Part 3 Materials and Methods (cont.)

Electron Microscopy and Single Particle Analyses-The samples were negatively stained with 2% uranyl acetate by the droplet method and imaged using a Philips CM100 electron microscope (EM) at 80 kV. Phase contrast was introduced by defocusing the microscope whereby the first minima of the contrast transfer function (23) was subsequently calculated to fall consistently in the range of 18-22 Å. Subsequent image analyses were conducted using the best 23 micrographs, imaged at 51,500. and displaying minimal or no discernible astigmatism and drift, of the C. reinhardtii LHCI-PSI supercomplex fraction. In addition, 11 micrographs for the PSI monomer/PSII dimer fraction were selected through similar criteria. The micrographs were digitized using a Leafscan 45 densitometer at a step-size of 10 .m (1.94 Å per pixel on the specimen scale). All image processing was performed within the Imagic-5 software environment (24, 25), running under the Linux operating system on dual processor PC computers. For the LHCI-PSI supercomplex a data set of 8,950 single particle images was obtained by interactively picking all discernible particles that were not overlapping or in close contact with other particles, and each image was floated into a 96 . 96 pixel box. A similar data set of 3300 images was picked from micrographs of the C. reinhardtii PSI/PSII fraction. Images of the particles were coarsened by a factor of 2-3.88 Å per pixel to aid the speed of subsequent processing. An analysis of these two data sets was made starting with the reference-free alignment-by-classification procedure (26). This identified several particle subpopulations of differing size and shape in each preparation. These subpopulations were in turn analyzed independently, with the reference-free alignment giving the initial class averages necessary for multireference alignments, eventually leading to averages with enhanced signal to noise ratios after iterative refinement (23, 25). Relative orientations were determined for the class averages by the angular reconstitution technique (27), resulting in initial three-dimensional reconstructions gained from implementation of the exact back projection technique (28). Reprojections were taken from each three-dimensional model and used to identify additional atypical views and to further refine the class averages within each subpopulation data set. The resolution was estimated by Fourier shell correlation between two independent three-dimensional reconstructions (23). PDB coordinate data sets derived for a monomer of LHCII, a kind gift of W. Ku¨ hlbrandt, Frankfurt, Germany, calculated at 3.4 Å resolution by electron crystallography (4) and the 2.5 Å resolution PSI monomer (PDB accession number 1JB0; http://www.rcsb.org) from the cyanobacterium S. <i>elongatus</i> (2) were modeled into the reconstructions using the "O" modeling software package (29).