Crystallography. The revolution of XFELs

13. The revolution of XFELs

Mode: full-screen / central-screen / help
Table of contents through the logo

Chapter written by Jose M. Martín García.

In the context of this chapter you will find answers to the following topics:

What is an XFEL?

Free electron lasers (FELs) have existed for over 50 years since they were introduced by John Madey in the early 70´s. Early FELs operated at infrared, visible and ultraviolet wavelengths. The first FEL operating in the X-ray regime (soft X-ray region) was the Deutsches Elektronen-Synchrotron’s (DESY) Free-electron LASer in Hamburg (FLASH), which began user operation in 2005. Shortly after, in 2009, the worlds´ first FEL operating with hard X-rays emerged, the SLAC’s Linac Coherent Light Source (LCLS) in the US. Since then, four more followed (SACLA in Japan, EuXFEL in Germany, PAL-XFEL in South Korea, and SwissFEL in Switzerland), two more are currently under construction (LCLS-II in USA, and SHINE in China), and one is on planned (UK-XFEL in UK).

X-ray Free-Electron Lasers (XFELs) are research facilities of a few Km long (~3 Km) that are able to routinely generate coherent, ultra-brilliant, tunable laser pulses of very short duration in the X-ray regime. Here are some of the most relevant ones:

Linac Coherent Light Source (LCLS, USA)

The LCLS facility is located at the SLAC National Accelerator Laboratory in Menlo Park, California, USA. LCLS was the first laser in the world, and one of just five now in operation, to produce “hard,” or very-high-energy X-rays. The LCLS currently delivers 120 X-ray pulses per second, each one lasting femtoseconds. Since the start of operations in 2009, the LCLS has more than 3,000 users, more than 9,000 experiments have been carried out in its seven instruments (TMO, ChemRIXS, XPP, XCS, FMX, CXI, and MEC), and more than 1,450 peer-reviewed articles have been published.

A major upgrade to the facility, known as LCLS-II, is underway. This will provide a revolutionary leap in capability by increasing the X-ray pulse repetition rate from 120 pulses per second to 1 million pulses per second. . LCLS-II will be a transformative tool for energy science, qualitatively changing the way that X-ray imaging, scattering and spectroscopy can be used to study how natural and artificial systems function. It will enable new ways to capture rare chemical events, study quantum materials with unprecedented resolution, and track the behavior of fluctuation biological systems.

Linac Coherent Light Source (LCLS, USA)

SPring-8 Angstrom Compact free electron LAser (SACLA, Japan)

SACLA is an X-ray free electron laser located in Harima Science Garden city in Japan. It is embedded in the Spring-8 accelerator and synchrotron complex. It came into play in 2011, being the second XFEL in the world after LCLS. It is a 700 m long facility that produces bright X-ray pulses with a repetition rate between 30 and 60 Hz. SACLA has three beamlines, two of them producing hard X-rays (BL2 and BL3) and one producing soft X-rays (BL1).

SPring-8 Angstrom Compact free electron LAser (SACLA, Japan)

European XFEL (Eu-XFEL)

The European XFEL (EuXFEL) is currently the largest XFEL facility in the world (3.4 Km). It runs from the DESY campus in Hamburg to the town of Schenefeld in Schleswig-Holstein. Its construction began in early 2009; user operation started in September 2017. The EuXFEL has been realized thanks to the big effort of 12 partner countries: Denmark, France, Germany, Hungary, Italy, Poland, Russia, Slovakia, Spain, Sweden, Switzerland, and the United Kingdom. It generates 27,000 ultrashort X-ray flashes per second and with a brilliance that is a billion times higher than that of the best conventional X-ray radiation sources. In addition, it has a unique pulse structure in which femtosecond pulses are delivered within trains with 10 Hz gaps between pulses.

European XFEL
The 3.4 kilometre-long facility runs from the DESY campus in Hamburg to the town of Schenefeld in Schleswig-Holstein

Pohang Accelerator Laboratory (PAL-XFEL, South Korea)

Pohang Accelerator Laboratory X-ray Free Electron Laser (PAL-XFEL) is located at Pohang in South Korea. It is operational since 2017, being the third XFEL that came into play. PAL-XFEL consists of a Hard X-ray (HX) and a Soft X-ray (SX) FEL line. The HX line includes a 780 m long accelerator line, a 250 m long undulator line, and 80 m long experimental halls. The SX line is branched at 260 m point from beginning. It includes a 170 m long accelerator line, a 130 m long undulator line, and a 30 m long experimental hall. The HX line generates between 2 and 15 keV FEL with over 1 mJ pulse energy, between 10 and 35 fs pulse duration, and under 20 fs arrival time jitter from 4 to 11 GeV electron beams. The SX line generates between 0.25 and 1.25 keV FEL with over 10¹² photons from 3 GeV e-beams.

Pohang Accelerator Laboratory (PAL-XFEL, South Korea)

Switzerland Free Electron Laser (SwissFEL, Switzerland)

The SwissFEL is a 740 m long X-ray free electron laser at the Paul Scherrer Institute (PSI) (Villigen, Switzerland), which was inaugurated in December 2016. The SwissFEL design is optimized to generate X-ray pulses in the wavelength range of 1 to 70 Å. Construction work for SwissFEL began in the spring of 2013. After completion of the building, installation of the technical components started at the beginning of 2015. The total construction cost is around 275 million Swiss francs. The first pilot experiments were carried out in 2017. In 2018 the first beamline, ARAMIS, was put into operation. The second beamline, ATHOS, is expected to follow by 2021. ARAMIS delivers very high-energy short-wave X-ray light, which can be used to follow how atoms behave during a fast-moving process. ATHOS delivers softer X-ray light with lower energy, making it possible to observe atoms and molecules as they form a new chemical bond.

Switzerland Free Electron Laser (SwissFEL, Switzerland)

At XFELs, electron bunches generated by an electron gun are accelerated to the speed of light in a linear accelerator of a few kilometers long before they reach the undulators. Once there, the electrons wiggle via the magnets allowing the electrons to interact with their emitted radiation, which causes microbunching of the electrons with spacing equal to the wavelength of the emitted X-rays. As the electrons proceed through the undulators, their emission becomes more and more coherent leading to highly coherent pulses with femtosecond duration. The extremely high number of electrons in a micro bunch (on the order of a billion), the high longitudinal coherence, and the ultrashort duration of the pulses, pushes the peak brilliance 10 orders of magnitude higher than that currently achievable at synchrotrons, making XFELs the Worlds’ most powerful X-ray light sources.

Undulator hall at the EuXFEL (Photo: EuXFEL)

Schematic representation of an undulator segment at an XFEL instrument.
Figure from Martin-Garcia et al., ABB (2016)

The figure above shows a relativistic electron beam (solid line) is brought to high energy in a linear accelerator (not pictured) prior to interaction with the undulator. The electrons then travel on a sinusoidal path, induced by a special arrangement of magnets called undulator, a periodic array of magnetic dipoles shown as red and blue boxes. Because the electrons move in curved paths via the magnets, the change in momentum causes the emission of monochromatic radiation the same way a synchrotron does (depicted as a red cone).

Animated gif illustrating the wiggling motion of the electron bunches passing through the undulator

Comparison of peak brightness as a function of photon energy between conventional lasers and higher harmonic generation sources, synchrotron sources, and X-ray free electron lasers. Figure from Boutet and Yabashi, in X-ray Free Electron Lasers, Springer (2018)

Interested readers should look at the video offered by SLAC (Stanford Linear Accelerator Center) explaining how an XFEL works.

Serial Femtosecond Crystallography (SFX)

XFELs have made the beginning of a new era of X-ray science. Indeed, the unique properties of XFELs (high brightness and short pulse duration) have facilitated the appearance of a new scientific-technological advance in the field of structural biology, the serial femtosecond crystallography (SFX).

SFX relies on the “diffraction before destruction” principle in which the photon scattering events by crystals when hit by XFEL pulses occur so rapid than diffraction from crystals can be recorded before being destroyed.

Explosion of T4 lysozyme (white: H; grey: C; blue: N; red: O; yellow: S) when exposed to an X-ray FEL pulse with an FWHM of 2 fs, and disintegration followed in time. Atomic positions in the first two structures (before and after the pulse) are practically identical at this pulse length induced by XFEL pulses. Figure from Neutze et al., Nature (2000)

In SFX, data sets are no longer collected from single, cryo-cooled crystals mounted on a goniometer; instead, hundreds of thousands of nanocrystals or microcrystals are delivered to the XFEL beam in a serial fashion, in random orientations, at room temperature, so that only one snapshot is recorded per crystal per X-ray pulse. Because the XFEL beam is so intense, crystals must be constantly replenished between X-ray pulses as the crystals are literally destroyed upon each shot. More experimental aspects on this new way of doing crystallography can be found in the following sections below.

Schematic representation of the experimental setup of a typical SFX experiment at XFELs. Randomly oriented nanocrystals (green) in their mother liquor are delivered into the focus of X-ray beam by a gas-focused liquid injector. The X-ray beam, which is transverse to the jet, hits the crystals in the interaction region and diffraction snapshots of single crystals are recorded on a detector. Figure from Martin-Garcia et al., ABB (2016)

Since the first crystal structure of a protein was determined at an XFEL, over 466 SFX structures have been deposited in the PDB as of October 2nd, 2021.

Because SFX is conducted at room temperature, it enables not only the structure determination of static structures, but also time-resolved studies to understand conformational dynamics by determining the structure of intermediate states in enzymatic reactions at timescales of 100 fs to minutes.

Moreover, the short duration of pulses produced by XFELs enables SFX to overcome the major limitations of macromolecular crystallography with synchrotrons: radiation damage. Thus, SFX enables studies of macromolecules under “native” conditions at room temperature in the mother liquor.

SFX to determine the structure of macromolecules

The possibility of collecting nearly damage-free X-ray data on macromolecules has greatly impacted two research areas in structural biology:

In vivo crystallography: Spontaneous growth of protein crystals inside living cells have been known for long time, being observed in a variety of living organisms, performing a diversity of functions from food storage to defense, as well as being identified in some human pathologies [Dogan et al., Head & Neck Pathology (2012), Doye & Poon, COCIS (2006), Lange et al., Cell & Tissue Res. (1982) , Pande et al., PNAS (2001)].

Recent advances in protein-expression systems, especially in the baculovirus-infected insect cell lines, have become in vivo crystallization as a plausible, interesting alternative to crystallize recombinant proteins that cannot be crystallized by conventional in vitro methods...

(a) Micrograph of Sf9 cells with needles-like crystals of CatB protein inside. (b) Scanning EM micrograph of a group of Sf9 cells showing crystals traversing the cell membrane. (c) TEM micrograph of an embedded and sectioned infected Sf9 cell with crystals. (d) TEM micrograph of a sectioned sample, showing a crystal cut perpendicular to the long axis of the needle. (e) TEM micrograph showing the lattice structure of a crystal. Figure taken from Koopmann, R. et al., Nat Methods (2012)

Animated gif displaying a microcrystal of luciferase protein growing inside a living insect cell
Original movie from Koopmann, R. et al., Nat Methods (2012)

In vivo grown crystals have two main limitations by which they are not suitable for X-ray measurements at synchrotron light sources: a) They tend to be very sensitive to radiation damage; and b) their size is proved to be too small (a few microns), being typically restricted to the size of the cells. The advent of high-brilliance XFELs and the establishment of SFX have opened new avenues in structural analysis using in vivo grown crystals to determine protein crystal structures. To date, a total of 14 structures have been determined by this approach.

G protein-coupled receptors (GPCRs): GPCRs belong to a large family of cell receptors involved in some of the most chronic human pathologies including cancer, cardiovascular diseases, diabetes, obesity, and immune disorders; being, likely, one of the most attractive targets for structure-based drug discovery. However, obtaining large, well-diffracting crystals for data collection at synchrotrons is very challenging. Crystallization of GPCRs typically leads to the formation of micrometer-sized crystals, which make them suitable for SFX experiments. In fact, SFX (along with Cryo-EM) has contributed notably to the rapid growth of structural studies of GPCRs. As of May 2020, a total of 13 structure have been determined by SFX since the first one came out in 2013 for the serotonin receptor 5-HT2B.

Timeline of GPCRs at XFELs. All GPCR structures determined as of May 2020
Figure from Fromme et al., eLS (2020)

Structural dynamics and molecular movies with XFELs

In addition to determining static structures of macromolecules, SFX with XFELs offers the possibility to capture functional dynamics.

The porous nature of macromolecular crystals with large solvent channels that allows ligands to diffuse into the crystals, along with the small size of the crystals used in SFX (in the nano- or micrometer range), becomes SFX ideal for Time-Resolved experiments (TR-SFX). Tiny crystals allow for a greater percentage of molecules in the crystals to be quick and evenly activated by either a laser or rapid mixing.

Illustration of the porous nature of macromolecular crystals. Middle shows how the molecules of human oxidoreductase NQO1 (PDB 5FUQ) are found in a real crystal. Left and right show the same crystal packing of NQO1 molecules, but as seen from other two sides of the crystals. Ligands can diffuse inside crystals through the spaces between individual molecules.

For a time resolved experiment to succeed, homogeneous and fast reaction initiation must occur. This can only be achievable by using very small crystals like those used in SFX. A study by Marius Schmidt in 2013 demonstrates that the small size of the crystals used in SFX, theoretically allow for diffusion times on microsecond time scales, allowing access to many reactions on the millisecond or even microsecond regimes. For instance, a crystal with dimensions of 0.5 x 0.5 x 0.5 µm³ has been modeled to exhibit a diffusion time of 17 µs, while a 3 x 4 x 5 µm³ crystal is estimated to take 1 ms, and a large 300 x 400 x 500 µm³ crystal would take 9.5 s.

There are two types of TR-SFX experiments that can be conducted at XFELs: 1) pump-probe TR-SFX experiments, in which crystals are injected into the path of a pump laser before being probed after a predetermined time delay by the X-rays, and 2) mix-and-inject TR-SFX experiments, where protein microcrystals are mixed with a substrate so that the substrate rapidly diffuses into the crystals and binds to the protein.

Typical schematic set-ups of TR-SFX experiments at XFELs. (a) Set-up of a time-resolved mix-and-inject experiment in which protein microcrystals are mixed with a substrate in the interaction region of the injector (black box) so that the substrate rapidly diffuses into the crystals and binds to the protein (red box). Time delays are probed by either varying the sample and buffer flow rates or by placing an expanded region after the constriction. (b) Set-up of a time-resolved pump–probe experiment in which crystals are injected into the path of a pump laser before being probed after a predetermined time delay (Δt) by the X-rays. Figure from Fromme et al., eLS (2020)

In both scenarios, one can access to not only the initial and final states of a biochemical reaction but also to the extremely fast states (intermediates) that form as the reaction proceeds. By elucidating all the intermediates formed in a reaction, a “molecular movie” can be obtained to reveal how biological molecules proceed in nature.

Since the first documented TR-SFX experiment was conducted at an XFEL to study the irreversible reaction of undocking of reduced ferredoxin from Photosystem I (PSI) upon light excitation with visible light [Aquila et al., Opt. Ex. (2012)], several reaction mechanisms have been either partially or fully elucidated successfully by using TR-SFX.

Examples of TR-SFX experiments of light-activated proteins at XFELs

Photoactive yellow protein (PYP): The molecular movie of the trans-to-cis isomerization of the photoactive center (a molecule of p-coumaric acid, pCA) of PYP, was “filmed” between 100 fs and 3 ps. The movie reveals that PYP remains in an excited state for about 600 fs upon light absorption before the chromophore atoms, still in trans configuration, begin to shift and continue over a 3ps time delay where the chromophore is in a cis configuration [see Tenboer et al., Science (2014) and Pande et al., Science (2016)].

Trans (pink) to cis (green) isomerization in PYP between 100 fs and 3 ps. The region highlighted in gray was not explored. Dashed line shows the transition time at about 590 fs. The inserts show the structures of PYP_trans (pink), PYP_cis (light green), and dark state structure PYP_ref (yellow circle). Difference electron density in red (−3σ) and blue (3σ). Figure from Pande et al., Science (2016)

Animated gif showing this isomerization process in PYP
The original molecular movie is also available

Bacteriorhodopsin (bR): The molecular movie of the photocycle of bR has been constructed at time delays from femtosecond to millisecond regimes. From these experiments the intermediates involved in the isomerization of the retinal molecule, the structural rearrangements undergone by bR, as well as the water molecule network acting as a proton wire, were all captured as a function of time [Nango et al., Science (2016); Nogly et al., Science (2018); Weinert et al., Science (2019)].

Isomerization cycle of the retinal molecule
A molecular movie is also available

Examples of TR-SFX experiments by rapid mixing/diffusion at XFELs

β-Lactamase C (BlaC): The proof-of-concept of a mixing experiment at XFELs was done for watching, in “real time”, the inactivation of an antibiotic (Ceftriaxone, CEF) by the enzyme BlaC from Mycobacterium tuberculosis. Reaction intermediates were captured by collecting data sets at time points between 5ms to 2s upon mixingthe protein crystals with CEF. From these experiments, it has been discovered that CEF binds BlaC shortly after mixing and that the nucleophilic attack by Serine70 and the subsequent opening of the β-lactam ring, occur 500ms later [Kupitz et al., Struc. Dynamics (2017); Olmos et al., BMC Biol. (2018), Pandey et al., IUCrJ (2021)].

Top. Overview of BlaC at 500 ms after mixing with CEF. The omit maps (maps calculated after excluding atoms in question from the structure) are shown for the covalently bound intermediate E-CFO* in green. Electron density of an additional, stacked ceftriaxone molecule near the active site is shown in dark green. Figure adapted from Olmos et al., BMC Biol. (2018)
Bottom. Simulated annealing omit maps of the active site structures (a–e) during the reaction at time points between 30 ms and 2 s upon mixing are shown. Figure adapted from Fromme et al., eLS (2020)

Animated gif of the above mentioned process
The original molecular movie is also available

Adenine riboswitch: The reaction of a riboswitch molecule with its adenine substrate was followed for a time delay of 10 min. During the first 10 seconds, the adenine molecule diffused into the crystals, bound and reacted with the RNA molecule, being able to capture the full mechanism. Then, if the reaction is left to evolve, between 10s and 10 min, a polymorphic transition occurs inside the crystals triggered by the substrate. Riboswitch molecules, initially in a P2₁ space group, underwent huge conformation changes leading to a P2₁2₁2₁ space group [Stagno et al., Nature (2017); Stagno et al., FEBS J. (2017)].

Structure model of the riboswitch rA71. Schematic of the reaction mechanism of adenine with riboswitch rA71 showing all species involved. Models, with hydrogen-bond interactions showing the key residues in the ligand-binding pockets of apo1 (blue), apo2 (cyan), IB (yellow), and ligand-bound (pink) states. Figure adapted from Stagno et al., Nature (2017)

Large conformational changes induced after 10 minutes of ligand mixing result in a polymorphic phase transition and lattice conversion from a monoclinic (apo1 and apo2: blue and cyan respectively) to an orthorhombic space group (B/ade: magenta), which was accommodated in micro/nanocrystals of rA71. A subset of 4 related molecules (white) between the two structures illustrates that half of the molecules rotate ~90° upon conversion. Figure from Stagno et al., FEBS J. (2017). A molecular movie is also available here.

Experimental considerations for conducting serial crystallography experiments

Preparation and characterization of crystal samples

A typical SFX experiment requires samples of small crystals (nanocrystals or microcrystals) at high crystal density (10⁹-10¹² crystals/ml), and highly homogeneous in size. This type of samples is never produced from scratch. Conventional commercial crystal screens are used to find those conditions that form nanocrystals or microcrystals instead. Alternatively, one could start from a condition in which large crystals have been obtained and apply the necessary modifications to produce the small crystals.

Many of the existing classical methods for producing large crystals (> 50 μm) could be applied for growing nano- or microcrystals with the necessary modifications. However, the density and homogeneity of the sample produced by these methods is insufficient. Instead, two techniques have been further developed from existing methods to produce large quantities of crystalline samples with high homogeneity...

Free interface diffusion (FID): In the common scenario, the protein (less dense solution) is placed on the bottom of a microcentrifuge tube, then the precipitant buffer (denser solution) is added dropwise allowing for a high nucleation rate at the interface, forming lots of nanocrystals over time, which will settle down due to gravity into the precipitant rich layer [Kupitz et al., Phil. Trans. Royal Soc.B (2014)].

Schematic of the set-up for crystallization experiments with FID (a,b) and FID centrifugation (c).

(a) Experimental set-up in which the protein solution is carefully layered on top of the precipitant solution, where only few crystals form at the interface.
(b) In the inverse set-up the precipitant solution is added dropwise to the protein solution, inducing increased transient nucleation at the drop–protein interface.
(c) The experiment shown in (b) is continued by centrifugation.
The nuclei formed in the protein solution are accelerated by centrifugation towards the interface zone, where they grow into nano- or microcrystals. When they reach a specific size they sediment into the precipitant zone, where they stop growing. Thereby nano- or microcrystals with a very narrow size distribution can be achieved. Figure from Kupitz et al., Phil. Trans. Royal Soc. (2014)

Example of an optimization experiment from large crystals to microcrystals. Large crystals (200-300 µm) of Taspase1 protein grown in 0.1 M sodium citrate pH4, 10% MPD, 1 M AmSO₄, ratio 1:1, and protein conc. 10 mg/ml, were optimized to grow microcrystals (5-10 µm) by FID in 0.1 M sodium citrate pH4, 15% MPD, 1.4 M AmSO₄, ratio 1:3, and protein conc. 20 mg/ml.

Batch: In this method, the protein solution is placed on the bottom of a microcentrifuge tube, then the precipitant buffer is added and quickly mixed with protein.

There exist a variety of techniques that can be used to characterize the tiny crystals used in SFX. When crystals size is on the order to the micrometer size, they can be identified by conventional methods such as polarized light microscopy in combination with UV-fluorescence microscopy. However, when crystals size is in the sub micrometer range, alternative methods are required. The Second Order of Non-linear Imaging of Chiral Crystals (SONICC), is the most powerful technique for identifying crystals as small as 100 nm [Wampler et al., J.Am.Chem.Soc. (2008)]. SONICC relies on the second-order harmonic generation (SHG) spectroscopy. To check crystal quality, the most trustworthy method is the classical X-ray powder diffraction, though, transmission electron microscopy (TEM) is another reliable method [Stevenson et al., Phil.Trans.Soc.B (2014)]. Dynamic light scattering (DLS) is commonly used in SFX experiments to assess nanocrystal size distribution, homogeneity, as well as optimizing the size of crystal seeds [Schubert et al., J. Appl. Cryst. (2015)].

To learn more about sample preparation and characterization please read other references [Beale et al., J. Appl. Cryst. (2019); Beale & Marsh, JoVE (2021); Kupitz et al., Phil. Trans. Royal Soc. (2014); Martin-Garcia et al., Arch.Biochem.& Biophys. (2016); Dorner et al., Cryst. Growth Des.(2016)].

Sample delivery methods

Early SFX experiments were carried out using the Gas Dynamic Virtual Nozzle (GDVN) injector to deliver microcrystalline samples to the XFEL pulses [Weierstall et al., Rev.Sci. Instrum. (2012); DePonte et al., J.Phys.D:App.Phys. (2008)]. Despite GDVN has been successfully used in many SFX experiments, it has a big limitation: high sample consumption. Sample delivery technology has improved very much in the past years allowing the emergence of novel sample delivery methods that consume a lot less sample. Below is a short description of all sample delivery devices currently available for SFX experiments.

Gas-focused liquid injectors: In this type of devices, crystals in their mother liquor are typically delivered through an inner capillary and focused by high-pressured sheath gas that flows between the inner and the outer capillary.

Left: Schematic of a GDVN (Gas Dynamic Virtual Nozzle) injector. The GDVN is assembled by placing a smaller, inner capillary, inside a larger, outer capillary. Crystals pass through the inner capillary while a focusing gas is passed through the outer. A thin, micrometer jet is produced when the co-flowing gas meets the crystals as they exit the inner capillary. Figure from DePonte et al., J.Phys.D:App.Phys. (2008)
Right. Minimovie of a liquid jet in action. Credits by SLAC

This type of injectors has two major limitations (clogging events, and high sample consumption), which were mitigated by the double flow focused nozzle (DFFN) [Oberthuer et al., Sci.Rep.(2017)]. DFFN uses a coaxially flowing liquid (typically an alcohol) to accelerate the flow of the crystal-containing solution, both of which are subsequently accelerated by a sheath gas. Another type of gas-focused liquid injectors is the mixing injectors used for time-resolved experiments. The most commonly known is the injector developed by Lois Pollack´s group at Cornell University, which consists of a microfluidic mixer of bonded, concentric glass capillaries coupled to regular GDVN [Calvey et al., Struc.Dyn.(2016)].

Schematic of the mixing injector developed by Pollack´s group at Cornell University. Black areas indicate regions bound by UV epoxy. Polyimide centering spacers are shown in orange. The inner sample line in the GDVN can be either 50 or 75 μm ID. The length of the device from the tip of the GDVN to the supply lines is approximately 10 cm. Figure from Calvey et al., Struc.Dyn.(2016)

High-viscosity injectors: Commonly referred to as the “LCP injector” [Weierstall et al., Nature Com.(2014)], the high-viscosity injector was initially developed to deliver membrane protein crystals grown in lipidic cubic phase (LCP). Protein soluble crystals can also be injected by mixing the crystals with a high viscosity medium. The LCP injector overcome the two major limitations of liquid injectors.

Schematic and actual photo of the “LCP injector” developed by Uwe Weierstall at ASU. Water (blue) and gas (green) lines are routed through the nozzle rod from the left, LCP (red) is extruded from the nozzle on the right. Water pushes the hydraulic plunger to drive LCP through a capillary. Two spherical Teflon beads are used to provide a tight. The co-flowing gas is necessary for reliable extrusion and to maintain co-axial flow. Figure adapted from Weierstall et al., Nature Com.(2014)

A minimovie of a viscous jet in action

Some common h-viscosity extrusion methods: Lipidic cubic phase (LCP), Hyaluronic acid (grease), agarose, PEO.

Electrospinning injectors: In this type of injectors, a high voltage is applied to crystal sample that flows through a capillary towards the XFEL beam. As the sample comes out, the polarizable liquid surface is deformed by the electric potential, resulting in a very narrow liquid stream. The most common device is the microfluidic electro-kinetic sample handling (MESH). MESH overcomes clogging issues, consumes a lot less sample that liquid injectors, and is compatible with highly viscous mother liquor solutions (e.g. glycerol, PEG, or sucrose).

Pulsed injectors: Since XFELs produce pulses of X-rays, a smart approach for saving sample when injecting samples is the idea to pulse the injection such that the sample is present when the X-ray pulse is “ON”, and absent when the X-ray pulse is “OFF”. There exist several methods of pulsed injection: 1) microfluidic droplet generator coupled with GDVN to segment the aqueous crystal suspension with an oil carrier phase [Echelmeier et al., Internat. Conf. Mini. Sys. Chem. & Life Sci. (2015)]; 2) piezoelectric droplet injection in which crystal suspension is pressure driven such that a meniscus of the sample is primed to be ejected when agitated by the piezoelectric element that is triggered by an electric pulse generator at the frequency and phase of the XFEL beam [Mafuné et al., Acta Cryst. (2016)]; 3) Acoustic droplet ejection (ADE) in which crystal suspension in ambient pressure conditions can be acoustically ejected as a droplet from a well plate into the path of the X-ray [Roessler et al., Structure (2016)]; and 4) droplet on tape device is an X-ray transparent tape (typically Kapton) coupled with ADE [Fuller et al., Nat. Methods (2017)].

All devices described above allow droplet generation either electrically or acoustically triggered, in a drop-on-demand mode, which is desirable for low sample waste.

Schematic representation of the segment flow droplet generation on-chip and off-chip interfaced with the XFEL instrument

Figure from Echelmeier et al., Internat. Conf. Mini. Sys. Chem. & Life Sci. (2015)

Left: Droplet generation in the PDMS microfluidic device

Middle: Transfer of crystal suspension droplets to a silica capillary

Right: The silica capillary is used as the inner portion of the GDVN injecting a liquid stream into the XFEL chamber

Left: Schematic representation of the piezoelectric sample delivery device. Figure from Zhao et al., FEBS J. (2019)
Middle: Minimovie of the piezoelectric in action. Credits by SLAC.
Right: Schematic overview of the ADE device. In the inverted system, the droplet is ejected downward out of a multi-well microplate. The modified Echo system is configured to eject droplets upwards out of a multi-well microplate. The pulses arrive at the interaction region concurrent with a crystal-containing droplet

Fixed sample holders: In these devices, the sample is pipetted onto a solid matrix, which vary from traditional loops, to windowed sample support wafers/grids (similar to those used in electro microscopy), and to microfluidic chips. Extremely low sample consumption and high hit rates are the most common advantages of this type of devices. On the other hand, crystal dehydration and slow data acquisition, are among their most common drawbacks.

Some of the most commonly used fixed target devices

a) and d) adapted from Cohen et al., PNAS (2014); b) adapted from Murray et al., Acta Cryst.D (2015); c) adapted from Mueller et al., Struc. Dyn.(2015)

Minimovie of how data collection from fixed targets is carried out at XFELs. Credits by SLAC

To learn more about sample delivery please read the following references: Zhao et al., FEBS J. (2019); Martiel et al., Acta Cryst.(2019); Echelmeier et al., Anal.& Bioanal.Chem.(2019).

Serial data collection and data processing in SFX

The high repetition rate at which XFELs operate [120 Hz at LCLS; 1 MHz at LCLS-II; 1.1 MHz (projected 4.5 MHz) at EuXFEL; 60 Hz at SACLA and PAL-XFEL; and 100 Hz at SwissFEL] has necessitated the development of new detector technologies capable of integrating all photons arriving within the time duration of just a few femtoseconds. Thus, detectors for SFX need to have a high repetition rate, high signal-to-noise ratio, as well as a large dynamic range.

The first detector suitable for SFX was the Cornell-SLAC pixel array detector (CSPAD), was developed by SLAC scientists. It consists of 64 separate modules (194 x 185 pixels each), allowing cost-effective replacement and experimental flexibility. The CSPAD is tiled to produce a 2.3 megapixels detector, with readout speeds matching the repetition rate of 120 Hz. The panel distribution leaves an adjustable sized hole in the middle to allow for adjustable incident beam focuses, preventing the beam from damaging the detector. Since then, three more module-based detectors for XFEL science have been developed: the AGIPD at EuXFEL; the JUNGFRAU at SwissFEL; and the ePIX at SLAC.

Physical layout of the CSPAD detector on the left (photo by SLAC), the AGIPD detector on the middle (photo by EuXFEL), and the ePIX detector on the right (photo by SLAC)

As mentioned before, data sets in SFX are collected from snapshot diffraction patters of randomly oriented crystals, which along with the high repetition rate of XFELs, results in hundreds of thousands of patterns collected per hour, thus creating terabytes of data. Thus, serial data collection/processing in SFX cannot be efficiently done with conventional crystallographic data processing methods, thus bringing new challenges with respect to the data sets typically collected at synchrotrons. Furthermore, because the SFX data sets consist of a combination of images collected of single crystal hits (diffraction from one crystal), blank patterns (no diffraction), and multi-hits (diffraction from more than one crystal), real-time analysis and data reduction have become a necessity to fast access data quality and optimize the data collection strategy.

Typical SFX data analysis workflow

Data processing steps and tools for SFX:

Data reduction: This is the very first step in the SFX data processing pipeline, also known as pre-processing step or “cleaning step”. In this step, all blank and multi-hit patterns are filtered out, detector calibrations and background subtraction are applied, Bragg peaks are identified (so called “peak finding”), and data collection statistics is performed. There are three programs currently available for data reduction: CASS, Cheetah (see also this link), and OM (OnDA before).

Indexing and integration: The background corrected and sorted diffraction patterns from previous step are then processed by programs such as CrystFEL (see also this link) or cctbx.xfel (see also this link). These programs identify the Bragg peaks in the hits, determine the unit cell parameters and the orientation of each crystal, and then perform indexing by widely used algorithms such as MOSFLM, DIRAX, and XDS, as well as new and more efficient autoindexing algorithms recently developed including XGANDALF, ASDF and FELIX, and TakeTwo. Two more programs, SPIND and EMC, have been developed for sparse and weak diffraction patterns that only contain a few of identifiable Bragg reflections. Finally, once each pattern has been successfully indexed, the intensities are merged and integrated using the Monte Carlo method.

New challenges for data collection, storage, and evaluation are about to come with the higher repetition rate (megahertz) of the upcoming XFELs, like the EuXFEL with a pulse train structure of 100,000 images/second or the LCLS-II, aiming at up to 1 million pulses/second, thus increasing data volume from terabytes to petabytes of data/hour.

Serial crystallography at synchrotron light sources

The advantages the serial crystallography approach offer and the new avenues it has open in protein crystallography research, has awaken the interest of the structural biology community to the point that XFELs facilities cannot longer accommodate such demand, being necessary to search for alternatives. The solution to this problem is the synchrotron radiation sources. Next-generation synchrotrons, upgraded for higher flux density and with beamlines using sophisticated focusing optics, submicron beam diameters and fast low-noise photon-counting detectors have become into real alternative.

At present, almost each one of the most powerful synchrotrons worldwide has at least one beamline dedicated, fully or partly, to serial crystallography approach. Examples of beamlines currently in operation or under construction are: The 23-ID-D beamline at the Advanced Photon Source (Chicago, USA); the 17-ID-2 beamline at the National Synchrotron Light Source II (New York, USA); the 12-1 beamline at Synchrotron Stanford Radiation Lightsource (Stanford, USA); the I24 beamline at Diamond Light Source (Oxford, UK) (Horrell et al., JoVE, 2021); the ID29 beamline at the ESRF (Grenoble, France); the MicroMAX beamline at MAX IV (Lund, Sweden); the TREXX beamline at PETRA III (Hamburg, Germany); and the BL13-XALOC and BL06-XAIRA beamlines at ALBA (Barcelona, Spain).

The adaptation of the serial crystallography approach at synchrotrons has been feasible thank to the recent advances in sample delivery technology with new and more efficient devices being developed to drastically reduced sample consumption. The most successful devices employed so far are the high-viscosity injectors, the fixed targets, and the drop-on-demand devices. Since the first serial crystallography was conducted in 2014 at a synchrotron by Gati and co-workers, numerous serial synchrotron crystallography experiments have successfully followed in the past years. Time-resolved experiments on time-scales currently limited to a few hundred milliseconds is now feasible at synchrotrons. However, future synchrotron upgrades to next-generation diffraction-limited sources will allow for time-resolved experiments in the microsecond and perhaps nanosecond range, which is highly relevant for the study of the majority of enzymatic processes in biology.

More information on serial crystallography at synchrotrons can be found in these two comprehensive reviews recently published [Martin-Garcia JM, Crystals (2021) 11, 521; Pearson and Mehrabi, Curr. Opin. Struct. Biol. (2020) 65, 168-174].

Compact pulsed X-ray light sources

The major limitation of XFELs is their accessibility, which is leading to an extreme bottleneck for cutting-edge research. Thus, with the main goal of doing XFEL science more accessible, the Center for Applied Structural Discovery (CASD) in coordination with William Graves (Professor in the College of Integrative Sciences and Arts and Director of Accelerator Science at the Biodesign Institute, of the Arizona State University) is working with international collaborators to build two compact X-ray sources, a compact X-ray light source (CXLS), and a compact X-ray free electron laser (CXFEL). By using the most recent advances in accelerator and laser technologies, the two instruments will be around 10 m long and cost about $25M, which is a lot less that the budget necessary to build a few Km long XFEL (over a $1 billion).

The key component of the compact machines where the hard X-rays will be produced, is the interaction region, which is based on the physical principle called “inverse Compton scattering (ICS)”. Relativistic electron bunches generated by an upstream electron gun, will travel through a short linear accelerator until they get to the interaction point where they collide with powerful infrared pulses. When the collision happens with the right geometry (near head-on), the conversion of low-energy laser photons to high-energy X-rays occurs and the scattered X-rays emerge in the same direction as the electrons.

Inverse Compton scattering (ICS)

Bill Graves (2019) explaining the CXLS machine (original source)

The first instrument currently being constructed at ASU is the CXLS. This instrument will be capable to provide 1000 X-ray pulses/s with a pulse duration of 100 fs, an average flux of 1 x 10¹¹ photons/s, a peak brilliance of 9 x 10¹⁸ photons/(s .1% mm²mrad²), and a beam focus of just 3 μm. Another advantage of the CXLS is that the X-rays will be tunable between 2 and 40 keV at two bandwidths, 0.1% and 5%. However, the X-rays will not be very coherent yet. The CXLS was turned on in February 2021 to produce the first electrons.

Layout of the compact X-ray light source (CXLS) at Arizona State University. Electron bunches generated by the photoinjector are accelerated and compressed by the linacs and compression chicanes devices, respectively. The interaction point, where electrons and laser collide to produce the X-rays by the ICS phenomenon, is highlighted by a red box. The total length from the photoinjector to ICS interaction point is about 10 m. Components in the experimental hutch (beam shaping and diagnostics, sample chamber and detector) are also illustrated. Figure adapted from Fromme et al. eLS (2020).

Left: Photoinjector
Middle: CXLS section corresponding to the photoinjector and linacs
Right: 1 m long linacs

Bill Graves (2021) explaining about the compact X-ray light source (CXLS)
See also the video of Bill Graves (2021) explaining the instrument

As for the CXFEL, this is still in its phase design and will become a reality in about 5 years. This instrument will exceed the properties of the CXLS, with fully coherent X-rays and the possibility to produce sub-femtosecond X-rays.

Peak brightness as a function of photon energy of the CXFEL compared to that of conventional lasers, synchrotron sources, and X-ray free electron lasers

Introducing the CXFEL technology (2019) (original source)

The compact light sources will revolutionize the field by making the technology available to many more scientists around the world and accelerate the rate at which we discover knowledge for helping solve critical global challenges. Below is a list of major impacts the CXLS could have:

Phase contrast imaging (PCI)
Macromolecular crystallography
Drug discovery
Radiation therapy
Cultural heritage studies
Time-resolved chemistry and biology
Advanced water purification membranes
Photovoltaic and energy storage materials
Semiconductor fabrication metrology
Mineral morphology for enhanced oil and gas extraction

Table of contents