Use of europium ions for SAD phasing of lysozyme at the Cu Kα wavelength

Europium(III) ions bound to the surface of hen egg-white lysozyme were found to exhibit good anomalous signal facilitating SAD phasing using laboratory-source data and automated model building. The europium ion-binding sites were observed up to the 15σ level.

Abstract

Europium is shown to be a good anomalous scatterer in SAD phasing for solving the structure of biological macromolecules. The large value of the anomalous contribution of europium, f′′ = 11.17 e⁻, at the Cu Kα wavelength is an advantage in de novo phasing and automated model building. Tetragonal crystals of hen egg-white lysozyme (HEWL) incorporating europium(III) chloride (50 mM) were obtained which diffracted to a resolution of 2.3 Å at a wavelength of 1.54 Å (Cu Kα). The master data set (360° frames) was split and analyzed for anomalous signal-to-noise ratio, multiplicity, completeness, SAD phasing and automated building. The structure solution and model building of the split data sets were carried out using phenix.autosol and phenix.autobuild. The contributions of the Eu ions to SAD phasing using in-house data collection are discussed. This study revealed successful lysozyme phasing by SAD using laboratory-source data involving Eu ions, which are mainly coordinated by the side chains of Asn46, Asp52 and Asp101 together with some water molecules.

1. Introduction  

The single-wavelength anomalous diffraction (SAD) method has been effectively established in structural biology in terms of data-collection techniques, accurate detectors and software and is used for data processing, phasing, density modification and model building for the determination of new macromolecular structures. Recent developments in the preparation of heavy-atom derivatives are more accurate and less disruptive than traditional heavy-atom soaking in high-throughput structure determination using X-ray crystallography (Moiseeva & Allaire, 2007 ▶). The incorporation of heavy atoms into protein crystals for phasing was pioneered by Green et al. (1954 ▶). Successful SAD phasing has been carried out using the anomalous signal from native protein atoms, including sulfur (Hendrickson & Teeter, 1981 ▶; Dauter et al., 1999 ▶), iron (Royer et al., 1989 ▶; Wu et al., 2001 ▶) and copper (Zheng et al., 1996 ▶; Harvey et al., 1998 ▶), or from incorporated heavy atoms such as caesium (Wallace et al., 1990 ▶; Burkhart et al., 1998 ▶), iodine (Chen et al., 1991 ▶), lead (Biou et al., 1995 ▶), mercury (Fan et al., 1990 ▶), selenium (Sha et al., 1995 ▶), platinum (Sabini et al., 2000 ▶), cadmium (Yogavel et al., 2010 ▶), cerium (Vennila & Velmurugan, 2011 ▶), gadolinium (Nagem et al., 2001 ▶), the [Eu(DPA)₃]³⁻ complex (Pompidor et al., 2010 ▶) and EuCl₃ (Ruggiero et al., 2011 ▶). In this study, diffraction data were collected to 2.3 Å resolution at the Cu Kα wavelength; the master data set (360°) was split and the anomalous pairs were separately merged as I ⁺ and I ⁻ in order to analyze the anomalous signal-to-noise ratio, multiplicity and completeness. Europium(III) ions were used to determine the structure by Eu-SAD phasing, and the use of the R _anom/R _p.i.m. ratio for estimating the significance of the anomalous signal is discussed.

Europium, with atomic number 63, is the seventh element of the lanthanide series. It has an L _III absorption edge at 1.77 Å. This wavelength is near to Cu Kα and is an easily accessible wavelength range on synchrotron beamlines. The anomalous scattering coefficients of europium are f′ = −7.838 e⁻ and f′′ = 11.17 e⁻ at the Cu Kα wavelength (http://skuld.bmsc.washington.edu/scatter/AS_periodic.html). The present work, SAD phasing at the Cu Kα wavelength, was successful owing to the high f′ value of the anomalous scatterer.

2. Materials and methods  

2.1. Crystallization and data collection  

Crystallization-grade hen egg-white lysozyme (HEWL) powder from Hampton Research was used for this study. The reservoir solution (1 ml) was made up of 1 M NaCl in 50 mM sodium acetate pH 4.5. The protein was cocrystallized with europium chloride in the protein-solution droplet (5 µl), which consisted of 2.5 µl protein solution (30 mg ml⁻¹ protein in 50 mM sodium acetate pH 4.5), 1.5 µl reservoir solution and 1 µl 50 mM europium chloride, using the hanging-drop vapour-diffusion method at 293 K. Tetragonal crystals were obtained within 2 d. A crystal was fished out using a copper magnetic loop (Hampton Research) and was immersed into 25% ethylene glycol for about 10–15 s. The crystal was quickly cooled in a liquid-nitrogen stream (Oxford Cryosystems) at 100 K. Data were collected at the Cu Kα wavelength (1.54 Å) using the home source at the GNR X-ray facility: a MICROSTAR rotating high-brilliance anode (HBA) generator (Bruker AXS) operating at 45 kV and 60 mA equipped with Helios focal optics. Diffraction data were collected from the crystal using 3 min exposure and 1° oscillation per frame (missetting angles 17.254°, 29.779°, −19.352°) and 360° images were collected to 2.3 Å resolution on a MAR345 dtb image-plate detector (MAR Research Inc.) at a distance of 200 mm from the crystal.

2.2. Software packages used  

The master images (360°) were split into five subsets of data, 45°, 90°, 180°, 270° and 360°, to analyze the effectiveness of the anomalous scattering of the europium in macromolecular phasing and automated model building. These data sets were integrated and scaled by the automar software (MAR Research GmbH, Germany). The anomalous pairs were merged separately as I ⁺ and I ⁻. The relevant data-collection statistics are summarized in Table 1 ▶. The standard SAD protocol was used in the PHENIX software (Adams et al., 2010 ▶) for substructure solution, SAD phasing, automatic model building and refinement. The phenix.xtriage (Zwart et al., 2005 ▶) program was used to analyze the anomalous measurability from the scaled intensity data sets. HySS (Hybrid Substructure Search) was used to find substructure sites, Phaser was used to calculate experimental phases and RESOLVE (Terwilliger, 2000 ▶) was used for density modification and partial model building in phenix.autosol (Terwilliger et al., 2009 ▶). These partial models were fed into phenix.autobuild (Terwilliger et al., 2008 ▶), a combination of density modification and chain tracing was performed using RESOLVE, and phenix.refine (Afonine et al., 2005 ▶) was used to generate a quality model. Coot (Emsley & Cowtan, 2004 ▶) was used to build missing residues and side chains from the autobuilt model. REFMAC5 (Murshudov et al., 2011 ▶) and phenix.refine were used to refine the manually built model. Figures were generated using Chimera (Pettersen et al., 2004 ▶) and PyMOL (DeLano, 2002 ▶). The quality of all atomic models was assessed with PROCHECK (Laskowski et al., 1993 ▶) and MolProbity (Chen et al., 2010 ▶).

Table 1. Summary of data-collection statistics, substructure solution, SAD phasing, model building and refinement.

Data set ()	45	90	180	270	360
Unit-cell parameters (, )	a = b = 77.48, c = 36.99, = = = 90
Space group	P43212
Resolution ()	30.02.3
Mosaicity ()	0.073	0.050	0.050	0.050	0.050
R merge † (%)	2.64 (4.12)	3.60 (6.53)	3.39 (6.73)	4.17 (7.31)	4.46 (7.46)
Completeness (%)	88.0 (91.6)	99.7 (99.8)	99.9 (99.3)	99.9 (100)	99.9 (100)
Multiplicity	3.31 (3.20)	6.31 (6.48)	12.4 (12.6)	18.6 (19.0)	24.7 (24.9)
I/(I)	20.7 (14.0)	15.8 (7.1)	28.5 (13.4)	25.0 (14.3)	25.2 (11.2)
R anom † (%)	5.41 (6.63)	5.15 (6.21)	4.97 (6.23)	4.80 (5.81)	4.77 (5.73)
R r.i.m † (%)	3.44 (5.73)	4.24 (7.72)	4.31 (7.28)	4.38 (7.70)	4.63 (7.75)
R p.i.m † (%)	2.17 (3.76)	2.22 (4.08)	1.62 (2.78)	1.35 (2.39)	1.23 (2.10)
Unique reflections	4838	5351	5354	5353	5353
Bijvoet pairs	3131	4211	4211	4211	4211
Lone Bijvoet pairs	836	0	0	0	0
Anomalous signal	0.079	0.076	0.068	0.066	0.066
phenix.autosol
FOM	0.39	0.46	0.47	0.49	0.49
Refined heavy atoms	3	4	4	6	6
Bayesian score‡	18.7	39.1	44.0	46.0	45.0
Modelmap CC	0.33	0.46	0.48	0.59	0.52
Residues built	47	62	81	97	84
phenix.autobuild
No. of residues/side chains built	81/0	112/112	119/113	113/113	119/119
Modelmap CC	0.46	0.72	0.72	0.73	0.76
Waters	0	56	49	50	51
R work	0.48	0.29	0.29	0.29	0.28
R free	0.53	0.35	0.34	0.33	0.33
Refinement
Resolution ()					26.752.30
R work/R free					0.17/0.23
No. of residues/waters					128/125
Eu ions					2
Overall mean B factor (2)					21.59
Stereochemistry
Bond lengths ()					0.007
Bond angles ()					1.057
Ramachandran plot, residues in (%)
Favoured regions					96.2
Allowed regions					3.8
Disallowed regions					0

^†Weiss (2001 ▶).

^‡Terwilliger et al. (2009 ▶).

3. Results and discussion  

3.1. Analysis of the data quality and anomalous signal  

Mean multiplicities of 3.31, 6.31, 12.4, 18.6 and 24.7 were observed in the 45°, 90°, 180°, 270° and 360° data sets, respectively. The 〈I/σ(I)〉 values for the 27.4–2.3 Å resolution range were 20.7, 15.8, 28.5, 25.0 and 25.2 for the respective data sets. The overall completeness was 88% for the 45° data set and 99% for remaining data sets. The R _merge and R _anom values were below 4.46% and 5.51%, respectively, in all rotation ranges (Table 1 ▶). The expected anomalous signal (〈ΔF〉/〈F〉 = 0.10) was estimated using the modified Hendrickson formula (Hendrickson & Teeter, 1981 ▶; Dauter et al., 2002 ▶), 〈ΔF〉/〈F〉_exp = 2^1/2[(f _A′′N _A ^1/2)]/[f _eff(θ)N _P ^1/2], where f _A′′ are the imaginary scattering contributions of europium ions and N _A and N _P are the number of Eu atoms and the total number of non-H atoms in the molecule, respectively. Here, N _A is 2 and N _P is 1001. f _eff(θ), which is the average number of electrons for protein atoms, is 6.7. The ratio of R _anom to R _p.i.m. has been used as a practical indicator of the level of anomalous signal (Mueller-Dieckmann et al., 2004 ▶, 2005 ▶; Weiss, 2001 ▶; Weiss et al., 2004 ▶, 2005 ▶; Weiss, Sicker, Djinovic-Carugo et al., 2001 ▶; Weiss, Sicker & Hilgenfeld, 2001 ▶), which can provide information on whether or not the data are sufficient for structure determination. As a rule of thumb, an R _anom/R _p.i.m. ratio of ≥1.5 is sufficient to determine the structure (Mueller-Dieckmann et al., 2005 ▶). Analysis of the R _anom/R _p.i.m. ratio for the europium data revealed that sufficient anomalous signal was exhibited in each data subset (Table 2 ▶).

Table 2. Statistics of the R _anom/R _p.i.m. ratio for different rotation ranges.

	Resolution ()
Total rotation ()	30.04.96	4.963.93	3.933.44	3.443.12	3.122.90	2.902.73	2.732.59	2.592.48	2.482.38	2.382.30
45	3.65	2.33	2.45	2.09	2.67	2.61	2.34	2.43	2.11	1.76
90	4.16	2.32	2.08	1.95	2.47	2.17	2.08	2.02	1.81	1.52
180	5.51	2.94	2.71	2.57	2.92	3.08	2.74	2.88	2.48	2.24
270	6.71	3.51	3.08	2.94	3.55	3.36	3.11	3.18	2.67	2.43
360	6.78	3.67	3.33	3.31	3.92	3.62	3.51	3.61	2.98	2.73

3.2. Substructure determination, phasing and model building  

The anomalous substructures were located in phenix.autosol using a 3.0 Å resolution cutoff. Between three and six heavy-atom sites (including S, Na and Cl atoms) were refined for phasing for the various ϕ rotation ranges and the highest two peaks corresponded to europium ions. The initial phases obtained from europium were sufficient for automatic tracing of the protein structure in preliminary model building. Figures of merit (FOMs) of 0.39–0.49 and correlation coefficients (CCs) of 0.33–0.52 were observed for the 45–360° data sets. In an asymmetric unit, preliminary models consisting of 47, 62, 81, 97 and 84 residues were built in the data sets. These partial models were fed into phenix.autobuild for iterative model building with refinement. The overall CC was 0.76 and 92% of the residues were built for the 360° data. The other data sets also showed good CCs with good automated model building (Figs. 1 ▶ a–1e). The 360° data set alone was used for manual building and refinement in Coot and phenix.refine, respectively. This model was finally refined to R _work and R _free values of 17.3% and 22.9%, respectively. The occupancy values of the bound Eu ions were refined using phenix.refine.

Figure 1.

Cartoon representations of the automatically built lysozyme-molecule model in the asymmetric unit: (a) 45°, (b) 90°, (c) 180°, (d) 270° and (e) 360° data sets. (f) Europium ions and sodium ion in the final refined model (360°) are shown in cyan and magenta, respectively.

3.3. Europium-binding sites  

Two europium ions are bound to the surface of the crystal structure without affecting the crystal lattice. The molecular structure remains the same after binding and is shown in Fig. 1 ▶(f). The ion-binding sites were validated by the anomalous difference Fourier map up to the 15σ level (Fig. 2 ▶ a). The europium ions are mainly coordinated by water molecules and by aspartic acid and asparagine side chains. Chloride and sodium ions were assigned based on the difference peak and binding environment of the protein. A distance of 16.7 Å is observed between the two europium ions. Eu-1 is coordinated by five water molecules and the Asp101 OD1 and Asp101 OD2 atoms at the solvent boundary. The occupancy of the ion is 0.57. Eu-2 is bound by six water molecules and two residues, Asn46 OD1 and Asp52 OD2, in the catalytic cleft; its occupancy is 0.69 (Figs. 2 ▶ b and 2 ▶ c). Both europium-binding sites are identical to those found for Ho³⁺ (Jakoncic et al., 2006 ▶) and Ce³⁺ (Vennila & Velmurugan, 2011 ▶) ions in HEWL.

Figure 2.

(a) Anomalous difference Fourier map of binding sites of europium ions (red) at the 15σ level. (b) The Eu-1 ion and (c) the Eu-2 ion (cyan), showing interactions with amino-acid residues and water molecules.

4. Conclusion  

In this study, the results using the five subsets of data show that europium ions can effectively be used as an anomalous scatterer for SAD phasing and model building at the Cu Kα wavelength. The 45° data set with a multiplicity of 3.31 was sufficient for SAD phasing and structure determination of the 14.4 kDa protein. The protein cocrystallized with a lower concentration of europium (50 mM) produced sufficient phasing power to allow automatic structure determination. The absorption edge of europium (1.77 Å) is near the Cu Kα wavelength. As a consequence, the europium ion has a large f′′ value which is sufficient for phasing and model building. Without altering the structure, the europium ions are bound to the surface of the protein and interact with the side chains of Asn and Asp residues and with water molecules. Eu³⁺, Ho³⁺ and Ce³⁺ ions interact with the same binding-site residues in HEWL. The results show that Eu can provide sufficient anomalous signal for phasing and model building to determine new macromolecular structures using the SAD technique.

Supplementary Material

PDB reference: lysozyme, europium derivative, 4h1p