Past decades have shown the impact of structural information derived from complexes of drug candidates with their protein targets to facilitate the discovery of safe and effective medicines. Despite recent developments in single particle cryo-electron microscopy, X-ray crystallography has been the main method to derive structural information. The unique properties of X-ray free electron laser (XFEL) with unmet peak brilliance and beam focus allow X-ray diffraction data recording and successful structure determination from smaller and weaker diffracting crystals shortening timelines in crystal optimization. To further capitalize on the XFEL advantage, innovations in crystal sample delivery for the X-ray experiment, data collection and processing methods are required. This development was a key contributor to serial crystallography allowing structure determination at room temperature yielding physiologically more relevant structures. Adding the time resolution provided by the femtosecond X-ray pulse will enable monitoring and capturing of dynamic processes of ligand binding and associated conformational changes with great impact to the design of candidate drug compounds.

Introduction

X-ray free electron lasers (XFELs) have transformed from a theoretical study into real research facilities, a dream that becomes a reality 50 years after the invention of the optical laser and nearly 40 years after the first publication [1]. Compared with optical lasers, FELs employ relativistic electron beams to amplify light in the entire spectral range from the far infrared up to hard X-rays (wavelength range from 1 mm to 0.05 nm). The durations of generated light pulses are ultra-short, highly coherent and of extremely high peak brilliance, several orders of magnitude higher than light generated by third generation synchrotron radiation facilities.

Theoretical considerations about ‘diffraction before destruction’ – collection of an X-ray diffraction pattern from a crystal before its destruction by radiation damage – and therefore the potential impact to improve protein X-ray experiments encouraged the construction of X-ray free-electron lasers for structural biology [2,3,4]. In 2009, the first hard X-ray free electron laser facility was opened in Stanford (U.S.A.) with the Linac Coherent Light Source (LCLS). This milestone was then followed by SPring-8 Angstrom Compact Free Electron Laser (SACLA) at Riken in Japan in 2011. Within the same year the structure of photosystem I, one of the largest membrane proteins, was reported after 3 million diffraction patterns were collected using a liquid jet stream of crystals with sizes of 200 nm to 2 μm3. Since then, methods for XFEL sample delivery, data collection and processing have developed significantly and more challenging systems have been investigated. This review gives an overview of the history and provides an outlook on developments in the field with an emphasis on the application to structure-based drug discovery (SBDD).

X-ray free electron laser

In a free-electron laser, the active medium is a beam of relativistic electrons. This beam moves in vacuum through a periodic magnet array forcing the electrons to follow a wiggling orbit centred on a straight line. The wiggling orbit introduces a transverse velocity component, which allows the electrons to exchange energy with a light wave co-linear with the electron beam. The electrons become accelerated or decelerated, depending on the phase of the transverse electric field of the light wave. At a particular wavelength of the light beam, this exchange becomes resonant for a single electron, leading to a continuous transfer of energy from the electron to the light wave and ultimately to so-called ‘bunching’ of the electron beam. As a result, the entire system acts analogously to a light amplifier in the optical laser with the fundamental advantage of the operation being independent of the transitions in a specific laser medium. Therefore, FEL amplification can be applied over a much wider photon wavelength range in comparison with conventional optical lasers (Figure 1). The FEL principle is the only proven method for providing coherent light pulses at hard X-ray wavelengths. An introduction to the theory of FELs can be found in Huang et al. [5].

Wavelength range of FEL facilities in comparison with optical lasers.

A list of XFEL facilities worldwide that provide hard X-ray radiation in the range of 10−10 m (1 Å) suitable for protein crystallography together with the most relevant parameters are presented in Table 1 (data compiled from Yabashi et al. [6], Ko et al. [7], Tschentscher et al. [8], Milne et al. [9] and White et al. [10]). An important difference of XFEL compared with synchrotron sources is the ‘single-user’ mode of measurement with corresponding higher beam time costs and limited access. One way to address the increasing demand for XFEL measurement time are developments of accelerator-based, so-called ‘table top’ facilities dedicated to structural biology with the emphasis to combine X-ray diffraction and spectroscopic techniques with extremely short pulses reaching the attosecond regime [11].

Table 1
Worldwide hard XFEL facilities and their key properties
FEL, Institution Location Start of operation Repitition rate (Hz) Photon wavelength (Å) Pulse length (fs [FWHM]) Photons/Pulse (×1012 at 1 Å) 
LCLS, SLAC Palo Alto, USA 2009 120 3.0–1.2 10–300 1.8 
SACLA, RIKEN Harima, Japan 2012 30/60 3.1–0.62 <10 0.25 
PAL_FEL Pohang, Korea 2016 60 2.5–1.0 <100 <1 
SwissFEL, PSI Würenlingen, Switzerland 2017 100 7.0–0.83 5–50 0.7 
EuXFEL Hamburg, Germany 2017 10 × 2700 4.0–0.5 2–100 0.9 
FEL, Institution Location Start of operation Repitition rate (Hz) Photon wavelength (Å) Pulse length (fs [FWHM]) Photons/Pulse (×1012 at 1 Å) 
LCLS, SLAC Palo Alto, USA 2009 120 3.0–1.2 10–300 1.8 
SACLA, RIKEN Harima, Japan 2012 30/60 3.1–0.62 <10 0.25 
PAL_FEL Pohang, Korea 2016 60 2.5–1.0 <100 <1 
SwissFEL, PSI Würenlingen, Switzerland 2017 100 7.0–0.83 5–50 0.7 
EuXFEL Hamburg, Germany 2017 10 × 2700 4.0–0.5 2–100 0.9 

FWHM, full width at half maximum.

FELs deliver photon pulses at energies that are ten orders of magnitude beyond state-of-the-art third generation synchrotrons such as APS, ESRF, Spring8 (electron energy of 6–8 GeV), SLS, Diamond and MAXIV (electron energy 2.4–3.0 GeV). Two technologies are used to construct electron accelerators for FEL applications: ambient temperature, normal conducting machines at low repetition rates of ~100–120 Hz (SACLA, LCLS, SwissFEL) and cold, super-conducting accelerators (EuXFEL). The latter allows acceleration of electron bunches at a much higher repetition rate of ~1–4.5 MHz, thereby boosting the average brilliance (Figure 2). Despite the different repetition rates, the pulse length and hence the ultimate time resolution for time-resolved experiments cover the range from a few femtoseconds to 100s of femtoseconds depending on accelerator parameters and is similar for all facilities (see Table 1). Due to the stochastic nature of light production, every single pulse has slightly different properties in terms of wavelength, flux density and arrival time at the sample. Therefore, sophisticated diagnostic tools have been developed and implemented to enable the characterization of the emitted photons on a pulse-to-pulse manner [8]. Full utilization of these parameters by software like Cheetah and CrystFEL [12] is an on-going effort to further improve the accuracy of XFEL data analysis.

Comparison of time pattern characteristics of FEL facilities

Figure 2
Comparison of time pattern characteristics of FEL facilities

The Burst mode is only used at the EuXFEL and all other facilities use the continuous wave (CW) mode. X-ray data collection in structural biology has to be adapted to the different time-pattern and repetition rate characteristics.

Figure 2
Comparison of time pattern characteristics of FEL facilities

The Burst mode is only used at the EuXFEL and all other facilities use the continuous wave (CW) mode. X-ray data collection in structural biology has to be adapted to the different time-pattern and repetition rate characteristics.

SwissFEL

The Paul Scherrer Institute has built the SwissFEL [9] and started commissioning of the instrument at low energy on 5 December, 2016. First user operation is anticipated at the end of 2017. The facility (Figure 3) consists of a high-brightness electron gun, a 5.8 GeV linear accelerator, permanent magnetic undulators and two photon beamlines Aramis and Athos covering the wavelength range 0.08–7 nm. X-ray pulses will simultaneously be available to both Aramis and Athos at 100 Hz repetition rate. Athos is designed for spectroscopic applications in the wavelength range 0.7–7 nm, and will offer the option of circularly polarized radiation. Aramis is designed for X-ray diffraction and spectroscopy applications in the wavelength range 0.08–0.7 nm. A system of offset mirrors in Aramis directs the X-ray beam to one of the three experimental hutches (ESA, ESB and ESC). At ESA the photon energy range, beam divergence and experimental chamber dimensions are optimized for jet-based serial crystallography experiments. The chamber can accommodate either liquid or viscous medium injectors for sample delivery and can be used either in vacuum, under helium or air. Promising research from fixed-target protein crystallography experiments has encouraged the construction of a fully dedicated fixed-target measurement station (ESB) with a high degree of automation.

Schematic plan of the SwissFEL facility, including the hard and soft X-ray beamlines Aramis and Athos respectively (gun and booster in red, linear accelerators in blue, undulators in grey, mirrors in green).

Both stations are equipped with an optical laser to trigger light-induced transition or conformational change in proteins, therefore allowing time-resolved studies with a time resolution in the 10 fs range. A new pixel detector (Jungfrau) has been designed to make best use of the properties of the XFEL for unprecedented data recording quality [13]. Planning for the construction of ESC as well as further improvements in the time resolution of the XFEL beam to a few femtoseconds has been initiated. The majority of beamtime is dedicated to academic research. However, the SwissFEL has a commercialization strategy in place to enable proprietary research at the facility as well.

Serial crystallography

Computer simulations [2] have predicted that the ultra-short, high photon flux (>1012 photons/pulse) femtosecond X-ray pulse would allow diffraction data to be collected before radiation damage destroys the sample (‘diffraction before destruction’). Even cryo-cooling of the crystal does not prevent its immediate destruction and consequently, after collecting a diffraction ‘still’ image, a new crystal needs to be delivered into the XFEL beam. Collected diffraction images are merged together from tens of thousands of crystals to provide a complete dataset. This method is now referred to as serial crystallography. Serial crystallography further differs from conventional goniometer-based data collection method where the crystal is rotated continuously in the beam. In the latter method, full reflections are recorded (or summed from partials) as the crystal continuously transverses the Ewald sphere. In serial crystallography, however, only partial reflections are recorded in still images from crystals in random orientations. This necessitated the development of new data processing algorithms [1417], since true intensities of Bragg candidates are less reliably estimated using only partial reflections. Moreover, indexing ambiguities arise from certain space groups as there are multiple options of indexing the same lattice. This poses another challenge as partial reflections from tens of thousands of crystals – which could be indexed differently – are merged. Algorithms have also been developed to resolve such indexing ambiguities during data merging [18,19]. In a serial crystallography experiment, most diffraction images generated show no diffraction pattern – when an XFEL pulse had not hit a passing crystal. A large volume of data (100s of gigabytes or terabytes) is generated, requiring sophisticated computing infrastructure (processing clusters) for storing and analysing and specialized software (Cheetah [12]) to identify images that contain diffraction patterns.

Serial crystallography demands a new research area in sample delivery technology for data collection. This includes injector-based [20,21] methods that handle either liquid (gas dynamic virtual nozzle (GDVN)) or viscous material (LCP, lipidic cubic phase), acoustic dispensing and levitation [22] technologies or fixed target methods based on using a traditional goniometer or other solid-support medium [23,24] (Figure 4). GDVN was the most commonly used liquid injector in early XFEL experiments. However, high sample consumption (∼10 μl/min) and crystallization of membrane proteins in viscous medium like LCP prompted the development of the LCP jet. This device handles viscous media at a slow flow rate (<1 μl/min) and thus reduced sample consumption. Crystals are typically extruded in a column flow using capillary with an inner diameter of 50–100 μm. This contributes a significant background signal to the diffraction from the matrix material. The choice of the capillary diameter used in the injector needs to be optimized with respect to the size of the crystals: too small a diameter will result in clogging of the injector thus wasting valuable experiment time, whereas too large a diameter will increase sample consumption and background contribution from the matrix material. Since not all protein crystals are compatible with the LCP matrix, researchers have investigated other materials compatible with the viscous jet including agarose [25], hydroxyethyl cellulose [26] and hydrogels such as sodium carboxymethyl cellulose and pluronic F-127 [27]. Hydroxyethyl cellulose and sodium carboxymethyl cellulose were reported to contribute much lower background to the diffraction as compared with LCP. These novel materials thus increase the options for optimization of data collection using the LCP jet.

Sample delivery systems for serial crystallography

Figure 4
Sample delivery systems for serial crystallography

The most frequently used systems can be classified in injector-based [20,21], acoustic dispensing and levitation [22] or fixed target methods [23,24]. All can be used for room temperature data collection, but the fixed target approach provides the option for cryo-crystallography as well.

Figure 4
Sample delivery systems for serial crystallography

The most frequently used systems can be classified in injector-based [20,21], acoustic dispensing and levitation [22] or fixed target methods [23,24]. All can be used for room temperature data collection, but the fixed target approach provides the option for cryo-crystallography as well.

The most critical parameter for successful XFEL data collection is microcrystal density that ensures passing crystals are hit by an XFEL pulse operating at a repetition rate of 60–120 Hz (at SACLA, LCLS and SwissFEL). Traditional crystallization theory applies to the generation of high density of microcrystals in syringes: manipulation of protein and precipitant concentration, controlling the equilibration rate between protein and mother liquor and microseeding to increase crystal nucleation. Crystal size can be monitored with an optical microscope and as they reach a size suitable for XFEL, further crystal growth can be stopped by removing the precipitant. Further analytical methods such as dynamic light scattering [28] and transmission electron microscopy [28,29] have been used to assess microcrystal size, density and homogeneity, and would be beneficial to optimize prior to an XFEL experiment. Different crystallization methods could also be explored. Using photosystem II as a model system, Kupitz et al. [30] compared different crystallization methods for the generation of microcrystals for XFEL experiments, and reported heterogeneous distribution of crystal sizes and density depending on the crystallization method. Seeding has also been successfully used to generate high densities of homogeneously sized microcrystals from macrocrystals of the photosynthetic reaction centre of Blastochloris viridis (RCvir) [31]. This could be applied to LCP crystallization as recently demonstrated [32].

Methods for serial crystallography are continuously in development and aim for minimization of sample consumption, improvement of measurement efficiency and data quality. Serial crystallography aims for data collection at room temperature and thus save time and efforts required to optimize freezing protocol. To preserve valuable XFEL beamtime, crystal samples and delivery systems should be thoroughly characterized for size, density, homogeneity, diffraction and jet flow properties employing a synchrotron beamline facility.

XFEL and serial crystallography provides advantages for drug discovery

How do the properties of the XFEL together with new sample preparation requirements influence the generation of experimental information for SBDD efforts? Table 2 gives a summary of key properties of the XFEL and their current or anticipated future impact to generate X-ray structural information for drug design.

Table 2
FEL X-ray data collection properties and their anticipated impact on the delivery of structural information to facilitate structure-based drug discovery
Property of XFEL Impact to structure-based drug discovery 
Unprecedented peak brilliance and μ-focus beam (~0.2–1.5 μm) Better resolution of the structural information, work with smaller crystals (<1 μm), shortening of crystallization optimization time 
Femtosecond X-ray pulses Enabling fs time-esolved measurements, observation of intermediate structural changes upon ligand binding, changes in water structure upon ligand binding 
Low/no radiation damage Radiation sensitive groups (metal complexes) are kept in physiological conditions 
Properties of serial crystallography at XFEL (SFX) and synchrotron (SMX)  
Room temperature No freezing protocol required with risk of crystal deterioration, more physiological, more realistic information on dynamic and conformation of structures 
Full automation of data collection Full liquid handling procedure, no crystal mounting 
Data redundancy Efficient screening of thousands of crystals, improved structure quality, native sulphur phasing 
Property of XFEL Impact to structure-based drug discovery 
Unprecedented peak brilliance and μ-focus beam (~0.2–1.5 μm) Better resolution of the structural information, work with smaller crystals (<1 μm), shortening of crystallization optimization time 
Femtosecond X-ray pulses Enabling fs time-esolved measurements, observation of intermediate structural changes upon ligand binding, changes in water structure upon ligand binding 
Low/no radiation damage Radiation sensitive groups (metal complexes) are kept in physiological conditions 
Properties of serial crystallography at XFEL (SFX) and synchrotron (SMX)  
Room temperature No freezing protocol required with risk of crystal deterioration, more physiological, more realistic information on dynamic and conformation of structures 
Full automation of data collection Full liquid handling procedure, no crystal mounting 
Data redundancy Efficient screening of thousands of crystals, improved structure quality, native sulphur phasing 

The XFEL triggered development in serial crystallography benefits both, synchrotron and XFEL data recording.

Unprecedented peak brilliance for small and weakly diffracting crystals

XFEL provides an increase in peak brilliance compared with synchrotron radiation by a factor of 109 [9]. Consequently, the diffraction power of small and weakly diffracting crystal can be maximized when combined with a microbeam (<1.5 μm at LCLS, SACLA and SwissFEL). This combination of properties results in X-ray structure determination from microcrystals which would otherwise be very difficult or impossible to obtain with conventional synchrotron radiation. For example, Zhang et al. [33] reported a room temperature structure of angiotensin II receptor at 2.8 Å using crystals of average size of 5 × 2 × 2 μm3, comparing with a 2.9 Å resolution structure obtained using cryo-cooled crystals of 70 × 40 × 20 μm3. Fenalti et al. [34] also reported a much higher resolution structure of the δ-opioid receptor (2.7 Å vs 3.3 Å) using microcrystals with sizes of ∼5 μm, whereas much larger crystals (∼50 μm) were used for the synchrotron derived structure. In our own group, we made similar observations with crystals of A2A adenosine receptor (1.7 Å SFX vs 1.95 Å cryo data) [35] (Figure 5, Table 3). While these studies showed that higher resolution could be achieved from XFEL despite using much smaller crystals, there are cases where XFEL enabled structure determination which was not possible at a synchrotron. For example, the high-resolution structure (1.75 Å) of CPV17 polyhedrin was obtained using sub-micron sized (∼0.5 μm) crystals, which was subsequently used to solve the same structure from a 2.2 Å dataset obtained from 768 crystals using conventional cryo-crystallography [36]. Although only partial reflections were used for data merging, serial crystallography has been demonstrated to produce very high quality data. For instance, a near atomic resolution (1.2 Å) structure was obtained with proteinase K [37], and the weak anomalous signal of sulphur atoms were used to determine the structures of lysozyme [38,39] and the A2A adenosine receptor [35,40]. The development of a viscous medium injector ensured that sample consumption is drastically reduced, and in many cases only hundreds of micrograms quantity of proteins were required for structure determination. Consequently, XFEL can enable high resolution structural determination of protein targets where the crystals’ diffraction power is limited due to size, without extensive sample consumption.

Serial Femtosecond X-ray crystallography provides improved quality of electron density

Figure 5
Serial Femtosecond X-ray crystallography provides improved quality of electron density

Comparison of electron density between cryo, SMX and SFX data [35].

Figure 5
Serial Femtosecond X-ray crystallography provides improved quality of electron density

Comparison of electron density between cryo, SMX and SFX data [35].

Table 3
Comparison of different membrane protein data collection parameters
 Indexed images Time (h) Resolution (Å) Redundancy 
SMX 128 086 2.14 1007 
SFX 3563 0.36 1.7 23 
Cryo 3500 1.95 
 Indexed images Time (h) Resolution (Å) Redundancy 
SMX 128 086 2.14 1007 
SFX 3563 0.36 1.7 23 
Cryo 3500 1.95 

Indexed images refers to the number of diffraction images used for data analysis. Time (h) corresponds to time used for data collection. Resolution (Å) to which the electron density was calculated. Redundancy is the average multiplicity of diffraction intensity recorded.

Cryo- versus room temperature data collection

About 25 years ago, data collection at cryogenic temperature was introduced to mitigate radiation damage and to prolong the useful diffraction lifetime of crystals in the X-ray beam up to 100-fold. It is also well recognized that protein molecules are highly conformational dynamic entities, with multiple side chain conformations and backbone motion that exist at physiological temperature. Freezing at cryogenic temperature traps certain conformational states in a more favourable energy landscape, disregarding the rest of the ensembles that are present at physiological condition. Hence, cryo-cooling introduces a bias in the understanding of protein function and ligand interaction. A comparison of 30 pairs of structures collected either under cryo-cooled or room temperature indicated a shift in the conformational equilibrium distribution of protein side chain, reduction of motion of flexible loops and in the case of the GTPase H-Ras, revealed a hidden catalytically active conformation of Q61 that is absent in the cryo-cooled structure [41] (Figure 6A). In another study, Fischer et al. [42] showed that transient binding pocket exists in cytochrome c peroxidase where fragment molecules were observed to bind at room temperature but not under cryo-conditions (Figure 6B). Such transient binding sites may constitute allosteric pockets that may be used to design subtype-selective inhibitors in drug discovery campaigns.

Comparison of structural differences observed at cryo- and room temperature (100K and 293K respectively)

Figure 6
Comparison of structural differences observed at cryo- and room temperature (100K and 293K respectively)

(A) Room temperature (red and orange) revealed a hidden catalytically active conformation of Q61 that is absent in the cryo-cooled structure (magenta). One of the room temperature conformations (orange) resembles the active transition state stabilized by GAP binding (pink). Figure reproduced from Fraser et al. [41]. (B) Fragment molecule benzimidazole was observed only in the room temperature structure of cytochrome c peroxidase, together with an alternative conformation of His96. Figure reproduced from Fischer et al. [42].

Figure 6
Comparison of structural differences observed at cryo- and room temperature (100K and 293K respectively)

(A) Room temperature (red and orange) revealed a hidden catalytically active conformation of Q61 that is absent in the cryo-cooled structure (magenta). One of the room temperature conformations (orange) resembles the active transition state stabilized by GAP binding (pink). Figure reproduced from Fraser et al. [41]. (B) Fragment molecule benzimidazole was observed only in the room temperature structure of cytochrome c peroxidase, together with an alternative conformation of His96. Figure reproduced from Fischer et al. [42].

Serial crystallography enables routine room temperature structure determination. This advantage gives researchers more opportunities to compare room temperature structure with their cryo-cooled equivalent to investigate if structural information is biased by freezing. An analysis of A2A adenosine receptor structures obtained under cryo-cooled and at room temperature in both synchrotron and XFEL [35] indicates that the structures are highly similar (r.m.s.d. of 0.22–0.33 Å), with similar B factor spreads along the protein polypeptide chains (data not shown). The orthosteric binding pocket is also almost identical, with the hydrogen-bonding network with water molecules preserved. Similar observations were reported for the δ-opioid receptor (r.m.s.d. of 0.5 Å) [34]. On the contrary, comparison of the XFEL and synchrotron structures of 5-HT2b receptor revealed differences in conformations for multiple side chains [43]. A similar comparison using thermolysin crystals soaked with small molecule ligand also revealed differences in ligand binding mode, protein side chain conformations and water network [44]. Clearly, whether structural deviations exist between cryo-cooled and room temperature structure depends highly on the system investigated. We therefore urge caution when investigators extract structural information from system of interest using data obtained under cryo-condition. Analysis of the PDB database indicates the number of room temperature crystal structure is very small compared with those obtained under cryogenic condition. As more structures are determined at room temperature using serial crystallography, we expect a more thorough comparison with structure under cryogenic conditions to reveal important structural dynamic information relevant to the understanding of proteins function and ligand interaction that is lost upon cryo-cooling.

Femtosecond X-ray pulses provides time resolution

Conformational dynamics are essential to protein function including solute transport, ligand binding, signal transduction and enzyme catalysis. Traditional cryo-crystallography has tremendously contributed to the understanding of the structure–function relationship of many different types of proteins. While each structure represents a single conformational snapshot of the molecule averaged over the crystal lattice and the course of data collection, details on the conformational transition between the two end states are not captured and can only be simulated in silico by molecular dynamics calculations. The extremely short pulse and the diffraction-before-destruction approach means that XFELs can provide structural information on the time dimension, and opens up a new domain in time-resolved X-ray crystallography (TR-SFX) [45]. This is a revolutionary technology as it significantly increases the level of structural information that could be derived from protein crystals, and makes feasible the generation of ‘molecular movies’ that detail the conformational transition during protein functions. TR-SFX has been demonstrated using bacteriorhodopsin (bR) [46] and photoactive yellow protein (PYP) [47,48] as model systems. In both cases the covalently attached chromophores all-trans retinal (bR) and p-coumaric acid (pCA) (PYP) were photoactivated by laser and data are then collected after a certain time-delay (‘pump–probe’ experiment). Photoactivation results in cistrans isomerization of the chromophores and structural changes in bR/PYP were observed on the femtosecond timescale. While these results demonstrated the feasibility of TR-SFX using model systems with covalently attached chromophores, experimental challenges remain for most biological systems where reactions are triggered by ligand binding. For instance, details of ligand entry and exit into a binding pocket would benefit structure-based drug design in the understanding of receptor function and enzymatic catalysis. Such experiments would require a different pump–probe setup, namely one that facilitates coherent diffusion of ligands into crystals to trigger the biological process on a synchronized timescale. Some success in this regard has been illustrated using a mixing injector [49] in which conformational changes in RNA riboswitches [50,51] could be observed by diffusing nucleotides into RNA crystals.

Radiation damage-free structure

During data collection, an optimal balance must be maintained between obtaining high resolution data with high X-ray dose and concurrent acceleration of the onset of radiation damage. A lot of work has been done in trying to understand [5254], mitigate and optimize [53,55,56] the dose limit and the effect of radiation damage to protein crystal. Radiation damage is particularly severe for high Z atoms with a large photoionization cross-section such as metal clusters. The diffraction before destruction principle for SFX has been validated when data collected with 40 fs pulses (which corresponds to the dose limit of 33 MGy for data collection at cryogenic temperature deposited per pulse) on lysozyme crystals showed no sign of radiation damage comparing with low-dose dataset (0.024 MGy) obtained at room temperature in a synchrotron [57]. The same dosage of 33 MGy deposited on crystals of photosynthetic reaction centre [58] preserved a covalent thioether bond between a cysteine and a diacylglycerol molecule which was cleaved when a dose of 77 MGy was used for data collection under cryogenic temperature. More recently, SFX enabled the observation of an intact Fe-CO bond between the heme a3 group of cytochrome c oxidase with bound substrate carbon monoxide whereas the synchrotron data showed the Fe-CO bond was cleaved [59]. These reports provide evidence that radiation damage can be avoided or at least minimized during XFEL data collection, and it would be ideal to provide structural information for radiation-sensitive biological systems. However, these studies were carried out using XFEL pulses of moderate energy (dosage in the MGy range), and it has since been demonstrated that with high intensity XFEL pulse (dosage in the GGy range) on a longer femtosecond timescale (>50 fs), local radiation damage effect can be observed for lysozyme [60] and at the [4Fe-4S] cluster of ferridoxin [61]. It has been suggested XFEL pulses as short as 10 fs may be required to minimize radiation damage [62]. Therefore, SFX data are not completely immune from the effect of radiation damage and care must be taken when interpreting data from SFX if radiation damage-free data are truly desired.

Key advantages for membrane proteins

It is estimated that 60% of pharmaceutically marketed drugs target membrane proteins [63]. Significant advances in protein engineering techniques (e.g. stabilizing mutations [64,65], fusion protein insertion [66,67]) and crystallization methodology such as lipidic cubic phase (LCP) [68] have contributed to an explosion in the number of atomic resolution structures of membrane proteins in the last decade. Macromolecular crystallography of membrane proteins require large crystals (typically >20 μm) with sufficient diffracting volume to be grown, potentially taking years of laborious optimization for particularly unstable or large membrane protein complexes. The requirement for large crystals has somewhat been mitigated by continuous improvements of the X-ray beam (increasing brilliance and advancement in X-ray optics which allows higher flux density to be focused into a smaller beam) in third generation synchrotrons. These advances combine with novel detector technology, such as hybrid pixel array/single photon counting detectors [69] that boosts excellent signal-to-noise ratio, high dynamic range and spatial resolution. Data collection at cryogenic temperature (<100 K) also substantially prolongs the useful diffraction lifetime of crystals by slowing down secondary damage events such as the diffusion of free radicals and thermal heat transfer. This increases the maximum tolerated X-ray dose per crystal from 1 MGy [54] at room temperature to 30 MGy [52]. These developments aid routine structure determination of membrane proteins. Crystal structures of more than 40 G-protein-coupled receptor (GPCR)s that cover a wide variety of members across different subfamilies have been published in the last decade and provide significant insight on the structural and functional mechanism of this important cell surface receptor family [63]. Low expression levels combined with high protein consumption for crystallization screening and optimization represent a significant time and cost burden for structure research.

Serial crystallography at XFEL facilities provides an exciting ‘second chance’ to derive structural data from membrane protein microcrystals that have failed with traditional, synchrotron-based data collection regimes. The diffraction power of small and weakly diffracting crystal is maximized by combining very high flux in a micro-beam while bypassing the effects of radiation damage. The need for continuous delivery of fresh crystals is circumvented by using injector or fixed target support with minimal manual intervention. This alternative experimental mode provides new opportunities for shorter time frames and better data quality (higher resolution with minimal radiation damage) in structure determination. For example, it would be extremely tedious and time consuming to harvest thousands of crystals manually using cryo-loops when a complete dataset has to be merged from a large numbers of crystals, especially when crystals exhibit heterogeneous diffraction quality common with membrane proteins. Moreover, data collection at room temperature avoids the need to optimize the freezing condition that is often detrimental to the lattice order of membrane protein crystals. The power of XFEL has been demonstrated with proteins structures determined using microcrystals (<10 μm) such as rabbit P-type ATPase [70], angiotensin II receptor [71], the rhodopsin/arrestin complex [72], the smoothened receptor [73] and the serotonin receptor 5-HT2b [43] (Table 4). The elimination of the need to grow large and/or well-diffracting crystals should speed up drug discovery projects by cutting out the laborious, time-consuming and expensive crystal optimization step. Further boosting the applicability of XFELs, successful structure determinations have also been demonstrated with crystals grown in vivo [7476], greatly simplifying sample preparation and optimization.

Table 4
Overview of structural investigation using XFEL facilities.
Protein Purpose of the study Protein class Size of crystals Resolution (Å) XFEL enabling Reference 
Photosystem I Proof of concept Membrane ∼200–2 μm 8.5 No Chapman et al. [4
Lysozyme Proof of concept Soluble 0.4 × 0.4 × 0.8 μm3 <7.6 No Lomb et al. [60
Cathepsin B (TbCatB) Proof of concept Soluble 0.5–1 × 10–15 μm >8 No Koopmann et al. [77
Photosystem I-Ferredoxin TR-SFX Membrane 500–2 µm Yes Aquila et al. [78
Thermolysin Proof of concept Soluble 1 × 2 × 3 µm3 >4 No Sierra et al. [79
Lysozyme Proof of concept Soluble <1 × 1 × 3 μm3 1.9 No Boutet et al. [57
Cathepsin B (TbCatB) Structure elucidation Soluble 1 × 1 × 11 μm3 2.1 Yes Redecke et al. [80
Lysozyme Proof of concept Soluble Microcrystals 3.2 No Barends et al. [38
30S ribosomal subunit Proof of concept Soluble 3 × 5 × 200 μm3 >6 No Demirci et al. [81
Photosynthetic reaction centre from Blastochloris viridis (RCvirComparison with room temperature structure Membrane Microcrystals 3.5 No Johansson et al. [58
Serotonin receptor 5-HT2B Proof of concept Membrane 5 × 5 × 5 μm3 2.8 No Liu et al. [43
Lysozyme Proof of concept Soluble <1 × 1 × 2 μm3 2.1 No Barends et al. [82
Smoothened receptor Proof of concept Membrane <5 μm 3.2 Yes Weierstall et al. [73
Cytochrome c oxidase Radiation damage free Membrane Not reported 1.9 Yes Hirata et al. [83
Photosystem II Thermolysin TR-SFX Membrane 5 × 5–10 μm 4.5 Yes Kern et al. [84
Soluble Not reported 1.8 No 
Cry3A toxin from Bacillus thuringiensis (Bt) Proof of concept Soluble 1.5 × 1.0 × 0.5 μm3 2.8 No Sawaya et al. [76
Photosystem II from Tricondyloides elongatus TR-SFX Membrane 1 μm Yes Kupitz et al. [85
Lysozyme Evaluation of grease matrix as carrier medium Soluble 7–10 μm No Sugahara et al. [86
Glucose isomerase Soluble 10–30 μm No 
Thaumatin Soluble 10–30 μm No 
FABP3 Soluble 10–20 μm No 
Cpl hydrogenase Proof of concept, goniometer-based Soluble >400 μm 1.6 No Cohen et al. [23
Myoglobin Soluble 1.4 No 
β2-Adrenoreceptor/ Complex with nanobody  Membrane 20–100 µm 2.8 No 
RNA polymerase II complex  Soluble 0.5–3 μm 3.3 No 
Photoactive yellow protein TR-SFX Soluble 1–5 μm 1.6 Yes Kupitz et al. [48
Clostridium ferredoxin Indication of radiation damage in XFEL Soluble 1–1.6 × 1–1.6 × 5–17 μm3 No Nass et al. [61
δ-opioid receptor Higher resolution achieved compared with synchrotron structure Membrane ∼5 µm 2.7 No Fenalti et al. [34
Angiotensin II Structure-based drug design Membrane 4 × 4 × 40 μm3 2.9 Yes Zhang et al. [71
CPV17 polyhedrin Structure determination Soluble   Yes Ginn et al. [87
Ca2+-ATPase SERCA Proof of concept Membrane 2 × 2 × 10–50 µm 2.8 No Bublitz et al. [70
Phycocyanin from T. elongatus Novel inert crystal delivery medium Soluble 1–5 µm 2.5 No Coquelle et al. [24
Lysozyme Proof of concept, LCP as carrier medium Soluble 1 × 1 × 2 μm3 1.9 No Fromme et al. [88
Phycocyanin Soluble 10 × 10 × 5 μm3 1.75  
Rhodospin/arrestin complex Structure determination Membrane 5–10 μm 3.8 along a and b axis and 3.3 along the c Yes Kang et al. [72
Luciferin-regenerating enzyme Proof of concept, de novo phasing Soluble 2–5 × 10–30 µm 1.5 No Yamashita et al. [89
Cyclophilin A (CypA) Room temperature Soluble  1.75 Yes Keedy et al. [90
Cytochrome c peroxidase (CCP) compound I (CmpI) Radiation-damage free structure Soluble 150 µ–1mm 1.5 Yes Chreifi et al. [91
Lysozyme Proof of concept, liquid-droplet injector Soluble 5 µm 2.3 No Miyajima et al. [92
Lysozyme Proof of concept, Acoustic levitator Soluble 5–100 µm 2.13 No Roessler et al. [93
Thermolysin 10–100 µm 2.52 
Stachydrine demethylase 25–300 µm 2.2 
A2A receptor Proof of concept, S-SAD phase Membrane 5 × 5 × 2 μm3 1.9 No Batyuk et al. [40
Bacteriorhodopsin TR-SFX Membrane Not reported 2.0 Yes Nango et al. [46
Granulovirus occlusion bodies Proof of concept Soluble <0.016 μm3 No Gati et al. [94
Photosystem II TR-SFX Membrane <100 µm in length 2.35 Yes Suga et al. [95
Proteinase K Proof of concept Soluble 8–12 µm 1.2 No Masuda et al. [37
Lysozyme Proof of concept, hydroxycellulose matrix Soluble 1 × 1 × 1 μm3 1.45 No Sugahara et al. [26
Thaumatin 2 × 2 × 4 μm3 1.55 
Proteinase K 4 × 4 × 4–5 × 5 × 7 μm3 1.5 
Thermolysin Proof of concept, soaking Soluble 4 × 4 × 8 μm3 No Naitow et al. [44
Picornavirus bovine enterovirus 2 polyhedrosis virus type 18 polyhedrin Proof of concept, fixed target Soluble 8 × 8 × 8 μm3<4 μM 2.3 2.4 No Roedig et al. [96
Cytochrome c oxidase TR-SFX Soluble 100 × 500 μm 2.2 Yes Shimada et al. [97
A2A receptor Serial crystallography at XFEL compared with synchrotron Membrane 30 × 30 × 5 μm3 1.7 No Weinert et al. [35
Protein Purpose of the study Protein class Size of crystals Resolution (Å) XFEL enabling Reference 
Photosystem I Proof of concept Membrane ∼200–2 μm 8.5 No Chapman et al. [4
Lysozyme Proof of concept Soluble 0.4 × 0.4 × 0.8 μm3 <7.6 No Lomb et al. [60
Cathepsin B (TbCatB) Proof of concept Soluble 0.5–1 × 10–15 μm >8 No Koopmann et al. [77
Photosystem I-Ferredoxin TR-SFX Membrane 500–2 µm Yes Aquila et al. [78
Thermolysin Proof of concept Soluble 1 × 2 × 3 µm3 >4 No Sierra et al. [79
Lysozyme Proof of concept Soluble <1 × 1 × 3 μm3 1.9 No Boutet et al. [57
Cathepsin B (TbCatB) Structure elucidation Soluble 1 × 1 × 11 μm3 2.1 Yes Redecke et al. [80
Lysozyme Proof of concept Soluble Microcrystals 3.2 No Barends et al. [38
30S ribosomal subunit Proof of concept Soluble 3 × 5 × 200 μm3 >6 No Demirci et al. [81
Photosynthetic reaction centre from Blastochloris viridis (RCvirComparison with room temperature structure Membrane Microcrystals 3.5 No Johansson et al. [58
Serotonin receptor 5-HT2B Proof of concept Membrane 5 × 5 × 5 μm3 2.8 No Liu et al. [43
Lysozyme Proof of concept Soluble <1 × 1 × 2 μm3 2.1 No Barends et al. [82
Smoothened receptor Proof of concept Membrane <5 μm 3.2 Yes Weierstall et al. [73
Cytochrome c oxidase Radiation damage free Membrane Not reported 1.9 Yes Hirata et al. [83
Photosystem II Thermolysin TR-SFX Membrane 5 × 5–10 μm 4.5 Yes Kern et al. [84
Soluble Not reported 1.8 No 
Cry3A toxin from Bacillus thuringiensis (Bt) Proof of concept Soluble 1.5 × 1.0 × 0.5 μm3 2.8 No Sawaya et al. [76
Photosystem II from Tricondyloides elongatus TR-SFX Membrane 1 μm Yes Kupitz et al. [85
Lysozyme Evaluation of grease matrix as carrier medium Soluble 7–10 μm No Sugahara et al. [86
Glucose isomerase Soluble 10–30 μm No 
Thaumatin Soluble 10–30 μm No 
FABP3 Soluble 10–20 μm No 
Cpl hydrogenase Proof of concept, goniometer-based Soluble >400 μm 1.6 No Cohen et al. [23
Myoglobin Soluble 1.4 No 
β2-Adrenoreceptor/ Complex with nanobody  Membrane 20–100 µm 2.8 No 
RNA polymerase II complex  Soluble 0.5–3 μm 3.3 No 
Photoactive yellow protein TR-SFX Soluble 1–5 μm 1.6 Yes Kupitz et al. [48
Clostridium ferredoxin Indication of radiation damage in XFEL Soluble 1–1.6 × 1–1.6 × 5–17 μm3 No Nass et al. [61
δ-opioid receptor Higher resolution achieved compared with synchrotron structure Membrane ∼5 µm 2.7 No Fenalti et al. [34
Angiotensin II Structure-based drug design Membrane 4 × 4 × 40 μm3 2.9 Yes Zhang et al. [71
CPV17 polyhedrin Structure determination Soluble   Yes Ginn et al. [87
Ca2+-ATPase SERCA Proof of concept Membrane 2 × 2 × 10–50 µm 2.8 No Bublitz et al. [70
Phycocyanin from T. elongatus Novel inert crystal delivery medium Soluble 1–5 µm 2.5 No Coquelle et al. [24
Lysozyme Proof of concept, LCP as carrier medium Soluble 1 × 1 × 2 μm3 1.9 No Fromme et al. [88
Phycocyanin Soluble 10 × 10 × 5 μm3 1.75  
Rhodospin/arrestin complex Structure determination Membrane 5–10 μm 3.8 along a and b axis and 3.3 along the c Yes Kang et al. [72
Luciferin-regenerating enzyme Proof of concept, de novo phasing Soluble 2–5 × 10–30 µm 1.5 No Yamashita et al. [89
Cyclophilin A (CypA) Room temperature Soluble  1.75 Yes Keedy et al. [90
Cytochrome c peroxidase (CCP) compound I (CmpI) Radiation-damage free structure Soluble 150 µ–1mm 1.5 Yes Chreifi et al. [91
Lysozyme Proof of concept, liquid-droplet injector Soluble 5 µm 2.3 No Miyajima et al. [92
Lysozyme Proof of concept, Acoustic levitator Soluble 5–100 µm 2.13 No Roessler et al. [93
Thermolysin 10–100 µm 2.52 
Stachydrine demethylase 25–300 µm 2.2 
A2A receptor Proof of concept, S-SAD phase Membrane 5 × 5 × 2 μm3 1.9 No Batyuk et al. [40
Bacteriorhodopsin TR-SFX Membrane Not reported 2.0 Yes Nango et al. [46
Granulovirus occlusion bodies Proof of concept Soluble <0.016 μm3 No Gati et al. [94
Photosystem II TR-SFX Membrane <100 µm in length 2.35 Yes Suga et al. [95
Proteinase K Proof of concept Soluble 8–12 µm 1.2 No Masuda et al. [37
Lysozyme Proof of concept, hydroxycellulose matrix Soluble 1 × 1 × 1 μm3 1.45 No Sugahara et al. [26
Thaumatin 2 × 2 × 4 μm3 1.55 
Proteinase K 4 × 4 × 4–5 × 5 × 7 μm3 1.5 
Thermolysin Proof of concept, soaking Soluble 4 × 4 × 8 μm3 No Naitow et al. [44
Picornavirus bovine enterovirus 2 polyhedrosis virus type 18 polyhedrin Proof of concept, fixed target Soluble 8 × 8 × 8 μm3<4 μM 2.3 2.4 No Roedig et al. [96
Cytochrome c oxidase TR-SFX Soluble 100 × 500 μm 2.2 Yes Shimada et al. [97
A2A receptor Serial crystallography at XFEL compared with synchrotron Membrane 30 × 30 × 5 μm3 1.7 No Weinert et al. [35

Initially, some reference proteins were investigated using an XFEL to establish sample delivery and X-ray diffraction analysis. Targets highlighted in bold are GPCR structural studies relevant to drug discovery which have been reported so far.

Summary and outlook

Two FEL facilities in the hard X-ray range have been in operation for several years (SACLA, LCLS) and their use in structure determination has been successfully demonstrated by determination of multiple high resolution X-ray structures that exceed the quality of conventional data collection methods. Areas for further significant improvements of the workflow of XFEL data collection are as follows: (1) sample delivery systems to optimize sample consumption and efficiency of data collection, (2) improved software, including experimental control, data management and analysis, taking the variation of the properties of X-ray pulses (wavelength and flux density) into account, (3) complete automation and increase in throughput for multiple structure determinations (e.g. compound screening), (4) better accessibility and establishment of an integrated concept from sample optimization to data analysis.

Synchrotrons will remain the source of choice for the vast majority of structure determinations, as well as to test crystal quality and optimize sample preparation for the XFEL experiment. On the other hand, structure determination by FELs already shows a number of advantages for structure-based drug discovery: (1) The ability to determine structures from poor quality crystals shortens the timelines for structure determination and makes some drug targets, such as challenging membrane proteins accessible; (2) higher resolution improves the accuracy of structure determination and interpretation of electron density (more reliable placement of ligand atoms); (3) structural information at room temperature (alone or in addition to cryo structures) gives better insight to protein and ligand conformational dynamics; (4) structures of proteins or ligands with radiation-sensitive groups benefit from minimal radiation damage; (5) the most exciting opportunity of FELs lies with time-resolved analysis of ligand binding and associated protein conformational change.

We are convinced that the greatest benefit of FEL application is in the area of very challenging but highly rewarding systems such as membrane protein-based drug discovery.

Summary

  • XFEL use in structure determination has matured from idea to reality.

  • XFEL advantages are unmet peak brilliance and focus, femtosecond pulses and low/no radiation damage.

  • Drug discovery will benefit from enablement of structures of challenging systems like membrane proteins, investigation of time-resolved ligand binding, new structural insights due to room temperature and full automation of crystal diffraction experiments.

The authors thank Peter Nollert for critical reading of the manuscript.

Competing Interests

The authors are associated with leadXpro AG, a company engaged in the application of XFEL to drug discovery and declare competing interests associated with the manuscript.

Abbreviations

     
  • CW

    continuous wave

  •  
  • EuXFEL

    European X-ray free electron laser

  •  
  • FEL

    free electron laser

  •  
  • GDVN

    Gas dynamic virtual nozzle

  •  
  • GPCR

    G-protein-coupled receptor

  •  
  • LCLS

    Linac Coherent Light Source

  •  
  • LCP

    lipidic cubic phase

  •  
  • SACLA

    Spring-8 Angstrom compact free electron lAser

  •  
  • SBDD

    structure-based drug discovery

  •  
  • SFX

    serial Femtosecond X-ray crystallography

  •  
  • SMX

    serial Millisecond X-ray crystallography

  •  
  • SwissFEL

    Swiss free electron laser

  •  
  • TR-SFX

    time resolved serial femtosecond crystallography

  •  
  • XFEL

    X-ray free electron laser

References

References
1
Maiman
T.H
(
1960
)
Stimulated optical radiation in ruby
.
Nature
188
,
929
930
2
Neutze
R.
,
Wouts
R.
,
van der Spoel
D.
,
Weckert
E.
and
Hajdu
J.
(
2000
)
Potential for biomolecular imaging with femtosecond X-ray pulses
.
Nature
406
,
752
757
3
Neutze
R.
,
Huldt
G.
,
Hajdu
J.
and
Van Der Spoel
D
(
2004
)
Potential impact of an X-ray free electron laser on structural biology
.
Radiat. Phys. Chem.
71
,
905
916
4
Chapman
H.N.
et al. 
(
2011
)
Femtosecond X-ray protein nanocrystallography
.
Nature
470
,
73
77
5
Huang
Z.
and
Kim
K.J.
(
2007
)
Review of X-ray free-electron laser theory
.
Phys. Rev. Spec. Top. – Accel. Beams
10
,
1
26
6
Yabashi
M.
,
Tanaka
H.
,
Tono
K.
and
Ishikawa
T.
(
2017
)
Status of the SACLA Facility
.
Appl. Sci.
7
,
604
7
Ko
I.
et al. 
(
2017
)
Construction and commissioning of PAL-XFEL facility
.
Appl. Sci.
7
,
479
8
Tschentscher
T.
et al. 
(
2017
)
Photon beam transport and scientific instruments at the european XFEL
.
Appl. Sci.
7
,
592
9
Milne
C.
et al. 
(
2017
)
SwissFEL: The Swiss X-ray free electron laser
.
Appl. Sci.
7
,
720
10
White
W.
,
Robert
A.
and
Dunne
M.
(
2015
)
The Linac Coherent Light Source
.
J. Synchrotron Radiat.
22
,
1
11
Kärtner
F.X.
,
Ahr
F.
,
Calendron
A.
,
Çankaya
H.
and
Carbajo
S.
(
2016
)
AXSIS: exploring the frontiers in attosecond X-ray science, imaging and spectroscopy
.
Nucl. Instrum. Methods Phys. Res. A
829
,
24
29
,
12
White
T.A.
et al. 
(
2013
)
Crystallographic data processing for free-electron laser sources
.
Acta Crystallogr., Sect. D: Biol. Crystallogr.
69
,
1231
1240
13
Mozzanica
A.
et al. 
(
2014
)
Prototype characterization of the JUNGFRAU pixel detector for SwissFEL
.
J. Instrum.
9
,
C05010
14
Ginn
H.M.
et al. 
(
2015
)
A revised partiality model and post-refinement algorithm for X-ray free-electron laser data
.
Acta Crystallogr., Sect. D: Biol. Crystallogr.
71
,
1400
1410
15
White
T.A.
(
2014
)
Post-refinement method for snapshot serial crystallography
.
Philos. Trans. R Soc. B Biol. Sci.
369
,
20130330
16
Lyubimov
A.Y.
et al. 
(
2016
)
Advances in X-ray free electron laser (XFEL) diffraction data processing applied to the crystal structure of the synaptotagmin-1 / SNARE complex
.
eLife
5
,
1
18
17
Sharma
A.
et al. 
(
2017
)
Asymmetry in serial femtosecond crystallography data
.
Acta Cryst.
A73
,
93
101
18
Brehm
W.
and
Diederichs
K.
(
2014
)
Breaking the indexing ambiguity in serial crystallography
.
Acta Crystallogr., Sect. D: Biol. Crystallogr.
70
,
101
109
19
White
T.A.
et al. 
(
2016
)
Recent developments in CrystFEL
.
J. Appl. Crystallogr.
49
,
680
689
20
Chavas
L. M.G.
,
Gumprecht
L.
and
Chapman
H.N.
(
2015
)
Possibilities for serial femtosecond crystallography sample delivery at future light sources
.
Struct. Dyn.
2, 041709
21
Weierstall
U
(
2014
)
Liquid sample delivery techniques for serial femtosecond crystallography
.
Philos. Trans. R Soc. Lond. B Biol. Sci.
369
,
20130337
22
Tsujino
S.
and
Tomizaki
T
(
2016
)
Ultrasonic acoustic levitation for fast frame rate X-ray protein crystallography at room tempreature
.
Sci. Rep.
6
,
25558
23
Cohen
A.E.
et al. 
(
2014
)
Goniometer-based femtosecond crystallography with X-ray free electron lasers
.
Proc. Natl. Acad. Sci. U.S.A.
111
,
17122
17127
24
Coquelle
N.
et al. 
(
2015
)
Raster-scanning serial protein crystallography using micro- and nano-focused synchrotron beams
.
Acta Crystallogr., Sect. D: Biol. Crystallogr.
71
,
1184
1196
25
Conrad
C.E.
et al. 
(
2015
)
A novel inert crystal delivery medium for serial femtosecond crystallography
.
IUCrJ
2
,
421
430
26
Sugahara
M.
et al. 
(
2017
)
Hydroxyethyl cellulose matrix applied to serial crystallography
.
Sci. Rep.
7
,
703
27
Kovácsová
G.
et al. 
(
2017
)
Viscous hydrophilic injection matrices for serial crystallography
.
IUCrJ
4
,
400
410
28
Darmanin
C.
et al. 
(
2016
)
Protein crystal screening and characterization for serial femtosecond nanocrystallography
.
Sci. Rep.
6
,
25345
29
Barnes
C.O.
et al. 
(
2015
)
Assessment of microcrystal quality by transmission electron microscopy for efficient serial femtosecond crystallography
.
Arch. Biochem. Biophys.
602
,
61
68
30
Kupitz
C.
et al. 
(
2014
)
Microcrystallization techniques for serial femtosecond crystallography using photosystem II from Thermosynechococcus elongatus as a model system
.
Phil. Trans. R. Soc. B.
369
,
20130316
31
Dods
R.
et al. 
(
2017
)
From macrocrystals to microcrystals: a strategy for resource from macrocrystals to microcrystals: a strategy for membrane protein serial crystallography
.
Structure
25
,
1
8
32
Kolek
S.A.
,
Bräuning
B.
and
Shaw Stewart
P.D.
(
2016
)
A novel microseeding method for the crystallization of membrane proteins in lipidic cubic phase
.
Acta Crystallogr. Sect. F Struct. Biol. Commun.
72
,
307
312
33
Zhang
H.
et al. 
(
2017
)
Structural basis for selectivity and diversity in angiotensin II receptors
.
Nature
544
,
327
332
34
Fenalti
G.
et al. 
(
2015
)
Structural basis for bifunctional peptide recognition at human δ-opioid receptor
.
Nat. Struct. Mol. Biol.
22
,
1
6
35
Weinert
T.
et al. 
(
2017
)
Serial millisecond crystallography for routine room-temperature structure determination at synchrotrons
.
Nat. Commun.
8
,
542
36
Ginn
H.M.
et al. 
(
2015
)
Structure of CPV17 polyhedrin determined by the improved analysis of serial femtosecond crystallographic data
.
Nat. Commun.
6
,
6435
37
Masuda
T.
et al. 
(
2017
)
Atomic resolution structure of serine protease proteinase K at ambient temperature
.
Sci. Rep.
7
,
45604
38
Barends
T.R.M.
et al. 
(
2013
)
Anomalous signal from S atoms in protein crystallographic data from an X-ray free-electron laser
.
Acta Crystallogr., Sect. D: Biol. Crystallogr.
69
,
838
842
39
Nakane
T.
et al. 
(
2015
)
Native sulfur/chlorine SAD phasing for serial femtosecond crystallography
.
Acta Crystallogr., Sect. D: Biol. Crystallogr.
71
,
2519
2525
40
Batyuk
A.
et al. 
(
2016
)
Native phasing of X-ray free-electron laser data for a G protein-coupled receptor
.
Sci. Adv.
2
,
e1600292
e1600292
41
Fraser
J.S.
et al. 
(
2011
)
Accessing protein conformational ensembles using room-temperature X-ray crystallography
.
Proc. Natl. Acad. Sci. U.S.A.
108
,
16247
16252
42
Fischer
M.
,
Shoichet
B.K.
and
Fraser
J.S.
(
2015
)
One crystal, two temperatures: cryocooling penalties alter ligand binding to transient protein sites
.
Chembiochem
16
,
1560
1564
43
Liu
W.
et al. 
(
2013
)
Serial femtosecond crystallography of G protein-coupled receptors
.
Science
342
,
1521
1524
44
Naitow
H.
et al. 
(
2017
)
Protein–ligand complex structure from serial femtosecond crystallography using soaked thermolysin microcrystals and comparison with structures from synchrotron radiation
.
Acta Crystallogr., Sect. D: Struct. Biol.
73
,
702
709
45
Neutze
R
(
2014
)
Opportunities and challenges for time-resolved studies of protein structural dynamics at X-ray free-electron lasers
.
Philos. Trans. R Soc. Lond. B. Biol. Sci.
369
,
20130318
46
Nango
E.
et al. 
(
2016
)
A three-dimensional movie of structural changes in bacteriorhodopsin
.
Science
354
,
1552
1557
47
Pande
K.
et al. 
(
2016
)
Femtosecond structural dynamics drives the trans/cis isomerization in photoactive yellow protein
.
Science
352
,
725
729
48
Kupitz
C.
et al. 
(
2015
)
Time-resolved serial crystallography captures high-resolution intermediates of photoactive yellow protein
.
Science
346
,
1242
1246
49
Calvey
G.D.
,
Katz
A.M.
,
Schaffer
C.B.
and
Pollack
L
(
2016
)
Mixing injector enables time-resolved crystallography with high hit rate at X-ray free electron lasers
.
Struct. Dyn.
3
,
1
19
50
Stagno
J.R.
et al. 
(
2016
)
Structures of riboswitch RNA reaction states by mix-and-inject XFEL serial crystallography
.
Nature
541
,
242
246
51
Stagno
J.R.
,
Bhandari
Y.R.
,
Conrad
C.E.
,
Liu
Y.
and
Wang
Y.-X.
(
2017
)
Real-time crystallographic studies of the adenine riboswitch using an X-ray free electron laser
.
FEBS J.
doi:10.1111/febs.14110
52
Owen
R.L.
,
Rudiño-Piñera
E.
and
Garman
E.F.
(
2006
)
Experimental determination of the radiation dose limit for cryocooled protein crystals
.
Proc. Natl. Acad. Sci. U.S.A.
103
,
4912
4917
53
Zeldin
O.B.
,
Brockhauser
S.
,
Bremridge
J.
,
Holton
J.M.
and
Garman
E.F.
(
2013
)
Predicting the X-ray lifetime of protein crystals
.
Proc. Natl. Acad. Sci. U.S.A.
110
,
20551
20556
54
Southworth-Davies
R.J.
,
Medina
M.A.
,
Carmichael
I.
and
Garman
E.F.
(
2007
)
Observation of decreased radiation damage at higher dose rates in room temperature protein crystallography
.
Structure
15
,
1531
1541
55
Sanishvili
R.
et al. 
(
2011
)
Radiation damage in protein crystals is reduced with a micron-sized X-ray beam
.
Proc. Natl. Acad. Sci. U.S.A.
108
,
6127
6132
56
Paithankar
K.S.
and
Garman
E.F.
(
2010
)
Know your dose: RADDOSE
.
Acta Crystallogr., Sect. D: Biol. Crystallogr.
66
,
381
388
57
Boutet
S.
et al. 
(
2012
)
High-resolution protein structure determination by serial femtosecond crystallography
.
Science
337
,
362
364
58
Johansson
L.C.
et al. 
(
2013
)
Structure of a photosynthetic reaction centre determined by serial femtosecond crystallography
.
Nat. Commun.
4
,
2911
59
Ishigami
I.
et al. 
(
2017
)
Crystal structure of CO-bound cytochrome c oxidase determined by serial femtosecond X-ray crystallography at room temperature
.
Proc. Natl. Acad. Sci. U.S.A.
114
,
8011
8016
60
Lomb
L.
et al. 
(
2011
)
Radiation damage in protein serial femtosecond crystallography using an x-ray free-electron laser
.
Phys. Rev. B. Condens. Matter Mater. Phys.
84
,
214111
61
Nass
K.
et al. 
(
2015
)
Indications of radiation damage in ferredoxin microcrystals using high-intensity X-FEL beams
.
J. Synchrotron Radiat.
22
,
225
238
62
Barty
A.
et al. 
(
2012
)
Self-terminating diffraction gates femtosecond X-ray nanocrystallography measurements
.
Nat. Photonics
6
,
35
40
63
Overington
J.P.
,
Al-Lazikani
B.
and
Hopkins
A.L.
(
2006
)
How many drug targets are there?
Nat. Rev. Drug Discov.
5
,
993
996
64
Serrano-Vega
M.J.
,
Magnani
F.
,
Shibata
Y.
and
Tate
C.G.
(
2008
)
Conformational thermostabilization of the beta1-adrenergic receptor in a detergent-resistant form
.
Proc. Natl. Acad. Sci. U.S.A.
105
,
877
882
65
Scott
D.J.
,
Kummer
L.
,
Tremmel
D.
and
Plückthun
A.
(
2013
)
Stabilizing membrane proteins through protein engineering
.
Curr. Opin. Chem. Biol.
17
,
427
435
66
Privé
G.G.
et al. 
(
1994
)
Fusion proteins as tools for crystallization: the lactose permease from Escherichia coli
.
Acta Crystallogr., Sect. D: Biol. Crystallogr.
50
,
375
379
67
Chun
E.
et al. 
(
2012
)
Fusion partner toolchest for the stabilization and crystallization of G protein-coupled receptors
.
Structure
20
,
967
976
68
Caffrey
M.
and
Porter
C.
(
2010
)
Crystallizing membrane proteins for structure determination using lipidic mesophases
.
J. Vis. Exp.
3
,
2
6
69
Johnson
I.
et al. 
(
2014
)
Eiger: a single-photon counting x-ray detector
.
J. Instrum.
9
,
C05032
70
Bublitz
M.
et al. 
(
2015
)
Structural studies of P-type ATPase-ligand complexes using an X-ray free-electron laser
.
IUCrJ
2
,
409
420
71
Zhang
H.
et al. 
(
2015
)
Structure of the angiotensin receptor revealed by serial femtosecond crystallography
.
Cell
161
,
833
844
72
Kang
Y.
et al. 
(
2015
)
Crystal structure of rhodopsin bound to arrestin by femtosecond X-ray laser
.
Nature
523
,
561
567
73
Weierstall
U.
et al. 
(
2014
)
Lipidic cubic phase injector facilitates membrane protein serial femtosecond crystallography
.
Nat. Commun.
5
,
3309
74
Gati
C.
et al. 
(
2014
)
Serial crystallography on in vivo grown microcrystals using synchrotron radiation
.
IUCrJ
1
,
87
94
75
Gallat
F.-X.
et al. 
(
2014
)
In vivo crystallography at X-ray free-electron lasers: the next generation of structural biology?
Philos. Trans. R. Soc. B Biol. Sci.
369
,
20130497
76
Sawaya
M.R.
et al. 
(
2014
)
Protein crystal structure obtained at 2.9 Å resolution from injecting bacterial cells into an X-ray free-electron laser beam
.
Proc. Natl. Acad. Sci. U.S.A.
111
,
12769
12774
77
Koopmann
R.
et al. 
(
2012
)
In vivo protein crystallization opens new routes in structural biology
.
Nat. Methods
9
,
259
262
78
Aquila
A.
et al. 
(
2012
)
Time-resolved protein nanocrystallography using an X-ray free-electron laser
.
Opt. Express
20
,
2706
79
Sierra
R.G.
et al. 
(
2012
)
Nanoflow electrospinning serial femtosecond crystallography
.
Acta Crystallogr., Sect. D: Biol. Crystallogr.
68
,
1584
1587
80
Redecke
L.
et al. 
(
2013
)
Natively Inhibited Trypanosoma brucei Cathepsin B structure determined by using an X-ray laser
.
Science
339
,
227
230
81
Demirci
H.
et al. 
(
2013
)
Serial femtosecond X-ray diffraction of 30S ribosomal subunit microcrystals in liquid suspension at ambient temperature using an X-ray free-electron laser
.
Acta Crystallogr., Sect. F: Struct. Biol. Cryst. Commun.
69
,
1066
1069
82
Barends
T.R.M.
et al. 
(
2013
)
De novo protein crystal structure determination from X-ray free-electron laser data
.
Nature
505
,
244
247
83
Hirata
K.
et al. 
(
2014
)
Determination of damage-free crystal structure of an X-ray–sensitive protein using an XFEL
.
Nat. Methods
11
,
734
736
84
Kern
J.
et al. 
(
2014
)
Taking snapshots of photosynthetic water oxidation using femtosecond X-ray diffraction and spectroscopy
.
Nat. Commun.
5
,
4371
85
Kupitz
C.
2016
Serial time-resolved crystallography of photosystem II using a femtosecond X-ray laser
.
Nature
513
,
261
265
86
Sugahara
M.
et al. 
(
2015
)
Grease matrix as a versatile carrier of proteins for serial crystallography
.
Nat. Methods
12
,
61
63
87
Ginn
H.M.
et al. 
(
2015
)
Structure of CPV17 polyhedrin determined by the improved analysis of serial femtosecond crystallographic data
.
Nat. Commun.
6
,
6435
88
Fromme
R.
et al. 
(
2015
)
Serial femtosecond crystallography of soluble proteins in lipidic cubic phase
.
IUCrJ
2
,
545
551
89
Yamashita
K.
et al. 
(
2015
)
An isomorphous replacement method for efficient de novo phasing for serial femtosecond crystallography
.
Sci. Rep.
5
,
14017
90
Keedy
D.A.
et al. 
(
2015
)
Mapping the conformational landscape of a dynamic enzyme by multitemperature and XFEL crystallography
.
eLife
4
,
1
26
91
Chreifi
G.
et al. 
(
2016
)
Crystal structure of the pristine peroxidase ferryl center and its relevance to proton-coupled electron transfer
.
Proc. Natl. Acad. Sci. U.S.A.
113
,
1226
1231
92
Miyajima
K.
et al. 
(
2016
)
Microcrystal delivery by pulsed liquid droplet for serial femtosecond crystallography research papers
.
Acta Cryst.
D72
,
520
523
93
Roessler
C.G.
et al. 
(
2016
)
Acoustic injectors for drop-on-demand serial femtosecond crystallography
.
Structure
24
,
631
640
94
Gati
C.
et al. 
(
2017
)
Atomic structure of granulin determined from native nanocrystalline granulovirus using an X-ray free-electron laser
.
Proc. Natl. Acad. Sci. U.S.A.
114
,
2247
2252
95
Suga
M.
et al. 
(
2017
)
Light-induced structural changes and the site of O = O bond formation in PSII caught by XFEL
.
Nature
543
,
131
135
96
Roedig
P.
et al. 
(
2017
)
High-speed fixed-target serial virus crystallography
.
Nat. Methods
14
,
805
810
97
Shimada
A.
et al. 
(
2017
)
A nanosecond time-resolved XFEL analysis of structural changes associated with CO release from cytochrome c oxidase
.
Sci. Adv.
3
,
e1603042