Cryo-electron microscopy (cryoEM) is an increasingly popular method for protein structure determination. However, identifying a sufficient number of particles for analysis (often >100,000) can take months of manual effort. Current computational approaches are limited by high false positive rates and require significant ad-hoc post-processing, especially for unusually shaped particles. To address this shortcoming, we develop Topaz, an efficient and accurate particle picking pipeline using neural networks trained with few labeled particles by newly leveraging the remaining unlabeled particles through the framework of positive-unlabeled (PU) learning. Remarkably, despite using minimal labeled particles, Topaz allows us to improve reconstruction resolution by up to 0.15 Å over published particles on three public cryoEM datasets without any post-processing. Furthermore, we show that our novel generalizedexpectation criteria approach to PU learning outperforms existing general PU learning approaches when applied to particle detection, especially for challenging datasets of nonglobular proteins. We expect Topaz to be an essential component of cryoEM analysis.

Summary Adherens junctions, which play a central role in intercellular adhesion, comprise clusters of type I classical cadherins that bind via extracellular domains extended from opposing cell surfaces. We show that a molecular layer seen in crystal structures of E- and N-cadherin ectodomains reported here and in the C-cadherin structure corresponds to the extracellular architecture of adherens junctions. In all three ectodomain crystals, cadherins dimerize through a trans adhesive interface and are connected by a second, cis, interface. Assemblies formed by E-cadherin ectodomains coated on liposomes also appear to adopt this structure. Fluorescent imaging of junctions formed from wild-type and mutant E-cadherins in cultured cells confirm conclusions derived from structural evidence. Mutations that interfere with the trans interface ablate adhesion, whereas cis interface mutations disrupt stable junction formation. Our observations are consistent with a model for junction assembly involving strong trans and weak cis interactions localized in the ectodomain.

Crystal structures of classical cadherins have revealed two dimeric configurations: in the first, Nterminal β-strands of EC1 domains "swap" between partner molecules. The second configuration (the "X-dimer"), also observed for T-cadherin, is mediated by residues near the EC1-2 calcium binding sites, and N-terminal β-strands of partner EC1 domains, though held adjacent, do not swap. Here we show that strand swapping mutants of type I and II classical cadherins form X-dimers. Mutant cadherins impaired for X-dimer formation show no binding in short timeframe surface plasmon resonance assays but in long timeframe experiments, have homophilic binding affinities close to wild-type. Further experiments show that exchange between monomers and dimers is slowed in these mutants. These results reconcile apparently disparate results from prior structural studies, and suggest that X-dimers are binding intermediates that facilitate the formation of strand swapped dimers.

Cadherins embody a superfamily of cell-surface glycoproteins whose ectodomains contain multiple repeats of β-sandwich EC (extracellular cadherin) domains that adopt a similar fold to immunoglobulin domains. The best characterized cadherins are the vertebrate “classical” cadherins, which mediate adhesion via trans homodimerization between their membrane-distal EC1 domains that extend from apposed cells, and assemble intercellular adherens junctions through cis clustering. To form mature trans adhesive dimers, cadherin domains from apposed cells dimerize in a “strand-swapped” conformation. This occurs in a two-step binding process involving a fast-binding intermediate called the “X-dimer”. Trans dimers are less flexible than cadherin monomers, a factor which drives junction assembly following cell-cell contact by reducing the entropic cost associated with the formation of lateral cis oligomers. Cadherins outside of the classical subfamily appear to have evolved distinct adhesive mechanisms which are just now beginning to be understood.