Snapshots of single particles from single cells using electron microscopy

Cryo-electron microscopy has become an indispensable tool for structural studies of biological macromolecules. There are two predominant methods for studying the architectures of multi-protein complexes: (1) single particle analysis of purified samples and (2) tomography of whole cells or cell sections. The former can produce high-resolution structures but is limited to highly purified samples, while the latter can capture proteins in their native state but is hindered by a low signal-to-noise ratio and results in lower-resolution structures. Here, we present a method combining microfluidic single cell extraction with single particle analysis by electron microscopy to characterize protein complexes from individual C. elegans embryos. Using this approach, we uncover three-dimensional structures of ribosomes directly from single embryo extracts. In addition, we investigate structural dynamics during development by counting the number of ribosomes per polysome in early and late embyros. This approach has significant potential applications for counting protein complexes and studying protein architectures from single cells in developmental, evolutionary and disease contexts.


26
Cell behavior is fundamentally dependent on the activities of macromolecular machines. These 27 machines, comprised of protein (and sometimes RNA) subunits, are responsible for catalytic, 28 structural and regulatory activities that allow cells to function. Structural biology, by revealing the 29 physical architecture of macromolecules and their assemblies, plays a critical role in efforts to 30 understand how molecular mechanisms contribute to cell behavior in vivo.

31
A crucial feature of most living cells is their ability to adjust their behavior in response to 32 their environment. In a developmental context, cells respond to chemical and mechanical cues

81
To reconcile single cell and single particle methods, we propose an alternative approach 82 which combines single cell lysis with EM to investigate individual C. elegans embryos. Using this 83 method, we are able to directly visualize the contents of a single C. elegans cell (the zygote). After 84 computationally classifying the particles from cell lysate, we uncover the structures of 40S and 85 60S ribosomes from disperse particles and the structure of an 80S ribosome from polysomes.

86
This approach additionally enables us to count individual particles from embryos at specific 87 developmental stages. In one application, we find that the number of ribosomes per polysome 88 remains consistent between early and late stage embryos. These results demonstrate the 89 potential of EM for structural characterization of unpurified macromolecular machines obtained 90 from samples as small as a single cell.

93
Extracting macromolecules from single embryos

94
Our primary goal in this study was to determine whether imaging of single cell lysates with EM 95 could yield sufficient, high-quality particles for 3D structure determination. To obtain intact, native 96 particles from single cells, C. elegans zygotes (i.e., 1-cell embryos) were trapped and lysed using 97 microfluidic chambers (Figure 1) (see Methods). We then transferred the cell lysates (a volume 98 of ~50 nL) from the microfluidic channels to EM grids using a glass needle (Supplemental Movie 99 1) (see Methods). Due to the small volume, which was insufficient to coat an entire grid, we used 6 reference grids containing alphanumeric markers to locate the placement of our samples under 101 both the dissecting scope and the electron microscope. Each reference grid was then 102 conventionally stained using 2% (w/v) uranyl acetate. We chose to use negative stain EM for its 103 high-signal-to-noise ratio in order to more accurately assess our ability to identify single particles 104 from individual cell lysates. Each grid was then examined by transmission EM to identify grid 105 squares that contained cellular protein particles embedded in stain.

106
To demonstrate our ability to capture small volumes of samples on EM grids, we first 107 transferred samples of a purified protein kinase from our microfluidic device to an EM grid for 108 visualization, resulting in successful detection of the kinase ( Figure S1). We then performed our  results unambiguously show that we were able to retrieve cellular contents from our microfluidic 120 lysis chips for subsequent imaging by EM (Figure 2A). We collected ~1,400 micrographs between 121 the 7 samples. While small particles were abundant in our micrographs ( Figure S1), we first chose 122 to analyze large particles (~150-300 Å in diameter), which were easily recognizable and appeared 123 relatively homogeneous. After manually selecting ~10,000 large particles, we generated 124 reference-free 2D class averages, which were subsequently used as templates for picking the 7 remainder of the particles. Using this template picking scheme, ~80,000 large particles were 126 selected from ~1,400 micrographs and used for reference-free 2D alignment and classification.

127
2D class averages with distinct structural features were generated from ~50,000 particles after 128 removing junk particles (e.g. detergent micelles, irregular small particles, two nearby particles, or

163
We then performed 3D classification of our large particles to determine if any distinct structures 164 could be obtained from lysates of single cells. Specifically, we were looking for structures of 165 ribosomes since they appeared as clear and abundant 2D class averages in our data. We first 166 combined two datasets from early-stage embryo samples for 3D classification using RELION 167 (Scheres, 2012). After removal of junk particles, ~14,000 particles were used for classification.

168
Initially, we used an unbiased approach for 3D classification by using an initial model of a 169 featureless 3D shape with uniform electron density. Using a model reconstructed from this initial 170 classification that resembled a previously determined 60S ribosome structure (Shen et al., 2015) 171 (EMDB-2811) as a reference, we then performed another round of 3D classification (see

189
With clear structures resembling a 40S and 60S ribosome, we next attempted to determine 190 the molecular architecture of the 80S ribosome directly from polysome clusters. Our data 191 contained ~9,000 particles within polysomes which were manually picked for single particle 192 analysis. We then performed 2D and 3D classification of the selected particles (see Methods).

193
The 2D class averages were approximately 250 Å in diameter, which is consistent with the size

209
Moving forward, a significant challenge will be to extend this approach beyond ribosomes, 210 to other macromolecular complexes. We focused here on ribosomes because they are large,      used for template-based particle picking with a template selected from reference-free 2D class 296 averages generated from ~10,000 large particles which were manually picked from the E1 dataset.

297
In total, ~81,600 particles were selected from template picking of all datasets. All image pre-14 processing was done in Appion (Lander et al., 2009). After removing junk particles, 17,070 299 particles remained for further processing. Particle box size was set to 576 Å x 576 Å. Reference-300 free 2D class averages were generated with 100 classes using RELION (Scheres, 2012

310
For our 60S ribosome reconstruction, a similar strategy was followed. Two independent 311 particle stacks from E1 and E2 were used. The contrast transfer function (CTF) of each 312 micrograph was estimated using CTFFIND4 (Rohou and Grigorieff, 2015). ~37,200 particles were 313 selected by template picking. After removing junk particles, 13,916 particles were left. Particle box 314 size was set to 432 Å x 432 Å. Reference-free 2D class averages were generated with 100 classes.

315
3D classification was performed to create 8 classes. The structure of a featureless 3D shape with 316 uniform electron density was chosen as an initial model after low-pass filtering to 60 Å. A 317 subsequent round of 3D classification was performed on the same data using a reconstructed 3D                    Single C. elegans embryos are trapped in a microfluidic device. After the embryo is crushed, the 434 lysate is extracted using a fine needle and applied to a specific area of a EM grid using a 435 stereoscope. The same area is then visualized using electron microscopy and single particle 436 analysis is applied for structure determination.