cytoNet: Spatiotemporal network analysis of cell communities.

We introduce cytoNet, a cloud-based tool to characterize cell populations from microscopy images. cytoNet quantifies spatial topology and functional relationships in cell communities using principles of network science. Capturing multicellular dynamics through graph features, cytoNet also evaluates the effect of cell-cell interactions on individual cell phenotypes. We demonstrate cytoNet's capabilities in four case studies: 1) characterizing the temporal dynamics of neural progenitor cell communities during neural differentiation, 2) identifying communities of pain-sensing neurons in vivo, 3) capturing the effect of cell community on endothelial cell morphology, and 4) investigating the effect of laminin α4 on perivascular niches in adipose tissue. The analytical framework introduced here can be used to study the dynamics of complex cell communities in a quantitative manner, leading to a deeper understanding of environmental effects on cellular behavior. The versatile, cloud-based format of cytoNet makes the image analysis framework accessible to researchers across domains.

A novel self-organizing embryonic stem cell system reveals signaling logic underlying the patterning of human ectoderm.

During development, the ectoderm is patterned by a combination of BMP and WNT signaling. Research in model organisms has provided substantial insight into this process; however, there are currently no systems in which to study ectodermal patterning in humans. Further, the complexity of neural plate border specification has made it difficult to transition from discovering the genes involved to deeper mechanistic understanding. Here, we develop an in vitro model of human ectodermal patterning, in which human embryonic stem cells self-organize to form robust and quantitatively reproducible patterns corresponding to the complete medial-lateral axis of the embryonic ectoderm. Using this platform, we show that the duration of endogenous WNT signaling is a crucial control parameter, and that cells sense relative levels of BMP and WNT signaling in making fate decisions. These insights allowed us to develop an improved protocol for placodal differentiation. Thus, our platform is a powerful tool for studying human ectoderm patterning and for improving directed differentiation protocols.This article has an associated 'The people behind the papers' interview.

Clinical relevance of proteomic profiling in de novo pediatric acute myeloid leukemia: a Children's Oncology Group study.

Pediatric acute myeloid leukemia (AML) remains a fatal disease for at least 30% of patients, stressing the need for improved therapies and better risk stratification. As proteins are the unifying feature of (epi)genetic and environmental alterations, and are often targeted by novel chemotherapeutic agents, we studied the proteomic landscape of pediatric AML. Protein expression and activation levels were measured in 500 bulk leukemic patients' samples and 30 control CD34+ cell samples, using reverse phase protein arrays with 296 strictly validated antibodies. The multistep MetaGalaxy analysis methodology was applied and identified nine protein expression signatures (PrSIG), based on strong recurrent protein expression patterns. PrSIG were associated with cytogenetics and mutational state, and with favorable or unfavorable prognosis. Analysis based on treatment (i.e., ADE vs. ADE plus bortezomib) identified three PrSIG that did better with ADE plus bortezomib than with ADE alone. When PrSIG were studied in the context of cytogenetic risk groups, PrSIG were independently prognostic after multivariate analysis, suggesting a potential value for proteomics in combination with current classification systems. Proteins with universally increased (n=7) or decreased (n=17) expression were observed across PrSIG. Certain proteins significantly differentially expressed from normal could be identified, forming a hypothetical platform for personalized medicine.

Differential proteome profile, biological pathways, and network relationships of osteogenic proteins in calcified human aortic valves.

Calcific aortic valve disease (CAVD) is the most common heart valve disease requiring intervention. Most research on CAVD has focused on inflammation, ossification, and cellular phenotype transformation. To gain a broader picture into the wide range of cellular and molecular mechanisms involved in this disease, we compared the total protein profiles between calcified and non-calcified areas from 5 human valves resected during surgery. The 1413 positively identified proteins were filtered down to 248 proteins present in both calcified and non-calcified segments of at least 3 of the 5 valves, which were then analyzed using Ingenuity Pathway Analysis. Concurrently, the top 40 differentially abundant proteins were grouped according to their biological functions and shown in interactive networks. Finally, the abundance of selected osteogenic proteins (osteopontin, osteonectin, osteocalcin, osteoprotegerin, and RANK) was quantified using ELISA and/or immunohistochemistry. The top pathways identified were complement system, acute phase response signaling, metabolism, LXR/RXR and FXR/RXR activation, actin cytoskeleton, mineral binding, nucleic acid interaction, structural extracellular matrix (ECM), and angiogenesis. There was a greater abundance of osteopontin, osteonectin, osteocalcin, osteoprotegerin, and RANK in the calcified regions than the non-calcified ones. The osteogenic proteins also formed key connections between the biological signaling pathways in the network model. In conclusion, this proteomic analysis demonstrated the involvement of multiple signaling pathways in CAVD. The interconnectedness of these pathways provides new insights for the treatment of this disease.

Inferring causal molecular networks: empirical assessment through a community-based effort.

It remains unclear whether causal, rather than merely correlational, relationships in molecular networks can be inferred in complex biological settings. Here we describe the HPN-DREAM network inference challenge, which focused on learning causal influences in signaling networks. We used phosphoprotein data from cancer cell lines as well as in silico data from a nonlinear dynamical model. Using the phosphoprotein data, we scored more than 2,000 networks submitted by challenge participants. The networks spanned 32 biological contexts and were scored in terms of causal validity with respect to unseen interventional data. A number of approaches were effective, and incorporating known biology was generally advantageous. Additional sub-challenges considered time-course prediction and visualization. Our results suggest that learning causal relationships may be feasible in complex settings such as disease states. Furthermore, our scoring approach provides a practical way to empirically assess inferred molecular networks in a causal sense.

Histone Modification Patterns Using RPPA-Based Profiling Predict Outcome in Acute Myeloid Leukemia Patients.

Posttranslational histone tail modifications are known to play a role in leukemogenesis and are therapeutic targets. A global analysis of the level and patterns of expression of multiple histone-modifying proteins (HMP) in acute myeloid leukemia (AML) and the effect of different patterns of expression on outcome and prognosis has not been investigated in AML patients. Here we analyzed 20 HMP by reverse phase protein array (RPPA) in a cohort of 205 newly diagnosed AML patients. Protein levels were correlated with patient and disease characteristics, including survival and mutational state. We identified different protein clusters characterized by higher (more on) or lower (more off) expression of HMP, relative to normal CD34+ cells. On state of HMP was associated with poorer outcome compared to normal-like and a more off state. FLT3 mutated AML patients were significantly overrepresented in the more on state. DNA methylation related mutations showed no correlation with the different HMP states. In this study, we demonstrate for the first time that HMP form recurrent patterns of expression and that these significantly correlate with survival in newly diagnosed AML patients.

Shrinkage Clustering: a fast and size-constrained clustering algorithm for biomedical applications.

BACKGROUND: Many common clustering algorithms require a two-step process that limits their efficiency. The algorithms need to be performed repetitively and need to be implemented together with a model selection criterion. These two steps are needed in order to determine both the number of clusters present in the data and the corresponding cluster memberships. As biomedical datasets increase in size and prevalence, there is a growing need for new methods that are more convenient to implement and are more computationally efficient. In addition, it is often essential to obtain clusters of sufficient sample size to make the clustering result meaningful and interpretable for subsequent analysis. RESULTS: We introduce Shrinkage Clustering, a novel clustering algorithm based on matrix factorization that simultaneously finds the optimal number of clusters while partitioning the data. We report its performances across multiple simulated and actual datasets, and demonstrate its strength in accuracy and speed applied to subtyping cancer and brain tissues. In addition, the algorithm offers a straightforward solution to clustering with cluster size constraints. CONCLUSIONS: Given its ease of implementation, computing efficiency and extensible structure, Shrinkage Clustering can be applied broadly to solve biomedical clustering tasks especially when dealing with large datasets.

Shining a light on cell signaling in leukemia through proteomics: relevance for the clinic.

INTRODUCTION: Although cure rates for acute leukemia have steadily improved over the past decades, leukemia remains a deadly disease. Enhanced risk stratification and new therapies are needed to improve outcome. Extensive genetic analyses have identified many mutations that contribute to the development of leukemia. However, most mutations occur infrequently and most gene alterations have been difficult to target. Most patients have more than one driver mutation in combination with secondary mutations, that result in a leukemic transformation via the alteration of proteins. The proteomics of acute leukemia could more directly identify proteins to facilitate risk stratification, predict chemoresistance and aid selection of therapy. Areas covered: This review discusses aberrantly expressed proteins identified by mass spectrometry and reverse phase protein arrays and their relationship to survival. In addition, we will discuss proteins in the context of functionally related protein groups. Expert commentary: Proteomics is a powerful tool to analyze protein abundance and functional alterations simultaneously for large numbers of patients. In the forthcoming years, validation of tools to quickly assess protein levels to enable routine rapid profiling of proteins with differential abundance and functional activation may be used as adjuncts to aid in therapy selection and to provide additional prognostic insights.

Reconstruction of Tissue-Specific Metabolic Networks Using CORDA.

Human metabolism involves thousands of reactions and metabolites. To interpret this complexity, computational modeling becomes an essential experimental tool. One of the most popular techniques to study human metabolism as a whole is genome scale modeling. A key challenge to applying genome scale modeling is identifying critical metabolic reactions across diverse human tissues. Here we introduce a novel algorithm called Cost Optimization Reaction Dependency Assessment (CORDA) to build genome scale models in a tissue-specific manner. CORDA performs more efficiently computationally, shows better agreement to experimental data, and displays better model functionality and capacity when compared to previous algorithms. CORDA also returns reaction associations that can greatly assist in any manual curation to be performed following the automated reconstruction process. Using CORDA, we developed a library of 76 healthy and 20 cancer tissue-specific reconstructions. These reconstructions identified which metabolic pathways are shared across diverse human tissues. Moreover, we identified changes in reactions and pathways that are differentially included and present different capacity profiles in cancer compared to healthy tissues, including up-regulation of folate metabolism, the down-regulation of thiamine metabolism, and tight regulation of oxidative phosphorylation.

A Crowdsourcing Approach to Developing and Assessing Prediction Algorithms for AML Prognosis.

Acute Myeloid Leukemia (AML) is a fatal hematological cancer. The genetic abnormalities underlying AML are extremely heterogeneous among patients, making prognosis and treatment selection very difficult. While clinical proteomics data has the potential to improve prognosis accuracy, thus far, the quantitative means to do so have yet to be developed. Here we report the results and insights gained from the DREAM 9 Acute Myeloid Prediction Outcome Prediction Challenge (AML-OPC), a crowdsourcing effort designed to promote the development of quantitative methods for AML prognosis prediction. We identify the most accurate and robust models in predicting patient response to therapy, remission duration, and overall survival. We further investigate patient response to therapy, a clinically actionable prediction, and find that patients that are classified as resistant to therapy are harder to predict than responsive patients across the 31 models submitted to the challenge. The top two performing models, which held a high sensitivity to these patients, substantially utilized the proteomics data to make predictions. Using these models, we also identify which signaling proteins were useful in predicting patient therapeutic response.

Simulation predicts IGFBP2-HIF1α interaction drives glioblastoma growth.

Tremendous strides have been made in improving patients' survival from cancer with one glaring exception: brain cancer. Glioblastoma is the most common, aggressive and highly malignant type of primary brain tumor. The average overall survival remains less than 1 year. Notably, cancer patients with obesity and diabetes have worse outcomes and accelerated progression of glioblastoma. The root cause of this accelerated progression has been hypothesized to involve the insulin signaling pathway. However, while the process of invasive glioblastoma progression has been extensively studied macroscopically, it has not yet been well characterized with regards to intracellular insulin signaling. In this study we connect for the first time microscale insulin signaling activity with macroscale glioblastoma growth through the use of computational modeling. Results of the model suggest a novel observation: feedback from IGFBP2 to HIF1α is integral to the sustained growth of glioblastoma. Our study suggests that downstream signaling from IGFI to HIF1α, which has been the target of many insulin signaling drugs in clinical trials, plays a smaller role in overall tumor growth. These predictions strongly suggest redirecting the focus of glioma drug candidates on controlling the feedback between IGFBP2 and HIF1α.

Abnormal expression of FLI1 protein is an adverse prognostic factor in acute myeloid leukemia.

Friend leukemia virus integration 1 (FLI1), an Ets transcription factor family member, is linked to acute myelogenous leukemia (AML) by chromosomal events at the FLI1 locus, but the biologic impact of FLI1 expression on AML is unknown. FLI1 protein expression was measured in 511 newly diagnosed AML patients. Expression was similar in peripheral blood (PB) and BM and higher at diagnosis than at relapse (P = .02). Compared with normal CD34(+) cells, expression in AML was above or below normal in 32% and 5% of patients, respectively. Levels were negatively correlated with an antecedent hematologic disorder (P = .002) but not with age or cytogenetics. Mutated NPM1 (P = .0007) or FLT3-ITD (P < .02) had higher expression. FLI1 levels were negatively correlated with 10 of 195 proteins associated with proliferation and stromal interaction, and positively correlated (R > 0.3) with 19 others. The FLI1 level was not predictive of remission attainment, but patients with low or high FLI1 expression had shorter remission duration (22.6 and 40.3 vs 51.1 weeks, respectively; P = .01) and overall survival (45.2 and 35.4 vs 59.4 weeks, respectively; P = .03). High FLI1 levels were adverse in univariate and multivariate analysis. FLI1 expression is frequently abnormal and prognostically adverse in AML. FLI1 and/or its response genes may be therapeutically targetable to interfere with AML cell biology.

Cells as state machines: cell behavior patterns arise during capillary formation as a function of BDNF and VEGF.

Cell behavior patterns that lead to distinct tissue or capillary phenotypes are difficult to identify using existing approaches. We present a strategy to characterize the form, frequency, magnitude and sequence of human endothelial cell activity when stimulated by vascular endothelial growth factor (VEGF) and brain-derived neurotrophic factor (BDNF). We introduce a "Rules-as-Agents" method for rapid comparison of cell behavior hypotheses to in vitro angiogenesis experiments. Endothelial cells are represented as machines that transition between finite behavior states, and their properties are explored by a search algorithm. We rank and quantify differences between competing hypotheses about cell behavior during the formation of unique capillary phenotypes. Results show the interaction of tip and stalk endothelial cells, and predict how migration, proliferation, branching, and elongation integrate to form capillary structures within a 3D matrix in the presence of varying VEGF and BDNF concentrations. This work offers the ability to understand - and ultimately control - human cell behavior at the microvasculature level.

Module-based multiscale simulation of angiogenesis in skeletal muscle.

BACKGROUND: Mathematical modeling of angiogenesis has been gaining momentum as a means to shed new light on the biological complexity underlying blood vessel growth. A variety of computational models have been developed, each focusing on different aspects of the angiogenesis process and occurring at different biological scales, ranging from the molecular to the tissue levels. Integration of models at different scales is a challenging and currently unsolved problem. RESULTS: We present an object-oriented module-based computational integration strategy to build a multiscale model of angiogenesis that links currently available models. As an example case, we use this approach to integrate modules representing microvascular blood flow, oxygen transport, vascular endothelial growth factor transport and endothelial cell behavior (sensing, migration and proliferation). Modeling methodologies in these modules include algebraic equations, partial differential equations and agent-based models with complex logical rules. We apply this integrated model to simulate exercise-induced angiogenesis in skeletal muscle. The simulation results compare capillary growth patterns between different exercise conditions for a single bout of exercise. Results demonstrate how the computational infrastructure can effectively integrate multiple modules by coordinating their connectivity and data exchange. Model parameterization offers simulation flexibility and a platform for performing sensitivity analysis. CONCLUSIONS: This systems biology strategy can be applied to larger scale integration of computational models of angiogenesis in skeletal muscle, or other complex processes in other tissues under physiological and pathological conditions.

Integrins and extracellular matrix proteins modulate adipocyte thermogenic capacity.

Obesity and the metabolic disease epidemic has led to an increase in morbidity and mortality. A rise in adipose thermogenic capacity via activation of brown or beige fat is a potential treatment for metabolic diseases. However, an understanding of how local factors control adipocyte fate is limited. Mice with a null mutation in the laminin α4 (LAMA4) gene (KO) exhibit resistance to obesity and enhanced expression of thermogenic fat markers in white adipose tissue (WAT). In this study, changes in WAT extracellular matrix composition in the absence of LAMA4 were evaluated using liquid chromatography/tandem mass spectrometry. KO-mice showed lower levels of collagen 1A1 and 3A1, and integrins α7 (ITA7) and ß1 (ITB1). ITA7-ITB1 and collagen 1A1-3A1 protein levels were lower in brown adipose tissue compared to WAT in wild-type mice. Immunohistochemical staining confirmed lower levels and different spatial distribution of ITA7 in KO-WAT. In culture studies, ITA7 and LAMA4 levels decreased following a 12-day differentiation of adipose-derived stem cells into beige fat, and knock-down of ITA7 during differentiation increased beiging. These results demonstrate that extracellular matrix interactions regulate adipocyte thermogenic capacity and that ITA7 plays a role in beige adipose formation. A better understanding of the mechanisms underlying these interactions can be used to improve systemic energy metabolism and glucose homeostasis.

Decoupling Lineage-Associated Genes in Acute Myeloid Leukemia Reveals Inflammatory and Metabolic Signatures Associated With Outcomes.

Acute myeloid leukemia (AML) is a heterogeneous disease with variable responses to therapy. Cytogenetic and genomic features are used to classify AML patients into prognostic and treatment groups. However, these molecular characteristics harbor significant patient-to-patient variability and do not fully account for AML heterogeneity. RNA-based classifications have also been applied in AML as an alternative approach, but transcriptomic grouping is strongly associated with AML morphologic lineages. We used a training cohort of newly diagnosed AML patients and conducted unsupervised RNA-based classification after excluding lineage-associated genes. We identified three AML patient groups that have distinct biological pathways associated with outcomes. Enrichment of inflammatory pathways and downregulation of HOX pathways were associated with improved outcomes, and this was validated in 2 independent cohorts. We also identified a group of AML patients who harbored high metabolic and mTOR pathway activity, and this was associated with worse clinical outcomes. Using a comprehensive reverse phase protein array, we identified higher mTOR protein expression in the highly metabolic group. We also identified a positive correlation between degree of resistance to venetoclax and mTOR activation in myeloid and lymphoid cell lines. Our approach of integrating RNA, protein, and genomic data uncovered lineage-independent AML patient groups that share biologic mechanisms and can inform outcomes independent of commonly used clinical and demographic variables; these groups could be used to guide therapeutic strategies.

Image-based Classification of Tumor Type and Growth Rate using Machine Learning: a preclinical study.

Medical images such as magnetic resonance (MR) imaging provide valuable information for cancer detection, diagnosis, and prognosis. In addition to the anatomical information these images provide, machine learning can identify texture features from these images to further personalize treatment. This study aims to evaluate the use of texture features derived from T1-weighted post contrast scans to classify different types of brain tumors and predict tumor growth rate in a preclinical mouse model. To optimize prediction models this study uses varying gray-level co-occurrence matrix (GLCM) sizes, tumor region selection and different machine learning models. Using a random forest classification model with a GLCM of size 512 resulted in 92%, 91%, and 92% specificity, and 89%, 85%, and 73% sensitivity for GL261 (mouse glioma), U87 (human glioma) and Daoy (human medulloblastoma), respectively. A tenfold cross-validation of the classifier resulted in 84% accuracy when using the entire tumor volume for feature extraction and 74% accuracy for the central tumor region. A two-layer feedforward neural network using the same features is able to predict tumor growth with 16% mean squared error. Broadly applicable, these predictive models can use standard medical images to classify tumor type and predict tumor growth, with model performance, varying as a function of GLCM size, tumor region, and tumor type.

LGALS3 is connected to CD74 in a previously unknown protein network that is associated with poor survival in patients with AML.

BACKGROUND: Galectin 3 (LGALS3) gene expression is associated with poor survival in acute myeloid leukemia (AML) but the prognostic impact of LGALS3 protein expression in AML is unknown. LGALS3 supports diverse survival pathways including RAS mediated cascades, protein expression and stability of anti-apoptotic BCL2 family members, and activation of proliferative pathways including those mediated by beta Catenin. CD74 is a positive regulator of CD44 and CXCR4 signaling and this molecule may be critical for AML stem cell function. At present, the role of LGALS3 and CD74 in AML is unclear. In this study, we examine protein expression of LGALS3 and CD74 by reverse phase protein analysis (RPPA) and identify new protein networks associated with these molecules. In addition, we determine prognostic potential of LGALS3, CD74, and their protein networks for clinical correlates in AML patients. METHODS: RPPA was used to determine relative expression of LGALS3, CD74, and 229 other proteins in 231 fresh AML patient samples and 205 samples were from patients who were treated and evaluable for outcome. Pearson correlation analysis was performed to identify proteins associated with LGALS3 and CD74. Progeny clustering was performed to generate protein networks. String analysis was performed to determine protein:protein interactions in networks and to perform gene ontology analysis. Kaplan-Meir method was used to generate survival curves. FINDINGS: LGALS3 is highest in monocytic AML patients and those with elevated LGALS3 had significantly shorter remission duration compared to patients with lower LGALS3 levels (median 21.9 vs 51.3 weeks, p = 0.016). Pearson correlation of LGALS3 with 230 other proteins identifies a distinct set of 37 proteins positively correlated with LGALS3 expression levels with a high representation of proteins involved in AKT and ERK signaling pathways. Thirty-one proteins were negatively correlated with LGALS3 including an AKT phosphatase. Pearson correlation of proteins associated with CD74 identified 12 proteins negatively correlated with CD74 and 16 proteins that are positively correlated with CD74. CD74 network revealed strong association with CD44 signaling and a high representation of apoptosis regulators. Progeny clustering was used to build protein networks based on LGALS3 and CD74 associated proteins. A strong relationship of the LGALS3 network with the CD74 network was identified. For AML patients with both the LGALS3 and CD74 protein cluster active, median overall survival was only 24.3 weeks, median remission duration was 17.8 weeks, and no patient survived beyond one year. INTERPRETATION: The findings from this study identify for the first time protein networks associated with LGALS3 and CD74 in AML. Each network features unique pathway characteristics. The data also suggest that the LGALS3 network and the CD74 network each support AML cell survival and the two networks may cooperate in a novel high risk AML population. FUND: Leukemia Lymphoma Society provided funds to SMK for RPPA study of AML patient population. Texas Leukemia provided funds to PPR and SMK to study CD74 and LGALS3 expression in AML patients using RPPA. No payment was involved in the production of this manuscript.

Proteomic Profiling of Acute Promyelocytic Leukemia Identifies Two Protein Signatures Associated with Relapse.

PURPOSE: Acute promyelocytic leukemia (APL) is the most prognostically favorable subtype of Acute myeloid leukemia (AML). Defining the features that allow identification of APL patients likely to relapse after therapy remains challenging. EXPERIMENTAL DESIGN: Proteomic profiling is performed on 20 newly diagnosed APL, 205 non-APL AML, and 10 normal CD34+ samples using Reverse Phase Protein Arrays probed with 230 antibodies. RESULTS: Comparison between APL and non-APL AML samples identifies 8.3% of the proteins to be differentially expressed. Proteins higher expressed in APL are involved in the pro-apoptotic pathways or are linked to higher proliferation. The "MetaGalaxy" approach that considers proteins in relation to other assayed proteins stratifies the APL patients into two protein signatures. All of the relapse patients (n = 4/4) are in protein signature 2 (S2). Comparison of proteins between the signatures shows significant differences in relative expression for 38 proteins. Protein expression summary plots suggest less translational activity in combination with a less proliferative character for S2 compared to signature 1. CONCLUSIONS AND CLINICAL RELEVANCE: This study provides a potential proteomic-based classification of APL patients that may be useful for risk stratification and therapeutic guidance. Validation in a larger independent cohort is required.

Three autocrine feedback loops determine HIF1 alpha expression in chronic hypoxia.

Hypoxia occurs in cancer, prolonged exercise, and long-term ischemia with durations of several hours or more, and the hypoxia-inducible factor 1 (HIF1) pathway response to these conditions differs from responses to transient hypoxia. We used computational modeling, validated by experiments, to gain a quantitative, temporal understanding of the mechanisms driving HIF1 response. To test the hypothesis that HIF1 alpha protein levels during chronic hypoxia are tightly regulated by a series of molecular feedbacks, we took into account protein synthesis and product inhibition, and analyzed HIF1 system changes in response to hypoxic exposures beyond 3 to 4 h. We show how three autocrine feedback loops together regulate HIF 1 alpha hydroxylation in different microenvironments. Results demonstrate that prolyl hydroxylase, succinate and HIF1 alpha feedback determine intracellular HIF1 alpha levels over the course of hours to days. The model provides quantitative insight critical for characterizing molecular mechanisms underlying a cell's response to long-term hypoxia.