Harnessing artificial intelligence to reduce phototoxicity in live imaging.

Fluorescence microscopy is essential for studying living cells, tissues and organisms. However, the fluorescent light that switches on fluorescent molecules also harms the samples, jeopardizing the validity of results - particularly in techniques such as super-resolution microscopy, which demands extended illumination. Artificial intelligence (AI)-enabled software capable of denoising, image restoration, temporal interpolation or cross-modal style transfer has great potential to rescue live imaging data and limit photodamage. Yet we believe the focus should be on maintaining light-induced damage at levels that preserve natural cell behaviour. In this Opinion piece, we argue that a shift in role for AIs is needed - AI should be used to extract rich insights from gentle imaging rather than recover compromised data from harsh illumination. Although AI can enhance imaging, our ultimate goal should be to uncover biological truths, not just retrieve data. It is essential to prioritize minimizing photodamage over merely pushing technical limits. Our approach is aimed towards gentle acquisition and observation of undisturbed living systems, aligning with the essence of live-cell fluorescence microscopy.

High-fidelity 3D live-cell nanoscopy through data-driven enhanced super-resolution radial fluctuation.

Live-cell super-resolution microscopy enables the imaging of biological structure dynamics below the diffraction limit. Here we present enhanced super-resolution radial fluctuations (eSRRF), substantially improving image fidelity and resolution compared to the original SRRF method. eSRRF incorporates automated parameter optimization based on the data itself, giving insight into the trade-off between resolution and fidelity. We demonstrate eSRRF across a range of imaging modalities and biological systems. Notably, we extend eSRRF to three dimensions by combining it with multifocus microscopy. This realizes live-cell volumetric super-resolution imaging with an acquisition speed of ~1 volume per second. eSRRF provides an accessible super-resolution approach, maximizing information extraction across varied experimental conditions while minimizing artifacts. Its optimal parameter prediction strategy is generalizable, moving toward unbiased and optimized analyses in super-resolution microscopy.

Live-cell imaging in the deep learning era.

Live imaging is a powerful tool, enabling scientists to observe living organisms in real time. In particular, when combined with fluorescence microscopy, live imaging allows the monitoring of cellular components with high sensitivity and specificity. Yet, due to critical challenges (i.e., drift, phototoxicity, dataset size), implementing live imaging and analyzing the resulting datasets is rarely straightforward. Over the past years, the development of bioimage analysis tools, including deep learning, is changing how we perform live imaging. Here we briefly cover important computational methods aiding live imaging and carrying out key tasks such as drift correction, denoising, super-resolution imaging, artificial labeling, tracking, and time series analysis. We also cover recent advances in self-driving microscopy.

Myosin-X recruits lamellipodin to filopodia tips.

Myosin-X (MYO10), a molecular motor localizing to filopodia, is thought to transport various cargo to filopodia tips, modulating filopodia function. However, only a few MYO10 cargoes have been described. Here, using GFP-Trap and BioID approaches combined with mass spectrometry, we identified lamellipodin (RAPH1) as a novel MYO10 cargo. We report that the FERM domain of MYO10 is required for RAPH1 localization and accumulation at filopodia tips. Previous studies have mapped the RAPH1 interaction domain for adhesome components to its talin-binding and Ras-association domains. Surprisingly, we find that the RAPH1 MYO10-binding site is not within these domains. Instead, it comprises a conserved helix located just after the RAPH1 pleckstrin homology domain with previously unknown functions. Functionally, RAPH1 supports MYO10 filopodia formation and stability but is not required to activate integrins at filopodia tips. Taken together, our data indicate a feed-forward mechanism whereby MYO10 filopodia are positively regulated by MYO10-mediated transport of RAPH1 to the filopodium tip.

Fast4DReg - fast registration of 4D microscopy datasets.

Unwanted sample drift is a common issue that plagues microscopy experiments, preventing accurate temporal visualization and quantification of biological processes. Although multiple methods and tools exist to correct images post acquisition, performing drift correction of three-dimensional (3D) videos using open-source solutions remains challenging and time consuming. Here, we present a new tool developed for ImageJ or Fiji called Fast4DReg that can quickly correct axial and lateral drift in 3D video-microscopy datasets. Fast4DReg works by creating intensity projections along multiple axes and estimating the drift between frames using two-dimensional cross-correlations. Using synthetic and acquired datasets, we demonstrate that Fast4DReg can perform better than other state-of-the-art open-source drift-correction tools and significantly outperforms them in speed. We also demonstrate that Fast4DReg can be used to register misaligned channels in 3D using either calibration slides or misaligned images directly. Altogether, Fast4DReg provides a quick and easy-to-use method to correct 3D imaging data before further visualization and analysis.

Mapping the Localization of Proteins Within Filopodia Using FiloMap.

Filopodia are fingerlike membrane protrusions extended by cells to sense their environment. Filopodia are widely used by migrating cells in vivo and directly contribute to several physiological processes and diseases. Due to the essential roles of filopodia in sensing the extracellular environment, there is a need to characterize their composition and ultrastructure further. This chapter highlights FiloMap, an image analysis pipeline that utilizes Fiji and R to map the localization of proteins within filopodia from microscopy images. I provide step-by-step protocols on (a) setting up FiloMap in Fiji and R, (b) extracting line intensity profiles from filopodia stainings in Fiji, (c) further analyzing line intensity profiles in R, and (d) creating filopodia maps to compare the localization of multiple proteins within filopodia. Notably, while FiloMap was written to analyze filopodia, the analysis pipeline described here can also analyze and compile any line intensity profiles.

MYO10-filopodia support basement membranes at pre-invasive tumor boundaries.

Ductal carcinoma in situ (DCIS) is a pre-invasive stage of breast cancer. During invasion, the encapsulating DCIS basement membrane (BM) is compromised, and tumor cells invade the surrounding stroma. The mechanisms that regulate functional epithelial BMs in vivo are poorly understood. Myosin-X (MYO10) is a filopodia-inducing protein associated with metastasis and poor clinical outcome in invasive breast cancer (IBC). We identify elevated MYO10 expression in human DCIS and IBC, and this suggests links with disease progression. MYO10 promotes filopodia formation and cell invasion in vitro and cancer-cell dissemination from progressively invasive human DCIS xenografts. However, MYO10-depleted xenografts are more invasive. These lesions exhibit compromised BMs, poorly defined borders, and increased cancer-cell dispersal and EMT-marker-positive cells. In addition, cancer spheroids are dependent on MYO10-filopodia to generate a near-continuous extracellular matrix boundary. Thus, MYO10 is protective in early-stage breast cancer, correlating with tumor-limiting BMs, and pro-invasive at later stages, facilitating cancer-cell dissemination.

DeepBacs for multi-task bacterial image analysis using open-source deep learning approaches.

This work demonstrates and guides how to use a range of state-of-the-art artificial neural-networks to analyse bacterial microscopy images using the recently developed ZeroCostDL4Mic platform. We generated a database of image datasets used to train networks for various image analysis tasks and present strategies for data acquisition and curation, as well as model training. We showcase different deep learning (DL) approaches for segmenting bright field and fluorescence images of different bacterial species, use object detection to classify different growth stages in time-lapse imaging data, and carry out DL-assisted phenotypic profiling of antibiotic-treated cells. To also demonstrate the ability of DL to enhance low-phototoxicity live-cell microscopy, we showcase how image denoising can allow researchers to attain high-fidelity data in faster and longer imaging. Finally, artificial labelling of cell membranes and predictions of super-resolution images allow for accurate mapping of cell shape and intracellular targets. Our purposefully-built database of training and testing data aids in novice users' training, enabling them to quickly explore how to analyse their data through DL. We hope this lays a fertile ground for the efficient application of DL in microbiology and fosters the creation of tools for bacterial cell biology and antibiotic research.

TrackMate 7: integrating state-of-the-art segmentation algorithms into tracking pipelines.

TrackMate is an automated tracking software used to analyze bioimages and is distributed as a Fiji plugin. Here, we introduce a new version of TrackMate. TrackMate 7 is built to address the broad spectrum of modern challenges researchers face by integrating state-of-the-art segmentation algorithms into tracking pipelines. We illustrate qualitatively and quantitatively that these new capabilities function effectively across a wide range of bio-imaging experiments.

Cargo-specific recruitment in clathrin- and dynamin-independent endocytosis.

Spatially controlled, cargo-specific endocytosis is essential for development, tissue homeostasis and cancer invasion. Unlike cargo-specific clathrin-mediated endocytosis, the clathrin- and dynamin-independent endocytic pathway (CLIC-GEEC, CG pathway) is considered a bulk internalization route for the fluid phase, glycosylated membrane proteins and lipids. While the core molecular players of CG-endocytosis have been recently defined, evidence of cargo-specific adaptors or selective uptake of proteins for the pathway are lacking. Here we identify the actin-binding protein Swiprosin-1 (Swip1, EFHD2) as a cargo-specific adaptor for CG-endocytosis. Swip1 couples active Rab21-associated integrins with key components of the CG-endocytic machinery-Arf1, IRSp53 and actin-and is critical for integrin endocytosis. Through this function, Swip1 supports integrin-dependent cancer-cell migration and invasion, and is a negative prognostic marker in breast cancer. Our results demonstrate a previously unknown cargo selectivity for the CG pathway and a role for specific adaptors in recruitment into this endocytic route.

Improving the knock-in efficiency of the MOF-encapsulated CRISPR/Cas9 system through controllable embedding structures.

Appropriate tuning of robust artificial coatings can not only enhance intracellular delivery but also preserve the biological functions of genetic molecules in gene based therapies. Here, we report a strategy to synthesize controllable nanostructures in situ by encapsulating CRISPR/Cas9 plasmids into metal-organic frameworks (MOFs) via biomimetic mineralization. The structure-functionality relationship studies indicate that MOF-coated nanostructures dramatically impact the biological features of the contained plasmids through different embedding structures. The plasmids are homogeneously distributed within the heterogeneous nanoarchitecture and protected from enzymatic degradation. In addition, the plasmid-MOF structure exhibits excellent loading capability, pH-responsive release, and affinity for plasmid binding. Through in vitro assays it was found that the superior MOF vector can greatly enhance cellular endocytosis and endo/lysosomal escape of sheltered plasmids, resulting in successful knock-in of GFP-tagged paxillin genomic sequences in cancer cell lines with high transfection potency compared to our previous studies. Thus, the development of new cost-effective approaches for MOF-based intracellular delivery systems offers an attractive option for overcoming the physiological barriers to CRISPR/Cas9 delivery, which shows great potential for investigating paxillin-associated focal adhesions and signal regulation.

SHANK3 conformation regulates direct actin binding and crosstalk with Rap1 signaling.

Actin-rich cellular protrusions direct versatile biological processes from cancer cell invasion to dendritic spine development. The stability, morphology, and specific biological functions of these protrusions are regulated by crosstalk between three main signaling axes: integrins, actin regulators, and small guanosine triphosphatases (GTPases). SHANK3 is a multifunctional scaffold protein, interacting with several actin-binding proteins and a well-established autism risk gene. Recently, SHANK3 was demonstrated to sequester integrin-activating small GTPases Rap1 and R-Ras to inhibit integrin activity via its Shank/ProSAP N-terminal (SPN) domain. Here, we demonstrate that, in addition to scaffolding actin regulators and actin-binding proteins, SHANK3 interacts directly with actin through its SPN domain. Molecular simulations and targeted mutagenesis of the SPN-ankyrin repeat region (ARR) interface reveal that actin binding is inhibited by an intramolecular closed conformation of SHANK3, where the adjacent ARR domain covers the actin-binding interface of the SPN domain. Actin and Rap1 compete with each other for binding to SHANK3, and mutation of SHANK3, resulting in reduced actin binding, augments inhibition of Rap1-mediated integrin activity. This dynamic crosstalk has functional implications for cell morphology and integrin activity in cancer cells. In addition, SHANK3-actin interaction regulates dendritic spine morphology in neurons and autism-linked phenotypes in vivo.

Myosin-X and talin modulate integrin activity at filopodia tips.

Filopodia assemble unique integrin-adhesion complexes to sense the extracellular matrix. However, the mechanisms of integrin regulation in filopodia are poorly defined. Here, we report that active integrins accumulate at the tip of myosin-X (MYO10)-positive filopodia, while inactive integrins are uniformly distributed. We identify talin and MYO10 as the principal integrin activators in filopodia. In addition, deletion of MYO10's FERM domain, or mutation of its ß1-integrin-binding residues, reveals MYO10 as facilitating integrin activation, but not transport, in filopodia. However, MYO10's isolated FERM domain alone cannot activate integrins, potentially because of binding to both integrin tails. Finally, because a chimera construct generated by swapping MYO10-FERM by talin-FERM enables integrin activation in filopodia, our data indicate that an integrin-binding FERM domain coupled to a myosin motor is a core requirement for integrin activation in filopodia. Therefore, we propose a two-step integrin activation model in filopodia: receptor tethering by MYO10 followed by talin-mediated integrin activation.

Imaging in focus: An introduction to denoising bioimages in the era of deep learning.

Fluorescence microscopy enables the direct observation of previously hidden dynamic processes of life, allowing profound insights into mechanisms of health and disease. However, imaging of live samples is fundamentally limited by the toxicity of the illuminating light and images are often acquired using low light conditions. As a consequence, images can become very noisy which severely complicates their interpretation. In recent years, deep learning (DL) has emerged as a very successful approach to remove this noise while retaining the useful signal. Unlike classical algorithms which use well-defined mathematical functions to remove noise, DL methods learn to denoise from example data, providing a powerful content-aware approach. In this review, we first describe the different types of noise that typically corrupt fluorescence microscopy images and introduce the denoising task. We then present the main DL-based denoising methods and their relative advantages and disadvantages. We aim to provide insights into how DL-based denoising methods operate and help users choose the most appropriate tools for their applications.

Talin rod domain-containing protein 1 (TLNRD1) is a novel actin-bundling protein which promotes filopodia formation.

Talin is a mechanosensitive adapter protein that couples integrins to the cytoskeleton. Talin rod domain-containing protein 1 (TLNRD1) shares 22% homology with the talin R7R8 rod domains, and is highly conserved throughout vertebrate evolution, although little is known about its function. Here we show that TLNRD1 is an α-helical protein structurally homologous to talin R7R8. Like talin R7R8, TLNRD1 binds F-actin, but because it forms a novel antiparallel dimer, it also bundles F-actin. In addition, it binds the same LD motif-containing proteins, RIAM and KANK, as talin R7R8. In cells, TLNRD1 localizes to actin bundles as well as to filopodia. Increasing TLNRD1 expression enhances filopodia formation and cell migration on 2D substrates, while TLNRD1 down-regulation has the opposite effect. Together, our results suggest that TLNRD1 has retained the diverse interactions of talin R7R8, but has developed distinct functionality as an actin-bundling protein that promotes filopodia assembly.

Talin mechanosensitivity is modulated by a direct interaction with cyclin-dependent kinase-1.

Talin (TLN1) is a mechanosensitive component of adhesion complexes that directly couples integrins to the actin cytoskeleton. In response to force, talin undergoes switch-like behavior of its multiple rod domains that modulate interactions with its binding partners. Cyclin-dependent kinase-1 (CDK1) is a key regulator of the cell cycle, exerting its effects through synchronized phosphorylation of a large number of protein targets. CDK1 activity maintains adhesion during interphase, and its inhibition is a prerequisite for the tightly choreographed changes in cell shape and adhesion that are required for successful mitosis. Using a combination of biochemical, structural, and cell biological approaches, we demonstrate a direct interaction between talin and CDK1 that occurs at sites of integrin-mediated adhesion. Mutagenesis demonstrated that CDK1 contains a functional talin-binding LD motif, and the binding site within talin was pinpointed to helical bundle R8. Talin also contains a consensus CDK1 phosphorylation motif centered on S1589, a site shown to be phosphorylated by CDK1 in vitro. A phosphomimetic mutant of this site within talin lowered the binding affinity of the cytoskeletal adaptor KANK and weakened the response of this region to force as measured by single molecule stretching, potentially altering downstream mechanotransduction pathways. The direct binding of the master cell cycle regulator CDK1 to the primary integrin effector talin represents a coupling of cell proliferation and cell adhesion machineries and thereby indicates a mechanism by which the microenvironment can control cell division in multicellular organisms.