Date Published: April 01, 2017
Publisher: International Union of Crystallography
Author(s): Robin L. Owen, Danny Axford, Darren A. Sherrell, Anling Kuo, Oliver P. Ernst, Eike C. Schulz, R. J. Dwayne Miller, Henrike M. Mueller-Werkmeister.
A time- and sample-efficient approach for the serial collection of room-temperature diffraction data using a fixed target at a synchrotron is demonstrated.
X-ray free-electron lasers have driven the emergence of serial crystallography in recent years. The short duration and high intensity of X-ray pulses from a free-electron laser mean that protein crystals are destroyed by a single pulse. The short duration (5–50 fs) and high intensity (>1012 photons) of the pulses have the benefit that useful data can be collected before the crystal is destroyed (Chapman et al., 2011 ▸), although the destructive nature of the interaction introduces the challenge of presenting a new crystal to each X-ray pulse. In order to address this challenge, a variety of sample-delivery methods have been developed, including, but not limited to, liquid jets, fixed-target arrays and goniometer-based approaches (all recently summarized by Chavas et al., 2015 ▸).
A summary of the data-collection parameters and associated scaling statistics is shown in Table 1 ▸. The crystals on chip 1 were SWMb-CO in space group P212121, while the crystals on chip 2 were SWMetMb in space group P6. While a desire to minimize the absorbed dose (though defocusing and attenuating the X-ray beam in the case of chip 1 and reducing the exposure time in the case of chip 2) resulted in a relatively low signal-to-noise level, the indexing and scaling of data was not problematic and the data extended to 1.9 Å resolution or better in each case. The use of the second intensity moment as a criterion for including data resulted in the inclusion of more weak reflections, and we observed that while the overall data quality improved in the approach described in §2.3, this also contributed to a low signal-to-noise level. Anecdotally, the individual diffraction patterns did not appear to be significantly weaker than those collected during a comparable ‘standard’ oscillation experiment. Indeed, the aperture-like nature of each chip feature acts to reduce diffuse scatter on each image, meaning that background levels are low. An additional contribution to low signal-to-noise values may therefore be owing to the challenges associated with post-refinement and estimating the partiality of data from individual still images collected using a monochromatic beam.
The low-dose structure solution from myoglobin crystals mounted on silicon chips at room temperature demonstrates the realisation of SSX in a sample- and time-efficient manner. The method by which the chips are loaded means that data are collected from a large fraction of all crystals prepared. Furthermore, those not successfully loaded can be recovered for loading onto subsequent chips. The maps obtained from chip 1 show that even a modestly loaded chip provides sufficient data for structure solution. The exposure time per aperture defines the throughput of our approach, with a significant advantage being that this can be easily varied according to the available flux or crystals to hand. On beamline I24, for ‘well diffracting’ samples we find that 25 ms exposures are sufficient for successful data collection using still images; this means that data can be collected from all 11 664 positions on a chip in less than 7 min. The time taken to exchange and align chips is somewhat less than this, and will become greatly reduced when stages that are capable of holding multiple chips and automated chip exchange are developed. In our approach only one alignment step is needed: alignment of the chip. With a good-quality on-axis viewing system this is relatively straightforward and allows the collection of data from many thousands of crystals in a few minutes.