Multiplexed DNA-PAINT Imaging of the Heterogeneity of Late Endosome/Lysosome Protein Composition.

Late endosomes/lysosomes (LELs) are crucial for numerous physiological processes and their dysfunction is linked to many diseases. Proteomic analyses have identified hundreds of LEL proteins, however, whether these proteins are uniformly present on each LEL, or if there are cell-type dependent LEL sub-populations with unique protein compositions is unclear. We employed a quantitative, multiplexed DNA-PAINT super-resolution approach to examine the distribution of six key LEL proteins (LAMP1, LAMP2, CD63, TMEM192, NPC1 and LAMTOR4) on individual LELs. While LAMP1 and LAMP2 were abundant across LELs, marking a common population, most analyzed proteins were associated with specific LEL subpopulations. Our multiplexed imaging approach identified up to eight different LEL subpopulations based on their unique membrane protein composition. Additionally, our analysis of the spatial relationships between these subpopulations and mitochondria revealed a cell-type specific tendency for NPC1-positive LELs to be closely positioned to mitochondria. Our approach will be broadly applicable to determining organelle heterogeneity with single organelle resolution in many biological contexts. Summary: This study develops a multiplexed and quantitative DNA-PAINT super-resolution imaging pipeline to investigate the distribution of late endosomal/lysosomal (LEL) proteins across individual LELs, revealing cell-type specific LEL sub-populations with unique protein compositions, offering insights into organelle heterogeneity at single-organelle resolution.

ActionCLIP: Adapting Language-Image Pretrained Models for Video Action Recognition.

The canonical approach to video action recognition dictates a neural network model to do a classic and standard 1-of-N majority vote task. They are trained to predict a fixed set of predefined categories, limiting their transferability on new datasets with unseen concepts. In this article, we provide a new perspective on action recognition by attaching importance to the semantic information of label texts rather than simply mapping them into numbers. Specifically, we model this task as a video-text matching problem within a multimodal learning framework, which strengthens the video representation with more semantic language supervision and enables our model to do zero-shot action recognition without any further labeled data or parameters' requirements. Moreover, to handle the deficiency of label texts and make use of tremendous web data, we propose a new paradigm based on this multimodal learning framework for action recognition, which we dub "pre-train, adapt and fine-tune." This paradigm first learns powerful representations from pre-training on a large amount of web image-text or video-text data. Then, it makes the action recognition task to act more like pre-training problems via adaptation engineering. Finally, it is fine-tuned end-to-end on target datasets to obtain strong performance. We give an instantiation of the new paradigm, ActionCLIP, which not only has superior and flexible zero-shot/few-shot transfer ability but also reaches a top performance on general action recognition task, achieving 83.8% top-1 accuracy on Kinetics-400 with a ViT-B/16 as the backbone. Code is available at https://github.com/sallymmx/ActionCLIP.git.

Interplay between stochastic enzyme activity and microtubule stability drives detyrosination enrichment on microtubule subsets.

Microtubules in cells consist of functionally diverse subpopulations carrying distinct post-translational modifications (PTMs). Akin to the histone code, the tubulin code regulates a myriad of microtubule functions, ranging from intracellular transport to chromosome segregation. However, how individual PTMs only occur on subsets of microtubules to contribute to microtubule specialization is not well understood. In particular, microtubule detyrosination, the removal of the C-terminal tyrosine on α-tubulin subunits, marks the stable population of microtubules and modifies how microtubules interact with other microtubule-associated proteins to regulate a wide range of cellular processes. Previously, we found that in certain cell types, only ∼30% of microtubules are highly enriched with the detyrosination mark and that detyrosination spans most of the length of a microtubule, often adjacent to a completely tyrosinated microtubule. How the activity of a cytosolic detyrosinase, vasohibin (VASH), leads to only a small subpopulation of highly detyrosinated microtubules is unclear. Here, using quantitative super-resolution microscopy, we visualized nascent microtubule detyrosination events in cells consisting of 1-3 detyrosinated α-tubulin subunits after nocodazole washout. Microtubule detyrosination accumulates slowly and in a dispersed pattern across the microtubule length. By visualizing single molecules of VASH in live cells, we found that VASH engages with microtubules stochastically on a short timescale, suggesting limited removal of tyrosine per interaction, consistent with the super-resolution results. Combining these quantitative imaging results with simulations incorporating parameters from our experiments, we provide evidence for a stochastic model for cells to establish a subset of detyrosinated microtubules via a detyrosination-stabilization feedback mechanism.

Learning SpatioTemporal and Motion Features in a Unified 2D Network for Action Recognition.

Recent methods for action recognition always apply 3D Convolutional Neural Networks (CNNs) to extract spatiotemporal features and introduce optical flows to present motion features. Although achieving state-of-the-art performance, they are expensive in both time and space. In this paper, we propose to represent both two kinds of features in a unified 2D CNN without any 3D convolution or optical flows calculation. In particular, we first design a channel-wise spatiotemporal module to present the spatiotemporal features and a channel-wise motion module to encode feature-level motion features efficiently. Besides, we provide a distinctive illustration of the two modules from the frequency domain by interpreting them as advanced and learnable versions of frequency components. Second, we combine these two modules and an identity mapping path into one united block that can easily replace the original residual block in the ResNet architecture, forming a simple yet effective network dubbed STM network by introducing very limited extra computation cost and parameters. Third, we propose a novel Twins Training framework for action recognition by incorporating a correlation loss to optimize the inter-class and intra-class correlation and a siamese structure to fully stretch the training data. We extensively validate the proposed STM on both temporal-related datasets (i.e., Something-Something v1 & v2) and scene-related datasets (i.e., Kinetics-400, UCF-101, and HMDB-51). It achieves favorable results against state-of-the-art methods in all the datasets.

CGG repeats trigger translational frameshifts that generate aggregation-prone chimeric proteins.

CGG repeat expansions in the FMR1 5'UTR cause the neurodegenerative disease Fragile X-associated tremor/ataxia syndrome (FXTAS). These repeats form stable RNA secondary structures that support aberrant translation in the absence of an AUG start codon (RAN translation), producing aggregate-prone peptides that accumulate within intranuclear neuronal inclusions and contribute to neurotoxicity. Here, we show that the most abundant RAN translation product, FMRpolyG, is markedly less toxic when generated from a construct with a non-repetitive alternating codon sequence in place of the CGG repeat. While exploring the mechanism of this differential toxicity, we observed a +1 translational frameshift within the CGG repeat from the arginine to glycine reading frame. Frameshifts occurred within the first few translated repeats and were triggered predominantly by RNA sequence and structural features. Short chimeric R/G peptides form aggregates distinct from those formed by either pure arginine or glycine, and these chimeras induce toxicity in cultured rodent neurons. Together, this work suggests that CGG repeats support translational frameshifting and that chimeric RAN translated peptides may contribute to CGG repeat-associated toxicity in FXTAS and related disorders.

Shared and divergent phase separation and aggregation properties of brain-expressed ubiquilins.

The brain-expressed ubiquilins, UBQLNs 1, 2 and 4, are highly homologous proteins that participate in multiple aspects of protein homeostasis and are implicated in neurodegenerative diseases. Studies have established that UBQLN2 forms liquid-like condensates and accumulates in pathogenic aggregates, much like other proteins linked to neurodegenerative diseases. However, the relative condensate and aggregate formation of the three brain-expressed ubiquilins is unknown. Here we report that the three ubiquilins differ in aggregation propensity, revealed by in-vitro experiments, cellular models, and analysis of human brain tissue. UBQLN4 displays heightened aggregation propensity over the other ubiquilins and, like amyloids, UBQLN4 forms ThioflavinT-positive fibrils in vitro. Measuring fluorescence recovery after photobleaching (FRAP) of puncta in cells, we report that all three ubiquilins undergo liquid-liquid phase transition. UBQLN2 and 4 exhibit slower recovery than UBQLN1, suggesting the condensates formed by these brain-expressed ubiquilins have different compositions and undergo distinct internal rearrangements. We conclude that while all brain-expressed ubiquilins exhibit self-association behavior manifesting as condensates, they follow distinct courses of phase-separation and aggregation. We suggest that this variability among ubiquilins along the continuum from liquid-like to solid informs both the normal ubiquitin-linked functions of ubiquilins and their accumulation and potential contribution to toxicity in neurodegenerative diseases.

The Fitting of a Fiber-Reinforced-Plastic Complex Curved Surface and Its Orbit Optimization Model with Belt Grinding Line Contact.

The surface quality and profile accuracy of a radar fiberglass radome are determined by the manufacturing of the fiber-reinforced-plastic (FRP) complex curved mold. The surface quality, thickness uniformity, and shape accuracy of the mold seriously affect the temperature and deformation control during the manufacturing process of the radome, thus affecting the antenna's serviceability, including its wave permeability and stability. Abrasive belt grinding is an effective method for processing FRP materials. However, issues regarding the profile fitting of the abrasive belt section line contact state and its influence on the precision of complex curved surfaces have not been solved, which seriously affects the processing quality. Here, an FRP complex curved surface mold surface based on the least-squares method was established. The local two-dimensional line contact and profile contour trajectory were obtained by the algorithm of optimal trajectory planning. Based on this, a grinding experiment was carried out. The experiments showed that the surface roughness based on this method was reduced from 0.503 to 0.289 µm, and the contour accuracy was improved by 16.9% compared with the conventional error. Through our analysis, the following conclusions can be drawn: the algorithm can effectively solve the problem of line contact surface fitting and significantly improve the precision of an FRP complex surface.