins followed a similar scheme with two major modifications: super-rich medium (15) was used in place of the defined medium, and biotin-agarose (Sigma) was used in place of iminobiotin-agarose as the affinity matrix. Dialyzed sample containing partially purified mutein was loaded to a 1-ml biotin-agarose column. Streptavidin muteins were eluted from the column using 20 mM D-biotin in phosphate-buffered saline (PBS; 50 mM sodium phosphate, 100 mM NaCl, pH 7.2). Concentration of purified streptavidin was determined spectrophotometrically using the known extinction coefficient at 280 nm (16,17) for each individual mutein.
Determination of the Molecular Size of Streptavidin-Molecular mass of purified streptavidin and its muteins was estimated by both gel filtration and dynamic light scattering studies. Gel filtration was performed on the Bio-Rad biologic work station equipped with a Bio-Prep SE 100/17 column that had been calibrated with molecular mass protein markers (Bio-Rad). Molecular mass was also estimated from the hydrodynamic radius of the mutein obtained using a DynaPro MS dynamic light scattering instrument (Protein Solutions) that had been calibrated with lysozyme. Protein samples (2-3 mg/ml in PBS) were passed through a 0.02-m filter (Whatman Anodisc 13) immediately prior to measurement. The size distribution profile was analyzed using the manufacturer's Dynamics V6 software.
Proteinase K Digestion of Streptavidin and Its Muteins-Purified streptavidin and its muteins (30 M monomer) were treated with proteinase K (Invitrogen, 5 M) for 15 min at 30°C in 50 mM Tris-HCl containing 5 mM CaCl 2 , pH 8.0. The reaction was stopped by precipitation with trichloroacetic acid (18). Boiled samples of precipitated proteins were resolved by reducing SDS-PAGE. The same analysis was performed with streptavidin samples treated with biotin (1 mM final concentration) prior to proteinase K digestion.
Cross-linking Reactions-Cross-linking of streptavidin and its muteins was carried out using ethylene glycol bis(sulfosuccinimidyl succinate) (sulfo-EGS) (Pierce) as the cross-linker. A typical reaction mixture (20 l) contained the purified mutein (0.25 mg/ml) and sulfo-EGS (10fold molar excess over the protein) in PBS. After 30 min at room temperature, the reaction was quenched with Tris-HCl (30 mM, pH 7.5). Aliquots of the cross-linking reaction samples were boiled and examined by SDS-PAGE. Lysozyme (Sigma, 0.25 mg/ml) was included in the study to help establish the optimal reaction conditions.
Kinetic Analysis of Streptavidin Muteins-The kinetic parameters (both on and off rates for interaction with biotin) of streptavidin muteins were determined in real time using the surface plasmon resonance-based BIAcoreX biosensor. Biotin-conjugated bovine serum albumin immobilized on a CM5 sensor chip was used to study the reversibility of biotin binding (12).
Computer Programs for Streptavidin Analyses-Swiss-pdb Viewer (19) was used to display streptavidin (Protein Data Bank code 1SWE (20)), analyze interfacial residues, measure distance between residues, and align the structures of streptavidin and avidin. Interfacial contact areas were calculated using the protein-protein interaction server (21) and the Formiga module in the Sting Millennium Suite (22). The plots of accessible surface area of individual residues in streptavidin in either the monomeric or tetrameric state were generated using the Protein Dossier module in the Sting Millennium Suite.

Selection of Key Residues in Streptavidin for Site-directed
Mutagenesis-Tetrameric streptavidin is arranged as a dimer of dimers (Fig. 1A). The interface between subunits A and B (and between C and D) has the most extensive subunit interactions. The interfacial contact area between A and B is ϳ1,557 Å 2 with 17 H-bonding interactions, two salt bridges, and numerous van der Waals interactions. The interface contact between A and D is also extensive with a contact area of 525 Å 2 and two interfacial H-bonding interactions. The weakest interface interaction is between subunits A and C with an interfacial contact area of 171 Å 2 . To engineer monomeric streptavidin with a minimal number of mutated residues, an attractive approach is to introduce both charge repulsion and steric hindrance at these interfaces. As protein has structural plasticity (23)(24)(25), it is vital to select interfacial residues located on a rigid surface to maximize the effects of charge repulsion and steric hindrance. Because streptavidin subunit forms an eightantiparallel stranded ␤-barrel structure (4, 5), the selected residues should be located on the ␤-strands rather than in the loop regions. Furthermore the selected residue in one subunit should be located very close to the equivalent residue or a charged residue in another subunit at the interface. Examination of interfacial residues (Fig. 1, B-D) shows that Thr-76, Val-125, and Val-55 meet the criteria. Hence they were selected for mutagenesis. Effects of Single Mutations on Monomerization of Streptavidin-Streptavidin muteins carrying a single amino acid change at the selected site were produced in their soluble form by B. subtilis via secretion. Analysis of non-boiled culture supernatants by SDS-PAGE offers a quick screen for the mutation effect (12). Weaker subunit interaction would result in a higher percentage of the sample in the monomeric state on the SDSpolyacrylamide gel. Because biotin can strengthen subunit interaction, samples were analyzed in the presence or absence of biotin (9,26). The impact of the mutation on weakening of the subunit interaction followed the order: T76R Ͼ V125R Ͼ V125T Ϸ V55R Ͼ V55T ( Fig. 2 and Table II). The T76R mutein (designated M1) existed 100% in the monomeric state on the SDS-polyacrylamide gel even in the presence of biotin. In contrast, V55T mutation had the lowest impact with the majority of molecules in the tetrameric state even in the absence of added biotin. Presence of biotin shifts the majority of the three remaining muteins (V125R, V125T, and V55R) to the tetrameric state. As expected, changing valine to arginine exerted greater impact than changing it to threonine. This is true for both Val-125 and Val-55.
Effects of Multiple Mutations on Monomerization of Streptavidin-To develop idealized monomeric streptavidin muteins that are more likely to remain in the monomeric state at high streptavidin concentrations and have excellent reversible biotin binding capability, two more muteins were created. M2 is the double mutant carrying both the T76R and V125R mutations. M4 is a quadruple mutant carrying T76R, V125R, V55T, and L109T mutations. In this combination, the three interfacial hydrophobic residues Val-125, Val-55, and Leu-109 were changed to hydrophilic ones. The last construct is AK, a double mutant (D61A,W120K) carrying two mutations (equivalent to those performed in avidin) that have been shown to convert tetrameric avidin to the monomeric state (10). As shown in Fig.  3 and Table II, just like M1, all these muteins existed in monomeric state on the SDS-polyacrylamide gel even in the presence of biotin.
Purification of Streptavidin Muteins-Purification of M4 was used as an example to illustrate the process (Fig. 4). Proteins partially purified by ion exchange chromatography (lane 2) were applied to a biotin-agarose column. M4 could be readily eluted off from the column using biotin-containing buffer as the eluant (lanes 5-7). Pure streptavidin mutein obtained by this simple procedure, after removal of biotin by dialysis, could be used for biochemical characterizations. To demonstrate that dialysis could effectively remove any bound biotin from the M4 mutein, the dialyzed sample was reloaded to the biotin-agarose matrix. Over 95% of the sample could be retained on the column and eluted off from the column using biotin (data not shown). Of all the muteins, M1 tended to have a long trailing tail during elution. This indicates that M1 may not have the desirable reversible biotin binding property. Therefore, it was not characterized further.
Determination of Apparent Molecular Mass of Streptavidin Muteins by Gel Filtration-Observation of 100% monomerization of the streptavidin mutein using a non-boiled sample for SDS-PAGE does not always truly reflect its existence in the monomeric state in solution because SDS can promote subunit dissociation (27,28). The apparent molecular masses of the purified wild-type streptavidin and the three muteins (M2, M4, and AK) were estimated by gel filtration (Supplemental Fig. 2A and Table III). The expected molecular mass of monomeric streptavidin is 16.5 kDa. M2 and M4 in the absence of biotin showed the apparent molecular masses of 19.95 and 21.87 kDa, respectively. These masses increased slightly in the presence of biotin. These data suggest that the muteins are monomeric in nature because their masses are less than that for the streptavidin dimer (33 kDa). In contrast, the AK mutein showed an apparent molecular mass of 45.66 kDa even in the absence of biotin. This indicates the oligomeric nature of this mutein. Supplemental Fig. 2B shows the elution profile of purified M4 (in the absence of biotin) from the gel filtration column. The sample (loaded at 2 mg/ml) was eluted as a single peak. There is no evidence for the presence of tetrameric streptavidin, which would be eluted at 30.5 min.

Determination of Apparent Molecular Mass of Streptavidin
Muteins by Dynamic Light Scattering-Because the apparent molecular mass of wild-type streptavidin in the absence of biotin is 10 kDa less than expected (56 instead of 66 kDa) as determined by gel filtration, dynamic light scattering (29) was used as a second method to estimate the apparent molecular masses. The apparent molecular mass of wild-type streptavidin obtained in this way (69 kDa) was closer to that expected (66 kDa) ( Table III). The apparent molecular masses for both M2 and M4 in the absence of biotin indicated that they were in the monomeric state. Addition of biotin caused only a slight increase in their apparent molecular masses. The AK mutein again was found to be oligomeric independent of the presence or absence of biotin.
Proteinase K Sensitivity of Streptavidin Muteins-Monomeric streptavidin is expected to be more susceptible to proteinase K digestion (10). Therefore, wild-type streptavidin and its muteins were treated with proteinase K (Fig. 5A). Wild-type streptavidin was converted to the core form independent of the presence or absence of biotin. Under the condition used, the core streptavidin was resistant to further degradation by proteinase K. In contrast, all three muteins including AK, M2, and M4 were much more susceptible to proteinase K digestion. Sensitivity to proteinase K is more apparent for M2 and M4, which were completely digested independent of the presence or absence of biotin. This property is consistent with the monomeric nature of these muteins. The AK mutein behaved differently. Although most of it was digested by proteinase K in the absence of biotin, it became much more resistant to proteinase K when biotin was present.
Cross-linking of Streptavidin and Its Muteins-To strengthen the idea that both M2 and M4 are monomeric whereas AK is oligomeric in nature, protein cross-linking was carried out using sulfo-EGS as the cross-linking agent. Sulfo-EGS reacts with both the accessible ␣-amino groups at the N termini and the surface-exposed ⑀-amino groups of the lysine side chains in proteins. Secreted wild-type streptavidin has eight lysine residues in each subunit. The three-dimensional structural model of streptavidin suggests that lysine 121 in subunit A is 14.1 Å from lysine 121 in subunit D. As the spacer arm in sulfo-EGS is 16.1 Å, subunits A and D (same for subunits B and C) should be easily cross-linked by sulfo-EGS. Also it is possible to have cross-linking between subunits A and B as the N-terminal region from subunit A, which contains two lysine residues, is likely to be positioned close to lysine 80 in subunit B. The same is true for subunits C and D. Therefore, one should be able to differentiate tetrameric streptavidin from the monomeric form with the observation of cross-linked tetrameric streptavidin using sulfo-EGS. Lysozyme, well known to be monomeric in solution (30,31), served as the negative control. Fig. 5B shows that the amount of dimeric lysozyme increased slightly in the presence of the cross-linking agent. This helped set the upper limit of the concentration of sulfo-EGS to be used under the experimental condition. The wildtype streptavidin subunit had an apparent molecular mass of 19 kDa on the SDS gel. After treatment with sulfo-EGS, most of these subunits were cross-linked to dimers and higher oligomers with small amounts remaining in the monomeric state. M2 and M4 muteins behaved very similarly (data for M2 are not shown). The majority of the M2 and M4 muteins after the cross-linking treatment migrated as monomers with small amounts in the dimeric form. These dimers may represent cross-linked monomeric subunits that were artificially generated in the same manner as with lysozyme. AK showed a cross-linking profile very similar to that of the wild-type streptavidin. These data strongly support the idea that M2 and Ϫve control, culture supernatant from WB800(pWB705HM) (34) that did not produce any streptavidin.
M4 muteins are monomeric, whereas the AK mutein is oligomeric in solution.
Reversible Interaction between Streptavidin Muteins and Biotin-The on rate and off rate of the interactions between streptavidin muteins and biotin were determined by surface plasmon resonance-based BIAcore biosensor (12). As shown in Table  IV (graphical plots for M4 are shown in Supplemental Fig. S3), M2, M4, and AK had their dissociation constant (K d ) in the range of 10 Ϫ7 M. The off rates (k d ) for these muteins were almost the same, whereas the on rate (k a ) for the AK mutein was slightly lower than the rest. One of the factors affecting the on rate is the diffusion coefficient (or molecular mass) of the streptavidin molecule. Because AK is oligomeric in nature, this may account for the lower on rate for this mutein-biotin interaction. DISCUSSION Although streptavidin and avidin have similar three-dimensional structures and biotin binding properties, development of monomeric streptavidin is much more challenging for two reasons. First, streptavidin has stronger subunit interfacial interactions than avidin (27,28). More potent mutations are required to weaken this strong interface interaction. Second, monomerization of streptavidin may result in the surface exposure of hydrophobic residues that normally would be buried at the interface in tetrameric streptavidin. This can potentially affect the solubility of the monomeric streptavidin and lead to reassociation of the monomers. The problem can be less dramatic for avidin, which is a glycosylated protein with a carbohydrate chain in each of the avidin subunits.
Despite the challenge, our study illustrates that, by selecting a critical residue located on a rigid surface for mutagenic study and the introduction of charge repulsion and steric hindrance at the interface, a single mutation (T76R) can be greatly effective in developing monomeric streptavidin. Besides the suggestion from SDS-PAGE analysis, gel filtration study of the M1 mutein also indicated that the majority of M1 was eluted at a position corresponding to the monomeric form (data not shown). The main drawback for this mutein is its elution behavior on the biotin-agarose column. The elution profile had a typical long trailing tail. Furthermore more M1 could be recovered by soaking the column overnight with buffer containing biotin. This suggests that some of the M1 population have higher affinity to the matrix.
To increase the efficiency of streptavidin monomerization, M2 mutein was developed by combining two potent mutations (T76R,V125R). Data from gel filtration study, dynamic light scattering, sensitivity to proteinase K, and cross-linking reac-   To ensure that the streptavidin mutein will stably remain in the monomeric state, more mutations were introduced to M2. The exposure of the AB (or CD) interface will expose three hydrophobic residues: Val-55, Leu-109, and Val-125. In the M2 mutein, Val-125 has been changed to arginine. We chose to further convert Val-55 and Leu-109 to threonine, a more hydrophilic residue, to develop the M4 mutein. Conversion to threonine instead of arginine is preferred because proteins with high pI are known to have nonspecific interactions via electrostatic interactions (32,33). The calculated pI of M2 is 8.1. If both Val-55 and Leu-109 are converted to arginine, the resulting mutein will have a pI of 9.2. This may increase the chance of charge-related nonspecific interactions. The conversion of these residues to negatively charged residues was not considered because both Val-55 and Leu-109 are located on the ␤-strands, and negatively charged residues are relatively poor ␤-strand formers.
M4 mutein shares with M2 mutein many desirable features of an idealized monomeric streptavidin. They both exist in the monomeric state at a reasonably high protein concentration (2 mg/ml or more as used in the dynamic light scattering study). Both have excellent reversible biotin binding capability as reflected by their on rate and off rate for biotin interaction. Both have a moderate pI value of 8.1 so that charge-related nonspecific interactions will be minimal. In addition, M4 has two features that make it even more attractive in practice. Molecules of M2 mutein tend to aggregate in solution. Filtration of M2 through a 0.02-m filter was essential for obtaining a good signal of the mutein for dynamic light scattering studies because of poor signal detection caused by the presence of small amounts of large aggregates in an unfiltered sample. On the other hand, a decent signal could at least be obtained with an unfiltered sample of similarly prepared M4. Thus, conversion of the two hydrophobic residues (Val-55 and Leu-109) to the more hydrophilic threonine residue did help minimize aggre-gation of the mutein. Another attractive feature is that M4 has a remarkably sharp elution profile with its purification using biotin-agarose. Over 95% of the mutein could be readily eluted off from the column using just 2 column volumes of the eluant, leading to a high rate of protein recovery.
In the site-directed mutagenesis study, it is not difficult to understand the impact of the mutations in the following order: T76R Ͼ V125R Ͼ V55R. Analysis of the solvent accessibility of individual amino acid residues with tetrameric streptavidin indicates that the solvent-accessible area of Thr-76 in subunit A is zero. Its close distances to both Thr-76 and Arg-59 in subunit B and location on a rigid surface of a ␤-barrel structure make it an ideal residue to be changed to arginine to achieve the maximal electrostatic repulsion and steric hindrance effects at the subunit interface (Fig. 1B). The surface-accessible area of Val-125 in subunit A is 1.78%. It has extensive interactions with Leu-109, Trp-120, Thr-123, and Val-125 in subunit D (Fig. 1C); Leu-109 in subunit B; and Gln-107 in subunit C. Its conversion to arginine results in charge repulsion in subunit D and potential steric hindrance for subunits B, C, and D. As for Val-55 in subunit A, its surface accessible area is 29.7%; and it is only close to Arg-59 in subunit B at the interface (Fig. 1D). Thus, V55R has the least impact on monomerization.
Although AK mutein shows reversible biotin binding property and monomeric behavior on the SDS-polyacrylamide gel, it clearly exists in the oligomeric state in solution as suggested by gel filtration studies, dynamic light scattering, and cross-linking pattern. This study illustrates the importance of selecting critical residues and effective approaches to achieve the maximal monomerization effect on tetrameric streptavidin. FIG. 5. Determination of the monomeric or oligomeric states of wild-type streptavidin and its muteins. Pictures show the Coomassie Blue-stained SDS-polyacrylamide gel. A, proteinase K digestion. B, cross-linking study using sulfo-EGS as the cross-linker. All samples were boiled prior to loading. M, molecular weight markers; wt, wild-type streptavidin; L, lysozyme. Numbering represents streptavidin molecules in monomeric (1), dimeric (2), and oligomeric (3)(4)(5) states, respectively.