Protein crystallography using x-ray free-electron lasers

Just over 100 years ago, Max von Laue reported the first x-ray diffraction patterns from crystals. Soon after, Lawrence Bragg interpreted diffraction spots. Their work provided the foundation for x-ray crystallography, which today allows us to build three-dimensional images of molecular structures. Several scientific fields were born as a result, including structural biology, where the molecular machinery of life is inspected with increasing complexity. Consequently, there are now in excess of 87,000 structure entries in the online protein databank.¹

To solve protein structures, scientists around the world have historically used the intense and tunable x-ray beams produced by synchrotron radiation facilities. Now there is a new source: the x-ray free-electron laser (FEL),² which is set to overcome the challenges in studying protein crystallography, vastly increasing the range of proteins whose structures can be solved.³

Figure 1. Schematic diagram of serial crystallography. X-ray pulses from the Linac Coherent Light Source (LCLS) are focused onto a continuously flowing microjet of crystals suspended in liquid. The diffraction is detected on one or two detector panels (p-n junction charge coupled devices⁴ or pixel-array detectors⁵), which read out a new pattern on every pulse.

X-ray FELs provide incredibly intense pulses of x-rays with up to 10¹³ photons in a duration of tens of femtoseconds (fs). The first hard x-ray FEL, the Linac Coherent Light Source at the SLAC National Accelerator Laboratory in California, went into operation in 2009. This facility routinely produces x-ray pulses of around 3mJ energy in a duration of 70fs or less. Focused beams can achieve x-ray intensities of greater than 10²⁰W/cm². This is many orders of magnitude greater than that required to form a plasma from any material, and indeed this beam vaporizes anything in its path.

Remarkably, these destructive pulses overcome the problem of radiation damage in protein crystallography and x-ray microscopy. With conventional sources, such as synchrotron particle accelerators (or storage rings), ionizing x-rays break bonds and cause changes to the structure as the exposure takes place. This limits the total dose that can be used to record images or diffraction patterns, since we cannot compensate for a weakly diffracting crystal by exposing the protein for longer. Therefore, crystallographers need to grow large, well-diffracting crystals of proteins, which can require months or years of sustained effort. Cryogenic cooling can increase the radiation dose tolerance by a factor of 30, by slowing down diffusion of x-ray-generated free radicals. But an x-ray FEL pulse overcomes these limitations by outrunning the radiation damage, and terminating before significant atomic motion can occur, so that our measurement is unaffected. By the end of a 100fs pulse, the average displacement of atoms is less than 1 angstrom. This allows us to collect data at doses in excess of 3 gigagray at room temperature, which is over 3000 times the conventional tolerable room-temperature dose.⁶

Figure 2. A full 3D set of x-ray structure factors (here³ from Photosystem I) is obtained by indexing individual diffraction patterns and summing intensities of indexed spots. This is akin to powder diffraction, where every grain is first oriented in 3D. (Figure courtesy of Thomas White, Center for Free-Electron Laser Science—CFEL—at DESY.)

Figure 3. A rendering of the molecular structure of the Trypanosoma brucei cathepsin B protein, showing the electron density of the naturally inhibiting propeptide structure in the green mesh and the atomic model. (Figure courtesy of Karol Nass, CFEL DESY.)

Using our technique we can characterize diffraction from protein crystals that are too small for conventional measurement. We quantified diffraction patterns from nanocrystals of the photosystem I complex (proteins involved in photosynthesis) that are smaller than 200nm in diameter—only six unit cells across.³ Because each pulse vaporizes the sample, we needed to replenish it in time for the next pulse, arriving at a rate of 120Hz. We did this by continuously flowing a suspension of nanocrystals in a microjet that intersected the beam path (see Figure 1). We collected 120 diffraction frames per second.⁷ The crystals are randomly oriented and not synchronized to the x-ray pulses, so the odds of hitting one depend on the concentration of crystals in the fluid suspension (the x-ray flash freezes all motion). We collected millions of frames, from which we extracted tens to hundreds of thousands of individual snapshot diffraction patterns. We individually indexed these and merged them into a set of 3D structure factors, using CrystFEL software.⁸ We can think of the dataset as a 3D powder diffraction pattern, where each grain in the powder pattern is individually captured and oriented before summing (see Figure 2).

We tested the method on several proteins, including the enzyme lysozyme and the challenging membrane-bound protein complex photosystem I. Careful scrutiny and comparison with low-dose diffraction data shows no sign of radiation-induced structural changes.⁷We collected almost 200,000 individual diffraction patterns from in vivo grown crystals of the protease cathepsin B from the Trypanosoma brucei parasite, which causes African sleeping sickness.⁹ When we artificially introduced the gene for that protein into insect cells by virus infection (to produce that protein in quantity) the crystals spontaneously formed in the cells, and we extracted them to run in the jet. We obtained the structure using the phasing method of molecular replacement, where a similar protein is used as a model. This showed the enzyme's precursor form with its natural inhibitor, which could offer insight for treatment of the disease (see Figure 3).

The method of serial femtosecond crystallography is well suited to studies of irreversible reactions that are time-resolved. We can 'pump' samples with an optical pulse synchronized to the x-ray to an accuracy of greater than 100fs, varying the delay between the pump and the x-ray measurement pulse to build up an ultrafast movie. The sub-micron-sized crystals make it possible to diffuse a reagent in a shorter time, increasing the accuracy. Future developments could include improved sample delivery to reduce the consumption of protein, and implementing anomalous diffraction schemes for de novo phasing, where no base model is required to solve the protein structure.¹⁰ The small number of unit cells in the crystals also offers intriguing possibilities for phasing, by measuring intensities between Bragg spots.¹¹ This approaches x-ray FEL's ultimate aim of measuring the smallest crystals of all: single molecules.¹²