Detecting Cerebral Ischemia From Electroencephalography During Carotid Endarterectomy Using Machine Learning.

Monitoring cerebral neuronal activity via electroencephalography (EEG) during surgery can detect ischemia, a precursor to stroke. However, current neurophysiologist-based monitoring is prone to error. In this study, we evaluated machine learning (ML) for efficient and accurate ischemia detection. We trained supervised ML models on a dataset of 802 patients with intraoperative ischemia labels and evaluated them on an independent validation dataset of 30 patients with refined labels from five neurophysiologists. Our results show moderate-to-substantial agreement between neurophysiologists, with Cohen's kappa values between 0.59 and 0.74. Neurophysiologist performance ranged from 58-93% for sensitivity and 83-96% for specificity, while ML models demonstrated comparable ranges of 63-89% and 85-96%. Random Forest (RF), LightGBM (LGBM), and XGBoost RF achieved area under the receiver operating characteristic curve (AUROC) values of 0.92-0.93 and area under the precision-recall curve (AUPRC) values of 0.79-0.83. ML has the potential to improve intraoperative monitoring, enhancing patient safety and reducing costs.

Time-Series Aware Metrics for the Evaluation of Intraoperative Electroencephalography-Based Ischemia Detection.

Continuous intraoperative monitoring with electroencephalo2 graphy (EEG) is commonly used to detect cerebral ischemia in high-risk surgical procedures such as carotid endarterectomy. Machine learning (ML) models that detect ischemia in real time can form the basis of automated intraoperative EEG monitoring. In this study, we describe and compare two time-series aware precision and recall metrics to the classical precision and recall metrics for evaluating the performance of ML models that detect ischemia. We trained six ML models to detect ischemia in intraoperative EEG and evaluated them with the area under the precision-recall curve (AUPRC) using time-series aware and classical approaches to compute precision and recall. The Support Vector Classification (SVC) model performed the best on the time-series aware metrics, while the Light Gradient Boosting Machine (LGBM) model performed the best on the classical metrics. Visual inspection of the probability outputs of the models alongside the actual ischemic periods revealed that the time-series aware AUPRC selected a model more likely to predict ischemia onset in a timely fashion than the model selected by classical AUPRC.

Obesity-Linked PPARγ S273 Phosphorylation Promotes Insulin Resistance through Growth Differentiation Factor 3.

The thiazolidinediones (TZDs) are ligands of PPARγ that improve insulin sensitivity, but their use is limited by significant side effects. Recently, we demonstrated a mechanism wherein TZDs improve insulin sensitivity distinct from receptor agonism and adipogenesis: reversal of obesity-linked phosphorylation of PPARγ at serine 273. However, the role of this modification hasn't been tested genetically. Here we demonstrate that mice encoding an allele of PPARγ that cannot be phosphorylated at S273 are protected from insulin resistance, without exhibiting differences in body weight or TZD-associated side effects. Indeed, hyperinsulinemic-euglycemic clamp experiments confirm insulin sensitivity. RNA-seq in these mice reveals reduced expression of Gdf3, a BMP family member. Ectopic expression of Gdf3 is sufficient to induce insulin resistance in lean, healthy mice. We find Gdf3 inhibits BMP signaling and insulin signaling in vitro. Together, these results highlight the diabetogenic role of PPARγ S273 phosphorylation and focus attention on a putative target, Gdf3.

Author Correction: γδ T cells producing interleukin-17A regulate adipose regulatory T cell homeostasis and thermogenesis.

In the version of this article initially published, three authors (Hui-Fern Kuoy, Adam P. Uldrich and Dale. I. Godfrey) and their affiliations, acknowledgments and contributions were not included. The correct information is as follows:Ayano C. Kohlgruber1,2, Shani T. Gal-Oz3, Nelson M. LaMarche1,2, Moto Shimazaki1, Danielle Duquette4, Hui-Fern Koay5,6, Hung N. Nguyen1, Amir I. Mina4, Tyler Paras1, Ali Tavakkoli7, Ulrich von Andrian2,8, Adam P. Uldrich5,6, Dale I. Godfrey5,6, Alexander S. Banks4, Tal Shay3, Michael B. Brenner1,10* and Lydia Lynch1,4,9,10*1Division of Rheumatology, Immunology and Allergy, Brigham and Women's Hospital, Boston, MA, USA. 2Division of Medical Sciences, Harvard Medical School, Boston, MA, USA. 3Department of Life Sciences, Ben-Gurion University of the Negev, Beersheba, Israel. 4Division of Endocrinology, Department of Medicine, Brigham and Women's Hospital, Boston, MA, USA. 5Department of Microbiology and Immunology, Peter Doherty Institute for Infection and Immunity, University of Melbourne, Parkville, Australia. 6ARC Centre of Excellence in Advanced Molecular Imaging, University of Melbourne, Parkville, Australia. 7Department of General and Gastrointestinal Surgery, Brigham and Women's Hospital, Boston, MA, USA. 8Department of Microbiology and Immunology, Harvard Medical School, Boston, MA, USA. 9School of Biochemistry and Immunology, Trinity College, Dublin, Ireland. 10These authors jointly supervised this work: Michael B. Brenner, Lydia Lynch. *e-mail: mbrenner@research.bwh.harvard.edu; llynch@bwh.harvard.eduAcknowledgementsWe thank A.T. Chicoine, flow cytometry core manager at the Human Immunology Center at BWH, for flow cytometry sorting. We thank D. Sant'Angelo (Rutgers Cancer Institute) for providing Zbtb16-/- mice and R. O'Brien (National Jewish Health) for providing Vg4/6-/- mice. Supported by NIH grant R01 AI11304603 (to M.B.B.), ERC Starting Grant 679173 (to L.L.), the National Health and Medical Research Council of Australia (1013667), an Australian Research Council Future Fellowship (FT140100278 for A.P.U.) and a National Health and Medical Research Council of Australia Senior Principal Research Fellowship (1117766 for D.I.G.).Author contributionsA.C.K., L.L., and M.B.B. conceived and designed the experiments, and wrote the manuscript. A.C.K., N.M.L., L.L., H.N.N., M.S., T.P., and D.D. performed the experiments. S.T.G.-O. and T.S. performed the RNA-seq analysis. A.S.B. and A.I.M. provided advice and performed the CLAMS experiments. A.T. provided human bariatric patient samples. Parabiosis experiments were performed in the laboratory of U.v.A. H.-F.K., A.P.U. and D.I.G provided critical insight into the TCR chain usage of PLZF+ γδ T cells. M.B.B., N.M.L., and L.L. critically reviewed the manuscript.The errors have been corrected in the HTML and PDF version of the article.Correction to: Nature Immunology doi:10.1038/s41590-018-0094-2 (2018), published online 18 April 2018.

Timing of Lymphedema After Treatment for Breast Cancer: When Are Patients Most At Risk?

PURPOSE: The purpose of the study was to determine when the risk of lymphedema is highest after treatment of breast cancer and which factors influence the time course of lymphedema development. METHODS AND MATERIALS: Between 2005 and 2017, 2171 women (with 2266 at-risk arms) who received surgery for unilateral or bilateral breast cancer at our institution were enrolled. Perometry was used to objectively assess limb volume preoperatively, and lymphedema was defined as a ≥10% relative arm-volume increase arising >3 months postoperatively. Multivariable regression was used to uncover risk factors associated with lymphedema, the Cox proportional hazards model was used to calculate lymphedema incidence, and the semiannual hazard rate of lymphedema was calculated. RESULTS: With a median follow-up of 4 years, the overall estimated 5-year cumulative incidence of lymphedema was 13.7%. Significant factors associated with lymphedema on multivariable analysis were high preoperative body mass index, axillary lymph node dissection (ALND), and regional lymph node radiation (RLNR). Patients receiving ALND with RLNR experienced the highest 5-year rate of lymphedema (31.2%), followed by those receiving ALND without RLNR (24.6%) and sentinel lymph node biopsy with RLNR (12.2%). Overall, the risk of lymphedema peaked between 12 and 30 months postoperatively; however, the time course varied as a function of therapy received. Early-onset lymphedema (<12 months postoperatively) was associated with ALND (HR [hazard ratio], 4.75; P < .0001) but not with RLNR (HR, 1.21; P = .55). In contrast, late-onset lymphedema (>12 months postoperatively) was associated with RLNR (HR, 3.86; P = .0001) and, to a lesser extent, ALND (HR, 1.86; P = .029). The lymphedema risk peaked between 6 and 12 months in the ALND-without-RLNR group, between 18 and 24 months in the ALND-with-RLNR group, and between 36 and 48 months in the group receiving sentinel lymph node biopsy with RLNR. CONCLUSIONS: The time course for lymphedema development depends on the breast cancer treatment received. ALND is associated with early-onset lymphedema, and RLNR is associated with late-onset lymphedema. These results can influence clinical practice to guide lymphedema surveillance strategies and patient education.

CalR: A Web-Based Analysis Tool for Indirect Calorimetry Experiments.

We report a web-based tool for analysis of experiments using indirect calorimetry to measure physiological energy balance. CalR simplifies the process to import raw data files, generate plots, and determine the most appropriate statistical tests for interpretation. Analysis using the generalized linear model (which includes ANOVA and ANCOVA) allows for flexibility in interpreting diverse experimental designs, including those of obesity and thermogenesis. Users also may produce standardized output files for an experiment that can be shared and subsequently re-evaluated using CalR. This framework will provide the transparency necessary to enhance consistency, rigor, and reproducibility. The CalR analysis software will greatly increase the speed and efficiency with which metabolic experiments can be organized, analyzed per accepted norms, and reproduced and will likely become a standard tool for the field. CalR is accessible at https://CalRapp.org/.

A selective gut bacterial bile salt hydrolase alters host metabolism.

The human gut microbiota impacts host metabolism and has been implicated in the pathophysiology of obesity and metabolic syndromes. However, defining the roles of specific microbial activities and metabolites on host phenotypes has proven challenging due to the complexity of the microbiome-host ecosystem. Here, we identify strains from the abundant gut bacterial phylum Bacteroidetes that display selective bile salt hydrolase (BSH) activity. Using isogenic strains of wild-type and BSH-deleted Bacteroides thetaiotaomicron, we selectively modulated the levels of the bile acid tauro-ß-muricholic acid in monocolonized gnotobiotic mice. B. thetaiotaomicron BSH mutant-colonized mice displayed altered metabolism, including reduced weight gain and respiratory exchange ratios, as well as transcriptional changes in metabolic, circadian rhythm, and immune pathways in the gut and liver. Our results demonstrate that metabolites generated by a single microbial gene and enzymatic activity can profoundly alter host metabolism and gene expression at local and organism-level scales.

Ablation of PM20D1 reveals N-acyl amino acid control of metabolism and nociception.

N-acyl amino acids (NAAs) are a structurally diverse class of bioactive signaling lipids whose endogenous functions have largely remained uncharacterized. To clarify the physiologic roles of NAAs, we generated mice deficient in the circulating enzyme peptidase M20 domain-containing 1 (PM20D1). Global PM20D1-KO mice have dramatically reduced NAA hydrolase/synthase activities in tissues and blood with concomitant bidirectional dysregulation of endogenous NAAs. Compared with control animals, PM20D1-KO mice exhibit a variety of metabolic and pain phenotypes, including insulin resistance, altered body temperature in cold, and antinociceptive behaviors. Guided by these phenotypes, we identify N-oleoyl-glutamine (C18:1-Gln) as a key PM20D1-regulated NAA. In addition to its mitochondrial uncoupling bioactivity, C18:1-Gln also antagonizes certain members of the transient receptor potential (TRP) calcium channels including TRPV1. Direct administration of C18:1-Gln to mice is sufficient to recapitulate a subset of phenotypes observed in PM20D1-KO animals. These data demonstrate that PM20D1 is a dominant enzymatic regulator of NAA levels in vivo and elucidate physiologic functions for NAA signaling in metabolism and nociception.

γδ T cells producing interleukin-17A regulate adipose regulatory T cell homeostasis and thermogenesis.

γδ T cells are situated at barrier sites and guard the body from infection and damage. However, little is known about their roles outside of host defense in nonbarrier tissues. Here, we characterize a highly enriched tissue-resident population of γδ T cells in adipose tissue that regulate age-dependent regulatory T cell (Treg) expansion and control core body temperature in response to environmental fluctuations. Mechanistically, innate PLZF+ γδ T cells produced tumor necrosis factor and interleukin (IL) 17 A and determined PDGFRα+ and Pdpn+ stromal-cell production of IL-33 in adipose tissue. Mice lacking γδ T cells or IL-17A exhibited decreases in both ST2+ Treg cells and IL-33 abundance in visceral adipose tissue. Remarkably, these mice also lacked the ability to regulate core body temperature at thermoneutrality and after cold challenge. Together, these findings uncover important physiological roles for resident γδ T cells in adipose tissue immune homeostasis and body-temperature control.

Phosphorylation of Beta-3 adrenergic receptor at serine 247 by ERK MAP kinase drives lipolysis in obese adipocytes.

OBJECTIVE: The inappropriate release of free fatty acids from obese adipose tissue stores has detrimental effects on metabolism, but key molecular mechanisms controlling FFA release from adipocytes remain undefined. Although obesity promotes systemic inflammation, we find activation of the inflammation-associated Mitogen Activated Protein kinase ERK occurs specifically in adipose tissues of obese mice, and provide evidence that adipocyte ERK activation may explain exaggerated adipose tissue lipolysis observed in obesity. METHODS AND RESULTS: We provide genetic and pharmacological evidence that inhibition of the MEK/ERK pathway in human adipose tissue, mice, and flies all effectively limit adipocyte lipolysis. In complementary findings, we show that genetic and obesity-mediated activation of ERK enhances lipolysis, whereas adipose tissue specific knock-out of ERK2, the exclusive ERK1/2 protein in adipocytes, dramatically impairs lipolysis in explanted mouse adipose tissue. In addition, acute inhibition of MEK/ERK signaling also decreases lipolysis in adipose tissue and improves insulin sensitivity in obese mice. Mice with decreased rates of adipose tissue lipolysis in vivo caused by either MEK or ATGL pharmacological inhibition were unable to liberate sufficient White Adipose Tissue (WAT) energy stores to fuel thermogenesis from brown fat during a cold temperature challenge. To identify a molecular mechanism controlling these actions, we performed unbiased phosphoproteomic analysis of obese adipose tissue at different time points following acute pharmacological MEK/ERK inhibition. MEK/ERK inhibition decreased levels of adrenergic signaling and caused de-phosphorylation of the ß3-adrenergic receptor (ß3AR) on serine 247. To define the functional implications of this phosphorylation, we showed that CRISPR/Cas9 engineered cells expressing wild type ß3AR exhibited ß3AR phosphorylation by ERK2 and enhanced lipolysis, but this was not seen when serine 247 of ß3AR was mutated to alanine. CONCLUSION: Taken together, these data suggest that ERK activation in adipocytes and subsequent phosphorylation of the ß3AR on S247 are critical regulatory steps in the enhanced adipocyte lipolysis of obesity.

Association Between Precautionary Behaviors and Breast Cancer-Related Lymphedema in Patients Undergoing Bilateral Surgery.

Purpose This study examined the lifestyle and clinical risk factors for lymphedema in a cohort of patients who underwent bilateral breast cancer surgery. Patients and Methods Between 2013 and 2016, 327 patients who underwent bilateral breast cancer surgery were prospectively screened for arm lymphedema as quantified by the weight-adjusted volume change (WAC) formula. Arm perometry and subjective data were collected preoperatively and at regular intervals postoperatively. At the time of each measurement, patients completed a risk assessment survey that reported the number of blood draws, injections, blood pressure readings, trauma to the at-risk arm, and number of flights since the previous measurement. Generalized estimating equations were applied to ascertain the association among arm volume changes, clinical factors, and risk exposures. Results The cohort comprised 327 patients and 654 at-risk arms, with a median postoperative follow-up that ranged from 6.1 to 68.2 months. Of the 654 arms, 83 developed lymphedema, defined as a WAC ≥ 10% relative to baseline. On multivariable analysis, none of the lifestyle risk factors examined through the risk assessment survey were significantly associated with increased WAC. Multivariable analysis demonstrated that having a body mass index ≥ 25 kg/m2 at the time of breast cancer diagnosis ( P = .0404), having undergone axillary lymph node dissection ( P = .0464), and receipt of adjuvant chemotherapy ( P = .0161) were significantly associated with increased arm volume. Conclusion Blood pressure readings, blood draws, injections, and number or duration of flights were not significantly associated with increases in arm volume in this cohort. These findings may help to guide patient education about lymphedema risk reduction strategies for those who undergo bilateral breast cancer surgery.

Cdkal1, a type 2 diabetes susceptibility gene, regulates mitochondrial function in adipose tissue.

OBJECTIVES: Understanding how loci identified by genome wide association studies (GWAS) contribute to pathogenesis requires new mechanistic insights. Variants within CDKAL1 are strongly linked to an increased risk of developing type 2 diabetes and obesity. Investigations in mouse models have focused on the function of Cdkal1 as a tRNALys modifier and downstream effects of Cdkal1 loss on pro-insulin translational fidelity in pancreatic ß-cells. However, Cdkal1 is broadly expressed in other metabolically relevant tissues, including adipose tissue. In addition, the Cdkal1 homolog Cdk5rap1 regulates mitochondrial protein translation and mitochondrial function in skeletal muscle. We tested whether adipocyte-specific Cdkal1 deletion alters systemic glucose homeostasis or adipose mitochondrial function independently of its effects on pro-insulin translation and insulin secretion. METHODS: We measured mRNA levels of type 2 diabetes GWAS genes, including Cdkal1, in adipose tissue from lean and obese mice. We then established a mouse model with adipocyte-specific Cdkal1 deletion. We examined the effects of adipose Cdkal1 deletion using indirect calorimetry on mice during a cold temperature challenge, as well as by measuring cellular and mitochondrial respiration in vitro. We also examined brown adipose tissue (BAT) mitochondrial morphology by electron microscopy. Utilizing co-immunoprecipitation followed by mass spectrometry, we performed interaction mapping to identify new CDKAL1 binding partners. Furthermore, we tested whether Cdkal1 loss in adipose tissue affects total protein levels or accurate Lys incorporation by tRNALys using quantitative mass spectrometry. RESULTS: We found that Cdkal1 mRNA levels are reduced in adipose tissue of obese mice. Using adipose-specific Cdkal1 KO mice (A-KO), we demonstrated that mitochondrial function is impaired in primary differentiated brown adipocytes and in isolated mitochondria from A-KO brown adipose tissue. A-KO mice displayed decreased energy expenditure during 4 °C cold challenge. Furthermore, mitochondrial morphology was highly abnormal in A-KO BAT. Surprisingly, we found that lysine codon representation was unchanged in Cdkal1 A-KO adipose tissue. We identified novel protein interactors of CDKAL1, including SLC25A4/ANT1, an inner mitochondrial membrane ADP/ATP translocator. ANT proteins can account for the UCP1-independent basal proton leak in BAT mitochondria. Cdkal1 A-KO mice had increased ANT1 protein levels in their white adipose tissue. CONCLUSIONS: Cdkal1 is necessary for normal mitochondrial morphology and function in adipose tissue. These results suggest that the type 2 diabetes susceptibility gene CDKAL1 has novel functions in regulating mitochondrial activity.

Genetic Depletion of Adipocyte Creatine Metabolism Inhibits Diet-Induced Thermogenesis and Drives Obesity.

Diet-induced thermogenesis is an important homeostatic mechanism that limits weight gain in response to caloric excess and contributes to the relative stability of body weight in most individuals. We previously demonstrated that creatine enhances energy expenditure through stimulation of mitochondrial ATP turnover, but the physiological role and importance of creatine energetics in adipose tissue have not been explored. Here, we have inactivated the first and rate-limiting enzyme of creatine biosynthesis, glycine amidinotransferase (GATM), selectively in fat (Adipo-Gatm KO). Adipo-Gatm KO mice are prone to diet-induced obesity due to the suppression of elevated energy expenditure that occurs in response to high-calorie feeding. This is paralleled by a blunted capacity for ß3-adrenergic activation of metabolic rate, which is rescued by dietary creatine supplementation. These results provide strong in vivo genetic support for a role of GATM and creatine metabolism in energy expenditure, diet-induced thermogenesis, and defense against diet-induced obesity.