Search

Single-cell technologies have elucidated mechanisms responsible for immune checkpoint inhibitor (ICI) response, but are not amenable to a clinical diagnostic setting. In contrast, bulk RNA sequencing (RNA-seq) is now routine for research and clinical applications. Our workflow uses transcription factor (TF)-directed co-expression networks (regulons) inferred from single-cell RNA-seq data to deconvolute immune functional states from bulk RNA-seq data. Regulons preserve the phenotypic variation in CD45+ immune cells from metastatic melanoma samples (n=19, discovery dataset) treated with ICIs, despite reducing dimensionality by > 100-fold. Four cell states, termed exhausted T cells, monocyte lineage cells, memory T cells, and B cells were associated with therapy response; and were characterized by differentially active and cell-state specific regulons. Clustering of bulk RNA-seq melanoma samples from four independent studies (n=209, validation dataset) according to regulon-inferred scores identified four groups with significantly different response outcomes (p < 0.001). An intercellular link was established between exhausted T cells and monocyte lineage cells, whereby their cell numbers were correlated, and exhausted T cells predicted prognosis as a function of monocyte lineage cell number. The ligand–receptor expression analysis suggested that monocyte lineage cells drive exhausted T cells into terminal exhaustion through programs that regulate antigen presentation, chronic inflammation, and negative co-stimulation. Together, our results demonstrate how regulon-based characterization of cell states provide robust and functionally informative markers that can deconvolve bulk RNA-seq data to identify ICI responders.

Purpose: KRAS wild-type status is an imperfect predictor of sensitivity to anti-EGF receptor (EGFR) monoclonal antibodies in colorectal cancer, motivating efforts to identify novel molecular aberrations driving RAS. This study aimed to build a quantitative readout of RAS pathway activity to (i) uncover molecular surrogates of RAS activity specific to colorectal cancer, (ii) improve the prediction of cetuximab response in patients, and (iii) suggest new treatment strategies.Experimental Design: A model of RAS pathway activity was trained in a large colorectal cancer dataset and validated in three independent colorectal cancer patient datasets. Novel molecular traits were inferred from The Cancer Genome Atlas colorectal cancer data. The ability of the RAS model to predict resistance to cetuximab was tested in mouse xenografts and three independent patient cohorts. Drug sensitivity correlations between our model and large cell line compendiums were performed.Results: The performance of the RAS model was remarkably robust across three validation datasets. (i) Our model confirmed the heterogeneity of the RAS phenotype in KRAS wild-type patients, and suggests novel molecular traits driving its phenotype (e.g., MED12 loss, FBXW7 mutation, MAP2K4 mutation). (ii) It improved the prediction of response and progression-free survival (HR, 2.0; P < 0.01) to cetuximab compared with KRAS mutation (xenograft and patient cohorts). (iii) Our model consistently predicted sensitivity to MAP–ERK kinase (MEK) inhibitors (P < 0.01) in two cell panel screens.Conclusions: Modeling the RAS phenotype in colorectal cancer allows for the robust interrogation of RAS pathway activity across cell lines, xenografts, and patient cohorts. It demonstrates clinical utility in predicting response to anti-EGFR agents and MEK inhibitors. Clin Cancer Res; 20(1); 265–72. ©2013 AACR.

PDF file - 35K, Characteristics of the 2 metastatic CRC datasets used for evaluation of prediction of PFS.

PDF file - 32K, PFS analysis within the French cohort.

PDF file - 33K, Univariate analysis.

PDF file - 219K, Supplemental Figure 1. A We evaluated 4 "RAS" signatures across 4 CRC data sets to assess their ability to discriminate KRAS mutant samples from KRAS wild-type. Supplemental Figure 2. The lower observed RIS of KRAS codon 13 mutated samples relative to KRAS codon 12 mutated samples may be a result of the imbalanced distribution of these mutations in the training data. Supplemental Figure 3. RIS distribution by mutation of KRAS, BRAF, NRAS mutations, and wild-type. Supplemental Figure 4. RAS index score (RIS) for PIK3CA mutant samples by exon in the TCGA CRC cohort. Supplemental Figure 5. Distribution of RIS for the subset of colorectal samples (n=19) within the Cancer Cell Line Encylopedia (CCLE), separated by KRAS codon, BRAF V600E, PIK3CA (KRAS/BRAF wild-type), or wild-type (no KRAS/BRAF/PIK3CA) mutation. Supplemental Figure 6. In the mouse early passage cohort (n=26), our RAS model outperforms the KRAS/BRAF mutation model (p=0.06 vs p > .1, respectively). Supplemental Figure 7. Association of RIS and cetuximab response for metastatic colorectal patients (Khambata-Ford, et al.) Patients annotated according to treatment response: Progressive Disease (PD), Stable Disease (SD) and Partial Response (PR) as well as KRAS mutation status.

PDF file - 35K, Data sets used for evaluation of the colorectal RAS model, and reported AUC in classification of KRAS mutant from wild-type samples.

This paper investigates statistical reproducibility of the [Formula: see text]-test. We formulate reproducibility as a predictive inference problem and apply the nonparametric predictive inference method. Within our research framework, statistical reproducibility provides inference on the probability that the same test outcome would be reached, if the test were repeated under identical conditions. We present an nonparametric predictive inference algorithm to calculate the reproducibility of the [Formula: see text]-test and then use simulations to explore the reproducibility both under the null and alternative hypotheses. We then apply nonparametric predictive inference reproducibility to a real-life scenario of a preclinical experiment, which involves multiple pairwise comparisons of test groups, where different groups are given a different concentration of a drug. The aim of the experiment is to decide the concentration of the drug which is most effective. In both simulations and the application scenario, we study the relationship between reproducibility and two test statistics, the Cohen’s [Formula: see text] and the [Formula: see text]-value. We also compare the reproducibility of the [Formula: see text]-test with the reproducibility of the Wilcoxon Mann–Whitney test. Finally, we examine reproducibility for the final decision of choosing a particular dose in the multiple pairwise comparisons scenario. This paper presents advances on the topic of test reproducibility with relevance for tests used in pharmaceutical research.

The unravelling of the complexity of cellular metabolism is in its infancy. Cancer-associated genetic alterations may result in changes to cellular metabolism that aid in understanding phenotypic changes, reveal detectable metabolic signatures, or elucidate vulnerabilities to particular drugs. To understand cancer-associated metabolic transformation we performed untargeted metabolite analysis of 173 different cancer cell lines from 11 different tissues under constant conditions for 1099 different species using liquid chromatography-mass spectrometry (LC-MS). We correlate known cancer-associated mutations and gene expression programs with metabolic signatures, generating novel associations of known metabolic pathways with known cancer drivers. We show that metabolic activity correlates with drug sensitivity and use metabolic activity to predict drug response and synergy. Finally, we study the metabolic heterogeneity of cancer mutations across tissues, and find that genes exhibit a range of context specific, and more general metabolic control.

BackgroundSingle-cell sequencing studies have elucidated some of the underlying mechanisms responsible for immune checkpoint inhibitor (ICI) response, but are difficult to implement as a general strategy or in a clinical diagnostic setting. In contrast, bulk RNAseq is now routine for both research and clinical applications. Therefore, our analysis extracts small transcription factor-directed co-expression networks (regulons) from single-cell RNA-seq data and uses them to deconvolute immune functional states from bulk RNA-seq data to characterize patient responses.MethodsRegulons were inferred in pre-treatment CD45+ cells from metastatic melanoma samples (n=19) treated with first-line ICI therapy (discovery dataset). A logistic regression-based classifier identified immune cell states associated with response, which were characterized according to differentially active, cell-state specific regulons. The complexity of these regulons was reduced and scored in bulk RNAseq melanoma samples from four independent studies (n=209, validation dataset). Patients were clustered according to their regulon scores, and the associations between cluster assignment, response, and survival were determined. Intercellular communication analysis of cell states was performed, and the resulting effector genes were analyzed by trajectory inference.ResultsRegulons preserved the information of gene expression data and accurately delineated immune cell phenotypes, despite reducing dimensionality by > 100-fold. Four cell states, termed exhausted T cells, monocyte lineage cells, memory T cells, and B cells, were associated with therapeutic responses in the discovery dataset. The cell states were characterized by seven differentially active and specific regulons that showed low specificity in non-immune cells. Four clusters with significantly different response outcomes (P ConclusionsRegulon-based characterization of cell states provides robust and functionally informative markers that can deconvolve bulk RNA-seq data to identify ICI responders.

The endothelial response to vascular endothelial growth factor A (VEGF-A) regulates many aspects of animal physiology in health and disease. Such VEGF-A-regulated phenomena include vasculogenesis, angiogenesis, tumor growth and progression. VEGF-A binding to receptor tyrosine kinases such as vascular endothelial growth factor receptor 2 (VEGFR2 ) activates multiple signal transduction pathways and changes in homeostasis, metabolism, gene expression, cell proliferation, migration, and survival. One such VEGF-A-regulated response is a rapid rise in cytosolic calcium ion levels which modulates different biochemical events and impacts on endothelial-specific responses. Here, we present a series of detailed and robust protocols for evaluating ligand-stimulated cytosolic calcium ion flux in endothelial cells. By monitoring an endogenous endothelial transcription factor (NFATc2 ) which displays calcium-sensitive redistribution, we can assess the relevance of cytosolic calcium to protein function. This protocol can be easily applied to both adherent and non-adherent cultured cells to evaluate calcium ion flux in response to exogenous stimuli such as VEGF-A.

Excessive production of reactive oxygen species (ROS) is characteristic of numerous diseases, but most studies in this area have not considered the impact of endogenous antioxidative defenses. Here, utilizing multi-omics, we demonstrate that in cardiomyocytes mitochondrial isocitrate dehydrogenase (IDH2) constitutes a major antioxidant defense. In both male and female mice and humans the paradoxical reduction in expression of IDH2 associated with heart failure is compensated for by an increase in the enzyme’s activity. We describe extensive mutual regulation of the antioxidant activities of IDH2 and NRF2 by a network involving 2-oxoglutarate and L2-hydroxyglutarate and mediated in part through unconventional hydroxymethylation of cytosine residues present in introns. Conditional targeting of ROS in a murine model of heart failure improves cardiac function. Together, these insights may explain why previous attempts to treat heart failure with antioxidants have been unsuccessful and open new approaches to personalizing and, thereby, improving such treatment.Graphical abstractHighlightsParadoxical downregulation of mitochondrial isocitrate dehydrogenase (IDH2) in response to oxidative stress leads to the discovery of a robust antioxidative defense in the heart.An antioxidative loop involving IDH2 coordinates other antioxidative defenses, such as NRF2.This loop produces epigenetic modifications that link oxidative stress to mitochondrial function.The conclusion that enhancing antioxidative capacity improves cardiac function only when the endogenous capacity is insufficient opens new approaches to individualized treatment of patients with heart failure.

A fundamental feature of cancer cells is genomic heterogeneity. It is a main driver of phenotypic differences, including the response to drugs, and therefore a key factor in therapy selection. Motivated by the increasing role attributed to metabolic reprogramming in tumor development, we wondered how genomic heterogeneity affects metabolic phenotype. To this end, we profiled the intracellular metabolome of 180 cancer cell lines grown in similar conditions to exclude environmental factors. For each cell line, we estimate activity for 49 pathways across the whole metabolic network. Upon clustering of activity data, we found a convergence into only two major metabolic types. These were further characterized by 13C-flux analysis, lipidomics, and analysis of sensitivity to perturbations. These experiments revealed differences in lipid, mitochondrial, and carbohydrate metabolism between the two major types. Finally, a thorough integration of our metabolic data with multiple omics data revealed a strong association with markers of epithelial-mesenchymal transition (EMT). Our analysis indicates that in absence of variations imposed by the microenvironment, the metabolism of cancer cell lines falls into only two major classes despite genetic heterogeneity.

The cellular thermal shift assay (CETSA) has been used extensively since its introduction to study drug-target engagement within both live cells and cellular lysate. This has proven to be a useful tool in early stage drug discovery and is used to study a wide range of protein classes. We describe the application of a single-cell CETSA workflow within a microfluidic affinity capture (MAC) chip. This has enabled us to quantitatively determine the active FOXO1 single-molecule count and observe FOXO1 stabilization and destabilization in the presence of three small molecule inhibitors, including demonstrating the determination of EC50. The successful use of the MAC chip for single-cell CETSA paves the way for the study of precious clinical samples owing to the low number of cells needed by the chip. It also provides a useful tool for studying any underlying population heterogeneity that exists within a cellular system, a feature that is usually masked when conducting ensemble measurements.

Immune checkpoint inhibitors (ICIs) targeting cytotoxic T lymphocyte-associated protein-4 (CTLA-4) and programmed cell death protein-1 (PD-1) have been hailed as major advances in cancer therapeutics; however, in many cancers response rates remain low. Extensive research efforts are underway to improve the efficacy of ICIs. The signaling pathways regulated by immune checkpoints (ICs) may be an important lever as they interfere with T-cell activation when activated by ICIs. Here, we review the current understanding of T-cell receptor signaling and their intersection with IC signaling pathways. As these signaling processes are highly dynamic and controlled by intricate spatiotemporal mechanisms, we focus on aspects of kinetic regulation that are modulated by ICs. Recent advances in computational modeling and experimental methods that can resolve spatiotemporal dynamics provide insights that reveal molecular mechanisms and new potential approaches for improving the design and application of ICIs.

An important part the absorption, distribution, metabolism and excretion of an oral therapeutic is the flux rate of drug compound crossing the mucus lining of the gut. To understand this part of the absorption process, we develop a mathematical model of advection, diffusion and binding of drug compounds within the mucus layer of the intestines. Analysis of this model yields simple, measurable criteria for the successful mucin layer traversal of drug compound.

In an age where the volume of data regarding biological systems exceeds our ability to analyse it, many researchers are looking towards systems biology and computational modelling to help unravel the complexities of gene and protein regulatory networks. In particular, the use of discrete modelling allows generation of signalling networks in the absence of full quantitative descriptions of systems, which are necessary for ordinary differential equation (ODE) models. In order to make such techniques more accessible to mainstream researchers, tools such as the BioModelAnalyzer (BMA) have been developed to provide a user-friendly graphical interface for discrete modelling of biological systems. Here we use the BMA to build a library of discrete target functions of known canonical molecular interactions, translated from ordinary differential equations (ODEs). We then show that these BMA target functions can be used to reconstruct complex networks, which can correctly predict many known genetic perturbations. This new library supports the accessibility ethos behind the creation of BMA, providing a toolbox for the construction of complex cell signalling models without the need for extensive experience in computer programming or mathematical modelling, and allows for construction and simulation of complex biological systems with only small amounts of quantitative data.

We present two applications of conformal prediction relevant to drug discovery. The first application is around interpretation of predictions and the second one around the selection of compounds to progress in a drug discovery project setting.

While deep learning has seen many recent applications to drug discovery, most have focused on predicting activity or toxicity directly from chemical structure. Phenotypic changes exhibited in cellular images are also indications of the mechanism of action (MoA) of chemical compounds. In this paper, we show how pre-trained convolutional image features can be used to assist scientists in discovering interesting chemical clusters for further investigation. Our method reduces the dimensionality of raw fluorescent stained images from a high throughput imaging (HTI) screen, producing an embedding space that groups together images with similar cellular phenotypes. Running standard unsupervised clustering on this embedding space yields a set of distinct phenotypic clusters. This allows scientists to further select and focus on interesting clusters for downstream analyses. We validate the consistency of our embedding space qualitatively with t-sne visualizations, and quantitatively by measuring embedding variance among images that are known to be similar. Results suggested the usefulness of our proposed workflow using deep learning and clustering and it can lead to robust HTI screening and compound triage.

Optimization of experimental conditions is critical in ensuring robust experimental reproducibility. Through detailed metabolomic analysis we found that cell culture conditions significantly impacted on glutaminase (GLS1) sensitivity resulting in variable sensitivity and irreproducibility in data. Baseline metabolite profiling highlighted that untreated cells underwent significant changes in metabolic status. Both the extracellular levels of glutamine and lactate and the intracellular levels of multiple metabolites changed drastically during the assay. We show that these changes compromise the robustness of the assay and make it difficult to reproduce. We discuss the implications of the cells’ metabolic environment when studying the effects of perturbations to cell function by any type of inhibitor. We then devised ‘metabolically rationalized standard’ assay conditions, in which glutaminase-1 inhibition reduced glutamine metabolism differently in both cell lines assayed, and decreased the proliferation of one of them. The adoption of optimized conditions such as the ones described here should lead to an improvement in reproducibility and help eliminate false negatives as well as false positives in these assays.

In an age where the volume of data regarding biological systems exceeds our ability to analyse it, many researchers are looking towards systems biology and computational modelling to help unravel the complexities of gene and protein regulatory networks. In particular, the use of discrete modelling allows generation of signalling networks in the absence of full quantitative descriptions of systems, which are necessary for ordinary differential equation (ODE) models. In order to make such techniques more accessible to mainstream researchers, tools such as the BioModelAnalyzer (BMA) have been developed to provide a user-friendly graphical interface for discrete modelling of biological systems. Here we use the BMA to build a library of discrete target functions of known canonical molecular interactions, translated from ordinary differential equations (ODEs). We then show that these BMA target functions can be used to reconstruct complex networks, which can correctly predict many known genetic perturbations. This new library supports the accessibility ethos behind the creation of BMA, providing a toolbox for the construction of complex cell signalling models without the need for extensive experience in computer programming or mathematical modelling, and allows for construction and simulation of complex biological systems with only small amounts of quantitative data.AUTHOR SUMMARYOrdinary differential equation (ODE) based models are a popular approach for modelling biological networks. A limitation of ODE models is that they require complete networks and detailed kinetic parameterisation. An alternative is the use of discrete, executable models, in which nodes are assigned discrete value ranges, and the relationship between them defined with simple mathematical operations. One tool for constructing such models is the BioModelAnalyzer (BMA), an open source and publicly available (https://www.biomodelanalyzer.org) software, aimed to be fully usable by researchers without extensive computational or mathematical experience. A fundamental question for executable models is whether the high level of abstraction substantially reduces expressivity relative to continuous approaches. Here, we present a canonical library of biological signalling motifs, initially defined by Tyson et al (2003), translated for the first time into the BMA. We show that; 1) these motifs are easily and fully translatable from continuous to discrete models, 2) Combining these motifs in a computationally naïve way generates a fully functional and predictive model of the yeast cell cycle.

Many antimicrobial and anti-tumour drugs elicit hormetic responses characterised by low-dose stimulation and high-dose inhibition. While this can have profound consequences for human health, with low drug concentrations actually stimulating pathogen or tumour growth, the mechanistic understanding behind such responses is still lacking. We propose a novel, simple but general mechanism that could give rise to hormesis in systems where an inhibitor acts on an enzyme. At its core is one of the basic building blocks in intracellular signalling, the dual phosphorylation-dephosphorylation motif, found in diverse regulatory processes including control of cell proliferation and programmed cell death. Our analytically-derived conditions for observing hormesis provide clues as to why this mechanism has not been previously identified. Current mathematical models regularly make simplifying assumptions that lack empirical support but inadvertently preclude the observation of hormesis. In addition, due to the inherent population heterogeneities, the presence of hormesis is likely to be masked in empirical population-level studies. Therefore, examining hormetic responses at single-cell level coupled with improved mathematical models could substantially enhance detection and mechanistic understanding of hormesis., Author Summary Hormesis is a highly controversial and poorly understood phenomenon. It describes the idea that an inhibitor molecule, like an anti-cancer or anti-microbial drug, can inadvertently stimulate cell growth instead of suppressing it. This can have a profound effect on human health leading to failures in clinical treatments. Therefore, getting at the mechanistic basis of hormesis is critical for drug development and clinical practice, however molecular mechanisms underpinning hormesis remain poorly understood. In this paper we use a mathematical model to propose a simple and yet general mechanism that could explain why we find hormesis so widely in living systems. In particular, we discover that hormesis is present within a fundamental structure that forms a basic building block of many intracellular signalling pathways found in diverse processes including control of cell reproduction and programmed cell death. The benefits of our study are two-fold. Having simple molecular understanding of the causes of hormetic responses can greatly improve the design of new drug compounds that avoid such responses. Moreover, due to the fundamental nature of the newly proposed mechanism, our findings have a potential broad applicability to both anti-cancer and anti-microbial drugs.

Understanding the therapeutic effect of drug dose and scheduling is critical to inform the design and implementation of clinical trials. The increasing complexity of both mono, and particularly combination therapies presents a substantial challenge in the clinical stages of drug development for oncology. Using a systems pharmacology approach, we have extended an existing PK-PD model of tumor growth with a mechanistic model of the cell cycle, enabling simulation of mono and combination treatment with the ATR inhibitor AZD6738 and ionizing radiation. Using AZD6738, we have developed multi-parametric cell based assays measuring DNA damage and cell cycle transition, providing quantitative data suitable for model calibration. Our in vitro calibrated cell cycle model is predictive of tumor growth observed in in vivo mouse xenograft studies. The model is being used for phase I clinical trial designs for AZD6738, with the aim of improving patient care through quantitative dose and scheduling prediction.

Understanding the therapeutic effect of drug dose and scheduling is critical to inform the design and implementation of clinical trials. The increasing complexity of both mono and particularly combination therapies presents a substantial challenge in the clinical stages of drug development for oncology. Using a systems pharmacology approach, we have extended an existing PK-PD model of tumor growth with a mechanistic model of the cell cycle, enabling simulation of mono and combination treatment with the ATR inhibitor AZD6738 and ionizing radiation. Using AZD6738, we have developed multi-parametric cell based assays measuring DNA damage and cell cycle transition, providing quantitative data suitable for model calibration. Our in vitro calibrated cell cycle model is predictive of tumor growth observed in in vivo mouse xenograft studies. The model is being used for phase I clinical trial designs for AZD6738, with the aim of improving patient care through quantitative dose and scheduling prediction.

Insulin- and contraction-stimulated increases in glucose uptake into skeletal muscle occur in part as a result of the translocation of glucose transporter 4 (GLUT4) from intracellular stores to the plasma membrane (PM). This study aimed to use immunofluorescence microscopy in human skeletal muscle to quantify GLUT4 redistribution from intracellular stores to the PM in response to glucose feeding and exercise. Percutaneous muscle biopsy samples were taken from the m. vastus lateralis of ten insulin-sensitive men in the basal state and following 30 min of cycling exercise (65% VO2 max). Muscle biopsy samples were also taken from a second cohort of ten age-, BMI- and VO2 max-matched insulin-sensitive men in the basal state and 30 and 60 min following glucose feeding (75 g glucose). GLUT4 and dystrophin colocalization, measured using the Pearson's correlation coefficient, was increased following 30 min of cycling exercise (baseline r = 0.47 ± 0.01; post exercise r = 0.58 ± 0.02; P