Iron-sequestering nanocompartments as multiplexed Electron Microscopy gene reporters

Felix Sigmund, Susanne Pettinger, Massimo Kube, Fabian Schneider, Martina Schifferer, Michaela Aichler, Steffen Schneider, Axel Walch, Thomas Misgeld, Hendrik Dietz, Gil G. WestmeyerbioRxiv 2019 · DOI

TL;DR

Researchers have developed a novel system of genetically encoded, iron-sequestering protein nanospheres called "EMcapsulins" that act as multiplexed reporters for electron microscopy (EM). These nanospheres can be precisely engineered and visualized, overcoming limitations of existing EM labeling techniques and enabling automated, high-resolution imaging of cellular processes.

What Problem Does This Paper Address?

Visualizing cellular structures and processes at the ultrastructural level is crucial for understanding complex biological mechanisms. Electron microscopy (EM) offers unparalleled resolution, but current methods for labeling specific cellular components for EM are limited. Fluorescent proteins, while excellent for light microscopy, don't translate well to EM. Existing EM labeling techniques often rely on precipitating exogenous chemicals, which can introduce artifacts, limit spatial precision, and affect tissue penetration. Even genetically encoded iron storage proteins like ferritin, while providing contrast, are small and difficult to distinguish from endogenous structures. The challenge is to create genetically encoded, highly visible, and multiplexable reporters that can be reliably detected and differentiated using conventional EM.

Methods

The core innovation is the use of bacterial encapsulins, specifically from Quasibacillus thermotolerans (QtEnc) and Myxococcus xanthus (MxEnc). These proteins self-assemble into spherical compartments. The researchers engineered these encapsulins to encapsulate a native cargo protein (QtIMEF) that possesses ferroxidase activity. This activity allows the encapsulins to sequester iron from the cell, forming an electron-dense core that provides strong contrast in Transmission Electron Microscopy (TEM).

Key techniques employed include:

  • Cryo-electron microscopy (Cryo-EM): Used to determine the high-resolution structure of the QtEnc nanospheres, revealing a novel T=4 icosahedral symmetry and a diameter of ~44 nm.
  • Blue Native PAGE (BN-PAGE) and SDS-PAGE: Used to analyze protein assembly and cargo loading in mammalian cells.
  • Dynamic Light Scattering (DLS): To measure the size of the assembled nanospheres.
  • Ferroxidase assay: To quantify the iron biomineralization efficiency of the encapsulin systems.
  • Deep Learning (Convolutional Neural Networks): A pre-trained residual network was adapted and trained to automatically classify TEM images, distinguishing between the different EMcapsulin variants and ferritin.
  • Cellular TEM: Used to visualize the EMcapsulins within mammalian cells and assess their contrast and localization.

Datasets

The study utilized HEK293T cells as the mammalian cell line for expressing the EMcapsulin systems. For the deep learning classification, a dataset of 56 EM images was acquired. This dataset was used for training and evaluating a deep learning model, with a 5-fold cross-validation scheme. The authors also mention releasing the source code for their deep learning model, suggesting it could serve as a benchmark for future studies.

Key Insights and Ideas

  • Novel T=4 Encapsulin Structure: The discovery of the T=4 icosahedral symmetry in QtEnc, leading to a larger ~44 nm diameter, is a significant structural finding that differentiates it from other known encapsulins (typically T=1 or T=3).
  • Efficient Iron Sequestration: The QtIMEF cargo exhibits robust ferroxidase activity, enabling efficient iron biomineralization within the QtEnc shell, creating a highly visible EM contrast agent.
  • Multiplexing Potential: The distinct sizes (~44 nm for QtEnc and ~32 nm for MxEnc) and potentially different contrast levels of the two encapsulin systems allow for multiplexed imaging, meaning multiple reporters can be tracked simultaneously.
  • Automated EM Image Classification: The successful application of a deep learning model to automatically differentiate between EMcapsulin variants in TEM images is a major step towards high-throughput EM analysis.
  • Superiority over Ferritin: EMcapsulins provide significantly better contrast and are more easily differentiated than endogenous ferritin, which is often upregulated by iron supplementation, leading to background noise.

Key Results

  • Structural Characterization: Cryo-EM revealed QtEnc nanospheres with T=4 symmetry and a diameter of ~44 nm, distinct from the ~32 nm MxEnc (T=3 symmetry).
  • Iron Loading Efficiency: QtEnc+QtIMEF showed significantly more efficient iron biomineralization than MxEnc+MxBCD, even at lower ferrous ammonium sulfate (FAS) concentrations. Total iron content in cells expressing QtEnc+QtIMEF was significantly higher than with MxEnc+MxBCD.
  • Quantitative Iron Content: Purified QtEnc+QtIMEF contained approximately 35,000 iron atoms per shell, compared to ~19,000 for MxEnc+MxBCD.
  • Ferroxidase Activity: QtEnc+QtIMEF demonstrated faster ferroxidase activity and higher absorbance plateau compared to MxEnc+MxBCD.
  • TEM Visualization: Both QtEnc and MxEnc were clearly detected as individual nanospheres in cellular TEM, with distinct size distributions. QtEnc nanospheres appeared darker, indicating higher contrast.
  • Deep Learning Performance: The automated image classification model achieved high accuracy in distinguishing between QtEnc, MxEnc, and ferritin, outperforming human performance in some aspects.
  • Membrane Targeting: A farnesylated variant of QtEnc (QtEncFarn) was successfully targeted to the plasma membrane, demonstrating the potential for spatial control.

Limitations and Future Work

  • Limited Cargo Diversity: While the study focused on iron sequestration, the authors suggest future work could involve encapsulating other cargo, such as fluorescent proteins or enzymes (like APEX), to expand the reporter repertoire for correlative imaging or other applications.
  • Spatial Resolution of Cargo: While the nanospheres themselves are well-defined, the precise spatial distribution of the iron core within the shell could be further investigated.
  • Optimization of Deep Learning: While the deep learning model showed strong performance, further optimization and validation on larger, more diverse datasets could improve robustness.
  • In Vivo Applications: The current study focuses on cell culture. Future work could explore the application of these EMcapsulins in more complex biological systems and in vivo.
  • Endogenous Competition: While QtEnc appears to compete effectively with endogenous ferritin, further studies could investigate its behavior in a wider range of cellular iron homeostasis contexts.

Summary updated

Abstract

Multi-colored gene reporters such as fluorescent proteins are indispensable for biomedical research, but equivalent tools for electron microscopy (EM), a gold standard for deciphering mechanistic details of cellular processes 1,2 and uncovering the network architecture of cell-circuits 3,4 , are still sparse and not easily multiplexable. Semi-genetic EM reporters are based on the precipitation of exogenous chemicals 5–9 which may limit spatial precision and tissue penetration and can affect ultrastructure due to fixation and permeabilization. The latter technical constraints also affect EM immunolabeling techniques 10–13 which may furthermore be complicated by limited epitope accessibility. The fully genetic iron storage protein ferritin generates contrast via its electron-dense iron core 14–16 , but its small size complicates differentiation of individual ferritin particles from cellular structures. To enable multiplexed gene reporter imaging via conventional transmission electron microscopy (TEM), we here introduce the encapsulin system of Quasibacillus thermotolerans (Qt) as a fully genetic iron-biomineralizing nanocompartment. We reveal by cryo-electron reconstructions that the Qt monomers (QtEnc) self-assemble to nanospheres with T=4 icosahedral symmetry and an ~44 nm diameter harboring two putative pore regions at the fivefold and threefold axes. We furthermore show that the native cargo (QtlMEF) auto-targets to the inner surface of QtEnc and exhibits ferroxidase activity leading to efficient iron sequestration inside mammalian cells. We then demonstrate that QtEnc can be robustly differentiated from the non-intermixing encapsulin of Myxococcus xanthus 17 (Mx, ~32 nm) via a deep-learning model, thus enabling automated multiplexed EM gene reporter imaging in mammalian cells.
BibTeX
@misc{sigmund2019iron,
  author = {Felix Sigmund and Susanne Pettinger and Massimo Kube and Fabian Schneider and Martina Schifferer and Michaela Aichler and Steffen Schneider and Axel Walch and Thomas Misgeld and Hendrik Dietz and Gil G. Westmeyer},
  title = {{Iron-sequestering nanocompartments as multiplexed Electron Microscopy gene reporters}},
  year = {2019},
  doi = {10.1101/516955},
  howpublished = {bioRxiv},
}
Figure