Microcrystal preparation for serial femtosecond X-ray crystallography of bacterial copper amine oxidase

Microcrystals of bacterial copper amine oxidase were prepared by combining micro-seeding and batch crystallization, and were used to determine a radiation-damage-free, nonfrozen structure by serial femtosecond X-ray crystallography.

Abstract

Recent advances in serial femtosecond X-ray crystallography (SFX) using X-ray free-electron lasers have paved the way for determining radiation-damage-free protein structures under nonfreezing conditions. However, the large-scale preparation of high-quality microcrystals of uniform size is a prerequisite for SFX, and this has been a barrier to its widespread application. Here, a convenient method for preparing high-quality microcrystals of a bacterial quinoprotein enzyme, copper amine oxidase from Arthrobacter globiformis, is reported. The method consists of the mechanical crushing of large crystals (5–15 mm³), seeding the crushed crystals into the enzyme solution and standing for 1 h at an ambient temperature of ∼26°C, leading to the rapid formation of microcrystals with a uniform size of 3–5 µm. The microcrystals diffracted X-rays to a resolution beyond 2.0 Å in SFX measurements at the SPring-8 Angstrom Compact Free Electron Laser facility. The damage-free structure determined at 2.2 Å resolution was essentially identical to that determined previously by cryogenic crystallography using synchrotron X-ray radiation.

1. Introduction  

X-ray free-electron lasers (XFELs) provide ultrahigh-intensity X-ray pulses with femtosecond duration and are available at five facilities currently in operation, the first two of which were at the Linac Coherent Light Source (LCLS) in Menlo Park, California, USA and the SPring-8 Angstrom Compact Free Electron Laser (SACLA) in Sayo-gun, Hyogo, Japan. Advances in XFELs have opened up new prospects for structural biology, allowing the determination of X-ray-damage-free structures at room temperature, as diffraction data can be collected before the crystals suffer from radiation damage (Neutze et al., 2000 ▸; Chapman et al., 2011 ▸). They can also visualize rapid structural changes associated with protein action as molecular movies by taking advantage of the femtosecond duration (Tenboer et al., 2014 ▸; Barends et al., 2015 ▸; Pande et al., 2016 ▸; Nango et al., 2016 ▸; Suga et al., 2017 ▸). Serial femtosecond X-ray crystallography (SFX) is a key method for collecting diffraction data from protein crystals with an XFEL, in which microcrystals suspended in grease or crystal mother liquor are serially delivered and flowed to the irradiation spot of the XFEL through an injector. Diffraction images are obtained from randomly oriented microcrystals when the pulse beam hits the crystal. Typically, the collection of over 10 000–30 000 images is required to obtain a data set. Therefore, the large-scale preparation of microcrystals is essential for SFX experiments, especially when ‘mix-and-inject’ time-resolved serial crystallography (Stagno et al., 2017 ▸; Olmos et al., 2018 ▸), which involves a high consumption rate of microcrystals, is performed. Serial femtosecond rotational crystallography (SF-ROX) using single crystals of submillimetre size also provides a damage-free structure both at cryogenic and room temperatures (Hirata et al., 2014 ▸; Halsted et al., 2018 ▸, 2019 ▸), although it is unsuitable for time-resolved SFX.

Copper amine oxidases catalyse the oxidative deamination of various primary amines to produce the corresponding aldehydes, hydrogen peroxide and ammonia, and occur widely from bacteria to plants and animals (McIntire & Hartmann, 1993 ▸). The enzymes commonly have a homodimeric subunit structure with a prosthetic Cu²⁺ ion and a protein-derived quinone cofactor, topaquinone (2,4,5-trihydroxyphenyl­alanine quinone; TPQ), in the active site of each subunit. Since the first demonstration of TPQ generation by Cu²⁺-dependent autooxidation of the precursor protein (Matsuzaki et al., 1994 ▸), we have been using the recombinant enzyme from the Gram-positive bacterium Arthrobacter globiformis (AGAO) for structural and mechanistic studies. The enzyme crystallized in the precursor form was successfully used to monitor the structural changes that occur during TPQ generation upon the addition of Cu²⁺ ions (Kim et al., 2002 ▸). High-quality crystals of the Cu²⁺/TPQ-containing active form of AGAO diffracted to 1.08 Å resolution in cryogenic X-ray crystallography (Murakawa et al., 2013 ▸). Moreover, we obtained extra-large crystals that were suitable for neutron crystallography (Murakawa et al., 2020 ▸). Thus, AGAO has proven to be useful in various crystallographic studies, despite its relatively large monomer size of ∼70 000 Da. Here, we report a convenient method for the large-scale preparation of high-quality microcrystals of AGAO that are applicable to SFX studies. The damage-free, nonfrozen AGAO structure was solved for the first time using SFX diffraction data.

2. Materials and methods  

2.1. Macromolecule production  

Recombinant AGAO was expressed and purified as described previously (Matsuzaki et al., 1994 ▸; Chiu et al., 2006 ▸). In brief, AGAO was expressed with a katGE-disrupted Escherichia coli BL21 (DE3) strain (designated CD03; Kishishita et al., 2003 ▸) that harbours an AGAO expression plasmid, pEPO-02 (Table 1 ▸). Expression was induced at 26°C by adding 0.1 mM isopropyl β-d-1-thiogalactopyranoside at the mid-log phase of the culture. AGAO was purified from the cells of the overnight culture as the Cu²⁺-free precursor form and was converted to the Cu²⁺/TPQ-containing active form as reported previously (Matsuzaki et al., 1994 ▸; Chiu et al., 2006 ▸). The purity was verified by sodium dodecyl sulfate polyacrylamide gel electrophoresis and protein concentrations were determined spectrophotometrically using the extinction coefficient at 280 nm (13.2) for a 1% solution of the active form (Matsuzaki et al., 1994 ▸).

Table 1. Macromolecule-production information.

Source organism	Arthrobacter globiformis
Expression vector	pEPO-02 (Kishishita et al., 2003 ▸)
Expression host	Escherichia coli CD03 (Kishishita et al., 2003 ▸)
Complete amino-acid sequence of the construct produced	MTPSTIQTASPFRLASAGEISEVQGILRTAGLLGPEKRIAYLGVLDPARGAGSEAEDRRFRVFIHDVSGARPQEVTVSVTNGTVISAVELDTAATGELPVLEEEFEVVEQLLATDERWLKALAARNLDVSKVRVAPLSAGVFEYAEERGRRILRGLAFVQDFPEDSAWAHPVDGLVAYVDVVSKEVTRVIDTGVFPVPAEHGNYTDPELTGPLRTTQKPISITQPEGPSFTVTGGNHIEWEKWSLDVGFDVREGVVLHNIAFRDGDRLRPIINRASIAEMVVPYGDPSPIRSWQNYFDTGEYLVGQYANSLELGCDCLGDITYLSPVISDAFGNPREIRNGICMHEEDWGILAKHSDLWSGINYTRRNRRMVISFFTTIGN(TPQ)DYGFYWYLYLDGTIEFEAKATGVVFTSAFPEGGSDNISQLAPGLGAPFHQHIFSARLDMAIDGFTNRVEEEDVVRQTMGPGNERGNAFSRKRTVLTRESEAVREADARTGRTWIISNPESKNRLNEPVGYKLHAHNQPTLLADPGSSIARRAAFATKDLWVTRYADDERYPTGDFVNQHSGGAGLPSYIAQDRDIDGQDIVVWHTFGLTHFPRVEDWPIMPVDTVGFKLRPEGFFDRSPVLDVPANPSQSGSHCHG

2.2. Crystallization  

AGAO microcrystals were prepared using a combination of micro-seeding and batch crystallization methods (Table 2 ▸). The seed stock was prepared as follows. Extra-large AGAO crystals, as used for neutron crystallography, were prepared as described previously (Murakawa et al., 2020 ▸). Several AGAO crystals with a total volume of 5–15 mm³ were washed with precipitant solution (1.05 M potassium/sodium tartrate in 25 mM HEPES buffer pH 7.4) and transferred to the new precipitant solution, in which the crystal:buffer volume ratio was around 1:3. The crystals were then ground for 1 min using a cell homogenizer (electric pestle; Fisher Scientific). The obtained slurry of crushed crystals (size ≫ 1 µm) was preserved as a seed stock at 4°C until use. To prepare a 0.9 ml suspension of the AGAO microcrystals, 10 µl seed stock and 100 µl AGAO solution (200 mg ml⁻¹ protein solution in 25 mM HEPES buffer pH 7.4) were carefully mixed with 800 µl precipitant solution in a 1.5 ml microtube and kept at ∼26°C for 1 h to allow the growth of microcrystals (size 3–5 µm). Approximately half of the microcrystal suspension was filtered through a nylon mesh (pore size 30 µm; Sysmex) by unforced sedimentation under atmospheric pressure for ∼10 min. Both filtered and unfiltered microcrystals were used in diffraction experiments in this study to evaluate the effects of the filtration treatment. For SFX data collection from both crystals, 50 µl of grease suspension containing approximately 5 × 10⁵ crystals was used. The size and concentration of the microcrystals were estimated by microscopic observation with a BZ-X710 all-in-one fluorescence microscope (Keyence).

Table 2. Crystallization condition.

Method	Micro-seeding and batch crystallization
Container type	Microtube (1.5 ml)
Temperature (°C)	∼26
Protein concentration (mg ml–1)	200
Buffer composition of protein solution	25 mM HEPES buffer pH 7.4
Composition of precipitant solution	1.05 M potassium/sodium tartrate in 25 mM HEPES buffer pH 7.4
Volume of protein solution (µl)	100
Volume of seed stock (µl)	10
Volume of precipitant solution (µl)	800

2.3. Data collection and processing  

SFX diffraction data were collected on BL2 at SACLA (Ishikawa et al., 2012 ▸) using the grease-matrix method (Sugahara et al., 2015 ▸, 2017 ▸) as follows: 20 µl filtered or unfiltered microcrystal suspension was mixed with 180 µl Super Lube nuclear-grade grease (No. 42150, Synco Chemical Co.) on a glass slide using a spatula (Sugahara et al., 2017 ▸). The grease/crystal mixture was loaded into a sample reservoir and run out from a high-viscosity sample injector (Shimazu et al., 2019 ▸) with a 75 µm internal diameter nozzle at a flow rate of ∼0.48 µl min^–1. Diffraction patterns were collected at ambient temperature (approximately 26°C) from randomly oriented crystals extruded from the injector in a DAPHNIS chamber filled with humid helium gas (Tono et al., 2015 ▸) using an MPCCD detector with a short-working-distance octal sensor arrangement (Kameshima et al., 2014 ▸) at a sample-to-detector distance of 50 mm. The XFEL beam was focused to a beam diameter of 1.5 µm full-width at half-maximum (FWHM) using a couple of elliptical mirrors in the Kirk­patrick–Baez geometry (Yumoto et al., 2013 ▸) and had a wavelength of 1.771 Å (7000 eV) with a repetition rate of 30 Hz, a temporal width within 10 fs duration (FWHM) and a pulse energy of approximately 640 µJ at the light source. Data collection was guided by real-time analysis using the SACLA data-processing pipeline (Nakane et al., 2016 ▸).

The data obtained were retrieved through the SACLA API (Joti et al., 2015 ▸) and filtered by Cheetah (Barty et al., 2014 ▸), under the criterion that each image with more than 20 spots was accepted as a hit. The hit images were indexed and integrated in a Monte Carlo-like fashion using CrystFEL (White et al., 2016 ▸). The detector background, estimated by averaging the dark images, was subtracted from the diffraction patterns. The detector metrology was refined with geoptimiser (Yefanov et al., 2015 ▸). Indexing was performed using XGANDALF (Gevorkov et al., 2019 ▸). The data-collection and processing statistics are summarized in Table 3 ▸. The high-resolution limit cutoff was determined by consulting the R _split (100%) and CC_1/2 (0.5) values.

Table 3. Data-collection and processing statistics.

Values in parentheses are for the outer shell.

	Filtered crystals	Unfiltered crystals	Merged
Diffraction source	SACLA BL2	SACLA BL2
Wavelength (Å)	1.771	1.771
Temperature (°C)	∼26	∼26
Detector	MPCCD	MPCCD
Crystal-to-detector distance (mm)	50.0	50.0
Space group	C2	C2	C2
a, b, c (Å)	158.75, 64.74, 93.35	158.79, 64.68, 93.40	158.75, 64.74, 93.55
α, β, γ (°)	90, 112.33, 90	90, 112.36, 90	90, 112.33, 90
Resolution range (Å)	20.0–2.30 (2.38–2.30)	20.0–2.30 (2.38–2.30)	20.0–2.20 (2.28–2.20)
No. of collected images	211233	192337	403570
No. of hits	44749	42282	87031
No. of indexed images	32128	30075	62203
Total No. of reflections	6212244 (432747)	5692661 (395203)	12835253 (700670)
No. of unique reflections	76311 (7691)	76254 (7664)	87284 (8717)
Completeness (%)	100 (100)	100 (100)	100 (100)
Multiplicity	81.4 (56.3)	74.7 (51.6)	147.3 (80.4)
〈I/σ(I)〉	6.32 (1.58)	6.29 (1.47)	7.76 (0.93)
Rsplit† (%)	15.11 (55.31)	14.61 (61.80)	11.11 (97.14)
CC1/2	0.9706 (0.7422)	0.9736 (0.7157)	0.9858 (0.6180)
Overall B factor from Wilson plot (Å2)	49.64	51.78	49.62

^†R_split = 2^{-1/2}(\textstyle\sum I_{\rm even}-I_{\rm odd}|/[1/2 \textstyle \sum (I_{\rm even}-I_{\rm odd})] (White et al., 2016 ▸).

2.4. Structure solution and refinement  

The initial phase was determined by molecular replacement using Phaser (McCoy et al., 2007 ▸). The search model was based on the coordinates of the AGAO monomer (PDB entry 3wa2; Murakawa et al., 2013 ▸) after removing all ligands, including water molecules and metal ions. Refinements, electron-density map calculations and solvent-molecule assignment were performed with Phenix version 1.82 (Liebschner et al., 2019 ▸). Manual rebuilding was performed with Coot (Emsley et al., 2010 ▸), and water molecules and other ligands, such as Cu²⁺ and TPQ, were added step by step to the model during the refinement process. PyMOL version 1.8 (Schrödinger) was used to generate the figures. The details and statistics pertaining to refinement are summarized in Table 4 ▸. The atomic coordinates and structure factors obtained from the merged data of the filtered and unfiltered crystal suspensions have been deposited in the Protein Data Bank with accession code 7f8k.

Table 4. Refinement statistics.

Values in parentheses are for the outer shell.

	Filtered crystals	Unfiltered crystals	Merged
Resolution range (Å)	19.96–2.30 (2.33–2.30)	19.96–2.30 (2.33–2.30)	19.96–2.20 (2.22–2.20)
Completeness (%)	100.0	100.0	99.9
No. of reflections, working set	72477 (2710)	72444 (2676)	82910 (2772)
No. of reflections, test set	3822 (143)	3790 (140)	4374 (147)
Final R cryst	0.160 (0.2909)	0.156 (0.2934)	0.154 (0.3276)
Final R free	0.193 (0.3346)	0.191 (0.3020)	0.179 (0.3355)
No. of non-H atoms
Protein	4868	4868	4868
Ligands	14	14	14
Waters	137	156	175
R.m.s. deviations
Bond lengths (Å)	0.007	0.007	0.007
Angles (°)	0.891	0.897	0.889
Average B factors (Å2)
Protein	53.49	51.83	51.84
Ligand (TPQ/Cu2+)	53.54	54.76	52.25
Water	49.59	50.13	50.32
Ramachandran plot
Most favoured (%)	95.2	95.5	95.2
Allowed (%)	4.6	4.5	4.8

3. Results and discussion  

To prepare the AGAO microcrystals necessary for SFX experiments, we first applied a simple batch method, in which the purified AGAO solution was directly mixed with the precipitant solution. However, this method resulted in the formation of only a few microcrystals. Therefore, we attempted to combine the micro-seeding method with batch crystallization to promote crystal nucleation as much as possible. The combined method yielded many rhombic plate microcrystals (1 × 10⁸ crystals per millilitre) with a uniform size (length 3–5 µm) that appeared after standing for 1 h (Fig. 1 ▸). It is likely that seed preparation from the extra-large crystals, with a smaller surface area per weight, resulted in the production of uniform seeds that provided high-quality microcrystals. To evaluate the quality of the microcrystals that were obtained, we conducted SFX diffraction measurements using the well established grease-matrix method (Sugahara et al., 2015 ▸). Before the SFX diffraction measurements, half of the microcrystal suspension (approximately 0.5 ml) was slowly filtered through a mesh (pore size 30 µm) to remove oversized crystals and debris, such as protein precipitates, that might clog the injector. Filtered and unfiltered crystal suspensions were independently subjected to SFX diffraction measurements to examine whether the microcrystals were damaged by filtration.

Figure 1.

Images of AGAO microcrystals grown by combining micro-seeding and batch crystallization. The right panel shows an enlarged view of the area indicated by a red rectangle in the left panel. The scale bars in the left and right panels represent 20 and 3 µm, respectively.

During the beam time of approximately 6 h, 211 233 and 192 337 diffraction images were collected for the filtered and unfiltered crystal suspensions, respectively (Table 3 ▸). Measurements with the unfiltered crystal suspension took a longer time than those with the filtered crystal suspension because of the necessity of removing large crystals and/or debris that was frequently stuck in the injector during the measurements (almost once every 30 min). Meanwhile, we were able to detect a diffraction spot with a maximum resolution of 1.98 Å (corresponding to the edge of the detector) from the filtered AGAO microcrystals (Fig. 2 ▸). During the measurements, the diffraction images were sorted by auto-excluding those containing less than 20 spots using Cheetah, in which the parameter set was optimized to SACLA. The hit rates obtained were 21.18% (44 749/211 233) and 21.98% (42 282/192 337) for the measurements with filtered and unfiltered crystals, respectively. The hit images were processed for structural determination. The space groups of the filtered and unfiltered crystals indexed using 32 128 and 30 075 images, respectively, were both assigned as C2, with nearly identical unit-cell parameters (Table 3 ▸). The asymmetric unit contained a monomer subunit of dimeric AGAO. The overall Wilson B factor (49.62 Ų) obtained from data processing (Table 3 ▸) is significantly higher than those obtained in previous studies using synchrotron X-ray crystallography (9.92 Ų; PDB entry 3wa2; Murakawa et al., 2013 ▸) and neutron crystallo­graphy (16.56 Ų; PDB entry 6l9c; Murakawa et al., 2020 ▸), although B factors are dependent on crystal resolution. This high B factor has also been reported in other SFX studies using microcrystals (Nakane et al., 2015 ▸; Lomelino et al., 2018 ▸) and might be attributable to the Monte Carlo integration of SFX diffraction data derived from a number of randomly oriented crystals, some of which include noise or very weak reflections. In addition, data collection for SFX is performed at room temperature, in which thermal fluctuation also makes the Wilson B factors higher than those in single-crystal crystallo­graphy performed at cryogenic temperature.

Figure 2.

(a) XFEL single diffraction pattern from an AGAO microcrystal. The identified reflection spots are indicated by small circles. (b) Enlarged view of the boxed region; this reflection image was estimated to have a spot at a resolution of 1.98 Å (red arrow) by CrystFEL.

Although the Wilson B factors estimated in this study were considerably high, the electron density obtained after refinement was very clear and allowed the models to be placed without difficulty (see Fig. 3 ▸ b). The crystallographic parameters, such as the R _cryst and R _free factors (Tables 3 ▸ and 4 ▸), and the overall structures obtained from the filtered crystals were essentially identical to those from the unfiltered crystals, with root-mean-square (r.m.s.) deviations for main-chain atoms of 0.078 Å. The filtration caused no damage to the microcrystals and did not affect the structural determination. Therefore, the diffraction data sets were merged. The resultant data set presented improved crystallographic statistics, including the high-resolution limit (from 2.30 to 2.20 Å), multiplicity (from 74.7 to 147.3) and final R _free (from 0.191 to 0.179), compared with those from the unfiltered crystals only (Tables 3 ▸ and 4 ▸). The AGAO structure determined with the merged data set provided the first SFX structure of the copper amine oxidase family, which essentially contained no X-ray radiation damage. The average B factor is still high (51.83 Ų for protein atoms; Table 4 ▸) but is comparable to the Wilson B factor (49.62 Ų for the merged data; Table 3 ▸), suggesting no overfitting. The overall structure of the monomer (Fig. 3 ▸ a) has almost the same conformation as those determined by synchrotron X-ray crystallography (PDB entry 3wa2; 1.08 Å resolution) and neutron crystallography (PDB entry 6l9c; 1.72 Å resolution), both of which were performed at cryogenic temperature, with r.m.s. deviations for main-chain atoms of 0.502 and 0.494 Å, respectively. The electron-density map obtained from the merged data set clearly showed the oxidized form of TPQ (Fig. 3 ▸ b), although its conformation was slightly different from those of the high-resolution X-ray and neutron structures in that it underwent a wedge-shaped movement (Figs. 3 ▸ c and 3 ▸ d) because of the high flexibility of the quinone ring (Murakawa et al., 2013 ▸). In addition, the active-site Cu²⁺ was ligated in the distorted pyramidal coordination, with three histidine residues (His431, His433 and His592) and two water molecules located at positions identical to those in the oxidized form of AGAO previously determined by X-ray and neutron crystallography (Figs. 3 ▸ c and 3 ▸ d). These results show that the bound Cu²⁺ in AGAO is resistant to X-ray photoreduction, which is accompanied by conformational changes of the metal coordination structure and has been observed in other copper proteins, including fungal laccase (De la Mora et al., 2012 ▸).

Figure 3.

Structure of AGAO determined from the merged data set from SFX. (a) The overall structure is shown as a green ribbon model. The counter subunit forming the dimer is also shown as a blue wire model. (b) Active-site structure; the assigned stick models are superimposed on F _o −F _c omit maps (grey mesh) for residue 382 (TPQ) and the Cu²⁺ and Cu²⁺ ligands (His431, His433, His592, W_ax and W_eq) contoured at 8σ, in which W_ax and W_eq are the axially and equatorially Cu²⁺-coordinating water molecules, respectively. (c, d) A stick model of the active-site structure determined by SFX, coloured green, is compared with those determined by high-resolution X-ray crystallography (PDB entry 3wa2; Murakawa et al., 2013 ▸), coloured magenta (c), and neutron crystallography (PDB entry 6l9c; Murakawa et al., 2020 ▸), coloured cyan (d).

In conclusion, our procedure for the large-scale preparation of high-quality AGAO microcrystals is useful for SFX studies. The present method of microcrystal preparation may also be applicable to other proteins. The availability of high-quality microcrystals in large quantities is also promising for studying the structural changes in nonfrozen crystals by ‘mix-and-inject’ time-resolved serial crystallography.

Supplementary Material

PDB reference: room-temperature structure of bacterial copper amine oxidase determined by serial femtosecond crystallography, 7f8k

Funding Statement

This work was funded by Japan Society for the Promotion of Science grants 20H05448, 19H05781, and 19K05694 to Takeshi Murakawa, Eriko Nango, and Toshihide Okajima; Japan Agency for Medical Research and Development grants JP20am0101070 and JP21am0101070 (support number 2302) to Takeshi Murakawa and So Iwata; Network Joint Research Center for Materials and Devices.