Faithful representation of similarities among three-dimensional shapes in human vision.

Efficient and reliable classification of visual stimuli requires that their representations reside a low-dimensional and, therefore, computationally manageable feature space. We investigated the ability of the human visual system to derive such representations from the sensory input-a highly nontrivial task, given the million or so dimensions of the visual signal at its entry point to the cortex. In a series of experiments, subjects were presented with sets of parametrically defined shapes; the points in the common high-dimensional parameter space corresponding to the individual shapes formed regular planar (two-dimensional) patterns such as a triangle, a square, etc. We then used multidimensional scaling to arrange the shapes in planar configurations, dictated by their experimentally determined perceived similarities. The resulting configurations closely resembled the original arrangements of the stimuli in the parameter space. This achievement of the human visual system was replicated by a computational model derived from a theory of object representation in the brain, according to which similarities between objects, and not the geometry of each object, need to be faithfully represented.

Projection structure of frog rhodopsin in two crystal forms.

Rhodopsin is the G protein-coupled receptor that upon light activation triggers the visual transduction cascade. Rod cell outer segment disc membranes were isolated from dark-adapted frog retinas and were extracted with Tween detergents to obtain two-dimensional rhodopsin crystals for electron crystallography. When Tween 80 was used, tubular structures with a p2 lattice (a = 32 A, b = 83 A, gamma = 91 degrees) were formed. The use of a Tween 80/Tween 20 mixture favored the formation of larger p22(1)2(1) lattices (a = 40 A, b = 146 A, gamma = 90 degrees). Micrographs from frozen hydrated frog rhodopsin crystals were processed, and projection structures to 7-A resolution for the p22(1)2(1) form and to 6-A resolution for the p2 form were calculated. The maps of frog rhodopsin in both crystal forms are very similar to the 9-A map obtained previously for bovine rhodopsin and show that the arrangement of the helices is the same. In a tentative topographic model, helices 4, 6, and 7 are nearly perpendicular to the plane of the membrane. In the higher-resolution projection maps of frog rhodopsin, helix 5 looks more tilted than it appeared previously. The quality of the two frog rhodopsin crystals suggests that they would be suitable to obtain a three-dimensional structure in which all helices would be resolved.

High-Voltage Electron Tomography of Spindle Pole Bodies and Early Mitotic Spindles in the Yeast Saccharomyces cerevisiaeV⃞

The spindle pole body (SPB) is the major microtubule-organizing center of budding yeast and is the functional equivalent of the centrosome in higher eukaryotic cells. We used fast-frozen, freeze-substituted cells in conjunction with high-voltage electron tomography to study the fine structure of the SPB and the events of early spindle formation. Individual structures were imaged at 5–10 nm resolution in three dimensions, significantly better than can be achieved by serial section electron microscopy. The SPB is organized in distinct but coupled layers, two of which show ordered two-dimensional packing. The SPB central plaque is anchored in the nuclear envelope with hook-like structures. The minus ends of nuclear microtubules (MTs) are capped and are tethered to the SPB inner plaque, whereas the majority of MT plus ends show a distinct flaring. Unbudded cells containing a single SPB retain 16 MTs, enough to attach to each of the expected 16 chromosomes. Their median length is ∼150 nm. MTs growing from duplicated but not separated SPBs have a median length of ∼130 nm and interdigitate over the bridge that connects the SPBs. As a bipolar spindle is formed, the median MT length increases to ∼300 nm and then decreases to ∼30 nm in late anaphase. Three-dimensional models confirm that there is no conventional metaphase and that anaphase A occurs. These studies complement and extend what is known about the three-dimensional structure of the yeast mitotic spindle and further our understanding of the organization of the SPB in intact cells.

Cell-Matrix Adhesions Differentially Regulate Fascin Phosphorylation

Cell adhesion to individual macromolecules of the extracellular matrix has dramatic effects on the subcellular localization of the actin-bundling protein fascin and on the ability of cells to form stable fascin microspikes. The actin-binding activity of fascin is down-regulated by phosphorylation, and we used two differentiated cell types, C2C12 skeletal myoblasts and LLC-PK1 kidney epithelial cells, to examine the hypothesis that cell adhesion to the matrix components fibronectin, laminin-1, and thrombospondin-1 differentially regulates fascin phosphorylation. In both cell types, treatment with the PKC activator 12-tetradecanoyl phorbol 13-acetate (TPA) or adhesion to fibronectin led to a diffuse distribution of fascin after 1 h. C2C12 cells contain the PKC family members α, γ, and λ, and PKCα localization was altered upon cell adhesion to fibronectin. Two-dimensional isoelectric focusing/SDS-polyacrylamide gels were used to determine that fascin became phosphorylated in cells adherent to fibronectin and was inhibited by the PKC inhibitors calphostin C and chelerythrine chloride. Phosphorylation of fascin was not detected in cells adherent to thrombospondin-1 or to laminin-1. LLC-PK1 cells expressing green fluorescent protein (GFP)-fascin also displayed similar regulation of fascin phosphorylation. LLC-PK1 cells expressing GFP-fascin S39A, a nonphosphorylatable mutant, did not undergo spreading and focal contact organization on fibronectin, whereas cells expressing a GFP-fascin S39D mutant with constitutive negative charge spread more extensively than wild-type cells. In contrast, C2C12 cells coexpressing S39A fascin with endogenous fascin remained competent to form microspikes on thrombospondin-1, and cells that expressed fascin S39D attached to thrombospondin-1 but did not form microspikes. Blockade of PKCα activity by TPA-induced down-regulation led to actin association of wild-type fascin in fibronectin-adherent C2C12 and LLC-PK1 cells but did not alter the distribution of S39A or S39D fascins. The association of fascin with actin in fibronectin-adherent cells was also evident in the presence of an inhibitory antibody to integrin α5 subunit. These novel results establish matrix-initiated PKC-dependent regulation of fascin phosphorylation at serine 39 as a mechanism whereby matrix adhesion is coupled to the organization of cytoskeletal structure.

Three-dimensional modeling of and ligand docking to vitamin D receptor ligand binding domain

The ligand binding domain of the human vitamin D receptor (VDR) was modeled based on the crystal structure of the retinoic acid receptor. The ligand binding pocket of our VDR model is spacious at the helix 11 site and confined at the β-turn site. The ligand 1α,25-dihydroxyvitamin D3 was assumed to be anchored in the ligand binding pocket with its side chain heading to helix 11 (site 2) and the A-ring toward the β-turn (site 1). Three residues forming hydrogen bonds with the functionally important 1α- and 25-hydroxyl groups of 1α,25-dihydroxyvitamin D3 were identified and confirmed by mutational analysis: the 1α-hydroxyl group is forming pincer-type hydrogen bonds with S237 and R274 and the 25-hydroxyl group is interacting with H397. Docking potential for various ligands to the VDR model was examined, and the results are in good agreement with our previous three-dimensional structure-function theory.

Three-dimensional image reconstruction of dephosphorylated smooth muscle heavy meromyosin reveals asymmetry in the interaction between myosin heads and placement of subfragment 2

Regulation of the actin-activated ATPase of smooth muscle myosin II is known to involve an interaction between the two heads that is controlled by phosphorylation of the regulatory light chain. However, the three-dimensional structure of this inactivated form has been unknown. We have used a lipid monolayer to obtain two-dimensional crystalline arrays of the unphosphorylated inactive form of smooth muscle heavy meromyosin suitable for structural studies by electron cryomicroscopy of unstained, frozen-hydrated specimens. The three-dimensional structure reveals an asymmetric interaction between the two myosin heads. The ATPase activity of one head is sterically “blocked” because part of its actin-binding interface is positioned onto the converter domain of the second head. ATPase activity of the second head, which can bind actin, appears to be inhibited through stabilization of converter domain movements needed to release phosphate and achieve strong actin binding. When the subfragment 2 domain of heavy meromyosin is oriented as it would be in an actomyosin filament lattice, the position of the heads is very different from that needed to bind actin, suggesting an additional contribution to ATPase inhibition in situ.

Novel phenomena in dilute electron systems in two dimensions

For the past two decades, all two-dimensional systems of electrons were believed to be insulating in the limit of zero temperature. Recent experiments provide evidence for an unexpected transition to a conducting phase at very low electron densities. The nature of this phase is not understood and is currently the focus of intense theoretical and experimental attention.

Three-dimensional structure of a cysteine-rich repeat from the low-density lipoprotein receptor.

The low-density lipoprotein (LDL) receptor plays a central role in mammalian cholesterol metabolism, clearing lipoproteins which bear apolipoproteins E and B-100 from plasma. Mutations in this molecule are associated with familial hypercholesterolemia, a condition which leads to an elevated plasma cholesterol concentration and accelerated atherosclerosis. The N-terminal segment of the LDL receptor contains a heptad of cysteine-rich repeats that bind the lipoproteins. Similar repeats are present in related receptors, including the very low-density lipoprotein receptor and the LDL receptor-related protein/alpha 2-macroglobulin receptor, and in proteins which are functionally unrelated, such as the C9 component of complement. The first repeat of the human LDL receptor has been expressed in Escherichia coli as a glutathione S-transferase fusion protein, and the cleaved and purified receptor module has been shown to fold to a single, fully oxidized form that is recognized by the monoclonal antibody IgG-C7 in the presence of calcium ions. The three-dimensional structure of this module has been determined by two-dimensional NMR spectroscopy and shown to consist of a beta-hairpin structure, followed by a series of beta turns. Many of the side chains of the acidic residues, including the highly conserved Ser-Asp-Glu triad, are clustered on one face of the module. To our knowledge, this structure has not previously been described in any other protein and may represent a structural paradigm both for the other modules in the LDL receptor and for the homologous domains of several other proteins. Calcium ions had only minor effects on the CD spectrum and no effect on the 1H NMR spectrum of the repeat, suggesting that they induce no significant conformational change.