It is clear that there is an urgent need for new digital detection devices that avoid the need for light conversion as required by the current generation of CCD detectors (Faruqi and Andrews, 1997
). Efforts to develop so-called direct detection devices have been the focus of intense interest from the cryoEM community over the past several years (Jin et al., 2008
, Faruqi, 2009
, Grigorieff and Harrison, 2011
, Battaglia et al., 2009
, Contarato et al., 2011
, Deptuch et al., 2007
, McMullan et al., 2009a
, Milazzo et al., 2005
). Several of these devices have been quite carefully characterized and demonstrated to have superior performance to CCD cameras (and in some cases film) (McMullan et al., 2009a
, Milazzo et al., 2010). Several other promising advantages make these devices a very attractive possibility for cryoEM, including the possibility of single electron counting (McMullan et al., 2009b
) and the ability to use sequential frames to correct for specimen movement or explore dose fractionation (Glaeser et al., 2011
The Direct Detection Device (DDD) is a camera developed for transmission electron microscopy based on a radiation hardened monolithic active pixel sensor design (Jin et al., 2008
, Milazzo et al., 2005
, Milazzo et al., 2010). Characterization of an earlier 1K×1K pixel prototype DDD showed improved high frequency performance compared to conventional CCD cameras (Milazzo et al., 2010). However, the DQE and MTF measurements were performed using a standard edge method and did not prove the ability of the detector to image a biological specimen under low dose conditions. To our knowledge, no results have yet been published that unequivocally illustrate that the new detector is suitable for imaging samples embedded in vitreous ice. We report here the results of initial testing of the DDD, using GroEL as a test specimen, that indicate the improved performance of the DDD based on a side by side comparison to a CCD camera, and the ability of the DDD to provide images suitable for reconstructing 3D electron density maps containing high resolution features.
The GroEL specimen was generously provided by Eli Chapman, Scripps Research Institute. Grid preparation and freezing conditions were similar to those reported previously (Stagg et al., 2006). All experiments were carried out on a Tecnai F20 Twin transmission electron microscope operated at 200 KeV, with a 50μm C2 aperture and a 100μm objective aperture. Prior to data acquisition, the instrument was adjusted for coma-free alignment of less than 0.1 mrad. A Gatan side-entry 626 cryo stage was used to maintain the specimen at a temperature below −170 °C.
The two cameras compared in this experiment were a Tietz F415 4K × 4K pixel (15μm pixel pitch) CCD camera and a retractable Direct Electron DE12 3K × 4K pixel (6μm pixel pitch) DDD camera. Nominal TEM magnification used for the acquisition was 62,000X for the CCD and 29,000X for the DDD, resulting in pixel sizes at the specimen of 1.37Å (CCD) and 1.38Å (DDD). For practical reasons the DDD images were stored as integrated 3K × 3K frames, with each image composed of a sum of 10 individual frames acquired at a rate of 40 frames/sec.
Data were collected using Leginon (Suloway et al., 2005
) in two separate experiments. The first experiment acquired data only to the DDD. In the second experiment, Leginon was used to randomly select either the DDD or the CCD cameras for acquiring images of targeted holes. Thus, the same grid was used under essentially identical conditions to acquire images on the two devices. Image exposure time was 400 ms, total dose per image was ~20e−
, defocus was randomly varied in a range from 0.8μm to 2.5μm.
A total of 1293 and 2966 DDD images were acquired in the first and second experiments, respectively. In the comparison experiment, 879 CCD images were acquired; fewer CCD images were required to achieve an equivalent number of particles from the CCD and DD images due to the larger field of view of the CCD camera.
After rejecting unsuitable micrographs, a total of 583 images from the CCD and 2313 images from the DDD were available for the comparison measurement, and 1981 images for the combined DDD reconstruction. Appion (Lander et al., 2009
) was used in all processing steps prior to projection-matching refinement. CTF estimation was performed using CTFFind (Mindell and Grigorieff, 2003
); particles were selected with FindEM (Roseman, 2004
) using templates of side, top and tilted views. Some additional filtering steps were used to reject classes of selected objects that were obviously not GroEL particles. Stacks were created by boxing out the selected particles from the micrographs and flipping the phases using the CTFFIND estimations. The initial model used for the reconstructions was a GroEL map low pass filtered to 30Å. XMIPP was used for the projection refinement analysis (Sorzano et al., 2004
) with D7 symmetry imposed; no masking was applied during the projection refinement. 14 iterations were performed with an inner radius for rotational correlation of 0 pixels and an outer radius for rotational correlation of 80 pixels. The angular distance between neighboring projection points for the first 4 iterations was 10 degrees. The next 10 iterations had the angular distance decremented to 5, 3, 2, 1, and 0.5 degrees every 2 iterations. The maximum change in rotation and tilt was unlimited for the first 4 iterations and constrained to 20 degrees for next 2 iterations, 9 degrees for the next 2 iterations, and 6 degrees for the final 6 iterations. The XMIPP weighted back projection method was used for the 3D reconstruction. FSC0.5
was calculated from even-odd volumes. Amplitude correction was applied to the final volumes using X-ray scattering data from GroEL as implemented in SPIDER where an enhancement curve from the X-ray data is applied as a filter to the EM volume (Frank et al., 1996
). After amplitude correction, a Gaussian low pass filter was applied. Data were visualized using Chimera (Pettersen et al., 2004
The cryo-EM density raw and post processed maps from the CCD camera micrographs are available from the EMDB as EMDB ID 5336 and EMDB ID 5340, respectively. The cryo-EM density raw and post processed maps from the DDD camera to compare to the CCD camera are available from the EMDB as EMDB ID 5337 and EMDB ID 5339, respectively. The cryo-EM density map from the combined DDD reconstruction is available from the EMDB as EMDB ID 5338.