3Corresponding author Communicated by R.Henderson Two-dimensional crystals from light-harvesting complex I (LHC I) of the purple non-sulfur bacterium Rhodospirillum rubrum have been reconstituted from detergent-solubilized protein complexes. Frozenhydrated samples have been analysed by electron microscopy. The crystals diffract beyond 8 A and a projection map was calculated to 8.5 A. The projection map shows 16 subunits in a 116 A diameter ring with a 68 A hole in the centre. These dimensions are sufficient to incorporate a reaction centre in vivo.Within each subunit, density for the a-and the Ppolypeptide chains is clearly resolved, and the density for the bacteriochlorophylls can be assigned. The experimentally determined structure contradicts models of the LHC I presented so far.

Recombinant membrane proteins in Escherichia coli are either expressed at relatively low level in the cytoplasmic membrane or they accumulate as inclusion bodies. Here, we report that the abundant over-production of subunit b of E. coli

Scanning force microscopy allows imaging of biological molecules in their native state in buffer solution. To this end samples have to be fixed to a flat solid support so that they cannot be displaced by the scanning tip. Here we describe a method to achieve the covalent binding of biological samples to glass surfaces. Coverslips were chemically modified with the photoactivatable cross-linker N-5-azido-2-nitrobenzoyloxysuccinimide. Samples are squeezed between derivatized coverslips and then cross-linked to the glass surface by irradiation with ultraviolet light. Such samples can be imaged repeatedly by the scanning force microscope without loss of image quality, whereas identical but not immobilized samples are pushed away by the stylus.

The atomic force microscope has the potential to monitor structural changsof a biological system In Its native environment. To For electron microscopy a small drop of 2% cadmium thioglycerol (9) was first applied to a carbon-coated grid made hydrophilic by glow-discharge. After 1 min, excess liquid was withdrawn and a small drop of the HPI layer suspension (1 mg/ml) was deposited. Sheets were allowed to adsorb for 1-2 min before most of the liquid was removed. Finally, another droplet of the cadmium thioglycerol solution was applied to ensure complete embedding of the specimen. All electron micrographs were recorded with a Philips EM 420 using low-dose procedures (9). Because cadmium thioglycerol is radiation-sensitive, the three-dimensional data set was collected over a tilt angle range between 0°and 80°in the form ofmany mini-tilt series, each ofthem comprising three to four micrographs, thus exposing the samples to a cumulative dose which was <4500 electrons per nm2.The Semper image-processing system was used to calculate correlation averages from AFM images and electron micrographs (10). Seventeen mini-tilt series were combined in performing the three-dimensional reconstruction. The micrographs were sorted into a "left-handed" and a "righthanded" data set corresponding to the two different orientations of the layer with respect to the supporting film; this differentiation was based on the handedness ofthe diffraction pattern. Identical areas within the mini-tilt series were digitized into 2048 by 2048 pixels of a size of 0.204 nm on the object by means of a Joyce-Loebl flat-bed densitometer. Correlation averages containing between 500 and 800 unit cells each were extracted from these projections. Effective tilt angle and tilt axis azimuth were evaluated from the base vectors of the projected lattices. Finally, the threedimensional density distribution of the HPI layer was calculated from these data by aligning and normalizing individual projections, eliminating those that exhibited a correlation coefficient of <0.9 to neighboring projections (11). The two individual three-dimensional maps thus comprised 25 and 27 projections, respectively. The HPI hexamer was surface rendered at 100% mass by assuming a protein density of 810 Da/nm3, and the topography ofthe side seen by the AFM was extracted.For quantitative evaluation of the AFM data the averaged topography composed of60 HPI hexamers was normalized to the layer thickness (7) and aligned with the topography ofthe HPI layer as obtained from electron microscopy to calculate the modulus of the difference. Further, the radial correlation function was determined from two independent averages to assess the lateral resolution (10), whereas the root-meanAbbreviations: AFM, atomic force microscope; HPI layer, hexagonally packed intermediate layer. 836The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact.