Cooperative hydrogen bond interactions in the streptavidin–biotin system

Abstract

The thermodynamic and structural cooperativity between the Ser45– and D128–biotin hydrogen bonds was measured by calorimetric and X-ray crystallographic studies of the S45A/D128A double mutant of streptavidin. The double mutant exhibits a binding affinity ~2 × 10⁷ times lower than that of wild-type streptavidin at 25°C. The corresponding reduction in binding free energy (ΔΔG) of 10.1 kcal/mol was nearly completely due to binding enthalpy losses at this temperature. The loss of binding affinity is 11-fold greater than that predicted by a linear combination of the single-mutant energetic perturbations (8.7 kcal/mol), indicating that these two mutations interact cooperatively. Crystallographic characterization of the double mutant and comparison with the two single mutant structures suggest that structural rearrangements at the S45 position, when the D128 carboxylate is removed, mask the true energetic contribution of the D128–biotin interaction. Taken together, the thermodynamic and structural analyses support the conclusion that the wild-type hydrogen bond between D128–OD and biotin–N2 is thermodynamically stronger than that between S45–OG and biotin–N1.

The binding of biotin (vitamin H) to streptavidin and avidin has been the subject of considerable fundamental and applied interest. This protein–ligand pair represents one of the strongest noncovalent affinities known. From the structure of the bound complex, it is known that the binding energy derives from multiple types of interactions between the protein and biotin (Hendrickson et al. 1989; Weber et al. 1989). There are large hydrophobic and van der Waals contributions arising from tryptophan contacts to biotin (Chilkoti et al. 1995; Sano and Cantor 1995; Dixon and Kollman 1999). There are also seven specific hydrogen bonding interactions, five of them deep within the pocket and shielded from competition with solvent, including three to a single oxygen on biotin (Klumb et al. 1998; Freitag et al. 1999; Hyre et al. 2000, 2002). Weber et al. (1992) postulated that a ure-ido-oxyanion resonance form of biotin is stabilized in the bound state, leading to the prediction that the hydrogen bonding contributions would be large.

Mutational analysis has demonstrated that the biotin–streptavidin hydrogen bonds indeed display exceptionally large mutational free energy alterations, among the largest mutational free energies ever observed (Klumb et al. 1998; Freitag et al. 1999; Hyre et al. 2000). Biophysical characterization of the S45A and D128A single mutants have shown that loss of hydrogen bonds to the ureido nitrogens decrease binding free energy by ~4.2 kcal/mol apiece at 37°C. Both mutants also displayed nearly identical equilibrium thermodynamic perturbations (Freitag et al. 1999; Hyre et al. 2000). The equilibrium binding enthalpy and entropy, and hence free energy, were within 0.2 kcal/mol of one another, as were the activation-barrier free energies. The only energetic difference was in the balance between enthalpic and entropic contributions to the dissociation barrier. The S45A/D128A double-mutant streptavidin (45/128) was created to determine the level of cooperativity between the interactions of these residues with biotin. Herein, we characterize the double mutant both thermodynamically and structurally, showing that subtle structural alterations in the D128A single mutant likely masks the full energetic contribution of the abrogated hydrogen bond.

Results

Structure

The structure of the biotin-bound S45A/D128A double mutant was determined by X-ray crystallography (Table 1). The 65 β-barrel Cα atoms of the double mutant were superimposed upon those of the wild-type and single-mutation structures (Fig. 1; Table 2). The structure revealed no gross changes from the wild-type protein, consistent with the tolerance to mutation observed in the core of our many previous mutants (Chilkoti et al. 1995; Klumb et al. 1998; Freitag et al. 1999; Hyre et al. 2000). The β-barrel backbone fold of the double mutant is most similar to that previously reported for the S45A single mutant (Hyre et al. 2000), and displays only the minor alterations previously reported for the D128A mutant (Freitag et al. 1999). However, there were noticeable structural changes in the biotin-binding pocket. The protein adjusts to the mutations by a combination of the alterations found in the individual mutants, with the main-chain near S45A shifting towards biotin and the main-chain near D128A shifting away from biotin. The biotin in the double mutant complex has shifted relative to that in the wild type, away from the ureido oxygen position at the bottom of the pocket as previously reported for the D128A–biotin complex.

Table 1.

Statistics for structural data collection and refinement

Data collection
Protein	S45A/D128A	Wild-type/biotin
Space group	P21	I222
Unit cell
a (Å)	50.6	46.1
b (Å)	97.8	93.3
c (Å)	52.2	103.8
β (8)	112.3
Monomers/au	4	2
Temperature	100 K	100 K
Resolution	1.65 (1.68–1.65)	1.40 (1.45–1.40)
Measured/unique reflections	1,009,902/57,334	383,001/44,171
Completeness (%)	99.6 (98.5)	98.8 (93.4)
Rmerge (%)	9.9 (54.2)	4.3 (20.0)
Source/λ (Å)	SSRL/0.97	ALS/1.00
Refinement
Resolution range (Å)	10.0–1.65	10.0–1.40
No. of unique reflections, 4s	41,655	38,979
No. of unique reflections, all	57,069	44,020
No. of parameters	35,047	18,964
No. of restraints	44,145	23,730
No. of atoms
Protein	3526	1844
Solvent	293	239
Heteroatoms	64	32
Rfactor (4σ cutoff) (%)	17.6	11.9
Rfree (4σ cutoff) (5%)	26.6	17.8
Rfactor (0σ cutoff) (%)	20.4	12.8
Rfree (0σ cutoff) (5%)	29.5	18.9
RMSD (Å)
Bond length	0.007	0.012
Bond angles	0.027	0.032

Figure 1.

Structural superposition of bound wild-type (white), S45A (blue), D128A (red), and S45A/D128A double-mutant (green) streptavidin structures. (A) Stereoview of the overall structures. (B). Details of the binding pocket described in the text. The bound structures were superimposed by least-squares fit of the 65 Cα atoms in the β-barrel core of each monomer. The nearly identical structure of these cores can be seen in the overlay of the β-sheet backbone in the β-barrel core. The protein adjusts to the two mutations by a combination of the shifts found in the individual mutants, with the main-chain near S45A shifting toward biotin, the main-chain near D128A shifting away from biotin, a water molecule replacing the missing D128 carboxylate, and the biotin shifting away from the bottom of the pocket (the new water in the D128A structure is hidden by that in the double mutant due to nearly complete overlap). Hydrogen bonds involving Q24, S45, D128, and biotin in wild-type streptavidin are shown by dotted lines.

Table 2.

Shifts in atom positions between variants, Å

Prot #1	Prot #2	bio-N1	bio-N2	bio-O3	β-Cα RMSD
45/128	wild-type	0.61	0.60	0.69	0.274
45/128	S45A	0.71	0.69	0.83	0.201
45/128	D128A	0.16	0.37	0.34	0.396
S45A	wild-type	0.26	0.28	0.29	0.223
D128A	wild-type	0.64	0.86	0.44	0.418
Wild-type	apo-wild-type	—	—	—	0.290

Determination of absolute K_A for wild-type and single mutants from wild-type kinetics

To allow calculation of more accurate affinity constants for the wild-type and single-mutant proteins as the basis for comparison with the double mutant, the association rate of streptavidin was measured directly by stopped-flow fluorescence (Fig. 2). Due to the rapid association rate, the measurement was made under second-order conditions. The association rate fit the data well at 7.5 × 10⁷ M⁻¹sec⁻¹, with little dependence on ionic strength (Table 3). This value is consistent with the association rate for avidin previously used for calculation of the wild-type streptavidin affinity from its dissociation rate (7 × 10⁷ M⁻¹sec⁻¹), though the updated K_A based on the newly determined association rate is calculated to be slightly higher, 1.9 ×10¹³ M⁻¹.

Figure 2.

Fluorescence emission of streptavidin during biotin association. The intrinsic tryptophan fluorescence decreases in second-order fashion as streptavidin is rapidly mixed with an equimolar biotin concentration in 110 mM NaCl, 10 mM phosphate buffer (pH 7.0), 25°C. The solid line indicates the fit of a second-order model with k_ON = 7.5×10⁷ M⁻¹sec⁻¹; note curvature of semilog plot.

Table 3.

Association rate constants for wild-type streptavidin

Ligand	NaCl M	k2nd× 108 M−1sec−1
Biotin	0a	0.87 ± 0.12
	0.11	0.75 ± 0.06
Biotinamide	0	1.43 ± 0.42
	0.11	1.29 ± 0.14

^a All reactions were in 10 mM sodium phosphate buffer (pH 7.0) with increasing NaCl at 25°C, and were performed under second-order conditions.

ITC characterization of double-mutant binding free energies

The biotin-binding affinity of the double mutant was predicted to be 10⁷ by a linear combination of the decreases in affinity of the individual S45A and D128A mutations (1400× and 1600×, respectively, at 25°C; 1.9 × 10¹³/(1400 × 1600) = 8.7 × 10⁶). This range is ideal for using ITC to determine K_A (Wiseman et al. 1989). Based on this, ITC experiments were performed using conditions optimized for K_A in the range of 105–10⁸, and ideal for 10⁶–10⁷. The data and model fits are shown in Figure 3, with both measured and derived thermodynamic parameters shown in Table 4. The midpoint of the transition indicates a stoichiometry of one biotin per streptavidin monomer, and thus 100% functional protein.

Figure 3.

Isothermal titrating calorimetry analysis of S45A/D128A streptavidin. (A) Raw data of biotin binding titration at 15°C. (B) Plot of three experiments, one each at 15°C, 25°C, and 32°C, with both the raw data (points) and associated global fits (lines). (C) Plot of globally fit model (solid line) onto titration at 32°C. Also shown are the fraction of streptavidin bound with biotin (long-dash line) and the curve expected if the individual mutations were additive in a simple noncooperative linear manner, i.e., the sum of the alterations in each mutant alone (short-dash line).

Table 4.

Basic thermodynamic parameters for S45A/D128A streptavidin

ΔG298	−8.1 ± 0.5 kcal mol−1
ΔH298	−14.2 ± 0.3 kcal mol−1
TΔS298	−6.1 ± 0.5 kcal mol−1
ΔCP	0.16 ± 0.02 kcal mol−1 C−1
KA,298	8.1 × 105 ± 2.2 × M−1

Multiple ITC titrations were done at each of three different temperatures to determine the complete equilibrium thermodynamics of the system, including ΔC_P by the temperature dependence of ΔH. All data sets were fit simultaneously by global parameters of ΔH₂₉₈, ΔS₂₉₈, and ΔC_P, and individual parameters of n for each experiment, thereby increasing the global degrees of freedom relative to individual fits. The results are given in Table 4. The recovered parameters are different than those predicted above, indicating that the mutations do not combine linearly. Indeed, the affinity is significantly weaker, although the binding enthalpy is more favorable, indicating a less favorable change in binding entropy.

Cooperativity

To examine cooperativity, additional quantities were derived from the relations governing the thermodynamic and kinetic quantities as described elsewhere for S45A (Hyre et al. 2000). These quantities are also listed in Table 5, along with the values predicted by linear combination of the mutational effects and the differences between these and the experimentally derived values. There is an ~11-fold cooperative effect in the double mutant, making it 18 times lower in binding affinity than is predicted by linear combination. From the ΔΔH and TΔΔS values of −1.6 and −3.0 kcal/mol, respectively, it is evident that the entropic effects are approximately twice as significant as the enthalpic effects. These differences are also evident on the thermodynamic cycle derived for creating the double mutant from wild type in single-mutation steps (Fig. 4).

Table 5.

Extended thermodynamic parameters for wild-type and mutant streptavidin

Equilibrium
Mutant	KA M−1	KD ratio factor	ΔG° kcal/mol	Δ ΔG° kcal/mol	ΔH°a kcal/mol	Δ ΔH° kcal/mol	Δ CPa eu	ΔSa eu	TΔS kcal/mol	TΔ ΔS kcal/mol
Wild-type	1.9E + 13	1	−18.1	0.0	−24.5	0.0	−345	−21	−6.4	0.0
S45A	1.4E + 10	1414	−13.8	4.3	−18.4	6.1	−223	−15	−4.6	1.8
D128A	1.2E + 10	1582	−13.8	4.4	−18.6	5.9	−238	−16	−4.8	1.5
S45A/D128A predicted	8.7E + 06	2.2E + 06	−9.5	8.7	−12.5	12.0	−116	−10	−3.1	3.3
S45A/D128A measured	8.2E + 05	2.4E + 07	−8.1	10.1	−14.2	10.3	−161	−20	−6.1	0.3
Cooperativityc	11	11	1.4	1.4	−1.6	−1.6	−45	−10	−3.0	−3.0
Kinetic

Mutant	kOFF 10−6 sec−1	ΔkOFF factor	T1/2 min	kON KA*kOFF	ΔG‡ kcal/mol	ΔH‡a kcal/mol	TΔS‡ kcal/mol	ΔS‡a cal/mol
a Reference parameter at 298 K.
b Estimated.
c Cooperativity: [measured − predicted] for energies, [measured/predicted] for all others.
Wild-type	3.8	1	3001	7.5E + 07	24.8	30.4	5.6	18.8
S45A	8796	2285	1.3	1.2E + 08	20.2	25.8	5.6	18.7
D128A	7004	1819	1.6	8.6E + 07	20.4	20.7	0.3	1.0
S45A/D128A predicted	1.6E + 07b	4.2E + 06b	7.2E − 04b	1.4E + 08b	15.8b	16.1	0.3	1.0
S45A/D128A measured	1.7E + 08b	4.4E + 07b	6.8E − 05b	9.4E + 07b	14.4b	15.6b	1.2b	4.0b
Cooperativityc	11	11	1/11	0.7	−1.4b	−0.5b	0.9b	3.1b

Figure 4.

Thermodynamic cycles for stepwise mutation of wild-type streptavidin to the S45A/D128A double mutant, showing ΔG₂₉₈, ΔH₂₉₈, and TΔS₂₉₈. The cycle starts with wild type in the upper left corner and proceeds to the single mutants, with the thermodynamic perturbation shown next to the arrows. The cycle continues to the double mutant, with the difference in thermodynamic parameters between each single mutant and the double mutant shown next to those arrows. The units on the quantities in the figure are kcal/mol.

Discussion

The S45A/D128A double-mutant streptavidin was constructed to examine the cooperativity between the two strongest hydrogen bonds to biotin that have been identified by site-directed mutagenesis (Freitag et al. 1999; Hyre et al. 2000). They are also interesting from a structural standpoint, as both positions are hydrogen bonded to the symmetric ureido nitrogen positions from either side of the binding pocket. The double mutant exhibits an interesting combination of structural alterations that are intermediate to those seen previously in the corresponding D128A and S45A single mutants. The crystallographic alterations in the D128A mutant were previously connected to those seen in early structural intermediates on a computational biotin dissociation trajectory, which provides a mechanistic context for the role of the D128 and S45 hydrogen bonds to biotin. The double mutant main-chain near the D128A mutation is shifted away from the biotin, with maximal displacement occurring at the Cα of residue 128. The side chain of Q24, which hydrogen bonds to the D128 side chain in wild type, is displaced away from biotin as well, as the interaction that constrains its position has been lost. These shifts are nearly identical to those observed in the D128A single mutant. Likewise, the entire biotin moiety, and particularly the ureido headgroup, is shifted in concert with the 46–51 flexible loop. As in the D128A structure, this shift is accompanied by the entry of a single water molecule into the binding site, which resides at the position of the hydrogen bonding carboxylate oxygen removed by the D128A mutation. The position of this water is within 0.5 Å of that observed in D128A. On the other side of the binding pocket at the S45 position, the protein shifts~0.5 Å to compensate for the lost oxygen atom, similar to the S45A structure. One difference from the individual-mutant structures is that the double mutant lacks a water molecule near the “rear channel” entrance for water identified from molecular dynamics studies (Hyre et al. 2002). In the MD studies, this water stochastically diffused into the binding site where it occasionally stabilized a displaced-biotin intermediate when accompanied by concerted fluctuations at N23. However, further away from the biotin ureido nitrogen near Asp128Ala, out in the “rear channel” but away from the entrance to the binding pocket, the double mutant exhibits water molecules in most of the locations noted in the MD studies, including one in the area immediately beyond that noted as lacking above.

Titration calorimetry analysis revealed that the affinity loss in the double mutant was in excess of what would be predicted in the absence of cooperative effects. The binding of biotin is 11 times weaker due to cooperative effects. These cooperative effects are dominated by the entropic component, which contributes an unfavorable binding energy perturbation roughly twice as much as the enthalpy component increases it (3.0 kcal/mol deviation from the linearly added ΔS, vs. 1.6 from ΔH). The entropic alterations are consistent with the loss of two strong bonding interactions, although it is possible that alterations in the unbound state also play an important role.

While this work was in progress, wild-type, S45A, D128A, and S45A/D128A forms of a modified streptavidin were independently constructed and analyzed by Qureshi et al. (2001). There are some discrepancies between those results and ours published here and previously. There are reported differences in the rates of both association and dissociation of biotin from streptavidin and in the dominant oligomeric species observed. The rates measured in our solution studies suggest that the SPR data of Qureshi does not accurately reflect the intrinsic binding kinetics. This is supported by a similar discrepancy observed between our solution results and the SPR results of Laitinen for the binding of iminobiotin to avidin (Laitinen et al. 2001). Surface-based methods such as SPR are sensitive to numerous phenomena that could cause differences similar to those observed, including mass transport limitations, altered diffusional characteristics, and multivalent binding, and ligand alteration (Schuck and Minton 1996; Edwards et al. 1998; Pérez-Luna et al. 1999; de Mol et al. 2000; Jung et al. 2000; Jeppesen et al. 2001). The kinetic rates measured herein and in our prior publication (Hyre et al. 2000), when applied in conjunction with the multivalent binding model for SPR proposed by Jung et al. (2000), are able to predict the apparent rates measured by Qureshi et al. (2001). This suggests that a primary reason for the discrepancy in measured rates is due to multivalent binding in the SPR experiments. Other possible factors are the additional residues at both termini of the streptavidin studied by Qureshi and accompanying differences in the oligomeric stability of their proteins. The double mutant with the core streptavidin sequence studied here existed as a stable tetramer in solution.

The high quality of the two single-mutant structures and the double-mutant structure provides a structural context to interpret the observed cooperativity between the D128 and S45 positions. It appears that the relatively large structural rearrangements in the single D128A mutant allow other interactions to be optimized to regain binding energy, particularly that of S45 itself and others on that side of the binding pocket opposite from D128. This is consistent with results obtained for ligands with altered hydrogen bond partners to the structure-related residues in avidin (T35, N118) (Green 1975). The alternative ligand D-hexyl imidazolidone lost 2.3 kcal/mol more binding energy when the hydrogen bond donor to N118 (equivalent to D128 in streptavidin) was replaced with oxygen than when the donor to T35 (equivalent to S45) was similarly replaced. That loss is of similar magnitude to what is reported here for alteration of the corresponding hydrogen bonding partners in the protein (1.7 kcal/mol). Streptavidin (and probably avidin as well) has thus likely exploited the natural symmetry of the biotin nitrogen atoms to evolve cooperative hydrogen bonding interactions. The structural compensation between the hydrogen bonds that are directed in opposite “pulling” directions at the symmetrical nitrogen atoms is nicely reflected in the energetic coupling free energies.

Materials and methods

Site-directed mutagenesis and protein production

The S45A/D128A double mutation was created in the synthetic core streptavidin gene using the QuikChange protocol (Stratagene) as previously reported (Hyre et al. 2000), with modifications as follows. The D128A mutant in pET21 (Freitag et al. 1999) was used as the substrate for the protocol, with the S45A mutational primers from a previous study (Hyre et al. 2000) used to create the second mutation. Sequencing confirmed existence of the two desired mutations and lack of any others. The protein was expressed in BL21 (DE3) Escherichia coli as previously described (Hyre et al. 2000), but purification deviated from the prior protocol. Inclusion bodies were isolated, resuspended, and dilution-refolded as before; but due to the very low binding affinity expected for this double mutant, the iminobiotin affinity column was not used as the final purification step. Instead, an ammonium sulfate precipitation was done similar to that of Schmidt and Skerra (1994). The concentrated, refolded streptavidin was brought to 40% saturation in ammonium sulfate (0.24 g/mL added), centrifuged 20 min at 14,000g, and the pellet discarded. The solution was then brought to 70% solution (0.20 g/mL added), centrifuged 30 min at 14,000g, and the supernatant discarded. The pellet was resuspended in a minimal volume of PBS buffer and dialyzed three times against buffer to remove the salt. The protein solution was centrifuged to remove unfolded and aggregated protein remaining (very minimal). The expected mass and purity were confirmed by electrospray mass spectrometry and SDS-PAGE electrophoresis, respectively.

Association kinetics

Stopped-flow fluorescence measurements of association rate were made using a Hi-Tech Scientific MX-2 instrument (Hi-Tech). The reaction was followed by the intrinsic fluorescence of streptavidin, which decreases by ~40% when biotin binds (Green 1990; Kurzban et al. 1990). The reactions were followed under second-order conditions using equimolar reactants at 1 μM. Stock solutions of streptavidin and of biotin were diluted ~100-fold with appropriate buffers to give final concentrations of 2 μM each in binding sites and ligand. Equal volumes (200 μL) of the protein and ligand were injected for each run, giving a final concentration of 1 μM for the second-order conditions. Fluorescence was excited at 280 nm and emission was monitored through a Schott WG320 cutoff filter. The dependence of rate on ionic strength was measured in 10 mM sodium phosphate buffer pH 7.0 with addition of 0–200 mM sodium chloride, at 25°C. At least five runs were made under each set of conditions. The dead time of the instrument was ~2.5 msec. Data were analyzed using the Hi-Tech software using a second-order model (Binns et al. 2000) and with DynaFit (BioKin; Kuzmic 1996) using an explicit reagent-limited model.

Calorimetry

Isothermal titration calorimetry was performed as reported with the individual mutants (Freitag et al. 1999; Hyre et al. 2000), but fit by a model in a different manner. Streptavidin (monomers) (30 μM) was titrated stepwise with 5-μL injections of 750 μM biotin, all in PBS buffer, at 15°C, 25°C, and 37°C, in a CSC ITC-4200 calorimeter in overfill mode. The integrated peak areas from all experiments were fit simultaneously in Microsoft Excel 97 by a global model that incorporated temperature in addition to the usual parameters of enthalpy, affinity, and stoichiometry. Confidence intervals were estimated by the support-plane method employed for the global fit of kinetic data for our S45A mutants (Hyre et al. 2000).

Crystallization of the double-mutant and wild-type streptavidins

The protein with twofold excess biotin was crystallized from solutions at 10 mg/mL (45/128) or 26 mg/mL (wild type) using sitting-drop vapor-diffusion methods; 45/128 crystals grew in 1.0 M sodium citrate and 0.1 M cacodylate (pH 6.5); wild type grew in 35% saturated ammonium sulphate, 0.1 M sodium acetate (pH 4.5), and 0.2 M NaCl. The cryosolution contained 35% glycerol. The double mutant crystallized in space group P2₁ with the unit cell a = 50.6 Å, b = 97.8 Å, c = 52.2 Å, and β = 112.3°. This cell has been observed before in the Streptavidin mono_b form (Freitag et al. 1997) and has one tetramer in the asymmetric unit. Wild type crystallized in space group I222, dimensions 46.1 × 93.3 × 103.8 Å, with two monomers in the asymmetric unit, similar to that observed for the S45A mutant (Hyre et al. 2000).

Data collection and processing

Data on 45/128 were collected at SSRL beamline 9-1 with a MAR Research image plate scanner. The wavelength was 0.97 Å, and the crystal diffracted to a resolution of 1.65 Å. Wild type data were collected at the ALS with a MAR Research CCD detector using 1.00 Å light, with diffraction to 1.40 Å. Data were processed with DENZO (Otwinowski and Minor 1997) and merged and scaled with SCALEPACK (Otwinowski and Minor 1997). Further information about the data sets is summarized in Table 1.

Structural refinement of the double mutant

The wild-type streptavidin model (PDB entry number 1SWE) without waters or biotin was used as a starting model for refinement. SHELXL97 (Sheldrick and Schneider 1997) was used as the refinement program and its auxiliary program SHELXPRO for map calculations, update of structure files for refinement, and data and model analysis. Graphical evaluation of the model and electron density maps were carried out with Xtalview (McRee 1992).

The structure was refined against squares of the structure factor amplitudes (F²). All parameters were refined simultaneously. Five percent of the data were used for calculating R_free (Brünger 1992). Distance, planarity, and chiral volume restraints were applied, as were anti-bumping restraints. The target values for 1–2 and 1–3 distances were based on the Engh and Huber (1991) study. A full-matrix least-squares rigid-body refinement was initially carried out for the whole tetramer and subsequently for each monomer. After that, conjugate-gradient least-squares refinement with constant isotropic B-values for all atoms was followed by B-factor refinement. In the next steps biotin and water molecules were added. After the refinement of anisotropic temperature factors, alternate conformations of side chains were added. Hydrogen atoms were not included because R and R_free went up with the riding hydrogen model. Despite the high R_merge of 9.9% and the high mosaicity of 1.27, the structural refinement went well. The final R is 17.6% for reflections with F > 4σ (F) and 20.4% for all data and the R_free ended up at 26.6% for reflection with F > 4σ (F) and 29.5% for all data. The double-mutant coordinates have been deposited in the PDB under ID 1MEP.

Structural refinement of the biotin complex of wild-type streptavidin

The wild-type streptavidin with biotin crystallized in the same space group and cell as the S45AB (PDB ID 1DF8) and the refined model was used without biotin, water, or alternate conformations as a starting model. The same steps were followed as above, except here hydrogen atoms were included in the refinements as a riding model. The final R-factor is 11.9% for reflections with F > 4σ (F), 12.8% for all data, and R_free is 17.8% for reflections with F > 4σ and 18.9% for all data. The wild type coordinates have been deposited in the PDB as file 1MK5.

Abbreviations

• ITC, isothermal calorimetry

• MD, molecular dynamics

• SPR, surface plasmon resonance

• StAv, streptavidin

• 45/128 S45A/D128A, double mutant of streptavidin

Article published online ahead of print. Article and publication date are at http://www.proteinscience.org/cgi/doi/10.1110/ps.051970306.