Conversion of an Engineered Potassium-binding Site into a Calcium-selective Site in Cytochrome c Peroxidase*

We have previously shown that the K+ site found in ascorbate peroxidase can be successfully engineered into the closely homologous peroxidase, cytochrome c peroxidase (CCP) (Bonagura, C. A., Sundaramoorthy, M., Pappa, H. S., Patterson, W. R., and Poulos, T. L. (1996) Biochemistry 35, 6107–6115; Bonagura, C. A., Sundaramoorthy, M., Bhaskar, B., and Poulos, T. L. (1999) Biochemistry 38, 5538–5545). All other peroxidases bind Ca2+ rather than K+. Using the K+-binding CCP mutant (CCPK2) as a template protein, together with observations from structural modeling, mutants were designed that should bind Ca2+ selectively. The crystal structure of the first generation mutant, CCPCA1, showed that a smaller cation, perhaps Na+, is bound instead of Ca2+. This is probably because the full eight-ligand coordination sphere did not form owing to a local disordering of one of the essential cation ligands. Based on these observations, a second mutant, CCPCA2, was designed. The crystal structure showed Ca2+ binding in the CCPCA2 mutant and a well ordered cation-binding loop with the full complement of eight protein to cation ligands. Because cation binding to the engineered loop results in diminished CCP activity and destabilization of the essential Trp191 radical as measured by EPR spectroscopy, these measurements can be used as sensitive methods for determining cation-binding selectivity. Both activity and EPR titration studies show that CCPCA2 binds Ca2+ more effectively than K+, demonstrating that an iterative protein engineering-based approach is important in switching protein cation selectivity.

CCP consists of a 294-residue polypeptide with a single heme group coordinated to His 175 . A unique feature of the CCP reaction cycle is the formation of a stable amino acid radical formed in compound I (7). This radical is centered on Trp 191 (8), which lies parallel to and in contact with the proximal heme ligand, His 175 , just beneath the heme (see Fig. 1). Trp 191 is essential for activity (9) and is considered to be a critical part of the electron transfer circuit leading from cytochrome c to CCP (3, 10 -12). The crystal structure of the CCP-cytochrome c noncovalent complex (13) supports the idea that Trp 191 lies along the shortest and most direct electron transfer route from cytochrome c to the CCP heme.
Various experimental approaches (3,14,15) as well as theoretical calculations (3,14,16) support the view that the Trp 191 radical is cationic, although recent theoretical assessments suggest a neutral radical (17). It has clearly been shown that CCP cavity mutants where Trp 191 is converted to Gly prefer to bind imidazolium cations in the pocket vacated by the missing Trp 191 indole ring (15). In addition, electrostatic calculations (3,16) show that the electrostatic environment of the protein surrounding Trp 191 is negative in CCP, which offers the possibility of stabilizing a cationic radical. Ascorbate peroxidase closely resembles the structure of CCP, including a homologue to Trp 191 (Trp 179 in ascorbate peroxidase). However, ascorbate peroxidase does not form a stable Trp-centered radical, but instead, a porphyrin-cation radical (18), as in other peroxidases. One hypothesis put forward to explain this difference is the electrostatic nature of CCP to more effectively stabilize the positive charge on the Trp 191 cationic radical. This view is supported by electrostatic calculations (3,16). An obvious factor helping to alter the surrounding electrostatic environment near the ascorbate peroxidase Trp 179 residue is a K ϩ metal cation bound ϳ8 Å from Trp 179 (Fig. 1). This cation site is conserved in all peroxidases for which structures are known, except CCP, where a water molecule occupies this position. We * This work was supported by grants from the National Institutes of Health and the National Science Foundation. 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  have suggested that the positively charged cation in ascorbate peroxidase helps to prevent formation of a stable Trp 179 cationic radical owing to electrostatic destabilization (3,18). This hypothesis has been tested by engineering the ascorbate peroxidase K ϩ site into CCP. The resulting CCP mutant, designated CCPK2, binds K ϩ , has ϳ1% wild-type activity in the presence of K ϩ , and exhibits a significantly weakened EPR signal associated with the Trp 191 radical (3,4). In addition, both enzyme activity and the Trp 191 EPR signal can be titrated with K ϩ , indicating that the binding of K ϩ to the engineered site is responsible for loss of activity and destabilization of the Trp 191 radical.
Ascorbate peroxidase is the only member of the peroxidase family for which x-ray structures are available that binds K ϩ rather than Ca 2ϩ . Owing to our success in engineering the ascorbate peroxidase K ϩ site into CCP, we next chose to convert this site to a Ca 2ϩ site in CCP. One reason is the relative ease of using various chelating agents for removal of calcium, which should enable easier manipulation of the engineered cation site, which was not possible with the engineered K ϩ site in CCP. This would potentially offer an easier means of controlling enzyme activity by providing a "molecular switch" that regulates the redox properties of Trp 191 . In addition, because it was possible to convert CCP into a K ϩ -binding protein, we viewed CCP as an excellent system for probing in greater detail the general problem of designing metal-binding sites in proteins. The close structural homology among various peroxidases in the proximal cation-binding loop and the ability to readily obtain high resolution x-ray crystal structures afford an excellent opportunity to understand, in some detail, factors that can contribute to the successful design of a protein divalent metal-binding site.

EXPERIMENTAL PROCEDURES
Materials-Enzymes and reagents for site-directed mutagenesis were purchased from Roche Molecular Biochemicals and New England Biolabs Inc. (Beverly, MA). Chromatography columns and media were purchased from Amersham Pharmacia Biotech. Horse heart cytochrome c, hydrogen peroxide (30%, w/v), EDTA, and EGTA were purchased from Sigma. 2-Methyl-2,4-pentanediol was purchased from Aldrich. All other chemicals were molecular biology grade or better and were purchased from Sigma or Fisher.
Site-directed Mutagenesis-Construction of CCPK2 has been previously reported (3,4,18). CCPK2 has the following mutations: A176T, G192T, A194N, T199D, and E201S, which alter the normal water environment in CCP to mimic the proximal cation-binding loop in ascorbate peroxidase. To change the proximal monovalent cation (potassium)-binding site into a CCP mutant that binds calcium, the following oligonucleotide was purchased from Operon Technologies, Inc. (Alameda, CA): GAA GGT CCA TGG GAC GCC AAT AAC AAC. The resulting amino acid side chain replaced on the CCPK2 template was T192D. Based on sequence alignment and molecular modeling, this alteration was intended to install a second formal negative charge and also to increase the ligand count to eight in the cation ligand coordination sphere. Site-directed mutagenesis was carried out by annealing this oligomer to the single-stranded uracylated CCPK2 template that was produced in Escherichia coli RZ1032 cells following the method of Kunkel et al. (19) as described previously (20,21). The mutagenic oligomer was phosphorylated with calf intestinal alkaline phosphatase; boiled with template at a ratio of 20 pM oligomer to 1 pM template; and annealed for 15 min each at 70°C, 37°C, and then at room temperature and finally on ice to add the reactants. Next, 1 mM ATP was added with T4 DNA polymerase and T4 DNA ligase in excess to the reaction mixture and incubated at room temperature for 3-4 h. The reaction mixture was then transformed by electroporation into JV30 cells, and the colonies grown were selected under ampicillin resistance. Transformation efficiency by this procedure was typically low, but the transformants were likely to be positive clones without spontaneous mutations. All mutations were sequenced at the DNA level (Promega thermal cycle sequencing) to ensure that the introduced mutations were only installed as predicted. The resulting mutant was called CCPCA1.
CCPCA1 was then changed to CCPCA2 using the same CCPK2 template and mutagenesis method described above with the following oligomer: GAA GGG CCA TGG GAC GCC ACT AAC AAC GTC TT. In addition to replacing Thr 192 with Asp, this oligomer replaces Asn 194 with Thr. Thr was chosen since this is the corresponding residue and Ca 2ϩ ligand in extracellular peroxidases. The Asn 194 residue on the CCPK2 template was installed to copy the corresponding ascorbate peroxidase cation ligand at position 194. Under "Discussion," we will explain why there is the need to again alter this residue.
Protein Expression and Purification-The recombinant CCP mutants were expressed in E. coli BL21(DE3) cells under the influence of the T7 promoter with isopropyl-␤-D-thiogalactopyranoside induction as described previously by Fishel et al. (22) and Choudhury et al. (21) using slightly modified conditions especially for the calcium-binding mutant CCPCA2. CCPCA1 was purified by column chromatography using the same conditions as used for wild-type CCP and CCPK2. CCPCA1 was stable in 50 mM potassium phosphate, pH 6.0. CCPCA1 was stored frozen at Ϫ80°C as microcrystalline precipitate after dialysis against water. CCPCA2 showed an obvious instability in potassium-containing buffers, and much of the protein tended to denature. Using calcium in phosphate buffers caused precipitation of calcium phosphate, which was detrimental to the stability of the mutant. As a result, CCPCA2 was purified in potassium phosphate buffer, pH 6.0; and after heme incorporation followed by DE52 anion-exchange chromatography, the protein was dialyzed against water followed by dialysis against water containing 100 M calcium to stabilize the protein. Dialysis of CCPCA2 against distilled water did not result in crystals typically observed during CCP purifications. Therefore, for long-term storage, CCPCA2 was dialyzed against 200 mM Tris/MES, pH 6.0, containing 50 M calcium and stored as aliquots at Ϫ80°C. The CCP concentration was estimated spectrophotometrically using an extinction coefficient (⑀) at 408 nm of 96 mM Ϫ1 cm Ϫ1 .
Steady-state Activity Assays-For substrate titration kinetic assays, the steady-state oxidation of dithionite-reduced horse heart cytochrome c (ferrocytochrome c) was measured at 24°C in a Cary 3E UV-visible spectrophotometer using ⌬⑀ 550 ϭ 19.6 mM Ϫ1 cm Ϫ1 . The final reaction conditions consisted of 180 M and 10 nM CCP and 30 M ferrocytochrome c in 5 mM Tris/MES, pH 6.0 (23). The initial linear region of the reaction slope was recorded and taken as least-squares fit using Cary Varian 01.00(6) software running on Windows 95. All data points reported in this paper reflect the average of at least three independent slopes averaged and plotted in Sigma Plot 4.0.
The following assay conditions were employed to measure steadystate activity as a function of cation concentration. CCPCA2 was added to 5 mM Tris/MES, pH 6.0, initially containing Ca 2ϩ or K ϩ ions or other ions at various concentrations, and then 180 and 40 M ferrocytochrome c were added in concert to begin the reaction. In the case of metal chelation experiments, the enzyme was preincubated at an end point concentration of 10 mM calcium in 5 mM Tris/MES, pH 6.0, before incubating with a metal chelator (EDTA or EGTA) at stoichiometric concentrations and then assaying for the remaining activity. Results are presented as the percentage activity remaining over the control in the absence of metal ions taken as 100%.
Rapid Reaction Kinetics-To evaluate the mutants for the efficiency of electron transfer rate, rapid reaction kinetics were performed at a single wavelength on a Hi-Tech Model SF-51 stopped-flow spectrophotometer equipped with a 1-cm path length. The output data were monitored and recorded on a Compaq PC for kinetic analysis using Hi-Tech IS-1 software suite version 1.0. CCPCA2 compound I in various concentrations of Ca 2ϩ or K ϩ was formed by the addition of an equal volume of 4.6 M H 2 O 2 and 4.0 M CCPCA2 in 5 mM Tris/MES, pH 6.0. CCPCA2 compound I (2 M) in one syringe was mixed with 2 M horse heart ferrocytochrome c in stoichiometric amounts. The oxidation of ferrocytochrome c was monitored at 416 nm, an isosbestic point for the resting CCP ferric and oxyferryl compound I. Results are presented as the percentage of electron transfer rate over the control enzyme in the absence of metal ion taken as 100%.
EPR Spectroscopy-EPR spectra were recorded on a Bruker ESP300 spectrometer equipped with an Air Products LTR3 liquid helium cryostat. Experimental conditions used to record Trp 191 cation radical for-mation of the wild-type protein and mutants were as follows: microwave frequency, 9.475 GHz; microwave power, 0.5 milliwatts; modulation amplitude, 4.57 G; modulation frequency, 1000 kHz; field sweep rate, 11.92 G/s; time constant, 0.0256 ms; and receiver gain, 1.0 ϫ 10 4 (wild-type CCP) and 2.5 ϫ 10 4 (CCPCA2). The resting state sample had wild-type CCP and CCPCA2 at 300 M in 5 mM Tris/MES, pH 6.0, in a total volume of 150 l. Compound I was formed by the addition of 360 M H 2 O 2 , and the samples were immediately frozen in quartz EPR tubes by submersion in liquid nitrogen over a period of 60 -80 s. Spectra were recorded at 8 K. The data obtained were an average of five scans.
Crystallization-Because of the requirement to soak crystals in Ca 2ϩ buffers, the usual use of potassium phosphate buffers could not be employed with CCPCA1 and CCPCA2. After several trials, crystals could be grown from 100 mM Trizma (Tris base), 50 mM acetic acid, and 50 mM MES, pH 6.0, or 50 mM Tris cacodylate, pH 5.6, using the usual 30% 2-methyl-2,4-pentanediol as the precipitant (24). CaCl 2 (10 and 2 mM) was included in the crystallization buffers for CCPCA1 and CCPCA2, respectively.
X-ray Data Collection and Refinement Statistics-Data for both CCPCA1 and CCPCA2 were obtained from one flash-frozen crystal each using an R-AXIS IV imaging plate and a Rigaku rotating anode x-ray source equipped with a Crystal Logic cryogenic N 2 delivery system. Initial image processing, indexing, and integration were performed with Denzo version 1.9.1, and the integrated data were scaled using ScalePack version 1.9.0 (25). The starting model for refinement was CCPK2. The models were refined with X-PLOR version 3.851 (26). Data collection statistics and final refinement parameters are given in Table I.

RESULTS AND DISCUSSION
Crystal Structures- Fig. 2 provides a sequence alignment of the proximal cation-binding loop in various peroxidases. Although the backbone polypeptide conformation of CCP and the proximal cation-binding loop of other peroxidases are nearly identical in this region, CCP lacks the side chain cation ligands that would bind the metal ion and hence binds water rather than a cation. CCPK2 is an engineered version of CCP that was designed to mimic the ascorbate peroxidase cation-binding site, and it does bind K ϩ (4). CCPCA1 was the first attempt to convert the K ϩ site in CCPK2 to a Ca 2ϩ -binding site. We reasoned that conversion of Thr 192 , a K ϩ ligand in CCPK2, to Asp would create a second negatively charged ligand, which, together with Asp 199 , would provide a charge-neutralizing coordination environment of eight oxygens for Ca 2ϩ . Changing Thr 192 to Asp in CCPK2 should install the bidentate bond at this side chain position that is observed in other peroxidase proximal cation loops such as the one found in lignin peroxidase (see Fig. 4) (27). This version of the CCP mutant was designated CCPCA1.
To see if Ca 2ϩ binds to CCPCA1, the crystal structure was solved and refined to 1.9-Å resolution. An omit 2F o Ϫ F c elec- The maps are contoured at 1. The models were subjected to one round of simulated annealing refinement using X-PLOR starting at 1000 K with the atoms shown excluded from the refinement. Note that Asn 194 in CCPCA1 is oriented toward the surface and is not a ligand as expected. In addition, there is a break in the electron density between Ala 193 and Asn 194 . In contrast, the loop is well ordered in CCPCA2, and Thr 194 coordinates the cation as in other calcium-binding peroxidases. tron density map around the cation-binding site is shown in Fig. 3. The omit F o Ϫ F c map shows a 5 peak at the position of the cation. This is far lower than would be expected for a Ca 2ϩ site since similar omit maps with other Ca 2ϩ -or K ϩ -binding peroxidases we have studied often give peaks of 10 or higher. This indicates that a smaller cation, most likely Na ϩ , is bound. The difference maps are flat when 0.55 K ϩ (equivalent to Na ϩ ) is modeled during refinement at the cation site in CCPCA1. In addition, one of the anticipated ligands, Asn 194 , is pointing away from the cation (Fig. 4). The side chain density for Asn 194 is not well defined, and there is a break in the electron density between Ala 193 and Asn 194 (Fig. 3). This failure of Asn 194 to form a cation ligand and the local disordering most likely account for the failure of Ca 2ϩ to bind in CCPCA1. The ligand count in the pocket now drops to six oxygens, which is consistent with a Na ϩ coordination sphere.
The reason why Asn 194 does not coordinate the cation is most likely due to steric and electrostatic factors not considered in the original design of CCPCA1. If Asn 194 were to rotate in toward the cation site, steric clashes would occur with Asp 192 , the newly installed cation ligand replacing Thr 192 . The new Asp 192 could come within 1.3 Å from the normal position of the carbonyl group of Asn 194 in the CCPK2 loop structure. We had anticipated sufficient adaptability of the site to relax such constraints, but apparently this assumption was incorrect. In addition, the hydrogen bond formed between Asn 194 and Asp 199 in CCPK2 (Fig. 4) was expected to remain in CCPCA1 and to provide a strong energetic incentive to correctly orient Asn 194 toward the cation. Obviously, this hydrogen bond was not sufficient to maintain Asn 194 in a coordination position. Since the only other Ca 2ϩ -binding protein listed in Fig. 2, lignin peroxidase, has Thr at this position, we next replaced Asn 194 with Thr and designated this mutant CCPCA2. We reasoned that removing the amide group from Asn 194 would allow sufficient local relaxation to accommodate Asp 192 . Fig. 3 shows the omit 2F o Ϫ F c maps for CCPCA2. Here, the F o Ϫ F c map shows a 14 peak at the cation site, compared with 5 for CCPCA1. In addition, Thr 194 now coordinates the cation, and the entire 192-199 section of polypeptide is well ordered as evidenced by strong continuous density in the omit 2F o Ϫ F c map. In addition, when the cation site is modeled as Ca 2ϩ (18 electrons), the temperature factor (B factor) refined to a value of 14.4 Å 2 , whereas in CCPCA1, the B factor was 24 Å 2 even when the site was modeled with only 0.5 Ca 2ϩ . Therefore, replacement of Asn 194 in CCPCA1 with Thr enables the new mutant, CCPCA2, to bind calcium. Another noticeable difference between CCPCA1 and CCPCA2 is the conformation of Asp 192 , a potential bidentate cation ligand. In CCPCA1, Asp 192 cannot adopt the same conformation as in CCPCA2 owing to the presence of Asn 194 in the cation coordination sphere (Fig. 4). Not too surprisingly, the conformation of Asp 192 in CCPCA2 is very close to the corresponding Ca 2ϩ ligand, Asp 194 , in lignin peroxidase. The C-␣-C-␤ and C-␤-C-␥ side chain torsion angles are 74 o and 4 o , respectively, in lignin peroxidase compared with 75 o and 4 o in CCPCA2 and Ϫ180 o and Ϫ122 o in CCPCA1. This enables both carboxylate oxygen atoms of Asp 192 to ligate the calcium in CCPCA2 as in lignin peroxidase, whereas in CCPCA1, only one carboxylate oxygen atom is able to coordinate the cation while the other hydrogen-bonds to the peptide amide group of Ala 193 . A comparison of the various peroxidase cation-binding sites is provided in Fig. 4.
Enzyme Activity and EPR-In our earlier work with CCPK2, we found that enzyme activity and the electron transfer rate could be titrated down by the addition of K ϩ , which gave K d ϳ 10 M (4). In addition, we found that the Trp 191 EPR signal could also be titrated in a similar way with added K ϩ ions. The close correlation between loss of EPR signal and loss of activity upon addition of K ϩ indicated that electrostatic destabilization of the Trp 191 radical was responsible for the loss of enzyme activity. For our present purposes, activity and EPR can be utilized to see if, indeed, CCPCA2 now specifically binds Ca 2ϩ .
For titrating enzyme activity, we again utilized both stopped-flow and steady-state kinetic assays. In the absence of added calcium, the activity of the CCPCA2 mutant is ϳ28% that of wild-type CCP. In our previous study, we showed that this effect is due to a destabilization of the calcium loop including Trp 191 (4). Despite this difference, a detailed kinetic analysis showed that CCPK2 behaves very much the same as wildtype CCP. As shown in Fig. 5A, Eadie-Hofstee plots for CCPCA2 give typical biphasic kinetics characteristic of wildtype CCP substrate turnover with very similar K m values (4,28). The two K m values for wild-type CCP determined in our previous study, 3.6 and 29.8 M (4), compare well with those for CCPCA2, 4.7 and 18.3 M (Fig. 5A). In addition, the break point between the two kinetic phases is 12 M ferrocytochrome c for CCPCA2 and 9 M ferrocytochrome c for wild-type CCP (4). This indicates that the overall catalytic machinery of CCPCA2 remains intact even though activity can be decreased by the addition of Ca 2ϩ . As shown in Fig. 5B, both K ϩ and Ca 2ϩ addition led to a loss of activity, but Ca 2ϩ was much more effective with an apparent K d of 27 M. Just the opposite was found with CCPK2, the mutant designed to bind K ϩ . In this case, K ϩ was more effective at decreasing activity than Ca 2ϩ (4). In the stopped-flow studies, CCP was treated with 1 eq of peroxide to give CCP compound I, followed by mixing with reduced horse heart cytochrome c. Electron transfer from cytochrome c to CCP was monitored by following the oxidation of cytochrome c. As shown in Fig. 5C, the results are very similar to the steady-state assay results in that Ca 2ϩ was much more effective at eliminating activity than K ϩ . In both steady-state and rapid kinetic assays, other monovalent and divalent cations were ineffective in inactivating the enzyme. However, in steady-state assays, it was observed that inactivation of the enzyme by the added calcium could be partially reversed by the addition of the divalent metal chelator EDTA or EGTA (specific for Ca 2ϩ ) in stoichiometric amounts (data not shown). This demonstrated that the reactivity of the Trp ϩ radical could be controlled by the addition of metal chelators.
In our earlier work on CCPK2, the EPR signal associated with the Trp 191 radical was decreased more effectively by K ϩ , and Ca 2ϩ had very little effect (4). Fig. 6 shows the effects of both Ca 2ϩ and K ϩ on the compound I EPR signal of CCPCA2. The EPR signal got progressively weaker as the CaCl 2 concentration increased. Note, however, that 10 mM KCl caused only a modest decrease in signal. This again is the reverse of what was observed in the CCPK2 mutant designed to bind K ϩ (4). Taken together with the enzyme activity measurements, the EPR experiments demonstrate that CCPCA2 is, indeed, more selective for Ca 2ϩ than K ϩ . CONCLUSIONS This work demonstrates that it is possible to convert the engineered K ϩ -binding site in CCP to a Ca 2ϩ -binding site. The crystal structure of CCPCA1 shows that CCPCA1 binds a smaller cation, possibly Na ϩ , with only six protein cation ligands formed by the partially disordered cation-binding loop. In contrast, the crystal structure of CCPCA2 shows that the cation loop is well ordered, binds Ca 2ϩ , and has eight protein cation ligands. Both EPR and enzyme activity titration data illustrate that the engineered cation site in CCPCA2 now prefers Ca 2ϩ over K ϩ . In effect, we have systematically converted the natural CCP H 2 O site first to bind K ϩ in CCPK2, then Na ϩ in CCPCA1, and finally Ca 2ϩ in CCPCA2.
Although it may have been possible to simply replace the entire cation-binding loop in CCP with the lignin peroxidase loop to achieve Ca 2ϩ selectivity, the iterative process of first engineering in a K ϩ site and then Ca 2ϩ has enabled a deeper appreciation for those structural factors that control protein metal site selectivity. The CCPCA1 structure demonstrated that providing a second Asp ligand was not sufficient to obtain Ca 2ϩ selectivity. The engineered site bound to a smaller cation, possibly Na ϩ . We had anticipated that ligands would adjust to accommodate Asn 194 as a ligand, but instead, Asn 194 oriented toward the molecular surface away from the cation. Replacing Asn 194 with the corresponding residue in lignin peroxidase, Thr, "repaired" the site, enabling the full complement of eight ligands in the correct geometry to coordinate Ca 2ϩ . Perhaps even more interestingly, another ligand, Asp 192 , adopted quite different conformations in CCPCA1 and CCPCA2, with the CCPCA2 conformation more closely resembling that found in lignin peroxidase. This means that the nature of the ligand at position 194 strongly influences the conformation of the ligand at position 192. A priori, this would have been very difficult to model correctly since simple steric considerations indicated that the ligands should be able to accommodate Asp at position 192 without grossly affecting cation coordination or favored side chain torsion angles. Therefore, the somewhat more conservative iterative approach adopted here has provided some important insights into protein metal-binding site design.