Fine Structure of a Bacterial Cell Envelope Protein: The RS-Layer of Sporosarcina ureae W. Baumeister, B. Emde, G. Flaskamp, M. Hahn, and R. Rachel Institut ffir Biophysik und Elektronenmikroskopie der Universitfit, D-4000 Dfisseldorf W.O. Saxton High Resolution Electron Microscope, University of Cambridge, Cambridge, CB2 3RQ, England Surface layers of regularly arrayed protein units (S- or RS-layers) have now been de- scribed for a multitude of bacterial species (see e.g. [1]). Although it is evident now that these layers are a rather common com- ponent of bacterial cell envelopes, their functional role is still merely a matter of speculation : protection against adverse en- vironmental factors such as lytic enzymes, retention of endogenous enzymes, and mo- lecular sieving, i.e. controlling the passage of small and medium-sized molecules, are amongst the functions attributed to RS- layers. Recently the existence of an RS- layer on cell envelopes of Sporosareina ure- ae has been established and its morphology studied on isolated cell wall fragments [2, 3]. Here we try to give a more detailed description of the structure of this RS- layer protein and in attempt to mark gen- eral design features we compare it with other bacterial surface layers. RS-layer fragments were isolated from Sporosarcina ureae (ATCC 13881) follow- ing essentially the procedure described in [2] and negatively stained with sodium-sili- cotungstate/uranyl acetate for electron microscopy (Fig. 1a). To recover clear im- ages free of spurious details micrographs of suitably stained crystalline patches (i.e. patches completely embedded in stain) were processed digitally employing the novel'correlation averaging' method [4, 5]. In essence, the positions of individual mo- lecular images are determined through their correlation with a reference image - a small patch of crystal, preferably as small as a single unit cell - which is moved around the whole field of view and com- pared with local image detail by cross-cor- relation. Rough molecular positions indi- cated by local correlation peaks are refined by peak centre-of-mass determinations and averages are then produced by superposing regions centred on the final peak positions. To avoid any bias towards the reference chosen the whole process is usually iterated using the first average as a reference in its turn. Correlation averaging, which can easily be implemented as a fully automatic procedure, has, we feel, several advantages over conventional Fourier-space filtering: it avoids the loss of resolution encountered with the conventional method when the specimen is not perfectly crystalline since it compensates for lattice distortions, it provides a convenient means of extending averages over several small and irregularly shaped crystal fragments and it allows the selection of the 'best' (i.e. the least dis- torted) molecules only for inclusion in an average using various criteria (correlation peak levels, low local strains) for their se- lection [6], Fig. lb shows a correlation average ob- tained from the boxed area of the RS-layer Fig. 1. (a) Isolated sheet of negatively stained RS-layer. The boxed area was used to produce the correlation average shown in (b), which includes a 4-fold rotational average Fig. 2. Electrophoresis of purified RS-layer preparations on SDS-polyacrylamide gels (lin- ear gradient from 7.5 to 15% and discontin- uous buffer system according to [11]). From left to right: RS-layer prepared at 20 °C show- ing extensiveautolytic digestion, RS-layer pre- pared in the cold showing a single band at 120000 dalton, and marker proteins shown in Fig. 1a. The tetragonal lattice (P4 symmetry) has a lattice constant of 12.8_+0.3 nm, and the unit cell is likely to be composed of 4 identical polypeptides with a molecular weight of 120000 dalton each (see Fig. 2). The most prominent stain-excluding feature is a tetragonal 'core' (diameter~Tnm) with a central stain-filled pit possibly the vestibule of a pore. Four centrcs of mass can be dis- cerned in the corners of the core, which further outwards smoothly changes from a tetragonal to an octagonal shape. Spokes emanate from the corners of the octagon so that pairs of them interconnect neigh- bouring cores in the lattice. They divide the periphery into domains heavily filled with staining material: only small amounts of protein mass appear to protrude into these interspaces creating a quite compli- cated dividing pattern. Remarkably, the te- tragonal core is positioned slightly askew to the lattice lines, the exact skew angle varying, however, from one crystal patch to another; concomitantly the amount of negative stain deposited in the central pit appears to change. Hence, it is tempting to speculate whether a change of the skew angle can actually influence the geometry of the pore. Systematic screening of several RS-layer sheets is necessary to verify or 626 Naturwissenschaften 68 (1981) © Springer-Verlag 1981