MemBrain: A deep learning-aided pipeline for detection of membrane proteins in Cryo-electron tomograms.

BACKGROUND AND OBJECTIVE: Cryo-electron tomography (cryo-ET) is an imaging technique that enables 3D visualization of the native cellular environment at sub-nanometer resolution, providing unpreceded insights into the molecular organization of cells. However, cryo-electron tomograms suffer from low signal-to-noise ratios and anisotropic resolution, which makes subsequent image analysis challenging. In particular, the efficient detection of membrane-embedded proteins is a problem still lacking satisfactory solutions. METHODS: We present MemBrain - a new deep learning-aided pipeline that automatically detects membrane-bound protein complexes in cryo-electron tomograms. After subvolumes are sampled along a segmented membrane, each subvolume is assigned a score using a convolutional neural network (CNN), and protein positions are extracted by a clustering algorithm. Incorporating rotational subvolume normalization and using a tiny receptive field simplify the task of protein detection and thus facilitate the network training. RESULTS: MemBrain requires only a small quantity of training labels and achieves excellent performance with only a single annotated membrane (F1 score: 0.88). A detailed evaluation shows that our fully trained pipeline outperforms existing classical computer vision-based and CNN-based approaches by a large margin (F1 score: 0.92 vs. max. 0.63). Furthermore, in addition to protein center positions, MemBrain can determine protein orientations, which has not been implemented by any existing CNN-based method to date. We also show that a pre-trained MemBrain program generalizes to tomograms acquired using different cryo-ET methods and depicting different types of cells. CONCLUSIONS: MemBrain is a powerful and annotation-efficient tool for the detection of membrane protein complexes in cryo-ET data, with the potential to be used in a wide range of biological studies. It is generalizable to various kinds of tomograms, making it possible to use pretrained models for different tasks. Its efficiency in terms of required annotations also allows rapid training and fine-tuning of models. The corresponding code, pretrained models, and instructions for operating the MemBrain program can be found at: https://github.com/CellArchLab/MemBrain.

Membrane-anchored HDCR nanowires drive hydrogen-powered CO2 fixation.

Filamentous enzymes have been found in all domains of life, but the advantage of filamentation is often elusive1. Some anaerobic, autotrophic bacteria have an unusual filamentous enzyme for CO2 fixation-hydrogen-dependent CO2 reductase (HDCR)2,3-which directly converts H2 and CO2 into formic acid. HDCR reduces CO2 with a higher activity than any other known biological or chemical catalyst4,5, and it has therefore gained considerable interest in two areas of global relevance: hydrogen storage and combating climate change by capturing atmospheric CO2. However, the mechanistic basis of the high catalytic turnover rate of HDCR has remained unknown. Here we use cryo-electron microscopy to reveal the structure of a short HDCR filament from the acetogenic bacterium Thermoanaerobacter kivui. The minimum repeating unit is a hexamer that consists of a formate dehydrogenase (FdhF) and two hydrogenases (HydA2) bound around a central core of hydrogenase Fe-S subunits, one HycB3 and two HycB4. These small bacterial polyferredoxin-like proteins oligomerize through their C-terminal helices to form the backbone of the filament. By combining structure-directed mutagenesis with enzymatic analysis, we show that filamentation and rapid electron transfer through the filament enhance the activity of HDCR. To investigate the structure of HDCR in situ, we imaged T. kivui cells with cryo-electron tomography and found that HDCR filaments bundle into large ring-shaped superstructures attached to the plasma membrane. This supramolecular organization may further enhance the stability and connectivity of HDCR to form a specialized metabolic subcompartment within the cell.

Deep learning improves macromolecule identification in 3D cellular cryo-electron tomograms.

Cryogenic electron tomography (cryo-ET) visualizes the 3D spatial distribution of macromolecules at nanometer resolution inside native cells. However, automated identification of macromolecules inside cellular tomograms is challenged by noise and reconstruction artifacts, as well as the presence of many molecular species in the crowded volumes. Here, we present DeepFinder, a computational procedure that uses artificial neural networks to simultaneously localize multiple classes of macromolecules. Once trained, the inference stage of DeepFinder is faster than template matching and performs better than other competitive deep learning methods at identifying macromolecules of various sizes in both synthetic and experimental datasets. On cellular cryo-ET data, DeepFinder localized membrane-bound and cytosolic ribosomes (roughly 3.2 MDa), ribulose 1,5-bisphosphate carboxylase-oxygenase (roughly 560 kDa soluble complex) and photosystem II (roughly 550 kDa membrane complex) with an accuracy comparable to expert-supervised ground truth annotations. DeepFinder is therefore a promising algorithm for the semiautomated analysis of a wide range of molecular targets in cellular tomograms.

Structural basis for VIPP1 oligomerization and maintenance of thylakoid membrane integrity.

Vesicle-inducing protein in plastids 1 (VIPP1) is essential for the biogenesis and maintenance of thylakoid membranes, which transform light into life. However, it is unknown how VIPP1 performs its vital membrane-remodeling functions. Here, we use cryo-electron microscopy to determine structures of cyanobacterial VIPP1 rings, revealing how VIPP1 monomers flex and interweave to form basket-like assemblies of different symmetries. Three VIPP1 monomers together coordinate a non-canonical nucleotide binding pocket on one end of the ring. Inside the ring's lumen, amphipathic helices from each monomer align to form large hydrophobic columns, enabling VIPP1 to bind and curve membranes. In vivo mutations in these hydrophobic surfaces cause extreme thylakoid swelling under high light, indicating an essential role of VIPP1 lipid binding in resisting stress-induced damage. Using cryo-correlative light and electron microscopy (cryo-CLEM), we observe oligomeric VIPP1 coats encapsulating membrane tubules within the Chlamydomonas chloroplast. Our work provides a structural foundation for understanding how VIPP1 directs thylakoid biogenesis and maintenance.

The state of oligomerization of Rubisco controls the rate of synthesis of the Rubisco large subunit in Chlamydomonas reinhardtii.

Ribulose 1,5-bisphosphate carboxylase/oxygenase (Rubisco) is present in all photosynthetic organisms and is a key enzyme for photosynthesis-driven life on Earth. Its most prominent form is a hetero-oligomer in which small subunits (SSU) stabilize the core of the enzyme built from large subunits (LSU), yielding, after a chaperone-assisted multistep assembly process, an LSU8SSU8 hexadecameric holoenzyme. Here we use Chlamydomonas reinhardtii and a combination of site-directed mutants to dissect the multistep biogenesis pathway of Rubisco in vivo. We identify assembly intermediates, in two of which LSU are associated with the RAF1 chaperone. Using genetic and biochemical approaches we further unravel a major regulation process during Rubisco biogenesis, in which LSU translation is controlled by its ability to assemble with the SSU, via the mechanism of control by epistasy of synthesis (CES). Altogether this leads us to propose a model whereby the last assembly intermediate, an LSU8-RAF1 complex, provides the platform for SSU binding to form the Rubisco enzyme, and when SSU is not available, converts to a key regulatory form that exerts negative feedback on the initiation of LSU translation.

Charting the native architecture of Chlamydomonas thylakoid membranes with single-molecule precision.

Thylakoid membranes scaffold an assortment of large protein complexes that work together to harness the energy of light. It has been a longstanding challenge to visualize how the intricate thylakoid network organizes these protein complexes to finely tune the photosynthetic reactions. Previously, we used in situ cryo-electron tomography to reveal the native architecture of thylakoid membranes (Engel et al., 2015). Here, we leverage technical advances to resolve the individual protein complexes within these membranes. Combined with a new method to visualize membrane surface topology, we map the molecular landscapes of thylakoid membranes inside green algae cells. Our tomograms provide insights into the molecular forces that drive thylakoid stacking and reveal that photosystems I and II are strictly segregated at the borders between appressed and non-appressed membrane domains. This new approach to charting thylakoid topology lays the foundation for dissecting photosynthetic regulation at the level of single protein complexes within the cell.

Direct visualization of degradation microcompartments at the ER membrane.

To promote the biochemical reactions of life, cells can compartmentalize molecular interaction partners together within separated non-membrane-bound regions. It is unknown whether this strategy is used to facilitate protein degradation at specific locations within the cell. Leveraging in situ cryo-electron tomography to image the native molecular landscape of the unicellular alga Chlamydomonas reinhardtii, we discovered that the cytosolic protein degradation machinery is concentrated within ∼200-nm foci that contact specialized patches of endoplasmic reticulum (ER) membrane away from the ER-Golgi interface. These non-membrane-bound microcompartments exclude ribosomes and consist of a core of densely clustered 26S proteasomes surrounded by a loose cloud of Cdc48. Active proteasomes in the microcompartments directly engage with putative substrate at the ER membrane, a function canonically assigned to Cdc48. Live-cell fluorescence microscopy revealed that the proteasome clusters are dynamic, with frequent assembly and fusion events. We propose that the microcompartments perform ER-associated degradation, colocalizing the degradation machinery at specific ER hot spots to enable efficient protein quality control.

The Arabidopsis DREB2 genetic pathway is constitutively repressed by basal phosphoinositide-dependent phospholipase C coupled to diacylglycerol kinase.

Phosphoinositide-dependent phospholipases C (PI-PLCs) are activated in response to various stimuli. They utilize substrates provided by type III-Phosphatidylinositol-4 kinases (PI4KIII) to produce inositol triphosphate and diacylglycerol (DAG) that is phosphorylated into phosphatidic acid (PA) by DAG-kinases (DGKs). The roles of PI4KIIIs, PI-PLCs, and DGKs in basal signaling are poorly understood. We investigated the control of gene expression by basal PI-PLC pathway in Arabidopsis thaliana suspension cells. A transcriptome-wide analysis allowed the identification of genes whose expression was altered by edelfosine, 30 µM wortmannin, or R59022, inhibitors of PI-PLCs, PI4KIIIs, and DGKs, respectively. We found that a gene responsive to one of these molecules is more likely to be similarly regulated by the other two inhibitors. The common action of these agents is to inhibit PA formation, showing that basal PI-PLCs act, in part, on gene expression through their coupling to DGKs. Amongst the genes up-regulated in presence of the inhibitors, were some DREB2 genes, in suspension cells and in seedlings. The DREB2 genes encode transcription factors with major roles in responses to environmental stresses, including dehydration. They bind to C-repeat motifs, known as Drought-Responsive Elements that are indeed enriched in the promoters of genes up-regulated by PI-PLC pathway inhibitors. PA can also be produced by phospholipases D (PLDs). We show that the DREB2 genes that are up-regulated by PI-PLC inhibitors are positively or negatively regulated, or indifferent, to PLD basal activity. Our data show that the DREB2 genetic pathway is constitutively repressed in resting conditions and that DGK coupled to PI-PLC is active in this process, in suspension cells and seedlings. We discuss how this basal negative regulation of DREB2 genes is compatible with their stress-triggered positive regulation.