AbstractIdentifying active compounds for a target is a time- and resource-intensive task in early drug discovery. Accurate bioactivity prediction using morphological profiles could streamline the process, enabling smaller, more focused compound screens. We investigate the potential of deep learning on unrefined single-concentration activity readouts and Cell Painting data, to predict compound activity across 140 diverse assays. We observe an average ROC-AUC of 0.744 ± 0.108 with 62% of assays achieving ≥0.7, 30% ≥0.8, and 7% ≥0.9. In many cases, the high prediction performance can be achieved using only brightfield images instead of multichannel fluorescence images. A comprehensive analysis shows that Cell Painting-based bioactivity prediction is robust across assay types, technologies, and target classes, with cell-based assays and kinase targets being particularly well-suited for prediction. Experimental validation confirms the enrichment of active compounds. Our findings indicate that models trained on Cell Painting data, combined with a small set of single-concentration data points, can reliably predict the activity of a compound library across diverse targets and assays while maintaining high hit rates and scaffold diversity. This approach has the potential to reduce the size of screening campaigns, saving time and resources, and enabling primary screening with more complex assays.

Abstract Current research on metabolic disorders and diabetes relies on animal models because multi-organ diseases cannot be well studied with standard in vitro assays. Here, we have connected cell models of key metabolic organs, the pancreas and liver, on a microfluidic chip to enable diabetes research in a human-based in vitro system. Aided by mechanistic mathematical modeling, we demonstrate that hyperglycemia and high cortisone concentration induce glucose dysregulation in the pancreas-liver microphysiological system (MPS), mimicking a diabetic phenotype seen in patients with glucocorticoid-induced diabetes. In this diseased condition, the pancreas-liver MPS displays beta-cell dysfunction, steatosis, elevated ketone-body secretion, increased glycogen storage, and upregulated gluconeogenic gene expression. Conversely, a physiological culture condition maintains glucose tolerance and beta-cell function. This method was reproducible in two laboratories and was effective in multiple pancreatic islet donors. The model also provides a platform to identify new therapeutic proteins, as demonstrated with a combined transcriptome and proteome analysis.

ABSTRACTCurrent research on metabolic disorders such as type 2 diabetes relies on animal models because multi-organ diseases cannot be well studied with the standardin vitroassays. Here, we connect models of key metabolism organs, pancreas and liver, on a microfluidic chip to enable diabetes research in a human-based preclinical system. Aided by mechanistic mathematical modelling, we developed a two-organ microphysiological system (MPS) that replicates clinically-relevant phenotypes of diabetic dysregulation both in the liver and pancreas compartments. Exposure to hyperglycemia and high cortisone created a diseased pancreas-liver MPS which displayed beta-cell dysfunction, steatosis, elevated ketone-body secretion, increased glycogen storage, and upregulated gluconeogenic machinery. In turn, normoglycemia and physiological cortisone concentration maintained glucose tolerance and stable liver and beta-cell functions. This method was evaluated for repeatability in two laboratories and was effective in multiple pancreatic islet donors. The model also provides a platform to identify new therapeutic targets as demonstrated with a liver-secreted IL-1R2 protein that induced islet proliferation.

High content imaging assays can capture rich phenotypic response data for large sets of compound treatments, aiding in the characterization and discovery of novel drugs. However, extracting representative features from high content images that can capture subtle nuances in phenotypes remains challenging. The lack of high-quality labels makes it difficult to achieve satisfactory results with supervised deep learning. Self-Supervised learning methods have shown great success on natural images, and offer an attractive alternative also to microscopy images. However, we find that self-supervised learning techniques underperform on high content imaging assays. One challenge is the undesirable domain shifts present in the data known as batch effects, which are caused by biological noise or uncontrolled experimental conditions. To this end, we introduce Cross-Domain Consistency Learning (CDCL), a self-supervised approach that is able to learn in the presence of batch effects. CDCL enforces the learning of biological similarities while disregarding undesirable batch-specific signals, leading to more useful and versatile representations. These features are organised according to their morphological changes and are more useful for downstream tasks -- such as distinguishing treatments and mechanism of action.

In order to avoid a prolonged pro-inflammatory neutrophil response, signaling downstream of an agonist-activated G protein-coupled receptor (GPCR) has to be rapidly terminated. Among the family of GPCR kinases (GRKs) that regulate receptor phosphorylation and signaling termination, GRK2, which is highly expressed by immune cells, plays an important role. The medium chain fatty acid receptor GPR84 as well as formyl peptide receptor 2 (FPR2), receptors expressed in neutrophils, play a key role in regulating inflammation. In this study, we investigated the effects of GRK2 inhibitors on neutrophil functions induced by GPR84 and FPR2 agonists. GRK2 was shown to be expressed in human neutrophils and analysis of subcellular fractions revealed a cytosolic localization. The GRK2 inhibitors enhanced and prolonged neutrophil production of reactive oxygen species (ROS) induced by GPR84- but not FPR2-agonists, suggesting a receptor selective function of GRK2. This suggestion was supported by β-arrestin recruitment data. The ROS production induced by a non β-arrestin recruiting GPR84 agonist was not affected by the GRK2 inhibitor. Termination of this β-arrestin independent response relied, similar to the response induced by FPR2 agonists, primarily on the actin cytoskeleton. In summary, we show that GPR84 utilizes GRK2 in concert with β-arrestin and actin cytoskeleton dependent processes to fine-tune the activity of the ROS generating NADPH-oxidase in neutrophils.

PROteolysis TArgeting Chimeras (PROTACs) use the ubiquitin-proteasome system to degrade a protein of interest for therapeutic benefit. Advances made in targeted protein degradation technology have been remarkable, with several molecules having moved into clinical studies. However, robust routes to assess and better understand the safety risks of PROTACs need to be identified, which is an essential step toward delivering efficacious and safe compounds to patients. In this work, we used Cell Painting, an unbiased high-content imaging method, to identify phenotypic signatures of PROTACs. Chemical clustering and model prediction allowed the identification of a mitotoxicity signature that could not be expected by screening the individual PROTAC components. The data highlighted the benefit of unbiased phenotypic methods for identifying toxic signatures and the potential to impact drug design.

Cyclin A2 is a key regulator of the cell cycle, implicated both in DNA replication and mitotic entry. Cyclin A2 participates in feedback loops that activate mitotic kinases in G2 phase, but why active Cyclin A2-CDK2 during the S phase does not trigger mitotic kinase activation remains unclear. Here, we describe a change in localisation of Cyclin A2 from being only nuclear to both nuclear and cytoplasmic at the S/G2 border. We find that Cyclin A2-CDK2 can activate the mitotic kinase PLK1 through phosphorylation of Bora, and that only cytoplasmic Cyclin A2 interacts with Bora and PLK1. Expression of predominately cytoplasmic Cyclin A2 or phospho-mimicking PLK1 T210D can partially rescue a G2 arrest caused by Cyclin A2 depletion. Cytoplasmic presence of Cyclin A2 is restricted by p21, in particular after DNA damage. Cyclin A2 chromatin association during DNA replication and additional mechanisms contribute to Cyclin A2 localisation change in the G2 phase. We find no evidence that such mechanisms involve G2 feedback loops and suggest that cytoplasmic appearance of Cyclin A2 at the S/G2 transition functions as a trigger for mitotic kinase activation.

As a result of stress, injury, or aging, cardiac fibrosis is characterized by excessive deposition of extracellular matrix (ECM) components resulting in pathological remodeling, tissue stiffening, ventricular dilatation, and cardiac dysfunction that contribute to heart failure (HF) and eventually death. Currently, there are no effective therapies specifically targeting cardiac fibrosis, partially due to limited understanding of the pathological mechanisms and the lack of predictive in vitro models for high-throughput screening of antifibrotic compounds. The use of more relevant cell models, three-dimensional (3D) models, and coculture systems, together with high-content imaging (HCI) and machine learning (ML)-based image analysis, is expected to improve predictivity and throughput of in vitro models for cardiac fibrosis. In this review, we present an overview of available in vitro assays for cardiac fibrosis. We highlight the potential of more physiological 3D cardiac organoids and coculture systems and discuss HCI and automated artificial intelligence (AI)-based image analysis as key methods able to capture the complexity of cardiac fibrosis in vitro. As 3D and coculture models will soon be sufficiently mature for application in large-scale preclinical drug discovery, we expect the combination of more relevant models and high-content analysis to greatly increase translation from in vitro to in vivo models and facilitate the discovery of novel targets and drugs against cardiac fibrosis.

Cells recovering from the G2/M DNA damage checkpoint rely more on Aurora A-PLK1 signaling than cells progressing through an unperturbed G2 phase, but the reason for this discrepancy is not known. Here, we devised a method based on a FRET reporter for PLK1 activity to sort cells in distinct populations within G2 phase. We employed mass spectroscopy to characterize changes in protein levels through an unperturbed G2 phase and validated that ATAD2 levels decrease in a proteasome-dependent manner. Comparing unperturbed cells with cells recovering from DNA damage, we note that at similar PLK1 activities, recovering cells contain higher levels of Cyclin B1 and increased phosphorylation of CDK1 targets. The increased Cyclin B1 levels are due to continuous Cyclin B1 production during a DNA damage response and are sustained until mitosis. Whereas partial inhibition of PLK1 suppresses mitotic entry more efficiently when cells recover from a checkpoint, partial inhibition of CDK1 suppresses mitotic entry more efficiently in unperturbed cells. Our findings provide a resource for proteome changes during G2 phase, show that the mitotic entry network is rewired during a DNA damage response, and suggest that the bottleneck for mitotic entry shifts from CDK1 to PLK1 after DNA damage.

A key feature in the pathogenesis of heart failure is cardiac fibrosis, but effective treatments that specifically target cardiac fibrosis are currently not available. A major impediment to progress has been the lack of reliable in vitro models with sufficient throughput to screen for activity against cardiac fibrosis. Here, we established cell culture conditions in micro-well format that support extracellular deposition of mature collagen from primary human cardiac fibroblasts - a hallmark of cardiac fibrosis. Based on robust biochemical characterization we developed a high-content phenotypic screening platform, that allows for high-throughput identification of compounds with activity against cardiac fibrosis. Our platform correctly identifies compounds acting on known cardiac fibrosis pathways. Moreover, it can detect anti-fibrotic activity for compounds acting on targets that have not previously been reported in in vitro cardiac fibrosis assays. Taken together, our experimental approach provides a powerful platform for high-throughput screening of anti-fibrotic compounds as well as discovery of novel targets to develop new therapeutic strategies for heart failure.

Identification of small molecules with the potential to selectively proliferate cardiac progenitor cells (CPCs) will aid our understanding of the signaling pathways and mechanisms involved and could ultimately provide tools for regenerative therapies for the treatment of post‐MI cardiac dysfunction. We have used an in vitro human induced pluripotent stem cell‐derived CPC model to screen a 10,000‐compound library containing molecules representing different target classes and compounds reported to modulate the phenotype of stem or primary cells. The primary readout of this phenotypic screen was proliferation as measured by nuclear count. We identified retinoic acid receptor (RAR) agonists as potent proliferators of CPCs. The CPCs retained their progenitor phenotype following proliferation and the identified RAR agonists did not proliferate human cardiac fibroblasts, the major cell type in the heart. In addition, the RAR agonists were able to proliferate an independent source of CPCs, HuES6. The RAR agonists had a time‐of‐differentiation‐dependent effect on the HuES6‐derived CPCs. At 4 days of differentiation, treatment with retinoic acid induced differentiation of the CPCs to atrial cells. However, after 5 days of differentiation treatment with RAR agonists led to an inhibition of terminal differentiation to cardiomyocytes and enhanced the proliferation of the cells. RAR agonists, at least transiently, enhance the proliferation of human CPCs, at the expense of terminal cardiac differentiation. How this mechanism translates in vivo to activate endogenous CPCs and whether enhancing proliferation of these rare progenitor cells is sufficient to enhance cardiac repair remains to be investigated.

After DNA damage, the cell cycle is arrested to avoid propagation of mutations. Arrest in G2 phase is initiated by ATM-/ATR-dependent signaling that inhibits mitosis-promoting kinases such as Plk1. At the same time, Plk1 can counteract ATR-dependent signaling and is required for eventual resumption of the cell cycle. However, what determines when Plk1 activity can resume remains unclear. Here, we use FRET-based reporters to show that a global spread of ATM activity on chromatin and phosphorylation of ATM targets including KAP1 control Plk1 re-activation. These phosphorylations are rapidly counteracted by the chromatin-bound phosphatase Wip1, allowing cell cycle restart despite persistent ATM activity present at DNA lesions. Combining experimental data and mathematical modeling, we propose a model for how the minimal duration of cell cycle arrest is controlled. Our model shows how cell cycle restart can occur before completion of DNA repair and suggests a mechanism for checkpoint adaptation in human cells.

In response to DNA damage, a cell can be forced to permanently exit the cell cycle and become senescent. Senescence provides an early barrier against tumor development by preventing proliferation of cells with damaged DNA. By studying single cells, we show that Cdk activity persists after DNA damage until terminal cell cycle exit. This low level of Cdk activity not only allows cell cycle progression, but also promotes cell cycle exit at a decision point in G2 phase. We find that residual Cdk1/2 activity is required for efficient p21 production, allowing for nuclear sequestration of Cyclin B1, subsequent APC/C-Cdh1-dependent degradation of mitotic inducers and induction of senescence. We suggest that the same activity that triggers mitosis in an unperturbed cell cycle enforces senescence in the presence of DNA damage, ensuring a robust response when most needed.

Immunofluorescence can be a powerful tool to detect protein levels, intracellular localization, and post-translational modifications. However, standard immunofluorescence provides only a still picture and thus lacks temporal information. Here, we describe a method to extract temporal information from immunofluorescence images of fixed cells. In addition, we provide an optional protocol that uses micropatterns, which increases the accuracy of the method. These methods allow assessing how protein levels, intracellular localization, and post-translational modifications change through the cell cycle.

The Spumaretrovirinae, or foamy viruses (FVs) are complex retroviruses that infect many species of monkey and ape. Despite little sequence homology, FV and orthoretroviral Gag proteins perform equivalent functions, including genome packaging, virion assembly, trafficking and membrane targeting. However, there is a paucity of structural information for FVs and it is unclear how disparate FV and orthoretroviral Gag molecules share the same function. To probe the functional overlap of FV and orthoretroviral Gag we have determined the structure of a central region of Gag from the Prototype FV (PFV). The structure comprises two all α-helical domains NtDCEN and CtDCEN that although they have no sequence similarity, we show they share the same core fold as the N- (NtDCA) and C-terminal domains (CtDCA) of archetypal orthoretroviral capsid protein (CA). Moreover, structural comparisons with orthoretroviral CA align PFV NtDCEN and CtDCEN with NtDCA and CtDCA respectively. Further in vitro and functional virological assays reveal that residues making inter-domain NtDCEN—CtDCEN interactions are required for PFV capsid assembly and that intact capsid is required for PFV reverse transcription. These data provide the first information that relates the Gag proteins of Spuma and Orthoretrovirinae and suggests a common ancestor for both lineages containing an ancient CA fold.

Unlike for other retroviruses, only a few host cell factors that aid the replication of foamy viruses (FVs) via interaction with viral structural components are known. Using a yeast-two-hybrid (Y2H) screen with prototype FV (PFV) Gag protein as bait we identified human polo-like kinase 2 (hPLK2), a member of cell cycle regulatory kinases, as a new interactor of PFV capsids. Further Y2H studies confirmed interaction of PFV Gag with several PLKs of both human and rat origin. A consensus Ser-Thr/Ser-Pro (S-T/S-P) motif in Gag, which is conserved among primate FVs and phosphorylated in PFV virions, was essential for recognition by PLKs. In the case of rat PLK2, functional kinase and polo-box domains were required for interaction with PFV Gag. Fluorescently-tagged PFV Gag, through its chromatin tethering function, selectively relocalized ectopically expressed eGFP-tagged PLK proteins to mitotic chromosomes in a Gag STP motif-dependent manner, confirming a specific and dominant nature of the Gag-PLK interaction in mammalian cells. The functional relevance of the Gag-PLK interaction was examined in the context of replication-competent FVs and single-round PFV vectors. Although STP motif mutated viruses displayed wild type (wt) particle release, RNA packaging and intra-particle reverse transcription, their replication capacity was decreased 3-fold in single-cycle infections, and up to 20-fold in spreading infections over an extended time period. Strikingly similar defects were observed when cells infected with single-round wt Gag PFV vectors were treated with a pan PLK inhibitor. Analysis of entry kinetics of the mutant viruses indicated a post-fusion defect resulting in delayed and reduced integration, which was accompanied with an enhanced preference to integrate into heterochromatin. We conclude that interaction between PFV Gag and cellular PLK proteins is important for early replication steps of PFV within host cells.

In response to DNA damage a cell can be forced to permanently exit the cell cycle and become senescent. Senescence provides an early barrier against tumor development by preventing proliferation of cells with damaged DNA. By studying single cells, we show that Cdk activity is retained after DNA damage until terminal cell cycle exit. The low level of Cdk activity not only allows cell cycle progression, but also forces cell cycle exit at a decision point in G2 phase. We find that Cdk activity stimulates p21 production, leading to nuclear sequestration of Cyclin B1, subsequent APC/CCdh1-dependent degradation of mitotic inducers and induction of senescence. We suggest that the same activity that triggers mitosis in an unperturbed cell cycle drives senescence in the presence of DNA damage, ensuring a robust response when most needed.