Label-free single-instance protein detection in vitrified cells

Abstract

A general method to map molecular interactions and conformational states in structurally intact cells would find wide application in biochemistry and cell biology. We used a library of imagescalculated on the basis of known structural data-as search templates to detect targets as small as the "head" domain (350 kDa) of the ribosome's small subunit in single-tilt electron cryomicrographs by cellular high resolution template matching (cHRTM). Atomically precise position and orientation estimates reveal the conformation of individual ribosomes and enable the detection of specifically bound ligands down to 24 kDa. We show that highly head-swivelled states are likely to play a role in mRNA translocation in living cells. cHRTM outperforms cryoelectron tomography three-fold in sensitivity and completely avoids the vicissitudes of exogenous labelling. Main:Finding molecular components in intact cells has driven the development of biological imaging techniques, some of which use specific labels, such as fluorescent antibodies, while others rely on a target's intrinsic properties such as its shape (Oliver 1973) and structure (Rickgauer, Grigorieff, and Denk 2017). Labelling is frequently incomplete or nonspecific, can alter the target's interactions with other molecules, provides only moderate precision of localization and

Locating Macromolecular Assemblies in Cells by 2D Template Matching withcisTEM

“…In contrast, the proportion of the 3DTM targets detected in the 2DTM search was variable and showed a negative correlation with sample thickness (Figure 5e). This is consistent with our and prior observations that 2DTM is particularly sensitive to sample thickness (Figure 5 -figure supplement 2) (Rickgauer et al, 2017(Rickgauer et al, , 2020, which likely contributes to the high false negative rate of 2DTM relative to 3DTM.…”

“…The simplifications, as well as deformation of the M. pneumoniae cells under the electron beam (Tegunov et al, 2021), leading to noticeable blurring in some of our images, likely affect the SNR values obtained in our template searches. Indeed, the SNR values we observe are lower on average than expected based on the molecular mass of our template (about 1.2 MDa), as well as previously observed values obtained with different samples and template structures (Rickgauer et al, 2017(Rickgauer et al, & 2020. Furthermore, the M. pneumoniae 50S atomic model built into a 3.5-Å map (Tegunov et al, 2021) may contain atomic coordinate errors that are larger than those in the 60S and 40S models used by Rickgauer et al (2020), which were built into 2.9-Å maps.…”

“…Indeed, the SNR values we observe are lower on average than expected based on the molecular mass of our template (about 1.2 MDa), as well as previously observed values obtained with different samples and template structures (Rickgauer et al, 2017(Rickgauer et al, & 2020. Furthermore, the M. pneumoniae 50S atomic model built into a 3.5-Å map (Tegunov et al, 2021) may contain atomic coordinate errors that are larger than those in the 60S and 40S models used by Rickgauer et al (2020), which were built into 2.9-Å maps. Coordinate errors will further decrease the SNR values obtained in a template search.…”

“…We conclude that including additional frames with exposure weighting does not diminish the 2DTM SNR, indicating the constructive contribution of lower-resolution features to the detected signal. Consistent with SNR values being sensitive to defocus errors (Rickgauer et al, 2017(Rickgauer et al, , 2020, searching each image with templates sampling a defocus range of 2400 Å in 200 Å steps (13 defocus planes) increased the number of detected targets by ~40% in a 100 nm thick sample and ~400% in a 220 nm thick sample (Figure 3 -figure supplement 1c). In addition to increasing the number of detected targets, the defocus search also provides a rough estimate of the out-of-plane position of each 50S.…”

Locating macromolecular assemblies in cells by 2D template matching with cisTEM

“…The large search space required for 2DTM makes this method computationally demanding; a single 1850 x pixel image required 1000 CPU-hours for a search in the proof-of-principle MATLAB implementation (Rickgauer et al, 2017) and was improved ~2-fold by GPUacceleration (Rickgauer et al, 2020) when comparing hardware of similar purchase price. To make 2DTM accessible to more users and a broader range of biological questions, we implemented 2DTM in cisTEM (Grant et al, 2018) using C++ (Figure 1a).…”