Distinguishing multiple chemotaxis Y protein conformations with laser-polarized ¹²⁹Xe NMR

Abstract

The chemical shift of the ¹²⁹Xe NMR signal has been shown to be extremely sensitive to the local environment around the atom and has been used to follow processes such as ligand binding by bacterial periplasmic binding proteins. Here we show that the ¹²⁹Xe shift can sense more subtle changes: magnesium binding, BeF₃⁻ activation, and peptide binding by the Escherichia coli chemotaxis Y protein. ¹H-¹⁵N correlation spectroscopy and X-ray crystallography were used to identify two xenon-binding cavities in CheY that are primarily responsible for the shift changes. One site is near the active site, and the other is near the peptide binding site.

Several properties of ¹²⁹Xe have made it an attractive NMR-based biomolecular probe. These include xenon’s affinity for small hydrophobic cavities in proteins, the remarkable chemical shift sensitivity of xenon to nonbonded local environment, and the signal-to-noise enhancement made possible by optical pumping (Goodson 2001). Laser-polarized ¹²⁹Xe nuclear magnetic resonance (¹²⁹Xe NMR) spectra are simple, have no background, and can be obtained in a single scan (Goodson 2001). ¹²⁹Xe NMR has been used to probe hydrophobic cavities in several proteins, including myoglobin, hemoglobin, T4 lysozyme, maltose binding protein, and recently an engineered cavity in the ribose binding protein (Tilton and Kuntz 1982; Bowers et al. 1999; Rubin et al. 2002; Lowery et al. 2004).

The possibility of using ¹²⁹Xe NMR as a means for detecting biomolecular conformational states was first demonstrated for maltose binding protein, MBP (Rubin et al. 2000). When MBP binds maltose, which changes it from the open to the closed conformation, the observed ¹²⁹Xe chemical shift moves 0.9 ppm mM⁻¹ upfield (Rubin et al. 2000). The dependence of ¹²⁹Xe chemical shift on MBP conformation arises from a single xenon binding cavity whose ¹²⁹Xe chemical shift and binding affinity depend on the protein’s conformation, a conformation-sensitive xenon-binding cavity (Rubin et al. 2002). Because such a cavity serves as a conformational reporter site, the protein does not need modification and is fully recoverable after analysis.

Here we report a further example in which ¹²⁹Xe NMR responds to several conformational states of the Escherichia coli chemotaxis Y protein (CheY). CheY (12 kDa) is a signal transduction protein required for chemotaxis that is activated by phosphorylation on a conserved aspartate residue (Asp⁵⁷). Phosphorylation induces a conformational change that increases CheY’s affinity for its downstream target protein, FliM (Falke et al. 1997). Here we demonstrate that CheY contains two xenon-binding cavities that confer ¹²⁹Xe chemical shift sensitivity to three different protein functional states as well as providing moderate sensitivity to peptide binding. ¹H-¹⁵N correlation spectroscopy (HSQC) and X-ray crystallography were used in identifying and characterizing the sites of xenon binding.

Results

¹²⁹Xe chemical shift sensitivity to CheY conformation

The change in ¹²⁹Xe chemical shift with increasing protein concentration (the concentration-normalized ¹²⁹Xe chemical shift) is a useful means of characterizing xenon–protein interactions. A significant difference between the concentration-normalized ¹²⁹Xe chemical shift of the folded and unfolded states of a protein indicates the presence of a xenon-binding cavity (Rubin et al. 2002). Figure 1 ▶ shows a plot of the ¹²⁹Xe chemical shift as a function of protein concentration for the folded and unfolded conformations of CheY. The concentration-normalized ¹²⁹Xe chemical shift values for the folded and unfolded CheY conformations are, respectively, 1.6 ± 0.1 ppm mM⁻¹ and 0.4 ± 0.1 ppm mM⁻¹. The strong dependence of xenon chemical shift on the folded state of CheY, twice the magnitude seen for MBP (Rubin et al. 2002), indicates the presence of at least one xenon-binding cavity.

Figure 1.

Change in ¹²⁹Xe chemical shift (ΔδXe) with protein concentration (C) for apo (○), magnesium-bound (▪), activated (•), FliM peptide-bound (□), and denatured (⋄) CheY. Δ δXe is the difference between the ¹²⁹Xe chemical shift of each titration point and buffer. ¹²⁹Xe chemical shift values and error bars (± 0.01 ppm) were obtained from peak fits. The concentration-normalized ¹²⁹Xe chemical shifts, or slopes of the lines, are 1.6 ± 0.1 ppm mM⁻¹, 1.3 ± 0.1 ppm mM⁻¹, 0.8 ± 0.1 ppm mM⁻¹, 0.7 ± 0.1 ppm mM⁻¹, and 0.4 ± 0.1 ppm mM⁻¹, for apo, magnesium-bound, activated, FliM-bound, and denatured CheY, respectively. ¹²⁹Xe chemical shift is sensitive to four different CheY functional states.

Figure 1 ▶ also shows that ¹²⁹Xe is affected differently by different states of CheY; the normalized shifts for the states—inactive, magnesium-bound, activated, and FliM-peptide bound—are 1.6 ± 0.1 ppm mM⁻¹, 1.3 ± 0.1 ppm mM⁻¹, 0.8 ± 0.1 ppm mM⁻¹, and 0.7 ± 0.1 ppm mM⁻¹, respectively. The sensitivity of ¹²⁹Xe chemical shift to four CheY conformations is the first example of ¹²⁹Xe responding to more than two different protein conformational states for a single protein and the first example of ¹²⁹Xe sensitivity to protein–peptide binding.

Identifying the xenon-binding cavities with ¹H-¹⁵N HSQC spectroscopy

The location of xenon binding cavities in proteins can be determined using amide ¹H-¹⁵N correlation (HSQC) spectra taken at different xenon concentrations (Rubin et al. 2002; Lowery et al. 2004). Assigned ¹H-¹⁵N HSQC resonances affected by increasing xenon concentration can be mapped onto the protein structure, identifying the region of the protein that is involved in xenon binding (Rubin et al. 2002; Gröger et al. 2003).

¹H-¹⁵N HSQC spectra were collected from both inactive and activated CheY conformations in the presence and absence of xenon. The total amide proton chemical shift difference between 0 mM and ~40 mM xenon for each residue of both the inactive and active conformations are plotted as histograms in Figure 2, A and B ▶. Three regions in the sequence are affected by xenon binding, residues near Asp¹², residues near Asp⁵⁷, and residues between Thr⁸⁷ and Tyr¹⁰⁶. Mapping the residues according to the magnitude of their xenon-induced shifts onto the inactive CheY backbone shows that the residues whose environments are perturbed most by increasing xenon concentration cluster about two cavities that were identified with the program VOIDOO (Fig. 3A ▶). One of these cavities (H1-β3) is located directly below residues Asp⁵⁷ and Asp¹², involving residues from H1, β1, β3, and β4. The second cavity (β4-H4) is located near Tyr¹⁰⁶, involving residues from β5, H4, and β5 (Cho et al. 2000; Lee et al. 2001a). The residues that significantly shift upon xenon binding for the activated conformation of CheY cluster about similar regions of the protein (Fig. 3B ▶).

Figure 2.

Histogram plots of the total xenon-induced shift for each assigned residue. (A) Shifts induced by 40 mM xenon for apo CheY. (B) Shifts induced by 45 mM xenon for activated CheY. Similar segments of the protein undergo xenon-induced shifts for both CheY conformations.

Figure 3.

Xenon-induced shifts from Figure 2 ▶ mapped on the backbone of apo CheY (A) (Simonovic and Volz 2001) and activated CheY (B) (this work). Residues are shaded according to the magnitude of their xenon-induced shifts. Unassigned residues are shaded white. For both conformations of CheY shifting residues cluster around two cavities identified by VOIDOO, the H1-β3 cavity (black mesh) and the β4-H4 cavity (gray mesh). Magnesium-binding and BeF₃⁻ activation occur in the active site, directly above the H1-β3 cavity. The side chains of Tyr¹⁰⁶ and Asp⁵⁷ are shown in both structures. Activation at Asp⁵⁷ results in Tyr¹⁰⁶ changing its position from “out” to “in,” filling the β4-H4 cavity. Cavities were calculated with VOIDOO (Jones et al. 1991) and images generated with PyMol (DeLano Scientific LLC).

Identifying the xenon-binding cavities with X-ray crystallography

The xenon binding sites in CheY were also analyzed using X-ray crystallography. The structure of CheY activated with the aspartyl phosphate mimic BeF₃⁻ was previously determined (Lee et al. 2001a) (crystallized in the space group P2₁2₁2₁ with two molecules in the asymmetric unit and diffracted to 2.4 Å). Crystals of BeF₃⁻ activated CheY were pressurized with 100 psi xenon for 1 min before freezing. Data collection and refinement statistics from the xenon-pressurized crystals are summarized in Table 1.

Table 1.

Crystallographic data collection and refinement statistics

Crystal
Space group	P212121
Cell dimensions a, b, c (Å)	53.7, 53.87, 161.1
Data collection (cryogenic)
Resolution limit (Å)	21.9–2.4
Measured reflections	224,801
Unique reflections	19,018
Rsym (%) (overall/last shell)	10.3/43.9
Completeness (%) (/last shell)	99.6/99.8
Wilson B-factor (Å 2)	57.6
Refinement
No. molecules in AU	2
No. amino acid residues per molecule	127
No. solvent molecules
No. Mn++ ions	2
No. water molecules	174
No. xenon molecules	2
Resolution used (Å)	20–2.4
σcutoff	0.0
No. of reflections work/test	17,103/1865
Final R-factor/Rfree (|F| > 0 σ) (%)	22.7/26.7
Average B-factor (Å 2)	52.3
RMS deviations from ideal geometry
Bond length (Å)	0.01
Bond angles (deg.)	1.5

Figure 4 ▶ shows a difference electron density omit map (F_obs–F_calc) with no xenon in the model. Electron density of the xenon atom (greater than 10σ) is evident directly below the active site, consistent with a single xenon atom in the H1-β3 cavity. The occupancy of xenon at this site is 0.3 in the final model. No excess electron density from xenon is evident in the β4-H4 region. Crystals were pressurized at a pressure of 150 psi in an attempt to observe xenon electron density in the β4-H4 region, but crystals cracked under this higher pressure.

Figure 4.

X-ray crystal structure of BeF₃⁻-activated CheY bound to xenon. (A) An F_obs–F_calc difference electron density omit map (mesh) contoured at 10σ with the xenon atom removed from the model. Electron density from the xenon atom is clearly visible just below the active site inside the H1-β3 cavity. (B) A 2F_obs–F_calc electron density map (mesh) of the H1-β3 cavity contoured at 1.5σ showing the xenon atom (sphere) and the surrounding residues (sticks).

Xenon affinity of the β4-H4 cavity is conformation dependent

Xenon binding affinities were measured by monitoring the changes in chemical shift of peaks in the ¹H-¹⁵N HSQC spectra of both inactive and activated CheY with increasing xenon concentration controlled by overpressure (Rubin et al. 2002). Those residues whose resonances underwent the most significant change in chemical shift for both conformers were Asp¹², Arg¹⁸, Ile²⁰, Asp⁵⁷, Ala⁸⁸, Tyr¹⁰⁶, and Val¹⁰⁷. The binding curves of these residues were fitted to a single-site binding model, as previously described (Rubin et al. 2002). Normalized data were obtained using the limiting shifts obtained from the initial fits. The normalized data for residues unique to the H1-β3 region, Asp¹², Arg¹⁸, Ile²⁰, and Asp⁵⁷, were averaged to obtain an average, normalized binding curve for the H1-β3 cavity (Fig. 5A ▶). Similarly, the normalized data for residues Tyr¹⁰⁶ and Val¹⁰⁷ were averaged to obtain an average, normalized binding curve for the β4-H4 cavity. Fitting these curves yielded xenon association constants for both cavities in the inactive and active conformations of CheY (Table 2). The binding affinity of the H1-β3 cavity in the activated state, ~27 M⁻¹ (Table 2), is comparable to that calculated from the occupancy of xenon in the activated CheY crystal structure, ~10 M⁻¹, assuming 4.4 mM xenon per atm (Rubin et al. 2002). The observed xenon binding affinities are similar to those previously reported for engineered RBP, MBP, and T4 Lysozyme mutants (Rubin et al. 2002; Lowery et al. 2004).

Figure 5.

(A) The average of the normalized change in chemical shift (Δ δ) vs. xenon concentration, C, for the amide proton resonances of residues Asp¹², Arg¹⁸, Ile²⁰, and Asp⁵⁷, which line the H1-β3 cavity, for the apo (○, solid line) and the activated (•, dashed line) forms of CheY. (B) The average of the normalized change in chemical shift (Δ δ) vs. xenon concentration, C, for the amide proton resonances of residues Tyr¹⁰⁶ and Val¹⁰⁷, which line the β4-H4 cavity, for the apo (□, solid line) and activated (▪, dashed line) conformations of CheY. Chemical shift changes were normalized according to the limiting shift obtained by an initial fit for each residue. These normalized values were averaged for residues unique to the two cavities to obtain the data and error bars shown in this figure. The respective binding constants obtained from these fits are shown in Table 2. Unlike the H1-β3 cavity, the xenon binding affinity of the β4-H4 cavity changes with activation.

Table 2.

Cavity volumes and xenon binding affinities for the inactive and active conformers of CheY

	Ka(Xe)a/M−1		Cavity volumeb/Å 3
	H1-β3 cavity	β4-H4 cavity	H1-β3 cavity	β4-H4 cavity
Inactive	27 ± 8	35 ± 8	74c	92c
Active	27 ± 8	22 ± 8	58d	116e

^a Xenon binding affinities were determined as described in the Materials and Methods section.

^b Cavity volumes were calculated from:

^cSimonovic and Volz 2001,

^d this work,

^e this work with Tyr¹⁰⁶ rotated to the out position.

Discussion

For maltose binding protein (MBP) ¹²⁹Xe NMR was used to discriminate open and closed conformations (ligand free and ligand bound forms), by the different ¹²⁹Xe shifts induced by the conformation-sensitive xenon-binding site (Rubin et al. 2000). An analogous conformation-sensitive xenon-binding site was engineered into ribose-binding protein (RBP) (Lowery et al. 2004). For both MBP and engineered RBP, a single xenon-binding site was reporting on the two conformational states. CheY is the first example in which ¹²⁹Xe NMR is sensitive to multiple protein conformation states as well as to peptide binding.

Conformation-dependent xenon binding cavities in CheY

There is a direct correlation between the locations of the xenon-binding cavities predicted by searching the crystal structure of apo CheY with VOIDOO and the clustering of xenon-sensitive residues in apo CheY measured with ¹H-¹⁵N correlation spectroscopy (Fig. 3 ▶). These cavities are near residues of CheY that are centrally involved in activation: Asp¹², Asp⁵⁷, Thr⁸⁷, and Tyr¹⁰⁶ (Cho et al. 2000). It is reasonable that changes in their positions should give rise to changes in ¹²⁹Xe chemical shift.

When CheY undergoes activation, Asp⁵⁷ and Thr⁸⁷ participate in binding phosphate (or equivalently BeF₃⁻). Consequently, the side chain of Thr⁸⁷ moves toward the active site, increasing the size of the β4-H4 cavity. This allows the side chain of Tyr¹⁰⁶ to rotate from a surface exposed position to a buried one (Fig. 3 ▶) (Cho et al. 2000). In the activated CheY structure the β4-H4 cavity has disappeared, and one would expect that xenon could no longer occupy this site. However, xenon-sensitive residues in this region are observed with ¹H-¹⁵N correlation spectroscopy, indicating that xenon does still bind in both the H1-β3 and the β4-H4 regions. Previous observations indicate that the functional states of CheY are a dynamic equilibrium of discrete conformational subpopulations (Cho et al. 2000). A change between two functional states arises from an increased stabilization of one set of conformational subpopulations and resulting population shift (Simonovic and Volz 2001). This suggests that in solution xenon binds to the β 4-H4 cavity of the activated state by displacing the Tyr¹⁰⁶ side chain to accommodate the xenon atom. The lack of binding seen in the crystals may be because the β4-H4 region is at an interface between the noncrystallographic symmetric dimer of activated CheY, sterically preventing Tyr¹⁰⁶ from rotating out.

The chemical shift response of ¹²⁹Xe to conformational changes in proteins arises from changes in the size and shape of cavities affecting both the bound xenon chemical shift and binding constant (Rubin et al. 2002; Lowery et al. 2004). Table 2 summarizes the cavity volumes calculated from the inactive and activated CheY crystal structures (Simonovic and Volz 2001 and this work, respectively) and measured xenon binding affinities. The xenon binding affinity of the H1-β3 cavity remains unchanged at 27 M⁻¹ when CheY is activated and the cavity volume decreases by 16 ų, indicating that change in ¹²⁹Xe chemical shift for this cavity arises only from a change in the bound chemical shift. In contrast, for the β4-H4 cavity CheY activation leads to an increase in cavity volume of 24 ų and a decrease in the xenon binding from 35 ± 8 M⁻¹ to 22 ± 8 M⁻¹ (Table 2), which is consistent with an energetic loss for the Tyr¹⁰⁶ side chain rotating out to allow xenon to occupy the cavity in the activated state.

The changes in cavity volumes and affinities shown in Table 2, together with the characteristics of the xenon binding cavities observed for MBP and engineered RBP (Rubin et al. 2002; Lowery et al. 2004), explain the dependence of the CheY-induced ¹²⁹Xe shift on CheY conformation. In the limit that a small fraction of xenon is bound at any site, the concentration normalized ¹²⁹Xe chemical shift, α_obs, is the sum of limiting chemical shift, δ_i, of ¹²⁹Xe at each site, weighted by its respective binding constants, K_i (Equation 1) (Rubin et al. 2002).

(1)

Contributions to the observed ¹²⁹Xe shift can be separated into nonspecific, nsp, and specific interactions, sp (Equation 2) (Rubin et al. 2001, 2002).

(2)

For the change in α_obs between two states, nonspecific binding sites can be considered to contribute equally to both conformations because of their low binding affinities. Using these assumptions and the binding affinity values in Table 2, Equation 2 can be rewritten using the change in the concentration-normalized ¹²⁹Xe chemical shift, Δα_i,a = 0.8 ppm mM⁻¹, between the inactive and activated conformations of CheY, i and a, respectively (Equation 3).

(3)

Previous work showed that the limiting xenon chemical shift is inversely proportional to cavity size (Rubin et al. 2002; Lowery et al. 2003). Considering changes in cavity volume only, the limiting shift of the H1-β3 cavity should increase upon activation, whereas that of the β4-H4 cavity should decrease upon activation. Therefore, the activation-induced effects of the H1-β3 cavity and the β4-H4 cavity on ¹²⁹Xe chemical shift partially offset each other. Because Δα_i,a is positive, the effects of the β4-H4 cavity on the observed ¹²⁹Xe shift must more than compensate for those from the H1-β3 cavity. This correlates with the relative magnitudes of the binding affinities (Equation 3), and volume changes (Table 2).

As mentioned above, both xenon-binding cavities are located in regions of CheY that are involved in a dynamic equilibrium between different conformational subpopulations. The sensitivity of ¹²⁹Xe chemical shift to the precise state of CheY reflects the sensitivity of these two cavities to shifts in conformational subpopulations that involve the the cavity-containing regions of CheY (Stock et al. 1993; Bellsolell et al. 1994; Djordjevic and Stock 1998).

As shown in Figure 1 ▶, there is also a moderate shift of ¹²⁹Xe upon peptide binding to CheY. Analogous to magnesium binding and activation, peptide binding shifts the conformational subpopulations of CheY in a manner that perturbs a xenon-binding cavity. CheY activation changes the H4-β5-H5 surface by rotating the Tyr¹⁰⁶ side chain from “out” to “in,” enhancing the affinity of CheY 20 fold for an N-terminal segment of its natural downstream target, the FliM protein (Bren and Eisenbach 1998; Lee et al. 2001b). With the FliM peptide bound, rearrangements that enable xenon binding to the β4-H4 cavity must be suppressed. The resulting decreased xenon occupancy of the β4-H4 cavity correlates with the further decrease in ¹²⁹Xe shift for the peptide complex relative to activated CheY alone.

Conclusions

The shift of ¹²⁹Xe in contact with CheY changes in response to transitions among four distinct CheY states: apo, magnesium-bound, activated, and FliM peptide-bound. The sensitivity of ¹²⁹Xe to these small conformational differences arises primarily from the subtle changes in the H1-β3 and β4-H4 cavities in the different CheY functional states. The response of ¹²⁹Xe to FliM peptide binding indicates that ¹²⁹Xe NMR can be used for sensing protein–protein interactions through a xenon-binding cavity near or coupled to the site of binding. The observations presented here, combined with the demonstration that reporter sites can be engineered into proteins (Lowery et al. 2004), indicate that ¹²⁹Xe NMR can be used to report almost any kind of protein conformational change or binding event.

Materials and methods

Both natural abundance and ¹⁵N-labeled E. coli CheY proteins were expressed and purified as described (Cho et al. 2000). Laser-polarized ¹²⁹Xe NMR spectroscopy for inactive, magnesium-bound, and activated CheY used 25 mM sodium phosphate buffer (pH 6.7), with the approach of sequential dilutions, as described (Lowery et al. 2004). Magnesium-bound CheY samples contained 5 mM MgCl₂. The magnesium dissociation constant was measured to be ~0.3 mM, which is close to previous estimates of ~0.5 mM (Lukat et al. 1990), by following the change in ¹H-¹⁵N chemical shift for Asp⁵⁷ versus total magnesium concentration and fitting the data to a single-site binding model. Therefore, ~90% of the protein was in the magnesium-bound state in the presence of 5 mM MgCl₂ (Moy et al. 1994). Activated CheY samples were prepared by adding MgCl₂, NaF, and BeCl₂ to final concentrations of 20 mM, 100 mM, and 16 mM, as previously described (Cho et al. 2000). Laser-polarized ¹²⁹Xe NMR spectroscopy for FliM-peptide-bound CheY was conducted as previously described for other proteins (Rubin et al. 2002) in 25 mM buffer (pH 6.7), 1.7 mM 16N FliM peptide (Lee et al. 2001b). VOIDOO calculations to identify cavities used a primary grid spacing of 0.5 Å and a probe radius of 1.4 Å (Jones et al. 1991; Rubin et al. 2002). VOIDOO calculations for apo CheY in Table 2 were done using coordinates from a previously reported crystal structure (Simonovic and Volz 2001). This structure of apo CheY does not contain bound xenon, but the similarity of the measured volumes for both cavities in activated CheY without xenon (Lee et al. 2001a) (62 ų for H1-β3 and 113 ų for H4-β4 with Tyr¹⁰⁶ rotated out) and with xenon (this work) (Table 2) indicate that xenon binding does not significantly alter these cavity volumes.

¹H-¹⁵N HSQC spectroscopy was conducted with 1.0 mM to 1.5 mM samples of ¹⁵N-labeled CheY in 25 mM sodium phosphate (pH 6.7). The ¹H-¹⁵N HSQC assignments for BeF₃⁻ activated CheY conformations have already been reported (Cho et al. 2000). The peak assignments for the inactive, or apo, conformer of CheY were obtained by measuring the ¹H-¹⁵N HSQC for 1.0 mM apo CheY, then following peaks as magnesium chloride was titrated from 0 mM to 10.0 mM. All resonances were in fast exchange between the magnesium-bound and magnesium-free forms for intermediate concentrations of magnesium. Resonances were tracked from the previously assigned magnesium-bound CheY spectrum (Moy et al. 1994). For both CheY conformers, xenon-induced shifts were assigned and reported for only those resonances that were clearly resolved (Figs. 2 ▶, 3 ▶). ¹H-¹⁵N HSQC xenon titrations were conducted as described (Rubin et al. 2002).

Crystals of the BeF₃⁻-activated CheY were prepared as previously described (Lee et al. 2001a). Crystals were pressurized with xenon at 100 psi for 1 min using a standard xenon chamber (Hampton Research) prior to freezing in liquid nitrogen (Rubin et al. 2002). Higher xenon pressures (≥150 psi) or longer exposure times (≥2 min) cracked the crystals. Diffraction data were collected at the Advanced Light Source on beamline 8.3.1 using an ADSC detector and a wavelength of 1.0 Å. The crystals were isomorphous with previously obtained crystals of activated CheY diffracting in spacegroup P2₁2₁2₁ with cell dimensions a = 53.73 Å, b = 53.82 Å, and c = 161.17 Å (Lee et al. 2001a). Crystallographic data were processed and scaled with MOSFLM (Leslie 1992; Bolotovsky et al. 1998) and the CCP4 suite (Vagin and Teplyakov 2000) using the ELVES interface (Holton and Alber 2004). Initial phases were obtained by molecular replacement with the CCP4 program MOLREP (Vagin and Teplyakov 2000) and the activated CheY structure (Lee et al. 2001a) as the starting model. A xenon atom in both molecules of the asymmetric unit was clearly visible (signal greater than 10σ) in an F_o–F_c omit electron density map. The structure was refined with CNS (Brunger et al. 1998) and the O software package (Jones et al. 1991). Occupancy of the xenon atoms was refined with five different initial B-factors between 40–60 Ų. The best fit occupancy always fell between 0.22–0.34.

Article published online ahead of print. Article and publication date are at http://www.proteinscience.org/cgi/doi/10.1110/ps.041231005.