Serial femtosecond crystallography

Serial femtosecond crystallography (SFX) is a form of X-ray crystallography developed for use at X-ray free-electron lasers (XFELs).^[1][2][3] Single pulses at free-electron lasers are bright enough to generate resolvable Bragg diffraction from sub-micron crystals. However, these pulses also destroy the crystals, meaning that a full data set involves collecting diffraction from many crystals. This method of data collection is referred to as serial, referencing a row of crystals streaming across the X-ray beam, one at a time. It can be performed at room temperature, allowing for the study of biochemical dynamics.^[4] It can be used to visualize samples prone to radiation damage, such as metalloproteins, and to observe transient structures, such as reaction intermediates, which would not be captured using conventional X-ray crystallography.^[5]

Serial Femtosecond Crystallography (SFX) schematic

History

While the idea of serial crystallography had been proposed earlier,^[6] it was first demonstrated with XFELs by Chapman et al.^[7] at the Linac Coherent Light Source (LCLS) in 2011. This method has since been extended to solve unknown structures, perform time-resolved experiments, and later even brought back to synchrotron X-ray sources.

Methods

In comparison to conventional crystallography, where a single (relatively large) crystal is rotated in order to collect a 3D data set, some additional methods have to be developed to measure in the serial mode. First, a method is required to efficiently stream crystals across the beam focus. The other major difference is in the data analysis pipeline. Here, each crystal is in a random, unknown orientation which must be computationally determined before the diffraction patterns from all the crystals can be merged into a set of 3D hkℓ intensities.

Sample Delivery

The first sample delivery system used for this technique was the Gas Dynamic Virtual Nozzle (GDVN) which generates a liquid jet in vacuum (accelerated by a concentric helium gas stream) containing crystals. Since then, many other methods have been successfully demonstrated at both XFELs and synchrotron sources. A summary of these methods along with their key relative features is given below:

• Gas Dynamic Virtual Nozzle (GDVN)^[8] - low background scattering, but high sample consumption. Only method available for high repetition rate sources.^[9]
• Lipidic Cubic Phase (LCP) injector^[10] - Low sample consumption, with relatively high background. Specially suited for membrane proteins
• Other viscous delivery media^[11][12] - Similar to LCP, low sample consumption with high background
• Fixed target scanning systems (wide variety of systems have been used with different features, with standard crystal loops,^[13] or silicon chips^[14]) - Low sample consumption, background depends on system, mechanically complex
• Tape drive (crystals auto-pipetted onto a Kapton tape and brought to X-ray focus) - Similar to fixed target systems, except with fewer moving parts

Data Analysis

In order to recover a 3D structure from the individual diffraction patterns, they must be oriented, scaled and merged to generate a list of hkℓ intensities. These intensities can then be passed to standard crystallographic phasing and refinement programs. The first experiments only oriented the patterns^[15] and obtained accurate intensity values by averaging over a large number of crystals (> 100,000). Later versions correct for variations in individual pattern properties such as overall intensity variations and B-factor variations as well as refining the orientations to fix the "partialities" of the individual Bragg reflections.^[16]