The crystal structure of the cis-proline to glycine variant (P114G) of ribonuclease A

Abstract

Replacement of a cis-proline by glycine at position 114 in ribonuclease A leads to a large decrease in thermal stability and simplifies the refolding kinetics. A crystallographic approach was used to determine whether the decrease in thermal stability results from the presence of a cis glycine peptide bond, or from a localized structural rearrangement caused by the isomerization of the mutated cis 114 peptide bond. The structure was solved at 2.0 Å resolution and refined to an R-factor of 19.5% and an R_free of 21.9%. The overall conformation of the protein was similar to that of wild-type ribonuclease A; however, there was a large localized rearrangement of the mutated loop (residues 110–117—a 9.3 Å shift of the Cα atom of residue 114). The peptide bond before Gly114 is in the trans configuration. Interestingly, a large anomalous difference density was found near residue 114, and was attributed to a bound cesium ion present in the crystallization experiment. The trans isomeric configuration of the peptide bond in the folded state of this mutant is consistent with the refolding kinetics previously reported, and the associated protein conformational change provides an explanation for the decreased thermal stability.

Kinetic studies of the folding of bovine pancreatic ribonuclease A (RNase A) have been carried out to identify the pathway that leads to the folded protein. Interpretation of the results from these studies has been complicated due to the cis ↔ trans isomerization of the peptide bonds preceding proline residues that are responsible for forming slowly refolding molecules, as outlined in the proline isomerization model of Brandts et al. (1975). In a number of systems, site-directed mutagenesis has been used to replace the cis prolines with other amino acids in order to simplify the refolding kinetics. In some cases (if a trans peptide bond is formed), this has simplified the refolding kinetics, as predicted by the proline isomerization model. Frequently, however, the refolding kinetics are still complex. One factor that may be partly responsible for the complexity is the persistence of a cis peptide bond preceding the substituted non-proline amino acid, due to conformational constraints imposed by the rest of the protein (Schultz and Baldwin 1992; Schultz et al. 1992; Dodge and Scheraga 1996). This bond would then still have to isomerize from the trans to the cis configuration during the folding process.

Replacement of cis-prolines can also have large effects on protein stability. A conformational change in the protein resulting from isomerization of the mutated cis peptide bond could cause destabilization. Alternatively, because the cis configuration of non-prolyl peptide bonds is unstable relative to that of cis-prolyl peptide bonds, the replacment of cis-proline by any other naturally occurring amino acid that remains cis in the folded state would be expected to destabilize a protein (Drakenberg et al. 1972; Ramachandran and Mitra 1976; Jorgensen and Gao 1988; Radzicka et al. 1988; Schultz et al. 1992).

The current study focuses on the structural consequences of cis-proline replacement in bovine pancreatic ribonuclease A (RNase A), and its relationship to protein folding, stability, flexibility, conformational change, and evolution. RNase A is 124 amino acids in length and contains two cis-prolines, Pro93 and Pro114 (Richards and Wyckoff 1971; Wlodawer and Sjolin 1983; Wlodawer et al. 1988; Robertson et al. 1989; Santoro et al. 1993). Pro114, the focus of this study, is located in a solvent-exposed type VIb-2 β-turn (Pal and Chakrabarti 1999), which is anchored by the Cys58–Cys110 disulfide bond, and at the other end by Pro117, whose mobility is restricted by tightly packed neighboring residues. Measurements of the isomeric ratio in unfolded RNase A (63%) favor the trans configuration of the Asn113–Pro114 peptide bond (Adler and Scheraga 1990). Thus, the cis isomer found for Pro114 in the folded protein must be held in place by constraints imposed by the rest of the protein. The loop containing the cis configuration may be preferentially stabilized relative to the trans configuration by hydrogen bonding, electrostatics, and van der Waals interactions. Alternatively, the formation of a low energy trans configuration of this loop may be prevented by steric clash between residues in the loop with the remainder of the folded protein. Replacement of a cis-proline by glycine at Pro114 in ribonuclease A leads to a large decrease in thermal stability and simplifies the refolding kinetics (Schultz and Baldwin 1992; Schultz et al. 1992; Dodge and Scheraga 1996).

A crystallographic structure determination of a cis Pro114 to Gly mutant (P114G) was undertaken to ascertain whether a trans configuration of the loop is adopted, and to determine whether the structural effects are localized to the loop or the rest of the protein imposes anchorage constraints that force the peptide linkage to remain cis. In addition, we examine the hypothesis (based on the results from kinetic experiments) that the Asn113–Gly114 peptide bond is trans in the folded P114G protein.

Results

Ribonuclease A P114G crystallization

Crystals of the ribonuclease A P114G variant were grown as described in Materials and Methods. The crystallization conditions (30% saturated ammonium sulfate, 3.0 M cesium chloride, 0.15 M sodium phosphate [pH 6.6], 25°C) are similar to those used for the crystallization of ribonuclease S and a high-salt crystal form of ribonuclease A (Kim et al. 1992; P. Harkins and H. Wyckoff, pers. comm.). However, the P114G crystals grew in a new space group, P4₃2₁2. This crystal form was prone to radiation decay unless stabilized in nearly saturating amounts of ammonium sulfate at 15°C. The crystals diffract to 2.0 Å resolution.

Refinement of the P114G crystal structure

The results of the final refinement using all data from 30.0 to 2.0 Å resolution are summarized in Table 1. Electron density was seen for all residues except parts of the side chains of Lys1, Lys91, and Tyr115. Although the side chains of several residues in RNAse A have been modeled in alternative conformations in highly refined structures (Svensson et al. 1986; Kim et al. 1992), we modeled all side chains in a single conformation. According to the PROCHECK program (Laskowski et al. 1993), the variations from ideality in bond properties lie within the expected ranges, 89% of the residues lie in the most favored regions of a Ramachandran plot, and 100% lie in the allowed regions.

Table 1.

Crystallographic and geometrical parameters at the end of the refinement

Space group	P43212
Unit cell lengths
a,b (Å )	40.87
c (Å )	129.25
Resolution range (Å )	30.0–2.0
No. of reflections	7592
Rsymmetrya	0.053
Rfactor (work)b	0.195
Rfree (test)c	0.219
No. of total atoms	1024
No. of protein atoms	956
No. of water molecules	62
No. of cesium atoms	1
No. of sulfate ions	1
Deviations from ideal geometry
RMSD on bond lengths (Å )	0.005
RMSD on bond angles (°)	1.20
RMSD on dihedrals angles (°)	24.2
RMSD on improper angles (°)	0.6

^a R_merge = ∑_hkl | I_hkl –〈I_hkl〉| /∑I_hkl, where 〈I_hkl〉 is the weighted

average of all the symmetry-related reflections.

^b R = ∑_hkl | |F_o| – |F_c| | /∑|F_o|, where F_o and F_c are observed and

calculated structure factors.

^c R_free is calculated as in b but with the test set of reflections.

Early in the refinement, before residues 110–116 had been included in the model, examination of the 2F_o–F_c electron density revealed unambiguous density for this loop, which could be modeled only with a trans peptide bond between Asn113 and the mutant Gly114. Further refinement (including residues 110–116, water, and sulfate and cesium ions) (see below) led to a final R-factor of 19.5%. As confirmation of the trans configuration, residues 110–116 and the cesium ion were removed from the model, followed by simulated annealing and Powell minimization to remove model bias (Hodel et al. 1992). The resulting 2F_o–F_c simulated annealing omit map clearly defined the trans peptide bond, including strong density for the carbonyl oxygen of Asn113 (Fig. 1 ▶).

Figure 1.

Electron density surrounding the site of the P114G substitution, with coordinates from the final refined structure. The configuration of the peptide bond preceding the mutated residue 114 is now trans. The fit of the carbonyl groups for residues 113 and 114 to the density confirms the trans configuration. The density corresponding to the bound cesium ion is also easily observed. Nitrogen, carbon, oxygen, and amide hydrogen atoms are colored blue, gray, red, and white, respectively.

Comparison of the P114G and wild-type structures

While several native RNase A structures have been previously reported (Kim et al. 1992), for this study we compared the mutant structure with a native structure crystallized under similar conditions (P. Harkins and H. Wyckoff, pers. comm.). A superposition of the C_α back-bone of this crystal form and P114G is shown in Figure 2A ▶. The differences between the two structures were investigated by the difference distance matrix approach (Kundrot and Richards 1987). The overall mean distance difference between the two structures is 0.35 Å (ES = 0.64 Å) for the C_α’s of all residues, and 0.25 Å (ESD = 0.30 Å) for C_α’s excluding residues 110–116. The latter deviation is not highly significant when compared with the predicted mean atomic coordinate error of 0.20 Å as calculated from a Luzzati plot (Luzzati 1952). The average temperature factors for the backbone and side-chain atoms of P114G and the wild-type structure are similar, further suggesting that only limited changes have occurred in the global protein structure. The position of residues forming the active site (His12, Lys41, Val43, Asn44, Thr45, His119, Phe120, Asp121, and Ser123) are the same in the mutant and the wild type, consistent with the retention of nearly wild-type catalytic activity (Schultz and Baldwin 1992; Schultz et al. 1992).

Figure 2.

(A) An alignment of the C_α atoms of the P114G and wild-type RNase A structures (Kim et al. 1992), colored gray and red, respectively. The overall structure is similar; however, there is a large structural rearrangement localized primarily to the loop containing the P114G substitution. The largest movement is the 9.3 Å shift of the C_α of residue 114. (B) A comparison of the P114G and wild-type structures (see Fig. 1 ▶), colored red and gray, respectively, showing the configuration of the mutated loop (residues 111–116) and the associated side chains. Wild-type Pro114 and all of the P114G side chains are labeled. In the wild-type structure, the peptide bond preceding Pro114 is cis, whereas in the P114G structure the peptide bond between Asn113 and Gly114 is trans.

Based on the difference distance plot (Fig. S1), the most significant structural changes are confined to the site of mutation (residues 110–116) (Fig. 2B ▶). Minor structural differences occurred in the 65–70, 35–40, and 91–96 turns. The region of residues 35–40 is known to be variable between different RNase structures (Kim et al. 1992), although the type of turn remains constant. In contrast, in the mutated loop, a new turn type was adopted, resulting in a 9.3 Å movement of the C_α atom of residue 114 (Fig. 2A ▶). The isomerization of the peptide bond preceding residue 114 from cis to trans altered the φ,ψ dihedral angles of the neighboring residues; Gly112 moved from the ~p region (wild type) to the I region (P114G), and the introduced Gly114 was found in the b region (Table S1). Nevertheless, all non-glycine/-proline residues in both structures had dihedral angles that fall in the allowed, low-energy areas of the Ramachandran plot. Along with the isomeric state of the 113–114 peptide bond, the most striking difference between the P114G and wild-type loop conformations is the rearrangement of the hydrogen-bond network. Three hydrogen bonds and two bifurcated hydrogen bonds present in the cis wild-type structure were replaced by four hydrogen bonds and two bifurcated hydrogen bonds in the trans P114G structure (Table S2). In the P114G protein, a new hydrogen bond was formed between Gly114 amide and Glu111 carbonyl oxygen, whereas in the wild-type protein, the proline side chain prevents hydrogen bonding to residue 114. An additional hydrogen bond between Cys110 and Asn113 was observed in the trans P114G structure. The observed dihedral angles and hydrogen-bonding pattern in P114G define the new β-turn as type IIB (Richardson 1981). Figure 2B ▶ shows the alignment of the wild-type and P114G structures, with cis and trans peptide bonds, respectively, preceding residue 114.

Identification of associated ligands using anomalous density

Anomalous difference maps were calculated to confirm the presence of a sulfate ion in the active site (Wlodawer et al. 1988) and to help identify the ligand giving rise to the strong electron density peak next to residue 114. Anomalous electron density (>3σ) was observed for the four disulfide bonds, and two of the five methionine residues. The density also showed a sulfate or phosphate present in the active site, and a very large (>15σ) peak observed in close proximity to residue 114 (Fig. 3A ▶). A comparison of the (relative) anomalous density peak heights suggested that the anomalous scattering was too large to be accounted for by a sulfate ion (f″ = 0.6 e⁻). Consideration of the ligand coordination geometry and strong anomalous density suggested that a cesium ion (f″ = 8.3 e⁻) was bound to the protein. While 3 M CsCl was present in the original crystallization conditions, most of it was removed when transferring the crystal into stabilizing solution. The anomalous density and associated ligands are shown in Figure 3B ▶. The predominant interactions and distances are listed in Table 2. The occupancy of the cesium ion (Q = 0.46) was calculated using CNS (Brünger et al. 1998). The B-factor of 24.8 Ų shows that the cesium is not extraordinarily mobile. The conformation of the loop must be similar with and without bound cesium as no additional density is observed for an alternative loop conformation, despite the partial occupancy of the cesium ion (Q = 0.46). This is further supported by the similarity in B-factors for the cesium and the loop, suggesting that the observed loop conformation is at full occupancy, while the cesium ion is only partially occupied.

Figure 3.

(A) The calculated anomalous difference map (F_o − F_c) for the P114G structure. Electron density is observed at the expected sites of the disulfide bonds, and also in the active site, where phosphate or sulfate is known to bind. The largest patch of anomalous electron density (noted by arrow) is localized near the site of the P114G mutation, and is attributed to the coordination of a cesium ion to the mutated loop. The side chains of all cysteines, methionines, and residues in close spatial proximity to the associated cesium ion are also displayed. (B) Structure of P114G, displaying residues in close proximity to the bound Cs⁺ ion. The anomalous electron densities for the cesium ion and five protein ligands are highlighted. The remaining coordination ligands are solvent- or symmetry molecule–based.

Table 2.

Cesium ligand distances for P114G ribonuclease A

Ligand	Atom	Ligand–Ion Distance (Å)
Cys58	Backbone oxygen	2.96
Tyr73	Oxygenη	3.32
Asn113	Backbone oxygen	3.18
Gly114	Backbone oxygen	3.07
Tyr115	Oxygenη	4.32

η, ring oxygen.

Discussion

Analysis of the P114G mutant structure

The crystal structure of the P114G RNase A mutant was solved in order to determine the structural consequence of replacing a cis proline, and, in particular, the role that Pro114 plays in influencing RNase A’s stability and refolding. We initially addressed the questions of whether the energetic penalty of placing the substituted amino acid into the cis configuration was larger than the energetic penalty caused by structural rearrangements that would result from the peptide bond’s isomerization, and whether the structural effects of the perturbation are localized or global.

Relationship between evolution and the occurrence of cis peptide bonds

The low frequency of direct replacement of cis-proline in proteins throughout evolution suggests that there is an energetic penalty. Ribonuclease sequences from different species show that although cis Pro114 is usually conserved, it is replaced by Leu in the capybara, and is deleted in the kangaroo, wallaby, and turtle (Beintema et al. 1986). This latter result coincides with the finding that, during evolution, cis-proline replacement is frequently accompanied by a neighboring insertion or deletion. The additional mutations may be necessary to provide increased flexibility to the substituted region so that it can rearrange to a stable structure with all trans peptide bonds. The frequency of cis non-proline peptide bonds is very low (0.03%–0.05%) based on an examination of known crystal structures (Stewart et al. 1990; Weiss et al. 1998; Jabs et al. 1999), whereas cis X-Pro peptide bonds occur commonly in folded proteins (5.2%–6.5%) (MacArthur and Thornton 1991). Therefore, itis not surprising that after cis-proline replacement, the peptide bond adopts the trans configuration (Stewart et al. 1990; Hynes et al. 1994). Further support for this expectation is provided by the unfolded state where the trans isomeric form is the dominant configuration for non-proline peptide bonds (99% trans), whereas for X-Pro peptide bonds (Grathwohl and Wüthrich 1976)the cis form is significantly populated. In the unfolded state of RNase A, the Asn113–Pro114bond is 37% cis (Adler and Scheraga 1990). So what are the structural consequences of simple cis-proline replacement? It necessarily creates either a disfavored cis non-proline peptide bond, or a change in backbone geometry with a trans peptide bond.

Structural effects of the mutation

The pivotal change in the structure is the isomerization of the cis peptide bond to trans. The other large movements in the loop probably are a consequence of the trans isomer. The differences between the structures of wild-type RNase A and P114G are largely confined to the loop containing the mutation, residues 110–117 (Figs. 2A ▶, S1). More subtle long-range effects of the mutation are minor, and we have made no attempt to interpret them because they may also result from crystal packing forces. Remarkably, a large adjustment in this loop (9.3 Å movement) can be accommodated without compromising the integrity of the rest of the protein. It should be noted that although there was no observed electron density (or other evidence) consistent with a cis peptide bond preceding Gly114, an extremely minor population of this isomeric form cannot be ruled out based solely on crystallographic analysis.

Why is the glycine variant trans, whereas the proline in wild-type RNase A is cis?

With the insight gained from the P114G structure we can now address a related question, why is the turn cis in wild-type RNase A when proline is present and what are the different factors that contribute to the isomeric preferences of the folded proteins? In the unfolded state of both P114G and the wild-type protein, the preferred isomeric configuration for the peptide bond preceding Pro114 is trans (wild type ~63% [Adler and Scheraga 1990] vs. 84%–94% in the mutant [Houry and Scheraga 1996]). Our finding that static substitution of the proline side chain of residue 114 in the P114G trans structure introduces steric clashes between the C_γ and C_δ of proline (1.5 Å and 1.4 Å respectively) and the carbonyl of Glu111 provides support for a model where the wild-type trans configuration is strained. Energy minimization of the proline trans structure relieves the steric clash but weakens the hydrogen bonding network, specifically, the hydrogen bond between E111N and Val116O (Table S2). However, when the 113–114 proline peptide bond is cis, less strained backbone conformations are available, which presumably compensate for the energy required for formation of a cis peptide bond. In contrast, in the P114G and P114A mutants, there is a larger driving force to remain trans due to the unfolded state equilibrium and the fact that both glycine and alanine are accommodated in the folded trans configuration without steric clash. A possible additional factor contributing to the formation of a trans peptide bond in the P114G variant is that in the wild-type protein the φψ values of Asn113 (−135.7°, 106.8°) and Pro114 (−58.9°, 151.8°) are not readily compatible with the average values of φψ for residues flanking non-proline cis peptide bonds (123°, 121° and 102°, 152°). Presumably, a more negative value of φ would be required to accommodate a non-proline cis peptide bond in the structure (Jabs et al. 1999; Pal and Chakrabarti 1999).

Influence of the cesium ligand on the cis/trans equilibrium configuration

Interpretation of the structural effects of the mutation must be examined carefully in light of the observation that a Cs+ ion is bound to the P114G variant through the side-chain oxygens of Tyr73 and Tyr115 and the backbone carbonyl oxygens of residues 58, 113, and 114 (Table 2). Several of these residues (113, 114, and 115) are present in the turn variant. The Cs+ ion appears to bind to a pre-existing loop conformation. If the Cs+ ion were driving a distinct conformation, we would expect an alternative conformation to have been visible in the density with an occupancy of up to 0.54. If the putative unbound form were disordered, the observed conformation would reflect this with a partial occupancy or higher B factors. The fact that the 0.46 occupancy Cs+ ion and the full occupancy loop have similar and moderate B factors indicates that the Cs+ ion is not perturbing the conformation of the loop, but is instead binding to a pre-existing conformation.

Comparison with other structural studies of cis-proline replacements in proteins

Interestingly, the generation of an Asn113–trans Gly114 peptide linkage is found here, in contrast with the X-ray structures of carbonic anhydrase II (P202A) (Tweedy et al. 1993), aspartate aminotransferase (P195A) (Birolo et al. 1999), Escherichia coli aspartate transcarbamoylase (P268A) (Jin et al. 2000), and the NMR structural analysis of ribonuclease T₁ (P39A), where cis peptide bonds are retained after the substitution (Mayr et al. 1994). Conformational changes associated with either the isomerization of pre-proline peptide linkages, or cis-proline replacement have also been structurally examined in bovine pancreatic ribonuclease A (Pro93 variants) (Pearson et al. 1998; Schultz et al. 1998; Xiong et al. 2000), staphylococcal nuclease variants (Evans et al. 1987, 1989; Hinck et al. 1993; Hodel et al. 1993, 1994, 1995a, Hodel et al. b; Hynes et al. 1994; Maki et al. 1999), and calbindin (Svensson et al. 1992). These studies demonstrate that either isomer can be accommodated, sometimes both isomers are found, or the isomer choice is dependent on the characteristics of the introduced amino acid. The cis peptide bond adopts a trans configuration in a P137A variant of aspartate aminotransferase (Birolo et al. 1999) and in a P76A variant of a protein fragment complementation system of thioredoxin (Yu et al. 2000).

Proline isomerization and effects on protein folding

The effects of cis-proline replacement on protein-folding kinetics have been examined in thioredoxin (Kelley and Richards 1987), RNase T1 (Kiefhaber et al. 1990a,b; Mayr and Schmid 1993; Mayr et al. 1993), aspartate aminotransferase (Birolo et al. 1999), pectate lyase C (Kamen and Woody 2002), and ribonuclease A (Schultz and Baldwin 1992; Schultz et al. 1992; Dodge and Scheraga 1996; Wedemeyer et al. 2002). In the case of RNase A (Pro93) and RNase T1 (Pro39), TEM-1 β-lactamase (Pro167) (Vanhove et al. 1996), α subunit of trp synthase (Pro28) (Wu and Matthews 2002), the folding kinetics were found to be consistent with the presence of introduced non-proline cis peptide bonds. This was later confirmed by structural analysis (Odefey et al. 1995; Pearson et al. 1998; Xiong et al. 2000). In contrast to the folding kinetic complexity that might occur if a non-Pro residue can adopt the cis isomer (in the examples above), the folding kinetics of the P114G variant are simplified. In addition, the result reported here that the P114G mutant has a trans peptide bond preceding residue 114 is consistent with the finding that the unfolded wild-type species U_F, which is postulated to contain a trans Asn113–Pro114 peptide bond, can fold into a stable trans conformation before isomerization of the peptide bond to the cis configuration (Dodge and Scheraga 1996; Houry and Scheraga 1996; Juminaga et al. 1997; Bhat et al. 2003). Our finding of a trans P114G peptide bond also agrees with the results of several theoretical and experimental studies showing that a nonnative Pro114 does not perturb the native conformation or interfere with the refolding of RNase A significantly (Stimson et al. 1982; Pincus et al. 1983; Oka et al. 1984; Ihara and Ooi 1985; Schultz and Baldwin 1992).

Materials and methods

Expression, purification, and crystal growth

The ribonuclease A P114G variant was expressed and purified as described previously (Schultz and Baldwin 1992; Schultz et al. 1992). The P114G protein (5 mg/mL) was crystallized from a solution containing sodium phosphate and cesium chloride (0.15 M Na₂HPO₄, 3 M CsCl₂ [pH 6.6]) using the precipitant ammonium sulfate (30% of saturation) at 25°C. Diamond-shaped crystals of dimensions 0.3 × 0.2 × 0.2 mm were obtained after a few days. The crystals were transferred to >95% ammonium sulfate, 0.1 M sodium phosphate (pH 6.6) for stabilization. Crystals were mounted in 0.8-mm glass capillaries with a small amount of mother liquor on both sides. The crystals obtained belonged to the tetragonal space group P4₁2₁2 or its enantiomorph (P4₃2₁2), with unit cell dimensions a = b = 40.87 Å, c = 129.25 Å.

Data collection and processing

Diffraction data were collected from a single crystal of P114G at 15°C using an R-axis II imaging plate system. CuKα radiation generated from a Rigaku RU300 rotating anode operating at 50 kV and 90 mA and focused with double Franks’ focusing mirrors. Each 1° oscillation image was collected for 10 min. Images were reduced using DENZO, and scaled and post-refined with SCALEPACK (Otwinowski and Minor 1997). Only frames that did not show significant radiation damage were included. Data were included to a d spacing of 2.0 Å, the highest resolution shell with an average intensity >2σ. 97,542 observations of 8030 reflections covering 97% of the possible reflections to 2.0 Å were collected. The R-merge was 10.7% and the R-factor between Friedel mates was 5.3%.

Structure solution and refinement

The refined structure of wild-type RNase A crystallized under similar high-salt conditions (Kim et al. 1992; P. Harkins and H. Wyckoff, pers. comm.), with solvent molecules, sulfate ligand, and residues 110–116 removed, was used as the molecular replacement search model. All molecular replacement was done with data between 8.0 and 3.5 Å resolution and F_o ≥ 2σ. Rotation functions were performed in XPLOR (Brünger 1987) via a Patterson search method. The orientation of the top solution (7.1σ) was refined using the Patterson correlation coefficient of the squared normalized structure factors as the target. The refined orientation was used in translation searches with the same target using the enantiomorphic space groups P4₃2₁2 and P4₁2₁2. Comparison of the solutions indicated the space group as P4₃2₁2 with a 15.6σ peak in the translation function. The R-factor was 38.1% after rigid body refinement.

Crystallographic least squares refinement was performed using the wild-type high-salt RNase A coordinates (Kim et al. 1992; P. Harkins and H. Wyckoff, pers. comm.) as the starting model. To minimize model bias, residues 110–116, which are the solvent-exposed loop and the site of mutation, were omitted from the starting model. Solvent molecules, the sulfate ligand, and the side chain of Lys1 and Lys91 were also excluded in the initial rounds of refinement. Additionally, the extra methionine at the N terminus, present as a consequence of heterogonous expression in E.coli, was not included in the model. Rigid body minimization, positional, and B-factor refinement were carried out with the program XPLOR (Brünger et al. 1990). The edited model, after application of the rotation and translation solutions, was rigid-body refined against the low-resolution diffraction data (8–3.5 Å, F_o ≥ 2σ) using the crystallographic residual as the target function. An overall temperature factor was then applied to the model. An active-site sulfate ion is typically found in high-salt RNase A and S crystals (Kim et al. 1992). Density was also observed in the active-site region of the P114G mutant and modeled as a sulfate. The model was then further refined against the high-resolution data (6–2 Å, F_o ≥ 2σ) by alternate cycles of positional and individual restrained B-factor refinement to an R-factor of 24.1%. The omitted loop residues (110–116) were manually fit to the electron density of an F_o–F_c map by visual inspection using FRODO (Jones 1985). The final refinement using CNS included all observations (30.0–2.0 Å) and utilized the PMB/CNS automated refinement interface (Scott et al. 2004). Sixty-two waters were placed in the map using CNS (Brünger et al. 1998) and edited by manual inspection using XtalView (McRee 1999). Occupancy of all waters was fixed at unity, waters with a B > 60 Ų after refinement were removed from the model. The N-terminal Met, originally omitted from the model, was located in a σ_A-weighted 2mF_o–nF_c map.

Identification of a bound cesium ion through anomalous differences

A large difference density peak was observed near the site of the P114G mutation in maps calculated with F_o–F_c amplitudes and phases from the refined protein model. The height of this peak was significantly greater than that of the active site sulfate. An anomalous difference map calculated with anomalous difference amplitudes between 8 and 3 Å resolution and using refined model phases, revealed significant density at the same site (12.8σ) and lower density at the sulfate and each disulfide site. Because of the strong anomalous signal, the missing density was attributed to an element with an anomalous scattering f″ component larger than sulfur (f″_sulfur = 0.6 e⁻ at CuK_α). Consideration of the elements in the crystallization and storage buffers indicated cesium (f″_cesium = 8 electrons at CuK_α) as the only element that could account for both the normal and anomalous electron density.

The occupancy (Q) and B-factor (B) of the cesium ion were refined as part of the final structure refinement in CNS (Q = 0.46, B = 24.8 Ų). The final model was refined using CNS with all data within 30.0–2.0 Å resolution, using a bulk solvent model. A final round of alternating positional and B-factor refinement (until the R_free no longer improved) resulted in an R-factor of 19.5% and an R_free of 21.9%.

Structural analysis

RMS differences between the structures were calculated after superimposing the backbone atoms (N, C, O, and C_α) of the wild-type (high-salt crystal form) (Kim et al. 1992) and P114G mutant using the least-squares methods in CNS (Brünger et al. 1998) and MIDAS (UCSF Graphics Laboratory). Ligands and residues 110–117 were excluded from the superposition. Figures were prepared using the program NEON in the MIDAS package (UCSF Graphics Laboratory). The structure was also analyzed using PROCHECK (Laskowski et al. 1996).

Coordinates

The coordinates and structure factors have been deposited in the Protein Data Bank (accession code 1KH8).

Article published online ahead of print. Article and publication date are at http://www.proteinscience.org/cgi/doi/10.1110/ps.051610505.