Multicolor single particle reconstruction of protein complexes

Abstract

13Single-particle reconstruction (SPR) from electron microscopy images is widely used in 14 structural biology, but lacks direct information on protein identity. To address this limitation, we 15 developed a computational and analytical framework that reconstructs and co-aligns multiple 16 proteins from 2D super-resolution fluorescence images. We demonstrate our method by 17 generating multi-color 3D reconstructions of several proteins within the human centriole and 18 procentriole, revealing their relative locations, dimensions and orientations. 22 or 23 Christian Sieben, christian.sieben@epfl.ch 24 25 2Macromolecular complexes within cells usually contain multiple protein species, whose precise 26 arrangement is required to generate properly functioning molecular machines. Single particle 27 analysis of electron microscopy (EM) images has been used to build 3D reconstructions of such 28 complexes, recently with near-atomic resolution 1,2 . To deduce the spatial organization of specific 29 proteins, computational methods have been used to dock structures deciphered from X-ray 30 crystallography or NMR within 3D reconstructed particles 1,3 . Alternatively, immunogold or 31 nanobody labelling can reveal the location and conformational state of target proteins 4,5 , 32 whereas electron density map differences can provide information on the position of mutated or 33 missing proteins 6 . Nevertheless, it remains challenging to locate native proteins within 3D 34 reconstructions, which is essential for deciphering the architecture underlying the assembly 35 mechanisms and functional modules of macromolecular complexes. 36Fluorescence-based single-molecule localization microscopy (SMLM) can help to 37 address this challenge, as demonstrated for the nuclear pore complex using 2D averaging 7 . 38Extending to 3D, a recent implementation of single-particle reconstruction (SPR) from 2D SMLM 39 images demonstrated isotropic reconstruction on DNA origami and simulated data 8 . However, 40 multi-color particle reconstruction of actual macromolecular complexes requires generating 41 large image libraries of multiple proteins and solving the complex problem of 3D multi-channel 42 alignment. Here, we developed a systematic framework that addresses both of these 43 challenges. We used a dedicated high-throughput SMLM setup 9 to generate the large multi-44 color particle data sets required for SPR, which we then processed using a semi-automated 45 computational workflow to reconstruct and align multiple proteins onto a single 3D particle. We 46 applied our method to the human centriole, and adapted it to accommodate off-axis 47 structures, such as those important during procentriole formation. Centrioles are evolutionarily 48 conserved sub-diffraction limited cylindrical organelles that seed the formation of cilia, flagella 49 and centrosomes 10 . The mature human centriole comprises nine-fold symmetrically arranged 50 microtubule triplets and contains >100 different proteins organized into distinct 51 substructures 11 . For instan...

3D Structure From 2D Microscopy Images Using Deep Learning

“…After training with these images, our network converged on a central torus for the CEP152 complex (Figures 10,11). This inferred structure is consistent with the confirmed structure of this protein complex (Sieben et al, 2018;Kim et al, 2019).…”

“…The first dataset was a super-resolution (STORM) microscopy data-set of CEP152, obtained and analysed as described in Sieben et al (2018). The structure of this centro-symmetric complex has been fitted with a toroid and found to be 400 nm in diameter (Sieben et al, 2018), which subsequent work confirmed (Kim et al, 2019). This yielded a list of localisations for each identified CEP152 structure which were reconstructed localisations into 2D images, rendering with a Gaussian.…”

“…In order to evaluate HOLLy, we selected a number of groundtruth point-clouds with different characteristics: a reduced version of the Stanford Bunny 1 , the Utah Teapot 2 and an approximation of the CEP152/HsSAS-6 complex (Sieben et al, 2018).…”

“…The third point-cloud used in these experiments is an approximation of the CEP152/HsSAS-6 complex (Sieben et al, 2018). The approximation consisted of two cylinders, one smaller and perpendicular to the other.…”

Nanoscale structural organization and stoichiometry of the budding yeast kinetochore

“…Super-resolution microscopy, and specifically Single-Molecule Localization Microscopy (SMLM) (Betzig et al, 2006;Hess et al, 2006;Rust et al, 2006), achieves nanometer resolution combined with molecular specificity, and has the potential to bridge this gap in our knowledge. It has been used to get structural insights into the organization of multi-protein complexes such as the nuclear pore complex (Szymborska et al, 2013), the endocytic machinery (Mund et al, 2018;Sochacki et al, 2017), centrioles (Sieben et al, 2018) or synaptic proteins (Dani et al, 2010). In this study, we use SMLM to determine the location of key proteins and their copy numbers with single kinetochore resolution in S. cerevisiae cells (Figure 1).…”