Patrick J. Trainor

Exercise-Induced Changes in Glucose Metabolism Promote Physiological Cardiac Growth

Background: Exercise promotes metabolic remodeling in the heart, which is associated with physiological cardiac growth; however, it is not known whether or how physical activity–induced changes in cardiac metabolism cause myocardial remodeling. In this study, we tested whether exercise-mediated changes in cardiomyocyte glucose metabolism are important for physiological cardiac growth. Methods: We used radiometric, immunologic, metabolomic, and biochemical assays to measure changes in myocardial glucose metabolism in mice subjected to acute and chronic treadmill exercise. To assess the relevance of changes in glycolytic activity, we determined how cardiac-specific expression of mutant forms of 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase affect cardiac structure, function, metabolism, and gene programs relevant to cardiac remodeling. Metabolomic and transcriptomic screenings were used to identify metabolic pathways and gene sets regulated by glycolytic activity in the heart. Results: Exercise acutely decreased glucose utilization via glycolysis by modulating circulating substrates and reducing phosphofructokinase activity; however, in the recovered state following exercise adaptation, there was an increase in myocardial phosphofructokinase activity and glycolysis. In mice, cardiac-specific expression of a kinase-deficient 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase transgene (GlycoLo mice) lowered glycolytic rate and regulated the expression of genes known to promote cardiac growth. Hearts of GlycoLo mice had larger myocytes, enhanced cardiac function, and higher capillary-to-myocyte ratios. Expression of phosphatase-deficient 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase in the heart (GlycoHi mice) increased glucose utilization and promoted a more pathological form of hypertrophy devoid of transcriptional activation of the physiological cardiac growth program. Modulation of phosphofructokinase activity was sufficient to regulate the glucose–fatty acid cycle in the heart; however, metabolic inflexibility caused by invariantly low or high phosphofructokinase activity caused modest mitochondrial damage. Transcriptomic analyses showed that glycolysis regulates the expression of key genes involved in cardiac metabolism and remodeling. Conclusions: Exercise-induced decreases in glycolytic activity stimulate physiological cardiac remodeling, and metabolic flexibility is important for maintaining mitochondrial health in the heart.

Proceedings of the 16th Annual UT-KBRIN Bioinformatics Summit 2016: bioinformatics

I1 Proceedings of the Sixteenth Annual UTKBRIN Bioinformatics Summit 2017 Eric C Rouchka, Julia H Chariker, David A Tieri, Juw Won Park Department of Computer Engineering and Computer Science, University of Louisville, Duthie Center for Engineering, Louisville, KY 40292, USA; Kentucky Biomedical Research Infrastructure (KBRIN) Bioinformatics Core, 522 East Gray Street, Louisville, KY 40292, USA; Department of Psychological and Brain Sciences, University of Louisville, Louisville, KY 40292, USA; Department of Anatomical Sciences and Neurobiology, University of Louisville, Louisville, KY 40292, USA Correspondence: Eric C Rouchka (eric.rouchka@louisville.edu) BMC Bioinformatics 2017, 18(Suppl 9):I1

Inferring metabolite interactomes via molecular structure informed Bayesian graphical model selection with an application to coronary artery disease

Introduction While the generation of reference genomes facilitates the elucidation of gene-phenome associations, reference models of the metabolome that are specific to organism, sample type (e.g. plasma, serum, urine, cell-culture), and state (including disease), remain uncommon. In studying heart disease in humans, a reference model describing the relationships between metabolites in plasma has not been determined but would have great utility as a reference for comparing acute disease states such as myocardial infarction. Materials and Methods We present a methodology for deriving probabilistic models that describe the partial correlation structure of metabolite distributions (“interactomes”) from metabolomics data. As determining partial correlation structures requires estimating p*(p-1)/2 parameters for p metabolites, the dimension of the search space for parameter values is immense. Consequently, we have developed a Bayesian methodology for the penalized estimation of model parameters in which the magnitude of penalization is drawn from probability distributions with hyperparameters linked to molecular structure similarity. In our work, structural similarity was determined as the Tanimoto coefficient of algorithmically-generated “atom colors” that capture the local structure around each atom within each structure. A Gibbs sampler (a Markov chain Monte Carlo technique) was implemented for simulating the posterior distribution of model parameters. We have made software for implementing this methodology publicly available via the R package BayesianGLasso. Results / Conclusions First, we demonstrate robust performance of our methodology (sensitivity, specificity, and measures of accuracy) for recovering the true underlying partial correlation structure over simulated datasets (with simulated metabolite abundances and simulated known structural similarity). We then present an interactome model for stable heart disease inferred from non-targeted mass spectrometry data via this methodology. Inspection of the local graph topology about cholate reveals probabilistic interactions with other primary bile acids, secondary bile acids, and many steroid hormones sharing the same precursors.

Viral DNA integration and methylation of human papillomavirus type 16 in high-grade oral epithelial dysplasia and head and neck squamous cell carcinoma

This study evaluated the integration and methlyation of human papillomavirus type 16 (HPV16) in head and neck squamous cell carcinoma (HNSCC) and its oral precursor, high-grade oral epithelial dysplasia (hgOED). Archival samples of HPV16-positive hgOED (N = 19) and HNSCC (N = 15) were evaluated, along with three HNSCC (UMSCC-1, -47 and -104) and two cervical cancer (SiHa and CaSki) cell lines. HgOED cases were stratified into three groups with increasing degrees of cytologic changes (mitosis, karyorrhexis and apoptosis). The viral load was higher and the E2/E6 ratio lower (indicating a greater tendency toward viral integration) in group 3 than in groups 1 or 2 (p = 0.002, 0.03). Methylation was not observed in hgOED cases and occurred variably in only three HNSCC cases (26.67%, 60.0% and 93.3%). In HNSCC cell lines, lower E7 expression correlated with higher levels of methylation. HgOED with increased cytologic change, now termed HPV-associated oral epithelial dysplasia (HPV-OED), exhibited an increased viral load and a tendency toward DNA integration, suggesting a potentially increased risk for malignant transformation. More detailed characterization and clinical follow-up of HPV-OED patients is needed to determine whether HPV-OED is a true precursor to HPV-associated HNSCC and to clarify the involvement of HPV in HNSCC carcinogenesis.

Knockout of human arylamine N-acetyltransferase 1 (NAT1) in MDA-MB-231 breast cancer cells leads to increased reserve capacity, maximum mitochondrial capacity, and glycolytic reserve capacity

Human arylamine N‐acetyltransferase 1 (NAT1) is a phase II xenobiotic metabolizing enzyme found in almost all tissues. NAT1 can also hydrolyze acetyl‐coenzyme A (acetyl‐CoA) in the absence of an arylamine substrate. Expression of NAT1 varies between individuals and is elevated in several cancers including estrogen receptor positive (ER+) breast cancers. To date, however, the exact mechanism by which NAT1 expression affects mitochondrial bioenergetics in breast cancer cells has not been described. To further evaluate the role of NAT1 in energy metabolism MDA‐MB‐231 breast cancer cells with parental, increased, and knockout levels of NAT1 activity were compared for bioenergetics profile. Basal oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured followed by programmed sequential injection of Oligomycin (ATP synthase inhibitor), FCCP (ETC uncoupler), Antimycin A (Complex III inhibitor), and Rotenone (Complex I inhibitor) to evaluate mitochondrial bioenergetics. Compared to the cell lines with parental NAT1 activity, NAT1 knockout MDA‐MB‐231 cell lines exhibited significant differences in bioenergetics profile, while those with increased NAT1 did not. Significant increases in reserve capacity, maximum mitochondrial capacity, and glycolytic reserve capacity were observed in NAT1 knockout MDA‐MB‐231 cell lines compared to those with parental and increased NAT1 activity. These data indicate that NAT1 knockout in MDA‐MB‐231 breast cancer cells may enhance adaptation to stress by increasing plasticity in response to energy demand.

Wisdom of artificial crowds feature selection in untargeted metabolomics: An application to the development of a blood-based diagnostic test for thrombotic myocardial infarction

INTRODUCTIONnHeart disease remains a leading cause of global mortality. While acute myocardial infarction (colloquially: heart attack), has multiple proximate causes, proximate etiology cannot be determined by a blood-based diagnostic test. We enrolled a suitable patient cohort and conducted a non-targeted quantification of plasma metabolites by mass spectrometry for developing a test that can differentiate between thrombotic MI, non-thrombotic MI, and stable disease. A significant challenge in developing such a diagnostic test is solving the NP-hard problem of feature selection for constructing an optimal statistical classifier.nnnOBJECTIVEnWe employed a Wisdom of Artificial Crowds (WoAC) strategy for solving the feature selection problem and evaluated the accuracy and parsimony of downstream classifiers in comparison with traditional feature selection techniques including the Lasso and selection using Random Forest variable importance criteria.nnnMATERIALS AND METHODSnArtificial Crowd Wisdom was generated via aggregation of the best solutions from independent and diverse genetic algorithm populations that were initialized with bootstrapping and a random subspaces constraint.nnnRESULTS/CONCLUSIONSnStrong evidence was observed that a statistical classifier utilizing WoAC feature selection can discriminate between human subjects presenting with thrombotic MI, non-thrombotic MI, and stable Coronary Artery Disease given abundances of selected plasma metabolites. Utilizing the abundances of twenty selected metabolites, a leave-one-out cross-validation estimated misclassification rate of 2.6% was observed. However, the WoAC feature selection strategy did not perform better than the Lasso over the current study.

When Given a Lemon, Make Lemonade: Revising Cardiovascular Risk Prediction Scores

Accurate estimation of cardiac risk is essential to balancing the risks and benefits of preventive therapies. Underestimation of risk results in preventable cardiovascular events, and overestimation results in unnecessary costs and exposure to treatment side effects. In 2013, the American College of Cardiology and American Heart Association published a new risk score for atherosclerotic cardiovascular disease (ASCVD) to guide risk-reducing therapy (1). However, the applicability of these new pooled cohort equations (PCEs) has been questioned because they have been found to overestimate risk (2, 3). Yadlowsky and colleagues evaluated 2 approaches for improving the PCEs: using the same derivation method as the 2013 PCEs with updated cohort data and using both updated data and new derivation methods (4). The first approach modestly improved discrimination but did not consistently improve calibration, whereas the second approach improved both calibration and discrimination. The authors novel approach to deriving a revised model for ASCVD risk prediction included use of elastic net regularization for variable selection and model estimation. Rather than maximizing the accuracy of the model on the derivation cohort, the elastic net seeks a more conservative model that favors more robust relationships, which may improve accuracy outside the derivation cohort. To increase generalization and limit overfitting of the model to the derivation cohort, Yadlowsky and colleagues did not construct separate race models as was done in derivation of the 2013 PCEs. They instead evaluated whether race modulates the association of each risk factor with overall ASCVD risk and included only interactions selected by the elastic net. Furthermore, they note that Cox proportional hazard models, used in the 2013 PCEs, require an assumption that risk factors only modulate risk in a proportional manner (as opposed to accelerating or decelerating the rate of ASCVD events). As an alternative, they used modified logistic regression to account for participants who did not complete follow-up or have an event before study completion (censoring). A logistic regression approach does not require a proportional risk assumption because it models only the occurrence versus nonoccurrence of ASCVD events over a specified period (10 years). Consequently, Yadlowsky and colleagues risk model does not distinguish between events on day 1 and those in year 10 and cannot be used to estimate risk over shorter or longer periods of time. Further research is warranted to determine how this method of accounting for censoring performs in comparison with other methods (5) or with survival models that do not require a proportional hazards assumption. The success of these approaches to revising the PCEs was evidenced by the finding that, compared with the original 2013 PCEs, the revised PCEs correctly reclassified 13 persons who did not have an ASCVD event as low risk for each 1 person who had an ASCVD event but was reclassified as low risk (10-year risk <7.5%). At a more conservative definition of low risk (10-year risk <5%), the ratio improved to 23:1, whereas at a more liberal definition (10-year risk <10%), the ratio was 8:1. Applying the proposed statistical methods to the updated cohort also produced a risk equation that substantially narrowed the discordance in risk prediction between black and white adults with comparable risk factor burdens. However, these data do not resolve the fundamental question: Does race change how the risk factors accounted for in the PCEs affect the propensity to have an ASCVD event? In our prior work using MESA (Multi-Ethnic Study of Atherosclerosis), we did not observe a difference in how risk factors impact ASCVD on the basis of race/ethnicity (6). This question is timely because a growing Hispanic population and 40% increase in Asian Americans account for more than half of the U.S. population growth between 2000 and 2010 (7). We clearly need an accurate risk assessment tool for these growing American populations. The most notable limitation of Yadlowsky and colleagues study is that results are from internal cross-validation and prospective hold-out validation. Although such validation is useful, results may differ when risk scores are applied to separate cohorts by independent investigators, as seen with the 2013 PCEs. Therefore, independent tests of these revised PCEs in independent cohorts are needed to determine appropriateness for clinical practice. Yadlowsky and colleagues highlight limitations of the 2013 PCEs, many of which can also be addressed with other statistical methods. For example, Bayesian models, like elastic net methods, reduce the risk for overfitting to random noise that may be present in the derivation cohort but do not reflect true relationships between risk factors and ASCVD events. In addition, a Bayesian risk prediction model would provide both estimated risk and its uncertainty for an individual patient (8). An estimated risk of 7.5% with little uncertainty (95% credible interval, 7.2% to 7.8%) may have different implications from the same risk with much greater uncertainty (95% credible interval, 7.0% to 20.5%). Risk prediction models that rely solely on risk factors are ultimately limited by how these factors are measured, factors not accounted for in the model, and how the included factors affect individual patients with differing susceptibility to cardiovascular injury. Measures of subclinical vascular disease (such as coronary artery calcium) can assess the cumulative effect of risk factors on a person and improve risk prediction beyond models that use ASCVD risk factors alone (9). However, even accurate measures of atherosclerotic burden have limitations. Although atherosclerosis is a requisite substrate for ASCVD events, not all atherosclerotic plaques result in ASCVD events. Factors beyond atherosclerotic burden must be identified to predict susceptibility to the transition from stable atherosclerosis to an ASCVD event. Risk assessment based on cardiovascular risk factors is a major medical advancement, but continued effort is needed to produce accurate risk assessment tools for specific patient populations. Risk prediction is an evolving science and will require continual updating through the study of contemporary data from various sources, including consortia of traditional cohorts and big data from electronic medical records (10). Yadlowsky and colleagues show us that contemporary cohorts and statistical methods beyond the purview of classical epidemiology are important for accomplishing this goal.

Circulating Prolidase Activity in Patients with Myocardial Infarction

Background Collagen is a major determinant of atherosclerotic plaque stability. Thus, identification of differences in enzymes that regulate collagen integrity could be useful for predicting susceptibility to atherothrombosis or for diagnosing plaque rupture. In this study, we sought to determine whether prolidase, the rate-limiting enzyme of collagen turnover, differs in human subjects with acute myocardial infarction (MI) versus those with stable coronary artery disease (CAD). Methods We measured serum prolidase activity in 15 patients with stable CAD and 49 patients with acute MI, of which a subset had clearly defined thrombotic MI (nu2009=u200922) or non-thrombotic MI (nu2009=u200912). Prolidase activity was compared across study time points (at cardiac catheterization, T0; 6u2009h after presentation, T6; and at a quiescent follow-up, Tf/u) in acute MI and stable CAD subjects. We performed subgroup analyses to evaluate prolidase activity in subjects presenting with acute thrombotic versus non-thrombotic MI. Results Although prolidase activity was lower at T0 and T6 versus the quiescent phase in acute MI and stable CAD subjects (pu2009<u20090.0001), it was not significantly different between acute MI and stable CAD subjects at any time point (T0, T6, and Tf/u) or between thrombotic and non-thrombotic MI groups. Preliminary data from stratified analyses of a small number of diabetic subjects (nu2009=u20098) suggested lower prolidase activity in diabetic acute MI subjects compared with non-diabetic acute MI subjects (pu2009=u20090.02). Conclusion Circulating prolidase is not significantly different between patients with acute MI and stable CAD or between patients with thrombotic and non-thrombotic MI. Further studies are required to determine if diabetes significantly affects prolidase activity and how this might relate to the risk of MI.