Phase Separation as a Missing Mechanism for Interpretation of Disease Mutations.

It is unclear how disease mutations impact intrinsically disordered protein regions (IDRs), which lack a stable folded structure. These mutations, while prevalent in disease, are frequently neglected or annotated as variants of unknown significance. Biomolecular phase separation, a physical process often mediated by IDRs, has increasingly appreciated roles in cellular organization and regulation. We find that autism spectrum disorder (ASD)- and cancer-associated proteins are enriched for predicted phase separation propensities, suggesting that IDR mutations disrupt phase separation in key cellular processes. More generally, we hypothesize that combinations of small-effect IDR mutations perturb phase separation, potentially contributing to "missing heritability" in complex disease susceptibility.

Condensation of Ded1p Promotes a Translational Switch from Housekeeping to Stress Protein Production.

Cells sense elevated temperatures and mount an adaptive heat shock response that involves changes in gene expression, but the underlying mechanisms, particularly on the level of translation, remain unknown. Here we report that, in budding yeast, the essential translation initiation factor Ded1p undergoes heat-induced phase separation into gel-like condensates. Using ribosome profiling and an in vitro translation assay, we reveal that condensate formation inactivates Ded1p and represses translation of housekeeping mRNAs while promoting translation of stress mRNAs. Testing a variant of Ded1p with altered phase behavior as well as Ded1p homologs from diverse species, we demonstrate that Ded1p condensation is adaptive and fine-tuned to the maximum growth temperature of the respective organism. We conclude that Ded1p condensation is an integral part of an extended heat shock response that selectively represses translation of housekeeping mRNAs to promote survival under conditions of severe heat stress.

Yeast Proteome Dynamics from Single Cell Imaging and Automated Analysis.

Proteomics has proved invaluable in generating large-scale quantitative data; however, the development of systems approaches for examining the proteome in vivo has lagged behind. To evaluate protein abundance and localization on a proteome scale, we exploited the yeast GFP-fusion collection in a pipeline combining automated genetics, high-throughput microscopy, and computational feature analysis. We developed an ensemble of binary classifiers to generate localization data from single-cell measurements and constructed maps of ∼3,000 proteins connected to 16 localization classes. To survey proteome dynamics in response to different chemical and genetic stimuli, we measure proteome-wide abundance and localization and identified changes over time. We analyzed >20 million cells to identify dynamic proteins that redistribute among multiple localizations in hydroxyurea, rapamycin, and in an rpd3Δ background. Because our localization and abundance data are quantitative, they provide the opportunity for many types of comparative studies, single cell analyses, modeling, and prediction. VIDEO ABSTRACT.

Cdk1 Controls Global Epigenetic Landscape in Embryonic Stem Cells.

The cyclin-dependent kinase 1 (Cdk1) drives cell division. To uncover additional functions of Cdk1, we generated knockin mice expressing an analog-sensitive version of Cdk1 in place of wild-type Cdk1. In our study, we focused on embryonic stem cells (ESCs), because this cell type displays particularly high Cdk1 activity. We found that in ESCs, a large fraction of Cdk1 substrates is localized on chromatin. Cdk1 phosphorylates many proteins involved in epigenetic regulation, including writers and erasers of all major histone marks. Consistent with these findings, inhibition of Cdk1 altered histone-modification status of ESCs. High levels of Cdk1 in ESCs phosphorylate and partially inactivate Dot1l, the H3K79 methyltransferase responsible for placing activating marks on gene bodies. Decrease of Cdk1 activity during ESC differentiation de-represses Dot1l, thereby allowing coordinated expression of differentiation genes. These analyses indicate that Cdk1 functions to maintain the epigenetic identity of ESCs.

Systematic identification of conditionally folded intrinsically disordered regions by AlphaFold2.

The AlphaFold Protein Structure Database contains predicted structures for millions of proteins. For the majority of human proteins that contain intrinsically disordered regions (IDRs), which do not adopt a stable structure, it is generally assumed that these regions have low AlphaFold2 confidence scores that reflect low-confidence structural predictions. Here, we show that AlphaFold2 assigns confident structures to nearly 15% of human IDRs. By comparison to experimental NMR data for a subset of IDRs that are known to conditionally fold (i.e., upon binding or under other specific conditions), we find that AlphaFold2 often predicts the structure of the conditionally folded state. Based on databases of IDRs that are known to conditionally fold, we estimate that AlphaFold2 can identify conditionally folding IDRs at a precision as high as 88% at a 10% false positive rate, which is remarkable considering that conditionally folded IDR structures were minimally represented in its training data. We find that human disease mutations are nearly fivefold enriched in conditionally folded IDRs over IDRs in general and that up to 80% of IDRs in prokaryotes are predicted to conditionally fold, compared to less than 20% of eukaryotic IDRs. These results indicate that a large majority of IDRs in the proteomes of human and other eukaryotes function in the absence of conditional folding, but the regions that do acquire folds are more sensitive to mutations. We emphasize that the AlphaFold2 predictions do not reveal functionally relevant structural plasticity within IDRs and cannot offer realistic ensemble representations of conditionally folded IDRs.

Improving the performance of supervised deep learning for regulatory genomics using phylogenetic augmentation.

MOTIVATION: Supervised deep learning is used to model the complex relationship between genomic sequence and regulatory function. Understanding how these models make predictions can provide biological insight into regulatory functions. Given the complexity of the sequence to regulatory function mapping (the cis-regulatory code), it has been suggested that the genome contains insufficient sequence variation to train models with suitable complexity. Data augmentation is a widely used approach to increase the data variation available for model training, however current data augmentation methods for genomic sequence data are limited. RESULTS: Inspired by the success of comparative genomics, we show that augmenting genomic sequences with evolutionarily related sequences from other species, which we term phylogenetic augmentation, improves the performance of deep learning models trained on regulatory genomic sequences to predict high-throughput functional assay measurements. Additionally, we show that phylogenetic augmentation can rescue model performance when the training set is down-sampled and permits deep learning on a real-world small dataset, demonstrating that this approach improves data efficiency. Overall, this data augmentation method represents a solution for improving model performance that is applicable to many supervised deep-learning problems in genomics. AVAILABILITY AND IMPLEMENTATION: The open-source GitHub repository agduncan94/phylogenetic_augmentation_paper includes the code for rerunning the analyses here and recreating the figures.

A New Phase of Networking: The Molecular Composition and Regulatory Dynamics of Mammalian Stress Granules.

Stress granules (SGs) are cytosolic biomolecular condensates that form in response to cellular stress. Weak, multivalent interactions between their protein and RNA constituents drive their rapid, dynamic assembly through phase separation coupled to percolation. Though a consensus model of SG function has yet to be determined, their perceived implication in cytoprotective processes (e.g., antiviral responses and inhibition of apoptosis) and possible role in the pathogenesis of various neurodegenerative diseases (e.g., amyotrophic lateral sclerosis and frontotemporal dementia) have drawn great interest. Consequently, new studies using numerous cell biological, genetic, and proteomic methods have been performed to unravel the mechanisms underlying SG formation, organization, and function and, with them, a more clearly defined SG proteome. Here, we provide a consensus SG proteome through literature curation and an update of the user-friendly database RNAgranuleDB to version 2.0 (http://rnagranuledb.lunenfeld.ca/). With this updated SG proteome, we use next-generation phase separation prediction tools to assess the predisposition of SG proteins for phase separation and aggregation. Next, we analyze the primary sequence features of intrinsically disordered regions (IDRs) within SG-resident proteins. Finally, we review the protein- and RNA-level determinants, including post-translational modifications (PTMs), that regulate SG composition and assembly/disassembly dynamics.

A flexible repertoire of transcription factor binding sites and a diversity threshold determines enhancer activity in embryonic stem cells.

Transcriptional enhancers are critical for development and phenotype evolution and are often mutated in disease contexts; however, even in well-studied cell types, the sequence code conferring enhancer activity remains unknown. To examine the enhancer regulatory code for pluripotent stem cells, we identified genomic regions with conserved binding of multiple transcription factors in mouse and human embryonic stem cells (ESCs). Examination of these regions revealed that they contain on average 12.6 conserved transcription factor binding site (TFBS) sequences. Enriched TFBSs are a diverse repertoire of 70 different sequences representing the binding sequences of both known and novel ESC regulators. Using a diverse set of TFBSs from this repertoire was sufficient to construct short synthetic enhancers with activity comparable to native enhancers. Site-directed mutagenesis of conserved TFBSs in endogenous enhancers or TFBS deletion from synthetic sequences revealed a requirement for 10 or more different TFBSs. Furthermore, specific TFBSs, including the POU5F1:SOX2 comotif, are dispensable, despite cobinding the POU5F1 (also known as OCT4), SOX2, and NANOG master regulators of pluripotency. These findings reveal that a TFBS sequence diversity threshold overrides the need for optimized regulatory grammar and individual TFBSs that recruit specific master regulators.

A functionally divergent intrinsically disordered region underlying the conservation of stochastic signaling.

Stochastic signaling dynamics expand living cells' information processing capabilities. An increasing number of studies report that regulators encode information in their pulsatile dynamics. The evolutionary mechanisms that lead to complex signaling dynamics remain uncharacterized, perhaps because key interactions of signaling proteins are encoded in intrinsically disordered regions (IDRs), whose evolution is difficult to analyze. Here we focused on the IDR that controls the stochastic pulsing dynamics of Crz1, a transcription factor in fungi downstream of the widely conserved calcium signaling pathway. We find that Crz1 IDRs from anciently diverged fungi can all respond transiently to calcium stress; however, only Crz1 IDRs from the Saccharomyces clade support pulsatility, encode extra information, and rescue fitness in competition assays, while the Crz1 IDRs from distantly related fungi do none of the three. On the other hand, we find that Crz1 pulsing is conserved in the distantly related fungi, consistent with the evolutionary model of stabilizing selection on the signaling phenotype. Further, we show that a calcineurin docking site in a specific part of the IDRs appears to be sufficient for pulsing and show evidence for a beneficial increase in the relative calcineurin affinity of this docking site. We propose that evolutionary flexibility of functionally divergent IDRs underlies the conservation of stochastic signaling by stabilizing selection.

Architect: A tool for aiding the reconstruction of high-quality metabolic models through improved enzyme annotation.

Constraint-based modeling is a powerful framework for studying cellular metabolism, with applications ranging from predicting growth rates and optimizing production of high value metabolites to identifying enzymes in pathogens that may be targeted for therapeutic interventions. Results from modeling experiments can be affected at least in part by the quality of the metabolic models used. Reconstructing a metabolic network manually can produce a high-quality metabolic model but is a time-consuming task. At the same time, current methods for automating the process typically transfer metabolic function based on sequence similarity, a process known to produce many false positives. We created Architect, a pipeline for automatic metabolic model reconstruction from protein sequences. First, it performs enzyme annotation through an ensemble approach, whereby a likelihood score is computed for an EC prediction based on predictions from existing tools; for this step, our method shows both increased precision and recall compared to individual tools. Next, Architect uses these annotations to construct a high-quality metabolic network which is then gap-filled based on likelihood scores from the ensemble approach. The resulting metabolic model is output in SBML format, suitable for constraints-based analyses. Through comparisons of enzyme annotations and curated metabolic models, we demonstrate improved performance of Architect over other state-of-the-art tools, notably with higher precision and recall on the eukaryote C. elegans and when compared to UniProt annotations in two bacterial species. Code for Architect is available at https://github.com/ParkinsonLab/Architect. For ease-of-use, Architect can be readily set up and utilized using its Docker image, maintained on Docker Hub.

Discovering molecular features of intrinsically disordered regions by using evolution for contrastive learning.

A major challenge to the characterization of intrinsically disordered regions (IDRs), which are widespread in the proteome, but relatively poorly understood, is the identification of molecular features that mediate functions of these regions, such as short motifs, amino acid repeats and physicochemical properties. Here, we introduce a proteome-scale feature discovery approach for IDRs. Our approach, which we call "reverse homology", exploits the principle that important functional features are conserved over evolution. We use this as a contrastive learning signal for deep learning: given a set of homologous IDRs, the neural network has to correctly choose a held-out homolog from another set of IDRs sampled randomly from the proteome. We pair reverse homology with a simple architecture and standard interpretation techniques, and show that the network learns conserved features of IDRs that can be interpreted as motifs, repeats, or bulk features like charge or amino acid propensities. We also show that our model can be used to produce visualizations of what residues and regions are most important to IDR function, generating hypotheses for uncharacterized IDRs. Our results suggest that feature discovery using unsupervised neural networks is a promising avenue to gain systematic insight into poorly understood protein sequences.

Efficacy and Safety of Degludec versus Glargine in Type 2 Diabetes.

BACKGROUND: Degludec is an ultralong-acting, once-daily basal insulin that is approved for use in adults, adolescents, and children with diabetes. Previous open-label studies have shown lower day-to-day variability in the glucose-lowering effect and lower rates of hypoglycemia among patients who received degludec than among those who received basal insulin glargine. However, data are lacking on the cardiovascular safety of degludec. METHODS: We randomly assigned 7637 patients with type 2 diabetes to receive either insulin degludec (3818 patients) or insulin glargine U100 (3819 patients) once daily between dinner and bedtime in a double-blind, treat-to-target, event-driven cardiovascular outcomes trial. The primary composite outcome in the time-to-event analysis was the first occurrence of an adjudicated major cardiovascular event (death from cardiovascular causes, nonfatal myocardial infarction, or nonfatal stroke) with a prespecified noninferiority margin of 1.3. Adjudicated severe hypoglycemia, as defined by the American Diabetes Association, was the prespecified, multiplicity-adjusted secondary outcome. RESULTS: Of the patients who underwent randomization, 6509 (85.2%) had established cardiovascular disease, chronic kidney disease, or both. At baseline, the mean age was 65.0 years, the mean duration of diabetes was 16.4 years, and the mean (±SD) glycated hemoglobin level was 8.4±1.7%; 83.9% of the patients were receiving insulin. The primary outcome occurred in 325 patients (8.5%) in the degludec group and in 356 (9.3%) in the glargine group (hazard ratio, 0.91; 95% confidence interval, 0.78 to 1.06; P<0.001 for noninferiority). At 24 months, the mean glycated hemoglobin level was 7.5±1.2% in each group, whereas the mean fasting plasma glucose level was significantly lower in the degludec group than in the glargine group (128±56 vs. 136±57 mg per deciliter, P<0.001). Prespecified adjudicated severe hypoglycemia occurred in 187 patients (4.9%) in the degludec group and in 252 (6.6%) in the glargine group, for an absolute difference of 1.7 percentage points (rate ratio, 0.60; P<0.001 for superiority; odds ratio, 0.73; P<0.001 for superiority). Rates of adverse events did not differ between the two groups. CONCLUSIONS: Among patients with type 2 diabetes at high risk for cardiovascular events, degludec was noninferior to glargine with respect to the incidence of major cardiovascular events. (Funded by Novo Nordisk and others; DEVOTE ClinicalTrials.gov number, NCT01959529 .).

YeastSpotter: accurate and parameter-free web segmentation for microscopy images of yeast cells.

SUMMARY: We introduce YeastSpotter, a web application for the segmentation of yeast microscopy images into single cells. YeastSpotter is user-friendly and generalizable, reducing the computational expertise required for this critical preprocessing step in many image analysis pipelines. AVAILABILITY AND IMPLEMENTATION: YeastSpotter is available at http://yeastspotter.csb.utoronto.ca/. Code is available at https://github.com/alexxijielu/yeast_segmentation. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Variational infinite heterogeneous mixture model for semi-supervised clustering of heart enhancers.

MOTIVATION: Mammalian genomes can contain thousands of enhancers but only a subset are actively driving gene expression in a given cellular context. Integrated genomic datasets can be harnessed to predict active enhancers. One challenge in integration of large genomic datasets is the increasing heterogeneity: continuous, binary and discrete features may all be relevant. Coupled with the typically small numbers of training examples, semi-supervised approaches for heterogeneous data are needed; however, current enhancer prediction methods are not designed to handle heterogeneous data in the semi-supervised paradigm. RESULTS: We implemented a Dirichlet Process Heterogeneous Mixture model that infers Gaussian, Bernoulli and Poisson distributions over features. We derived a novel variational inference algorithm to handle semi-supervised learning tasks where certain observations are forced to cluster together. We applied this model to enhancer candidates in mouse heart tissues based on heterogeneous features. We constrained a small number of known active enhancers to appear in the same cluster, and 47 additional regions clustered with them. Many of these are located near heart-specific genes. The model also predicted 1176 active promoters, suggesting that it can discover new enhancers and promoters. AVAILABILITY AND IMPLEMENTATION: We created the 'dphmix' Python package: https://pypi.org/project/dphmix/. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Whence Blobs? Phylogenetics of functional protein condensates.

What do we know about the molecular evolution of functional protein condensation? The capacity of proteins to form biomolecular condensates (compact, protein-rich states, not bound by membranes, but still separated from the rest of the contents of the cell) appears in many cases to be bestowed by weak, transient interactions within one or between proteins. Natural selection is expected to remove or fix amino acid changes, insertions or deletions that preserve and change this condensation capacity when doing so is beneficial to the cell. A few recent studies have begun to explore this frontier of phylogenetics at the intersection of biophysics and cell biology.

Learning unsupervised feature representations for single cell microscopy images with paired cell inpainting.

Cellular microscopy images contain rich insights about biology. To extract this information, researchers use features, or measurements of the patterns of interest in the images. Here, we introduce a convolutional neural network (CNN) to automatically design features for fluorescence microscopy. We use a self-supervised method to learn feature representations of single cells in microscopy images without labelled training data. We train CNNs on a simple task that leverages the inherent structure of microscopy images and controls for variation in cell morphology and imaging: given one cell from an image, the CNN is asked to predict the fluorescence pattern in a second different cell from the same image. We show that our method learns high-quality features that describe protein expression patterns in single cells both yeast and human microscopy datasets. Moreover, we demonstrate that our features are useful for exploratory biological analysis, by capturing high-resolution cellular components in a proteome-wide cluster analysis of human proteins, and by quantifying multi-localized proteins and single-cell variability. We believe paired cell inpainting is a generalizable method to obtain feature representations of single cells in multichannel microscopy images.

Development of a hypoglycaemia risk score to identify high-risk individuals with advanced type 2 diabetes in DEVOTE.

AIMS: The ability to differentiate patient populations with type 2 diabetes at high risk of severe hypoglycaemia could impact clinical decision making. The aim of this study was to develop a risk score, using patient characteristics, that could differentiate between populations with higher and lower 2-year risk of severe hypoglycaemia among individuals at increased risk of cardiovascular disease. MATERIALS AND METHODS: Two models were developed for the risk score based on data from the DEVOTE cardiovascular outcomes trials. The first, a data-driven machine-learning model, used stepwise regression with bidirectional elimination to identify risk factors for severe hypoglycaemia. The second, a risk score based on known clinical risk factors accessible in clinical practice identified from the data-driven model, included: insulin treatment regimen; diabetes duration; sex; age; and glycated haemoglobin, all at baseline. Both the data-driven model and simple risk score were evaluated for discrimination, calibration and generalizability using data from DEVOTE, and were validated against the external LEADER cardiovascular outcomes trial dataset. RESULTS: Both the data-driven model and the simple risk score discriminated between patients at higher and lower hypoglycaemia risk, and performed similarly well based on the time-dependent area under the curve index (0.63 and 0.66, respectively) over a 2-year time horizon. CONCLUSIONS: Both the data-driven model and the simple hypoglycaemia risk score were able to discriminate between patients at higher and lower risk of severe hypoglycaemia, the latter doing so using easily accessible clinical data. The implementation of such a tool (http://www.hyporiskscore.com/) may facilitate improved recognition of, and education about, severe hypoglycaemia risk, potentially improving patient care.

Risk of severe hypoglycaemia and its impact in type 2 diabetes in DEVOTE.

AIMS: To undertake a post-hoc analysis, utilizing a hypoglycaemia risk score based on DEVOTE trial data, to investigate if a high risk of severe hypoglycaemia was associated with an increased risk of cardiovascular events, and whether reduced rates of severe hypoglycaemia in patients identified as having the highest risk affected the risk of cardiovascular outcomes. MATERIALS AND METHODS: The DEVOTE population was divided into quartiles according to patients' individual hypoglycaemia risk scores. For each quartile, the observed incidence and rate of severe hypoglycaemia, major adverse cardiovascular event (MACE) and all-cause mortality were determined to investigate whether those with the highest risk of hypoglycaemia were also at the greatest risk of MACE and all-cause mortality. In addition, treatment differences within each risk quartile [insulin degludec (degludec) vs. insulin glargine 100 units/mL (glargine U100)] in terms of severe hypoglycaemia, MACE and all-cause mortality were investigated. RESULTS: Patients with the highest risk scores had the highest rates of severe hypoglycaemia, MACE and all-cause mortality. Treatment ratios between degludec and glargine U100 in the highest risk quartile were 95% confidence interval (CI) 0.56 (0.39; 0.80) (severe hypoglycaemia), 95% CI 0.76 (0.58; 0.99) (MACE) and 95% CI 0.77 (0.55; 1.07) (all-cause mortality). CONCLUSIONS: The risk score demonstrated that a high risk of severe hypoglycaemia was associated with a high incidence of MACE and all-cause mortality and that, in this high-risk group, those treated with degludec had a lower incidence of MACE. These observations support the hypothesis that hypoglycaemia is a risk factor for cardiovascular events.

Selection maintains signaling function of a highly diverged intrinsically disordered region.

Intrinsically disordered regions (IDRs) are characterized by their lack of stable secondary or tertiary structure and comprise a large part of the eukaryotic proteome. Although these regions play a variety of signaling and regulatory roles, they appear to be rapidly evolving at the primary sequence level. To understand the functional implications of this rapid evolution, we focused on a highly diverged IDR in Saccharomyces cerevisiae that is involved in regulating multiple conserved MAPK pathways. We hypothesized that under stabilizing selection, the functional output of orthologous IDRs could be maintained, such that diverse genotypes could lead to similar function and fitness. Consistent with the stabilizing selection hypothesis, we find that diverged, orthologous IDRs can mostly recapitulate wild-type function and fitness in S. cerevisiae We also find that the electrostatic charge of the IDR is correlated with signaling output and, using phylogenetic comparative methods, find evidence for selection maintaining this quantitative molecular trait despite underlying genotypic divergence.

Parallel reorganization of protein function in the spindle checkpoint pathway through evolutionary paths in the fitness landscape that appear neutral in laboratory experiments.

Regulatory networks often increase in complexity during evolution through gene duplication and divergence of component proteins. Two models that explain this increase in complexity are: 1) adaptive changes after gene duplication, such as resolution of adaptive conflicts, and 2) non-adaptive processes such as duplication, degeneration and complementation. Both of these models predict complementary changes in the retained duplicates, but they can be distinguished by direct fitness measurements in organisms with short generation times. Previously, it has been observed that repeated duplication of an essential protein in the spindle checkpoint pathway has occurred multiple times over the eukaryotic tree of life, leading to convergent protein domain organization in its duplicates. Here, we replace the paralog pair in S. cerevisiae with a single-copy protein from a species that did not undergo gene duplication. Surprisingly, using quantitative fitness measurements in laboratory conditions stressful for the spindle-checkpoint pathway, we find no evidence that reorganization of protein function after gene duplication is beneficial. We then reconstruct several evolutionary intermediates from the inferred ancestral network to the extant one, and find that, at the resolution of our assay, there exist stepwise mutational paths from the single protein to the divergent pair of extant proteins with no apparent fitness defects. Parallel evolution has been taken as strong evidence for natural selection, but our results suggest that even in these cases, reorganization of protein function after gene duplication may be explained by neutral processes.