Comment on Flash-cooling of macromolecular crystals to overcome increased mosaicity by Yao, Yasutake & Tanaka (2004).

# 2004 International Union of Crystallography Printed in Denmark ± all rights reserved In their article on the use of capillary versus loop-mounted crystals to obtain slower cooling rates to cryotemperatures, Yao et al. (2004) refer to our use of capillaries for cooling nucleosome core particle crystals (Luger et al., 1997). They indicate that the cooling process we used is even slower than theirs. We note, however, that although our process takes 20 min to complete, the cooling rates achieved for vitri®cation are actually very fast. Considering the importance of ̄ash-cooling in macromolecular crystallography, we describe our procedure in greater detail here. Following transfer of nucleosome core particle crystals to a 24% solution of the cryoprotectant 2-methyl-2,4-pentanediol, our cooling procedure using capillary mounts involved four distinct steps: (1) transfer to 277 K, (2) cooling to 255 K over 5 min, (3) ̄ash-cooling into liquid propane at 153 K, and (4) transfer into a cold gas stream at 103 5 K. While the ®rst two steps in the liquid phase occurred relatively slowly, the cooling rate in the ̄ash-cooling step was limited only by the heat-transfer physics of the system when using a mechanical device (e.g. Hope et al., 1989) capable of plunging the mounted crystal rapidly into the cryobath. Rod-shaped crystals of about 0.2 mm in cross-section were mounted in thin-walled tapered capillaries, with the crystals lodged in the taper. Excess solution above and below the crystal was removed, leaving only the liquid held by capillary attraction between the crystal and the capillary wall. With this residual liquid ensuring optimal thermal contact with the bath, our calculations indicate that the expected initial cooling rates would be greater than 1000 K sÿ1 for the jump from 255 to 153 K. Experimental tests using a ®ne thermocouple indicated that the crystals reach 173 K within 50 ms and the ®nal temperature within 120 ms (unpublished work). Cooling rates with different ̄ash-cooling strategies vary over ten to 100-fold, and as Yao et al. (2004) note, the optimal cooling rate will depend on the material in the crystal, crystal size and the cryoprotectant. Our observations indicate that the cryobath temperature can also be critical for optimal diffraction quality. For several protein/DNA complexes, including the nucleosome core particle, we have found that the optimal bath temperature for the vitri®cation step is always near 153 K. At higher temperatures, the intensity of the diffuse water ring increased signi®cantly. At lower temperatures, mosaicity increased and the diffraction limit deteriorated. Once vitri®ed, the rate of transfer to 103 K had no apparent effect on diffraction quality. With nucleosome core particle crystals prepared in capillaries as described in Luger et al. (1997), or later with loops (Davey et al., 2002), no improvement in diffraction quality could be achieved with the annealing procedure reported by Harp et al. (1998), who ̄ashcooled crystals by direct transfer into a cold gas stream. Crystals that have suffered a substantial increase in mosaicity because of excessive thermal or osmotic shock may partially recover through a melting and revitri®cation cycle. We have seen that crystals that have developed cracks during transfer to cryoprotectant in many cases reanneal. However, use of smaller crystals that are less susceptible to cracking generally yields higher resolution singlecrystal diffraction. We suggest that where cryo-annealing procedures result in improved diffraction quality, an optimized cooling protocol would be likely to yield better diffraction quality without a cryo-annealing step. In our experience, optimization of the cryocooling protocol used for different crystal types has generally resulted in improved data quality, and the capillary-mount method described by Yao et al. (2004) is a useful tool for this purpose.