SuperCUT, an unsupervised multimodal image registration with deep learning for biomedical microscopy.

Numerous imaging techniques are available for observing and interrogating biological samples, and several of them can be used consecutively to enable correlative analysis of different image modalities with varying resolutions and the inclusion of structural or molecular information. Achieving accurate registration of multimodal images is essential for the correlative analysis process, but it remains a challenging computer vision task with no widely accepted solution. Moreover, supervised registration methods require annotated data produced by experts, which is limited. To address this challenge, we propose a general unsupervised pipeline for multimodal image registration using deep learning. We provide a comprehensive evaluation of the proposed pipeline versus the current state-of-the-art image registration and style transfer methods on four types of biological problems utilizing different microscopy modalities. We found that style transfer of modality domains paired with fully unsupervised training leads to comparable image registration accuracy to supervised methods and, most importantly, does not require human intervention.

Chloroplast phosphate transporter CrPHT4-7 regulates phosphate homeostasis and photosynthesis in Chlamydomonas.

In eukaryotic cells, phosphorus is assimilated and utilized primarily as phosphate (Pi). Pi homeostasis is mediated by transporters that have not yet been adequately characterized in green algae. This study reports on PHOSPHATE TRANSPORTER 4-7 (CrPHT4-7) from Chlamydomonas reinhardtii, a member of the PHT4 transporter family, which exhibits remarkable similarity to AtPHT4;4 from Arabidopsis (Arabidopsis thaliana), a chloroplastic ascorbate transporter. Using fluorescent protein tagging, we show that CrPHT4-7 resides in the chloroplast envelope membrane. Crpht4-7 mutants, generated by the CRISPR/Cas12a-mediated single-strand templated repair, show retarded growth, especially in high light, reduced ATP level, strong ascorbate accumulation, and diminished non-photochemical quenching in high light. On the other hand, total cellular phosphorous content was unaffected, and the phenotype of the Crpht4-7 mutants could not be alleviated by ample Pi supply. CrPHT4-7-overexpressing lines exhibit enhanced biomass accumulation under high light conditions in comparison with the wild-type strain. Expressing CrPHT4-7 in a yeast (Saccharomyces cerevisiae) strain lacking Pi transporters substantially recovered its slow growth phenotype, demonstrating that CrPHT4-7 transports Pi. Even though CrPHT4-7 shows a high degree of similarity to AtPHT4;4, it does not display any substantial ascorbate transport activity in yeast or intact algal cells. Thus, the results demonstrate that CrPHT4-7 functions as a chloroplastic Pi transporter essential for maintaining Pi homeostasis and photosynthesis in C. reinhardtii.

Image-based and machine learning-guided multiplexed serology test for SARS-CoV-2.

We present a miniaturized immunofluorescence assay (mini-IFA) for measuring antibody response in patient blood samples. The method utilizes machine learning-guided image analysis and enables simultaneous measurement of immunoglobulin M (IgM), IgA, and IgG responses against different viral antigens in an automated and high-throughput manner. The assay relies on antigens expressed through transfection, enabling use at a low biosafety level and fast adaptation to emerging pathogens. Using severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) as the model pathogen, we demonstrate that this method allows differentiation between vaccine-induced and infection-induced antibody responses. Additionally, we established a dedicated web page for quantitative visualization of sample-specific results and their distribution, comparing them with controls and other samples. Our results provide a proof of concept for the approach, demonstrating fast and accurate measurement of antibody responses in a research setup with prospects for clinical diagnostics.

Development of a Laser Microdissection-Coupled Quantitative Shotgun Lipidomic Method to Uncover Spatial Heterogeneity.

Lipid metabolic disturbances are associated with several diseases, such as type 2 diabetes or malignancy. In the last two decades, high-performance mass spectrometry-based lipidomics has emerged as a valuable tool in various fields of biology. However, the evaluation of macroscopic tissue homogenates leaves often undiscovered the differences arising from micron-scale heterogeneity. Therefore, in this work, we developed a novel laser microdissection-coupled shotgun lipidomic platform, which combines quantitative and broad-range lipidome analysis with reasonable spatial resolution. The multistep approach involves the preparation of successive cryosections from tissue samples, cross-referencing of native and stained images, laser microdissection of regions of interest, in situ lipid extraction, and quantitative shotgun lipidomics. We used mouse liver and kidney as well as a 2D cell culture model to validate the novel workflow in terms of extraction efficiency, reproducibility, and linearity of quantification. We established that the limit of dissectible sample area corresponds to about ten cells while maintaining good lipidome coverage. We demonstrate the performance of the method in recognizing tissue heterogeneity on the example of a mouse hippocampus. By providing topological mapping of lipid metabolism, the novel platform might help to uncover region-specific lipidomic alterations in complex samples, including tumors.

Correlative Fluorescence and Raman Microscopy to Define Mitotic Stages at the Single-Cell Level: Opportunities and Limitations in the AI Era.

Nowadays, morphology and molecular analyses at the single-cell level have a fundamental role in understanding biology better. These methods are utilized for cell phenotyping and in-depth studies of cellular processes, such as mitosis. Fluorescence microscopy and optical spectroscopy techniques, including Raman micro-spectroscopy, allow researchers to examine biological samples at the single-cell level in a non-destructive manner. Fluorescence microscopy can give detailed morphological information about the localization of stained molecules, while Raman microscopy can produce label-free images at the subcellular level; thus, it can reveal the spatial distribution of molecular fingerprints, even in live samples. Accordingly, the combination of correlative fluorescence and Raman microscopy (CFRM) offers a unique approach for studying cellular stages at the single-cell level. However, subcellular spectral maps are complex and challenging to interpret. Artificial intelligence (AI) may serve as a valuable solution to characterize the molecular backgrounds of phenotypes and biological processes by finding the characteristic patterns in spectral maps. The major contributions of the manuscript are: (I) it gives a comprehensive review of the literature focusing on AI techniques in Raman-based cellular phenotyping; (II) via the presentation of a case study, a new neural network-based approach is described, and the opportunities and limitations of AI, specifically deep learning, are discussed regarding the analysis of Raman spectroscopy data to classify mitotic cellular stages based on their spectral maps.

Lipid Metabolic Reprogramming Extends beyond Histologic Tumor Demarcations in Operable Human Pancreatic Cancer.

Pancreatic ductal adenocarcinoma (PDAC) is among the deadliest malignancies and potentially curable only with radical surgical resection at early stages. The tumor microenvironment has been shown to be central to the development and progression of PDAC. A better understanding of how early human PDAC metabolically communicates with its environment and differs from healthy pancreas could help improve PDAC diagnosis and treatment. Here we performed deep proteomic analyses from diagnostic specimens of operable, treatment-naïve PDAC patients (n = 14), isolating four tissue compartments by laser-capture microdissection: PDAC lesions, tumor-adjacent but morphologically benign exocrine glands, and connective tissues neighboring each of these compartments. Protein and pathway levels were compared between compartments and with control pancreatic proteomes. Selected targets were studied immunohistochemically in the 14 patients and in additional tumor microarrays, and lipid deposition was assessed by nonlinear label-free imaging (n = 16). Widespread downregulation of pancreatic secretory functions was observed, which was paralleled by high cholesterol biosynthetic activity without prominent lipid storage in the neoplastic cells. Stromal compartments harbored ample blood apolipoproteins, indicating abundant microvasculature at the time of tumor removal. The features best differentiating the tumor-adjacent exocrine tissue from healthy control pancreas were defined by upregulation of proteins related to lipid transport. Importantly, histologically benign exocrine regions harbored the most significant prognostic pathways, with proteins involved in lipid transport and metabolism, such as neutral cholesteryl ester hydrolase 1, associating with shorter survival. In conclusion, this study reveals prognostic molecular changes in the exocrine tissue neighboring pancreatic cancer and identifies enhanced lipid transport and metabolism as its defining features. SIGNIFICANCE: In clinically operable pancreatic cancer, regions distant from malignant cells already display proteomic changes related to lipid transport and metabolism that affect prognosis and may be pharmacologically targeted.

Deep Visual Proteomics defines single-cell identity and heterogeneity.

Despite the availabilty of imaging-based and mass-spectrometry-based methods for spatial proteomics, a key challenge remains connecting images with single-cell-resolution protein abundance measurements. Here, we introduce Deep Visual Proteomics (DVP), which combines artificial-intelligence-driven image analysis of cellular phenotypes with automated single-cell or single-nucleus laser microdissection and ultra-high-sensitivity mass spectrometry. DVP links protein abundance to complex cellular or subcellular phenotypes while preserving spatial context. By individually excising nuclei from cell culture, we classified distinct cell states with proteomic profiles defined by known and uncharacterized proteins. In an archived primary melanoma tissue, DVP identified spatially resolved proteome changes as normal melanocytes transition to fully invasive melanoma, revealing pathways that change in a spatial manner as cancer progresses, such as mRNA splicing dysregulation in metastatic vertical growth that coincides with reduced interferon signaling and antigen presentation. The ability of DVP to retain precise spatial proteomic information in the tissue context has implications for the molecular profiling of clinical samples.

A versatile transposon-based technology to generate loss- and gain-of-function phenotypes in the mouse liver.

BACKGROUND: Understanding the contribution of gene function in distinct organ systems to the pathogenesis of human diseases in biomedical research requires modifying gene expression through the generation of gain- and loss-of-function phenotypes in model organisms, for instance, the mouse. However, methods to modify both germline and somatic genomes have important limitations that prevent easy, strong, and stable expression of transgenes. For instance, while the liver is remarkably easy to target, nucleic acids introduced to modify the genome of hepatocytes are rapidly lost, or the transgene expression they mediate becomes inhibited due to the action of effector pathways for the elimination of exogenous DNA. Novel methods are required to overcome these challenges, and here we develop a somatic gene delivery technology enabling long-lasting high-level transgene expression in the entire hepatocyte population of mice. RESULTS: We exploit the fumarylacetoacetate hydrolase (Fah) gene correction-induced regeneration in Fah-deficient livers, to demonstrate that such approach stabilizes luciferase expression more than 5000-fold above the level detected in WT animals, following plasmid DNA introduction complemented by transposon-mediated chromosomal gene transfer. Building on this advancement, we created a versatile technology platform for performing gene function analysis in vivo in the mouse liver. Our technology allows the tag-free expression of proteins of interest and silencing of any arbitrary gene in the mouse genome. This was achieved by applying the HADHA/B endogenous bidirectional promoter capable of driving well-balanced bidirectional expression and by optimizing in vivo intronic artificial microRNA-based gene silencing. We demonstrated the particular usefulness of the technology in cancer research by creating a p53-silenced and hRas G12V-overexpressing tumor model. CONCLUSIONS: We developed a versatile technology platform for in vivo somatic genome editing in the mouse liver, which meets multiple requirements for long-lasting high-level transgene expression. We believe that this technology will contribute to the development of a more accurate new generation of tools for gene function analysis in mice.

Regression plane concept for analysing continuous cellular processes with machine learning.

Biological processes are inherently continuous, and the chance of phenotypic discovery is significantly restricted by discretising them. Using multi-parametric active regression we introduce the Regression Plane (RP), a user-friendly discovery tool enabling class-free phenotypic supervised machine learning, to describe and explore biological data in a continuous manner. First, we compare traditional classification with regression in a simulated experimental setup. Second, we use our framework to identify genes involved in regulating triglyceride levels in human cells. Subsequently, we analyse a time-lapse dataset on mitosis to demonstrate that the proposed methodology is capable of modelling complex processes at infinite resolution. Finally, we show that hemocyte differentiation in Drosophila melanogaster has continuous characteristics.

Neuroinflammatory processes are augmented in mice overexpressing human heat-shock protein B1 following ethanol-induced brain injury.

BACKGROUND: Heat-shock protein B1 (HSPB1) is among the most well-known and versatile member of the evolutionarily conserved family of small heat-shock proteins. It has been implicated to serve a neuroprotective role against various neurological disorders via its modulatory activity on inflammation, yet its exact role in neuroinflammation is poorly understood. In order to shed light on the exact mechanism of inflammation modulation by HSPB1, we investigated the effect of HSPB1 on neuroinflammatory processes in an in vivo and in vitro model of acute brain injury. METHODS: In this study, we used a transgenic mouse strain overexpressing the human HSPB1 protein. In the in vivo experiments, 7-day-old transgenic and wild-type mice were treated with ethanol. Apoptotic cells were detected using TUNEL assay. The mRNA and protein levels of cytokines and glial cell markers were examined using RT-PCR and immunohistochemistry in the brain. We also established primary neuronal, astrocyte, and microglial cultures which were subjected to cytokine and ethanol treatments. TNFα and hHSPB1 levels were measured from the supernates by ELISA, and intracellular hHSPB1 expression was analyzed using fluorescent immunohistochemistry. RESULTS: Following ethanol treatment, the brains of hHSPB1-overexpressing mice showed a significantly higher mRNA level of pro-inflammatory cytokines (Tnf, Il1b), microglia (Cd68, Arg1), and astrocyte (Gfap) markers compared to wild-type brains. Microglial activation, and 1 week later, reactive astrogliosis was higher in certain brain areas of ethanol-treated transgenic mice compared to those of wild-types. Despite the remarkably high expression of pro-apoptotic Tnf, hHSPB1-overexpressing mice did not exhibit higher level of apoptosis. Our data suggest that intracellular hHSPB1, showing the highest level in primary astrocytes, was responsible for the inflammation-regulating effects. Microglia cells were the main source of TNFα in our model. Microglia isolated from hHSPB1-overexpressing mice showed a significantly higher release of TNFα compared to wild-type cells under inflammatory conditions. CONCLUSIONS: Our work provides novel in vivo evidence that hHSPB1 overexpression has a regulating effect on acute neuroinflammation by intensifying the expression of pro-inflammatory cytokines and enhancing glial cell activation, but not increasing neuronal apoptosis. These results suggest that hHSPB1 may play a complex role in the modulation of the ethanol-induced neuroinflammatory response.

The Activities of the Gelsolin Homology Domains of Flightless-I in Actin Dynamics.

Flightless-I is a unique member of the gelsolin superfamily alloying six gelsolin homology domains and leucine-rich repeats. Flightless-I is an established regulator of the actin cytoskeleton, however, its biochemical activities in actin dynamics are still largely elusive. To better understand the biological functioning of Flightless-I we studied the actin activities of Drosophila Flightless-I by in vitro bulk fluorescence spectroscopy and single filament fluorescence microscopy, as well as in vivo genetic approaches. Flightless-I was found to interact with actin and affects actin dynamics in a calcium-independent fashion in vitro. Our work identifies the first three gelsolin homology domains (1-3) of Flightless-I as the main actin-binding site; neither the other three gelsolin homology domains (4-6) nor the leucine-rich repeats bind actin. Flightless-I inhibits polymerization by high-affinity (∼nM) filament barbed end capping, moderately facilitates nucleation by low-affinity (∼µM) monomer binding, and does not sever actin filaments. Our work reveals that in the presence of profilin Flightless-I is only able to cap actin filament barbed ends but fails to promote actin assembly. In line with the in vitro data, while gelsolin homology domains 4-6 have no effect on in vivo actin polymerization, overexpression of gelsolin homology domains 1-3 prevents the formation of various types of actin cables in the developing Drosophila egg chambers. We also show that the gelsolin homology domains 4-6 of Flightless-I interact with the C-terminus of Drosophila Disheveled-associated activator of morphogenesis formin and negatively regulates its actin assembly activity.

nucleAIzer: A Parameter-free Deep Learning Framework for Nucleus Segmentation Using Image Style Transfer.

Single-cell segmentation is typically a crucial task of image-based cellular analysis. We present nucleAIzer, a deep-learning approach aiming toward a truly general method for localizing 2D cell nuclei across a diverse range of assays and light microscopy modalities. We outperform the 739 methods submitted to the 2018 Data Science Bowl on images representing a variety of realistic conditions, some of which were not represented in the training data. The key to our approach is that during training nucleAIzer automatically adapts its nucleus-style model to unseen and unlabeled data using image style transfer to automatically generate augmented training samples. This allows the model to recognize nuclei in new and different experiments efficiently without requiring expert annotations, making deep learning for nucleus segmentation fairly simple and labor free for most biological light microscopy experiments. It can also be used online, integrated into CellProfiler and freely downloaded at www.nucleaizer.org. A record of this paper's transparent peer review process is included in the Supplemental Information.

Microtubule organization in presynaptic boutons relies on the formin DAAM.

Regulation of the cytoskeleton is fundamental to the development and function of synaptic terminals, such as neuromuscular junctions. Despite the identification of numerous proteins that regulate synaptic actin and microtubule dynamics, the mechanisms of cytoskeletal control during terminal arbor formation have remained largely elusive. Here, we show that DAAM, a member of the formin family of cytoskeleton organizing factors, is an important presynaptic regulator of neuromuscular junction development in Drosophila We demonstrate that the actin filament assembly activity of DAAM plays a negligible role in terminal formation; rather, DAAM is necessary for synaptic microtubule organization. Genetic interaction studies consistently link DAAM with the Wg/Ank2/Futsch module of microtubule regulation and bouton formation. Finally, we provide evidence that DAAM is tightly associated with the synaptic active zone scaffold, and electrophysiological data point to a role in the modulation of synaptic vesicle release. Based on these results, we propose that DAAM is an important cytoskeletal effector element of the Wg/Ank2 pathway involved in the determination of basic synaptic structures, and, additionally, that DAAM may couple the active zone scaffold to the presynaptic cytoskeleton.

The activities of the C-terminal regions of the formin protein disheveled-associated activator of morphogenesis (DAAM) in actin dynamics.

Disheveled-associated activator of morphogenesis (DAAM) is a diaphanous-related formin protein essential for the regulation of actin cytoskeleton dynamics in diverse biological processes. The conserved formin homology 1 and 2 (FH1-FH2) domains of DAAM catalyze actin nucleation and processively mediate filament elongation. These activities are indirectly regulated by the N- and C-terminal regions flanking the FH1-FH2 domains. Recently, the C-terminal diaphanous-autoregulatory domain (DAD) and the C terminus (CT) of formins have also been shown to regulate actin assembly by directly interacting with actin. Here, to better understand the biological activities of DAAM, we studied the role of DAD-CT regions of Drosophila DAAM in its interaction with actin with in vitro biochemical and in vivo genetic approaches. We found that the DAD-CT region binds actin in vitro and that its main actin-binding element is the CT region, which does not influence actin dynamics on its own. However, we also found that it can tune the nucleating activity and the filament end-interaction properties of DAAM in an FH2 domain-dependent manner. We also demonstrate that DAD-CT makes the FH2 domain more efficient in antagonizing with capping protein. Consistently, in vivo data suggested that the CT region contributes to DAAM-mediated filopodia formation and dynamics in primary neurons. In conclusion, our results demonstrate that the CT region of DAAM plays an important role in actin assembly regulation in a biological context.

The formin DAAM is required for coordination of the actin and microtubule cytoskeleton in axonal growth cones.

Directed axonal growth depends on correct coordination of the actin and microtubule cytoskeleton in the growth cone. However, despite the relatively large number of proteins implicated in actin-microtubule crosstalk, the mechanisms whereby actin polymerization is coupled to microtubule stabilization and advancement in the peripheral growth cone remained largely unclear. Here, we identified the formin Dishevelled-associated activator of morphogenesis (DAAM) as a novel factor playing a role in concerted regulation of actin and microtubule remodeling in Drosophilamelanogaster primary neurons. In vitro, DAAM binds to F-actin as well as to microtubules and has the ability to crosslink the two filament systems. Accordingly, DAAM associates with the neuronal cytoskeleton, and a significant fraction of DAAM accumulates at places where the actin filaments overlap with that of microtubules. Loss of DAAM affects growth cone and microtubule morphology, and several aspects of microtubule dynamics; and biochemical and cellular assays revealed a microtubule stabilization activity and binding to the microtubule tip protein EB1. Together, these data suggest that, besides operating as an actin assembly factor, DAAM is involved in linking actin remodeling in filopodia to microtubule stabilization during axonal growth.

Biochemical Activities of the Wiskott-Aldrich Syndrome Homology Region 2 Domains of Sarcomere Length Short (SALS) Protein.

Drosophila melanogaster sarcomere length short (SALS) is a recently identified Wiskott-Aldrich syndrome protein homology 2 (WH2) domain protein involved in skeletal muscle thin filament regulation. SALS was shown to be important for the establishment of the proper length and organization of sarcomeric actin filaments. Here, we present the first detailed characterization of the biochemical activities of the tandem WH2 domains of SALS (SALS-WH2). Our results revealed that SALS-WH2 binds both monomeric and filamentous actin and shifts the monomer-filament equilibrium toward the monomeric actin. In addition, SALS-WH2 can bind to but fails to depolymerize phalloidin- or jasplakinolide-bound actin filaments. These interactions endow SALS-WH2 with the following two major activities in the regulation of actin dynamics: SALS-WH2 sequesters actin monomers into non-polymerizable complexes and enhances actin filament disassembly by severing, which is modulated by tropomyosin. We also show that profilin does not influence the activities of the WH2 domains of SALS in actin dynamics. In conclusion, the tandem WH2 domains of SALS are multifunctional regulators of actin dynamics. Our findings suggest that the activities of the WH2 domains do not reconstitute the presumed biological function of the full-length protein. Consequently, the interactions of the WH2 domains of SALS with actin must be tuned in the cellular context by other modules of the protein and/or sarcomeric components for its proper functioning.

The Formin DAAM Functions as Molecular Effector of the Planar Cell Polarity Pathway during Axonal Development in Drosophila.

Recent studies established that the planar cell polarity (PCP) pathway is critical for various aspects of nervous system development and function, including axonal guidance. Although it seems clear that PCP signaling regulates actin dynamics, the mechanisms through which this occurs remain elusive. Here, we establish a functional link between the PCP system and one specific actin regulator, the formin DAAM, which has previously been shown to be required for embryonic axonal morphogenesis and filopodia formation in the growth cone. We show that dDAAM also plays a pivotal role during axonal growth and guidance in the adult Drosophila mushroom body, a brain center for learning and memory. By using a combination of genetic and biochemical assays, we demonstrate that Wnt5 and the PCP signaling proteins Frizzled, Strabismus, and Dishevelled act in concert with the small GTPase Rac1 to activate the actin assembly functions of dDAAM essential for correct targeting of mushroom body axons. Collectively, these data suggest that dDAAM is used as a major molecular effector of the PCP guidance pathway. By uncovering a signaling system from the Wnt5 guidance cue to an actin assembly factor, we propose that the Wnt5/PCP navigation system is linked by dDAAM to the regulation of the growth cone actin cytoskeleton, and thereby growth cone behavior, in a direct way.

Cdc42 and formin activity control non-muscle myosin dynamics during Drosophila heart morphogenesis.

During heart formation, a network of transcription factors and signaling pathways guide cardiac cell fate and differentiation, but the genetic mechanisms orchestrating heart assembly and lumen formation remain unclear. Here, we show that the small GTPase Cdc42 is essential for Drosophila melanogaster heart morphogenesis and lumen formation. Cdc42 genetically interacts with the cardiogenic transcription factor tinman; with dDAAM which belongs to the family of actin organizing formins; and with zipper, which encodes nonmuscle myosin II. Zipper is required for heart lumen formation, and its spatiotemporal activity at the prospective luminal surface is controlled by Cdc42. Heart-specific expression of activated Cdc42, or the regulatory formins dDAAM and Diaphanous caused mislocalization of Zipper and induced ectopic heart lumina, as characterized by luminal markers such as the extracellular matrix protein Slit. Placement of Slit at the lumen surface depends on Cdc42 and formin function. Thus, Cdc42 and formins play pivotal roles in heart lumen formation through the spatiotemporal regulation of the actomyosin network.

DAAM is required for thin filament formation and Sarcomerogenesis during muscle development in Drosophila.

During muscle development, myosin and actin containing filaments assemble into the highly organized sarcomeric structure critical for muscle function. Although sarcomerogenesis clearly involves the de novo formation of actin filaments, this process remained poorly understood. Here we show that mouse and Drosophila members of the DAAM formin family are sarcomere-associated actin assembly factors enriched at the Z-disc and M-band. Analysis of dDAAM mutants revealed a pivotal role in myofibrillogenesis of larval somatic muscles, indirect flight muscles and the heart. We found that loss of dDAAM function results in multiple defects in sarcomere development including thin and thick filament disorganization, Z-disc and M-band formation, and a near complete absence of the myofibrillar lattice. Collectively, our data suggest that dDAAM is required for the initial assembly of thin filaments, and subsequently it promotes filament elongation by assembling short actin polymers that anneal to the pointed end of the growing filaments, and by antagonizing the capping protein Tropomodulin.