# PERIPHERAL REGULATORS OF OBESITY

EDITED BY : Andrew J. McAinch, Deanne H. Hryciw and Sudip Bajpeyi PUBLISHED IN : Frontiers in Endocrinology

#### Frontiers Copyright Statement

© Copyright 2007-2019 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.

The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.

Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.

Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.

As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.

All copyright, and all rights therein, are protected by national and international copyright laws.

The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use. ISSN 1664-8714 ISBN 978-2-88963-020-2 DOI 10.3389/978-2-88963-020-2

#### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

#### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

#### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

#### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

# PERIPHERAL REGULATORS OF OBESITY

Topic Editors:

Andrew J. McAinch, Victoria University, Australia Deanne H. Hryciw, Griffith University and Victoria University, Australia Sudip Bajpeyi, University of Texas at El Paso, United States

Citation: McAinch, A. J., Hryciw, D. H., Bajpeyi, S., eds. (2019). Peripheral Regulators of Obesity. Lausanne: Frontiers Media. doi: 10.3389/978-2-88963-020-2

# Table of Contents

	- *20 Deciphering the Roles of PPAR*γ *in Adipocytes via Dynamic Change of Transcription Complex*

Xinran Ma, Dongmei Wang, Wenjun Zhao and Lingyan Xu


Alexandre G. Oliveira, Tiago G. Araújo, Bruno de Melo Carvalho, Guilherme Z. Rocha, Andrey Santos and Mario J. A. Saad


Jing Cui, Xiufen Sun, Xiaojing Li, Ma Ke, Jianping Sun, Nafeesa Yasmeen, Jamal Muhammad Khan, Hualei Xin, Shouyong Xue and Zulqarnain Baloch

*75 Fat Mass Follows a U-Shaped Distribution Based on Estradiol Levels in Postmenopausal Women*

Georgia Colleluori, Rui Chen, Nicola Napoli, Lina E. Aguirre, Clifford Qualls, Dennis T. Villareal and Reina Armamento-Villareal


Ki-Hoon Park, Dhananjay Yadav, Suk-Jeong Kim, Jae-Ryong Kim and Kyung-Hyun Cho

*101 Twelve Weeks of Yoga or Nutritional Advice for Centrally Obese Adult Females*

Shirley Telles, Sachin K. Sharma, Niranjan Kala, Sushma Pal, Ram K. Gupta and Acharya Balkrishna

# Editorial: Peripheral Regulators of Obesity

#### Andrew J. McAinch1,2 \*, Deanne H. Hryciw1,3 and Sudip Bajpeyi <sup>4</sup>

*1 Institute for Health and Sport, College of Health and Biomedicine, Victoria University, Melbourne, VIC, Australia, <sup>2</sup> Australian Institute for Musculoskeletal Science, Victoria University, Melbourne, VIC, Australia, <sup>3</sup> School of Environment and Science, Griffith Sciences, Griffith University, Nathan, QLD, Australia, <sup>4</sup> Department of Kinesiology, College of Health Sciences, University of Texas at El Paso, El Paso, TX, United States*

Keywords: obesity, diabetes, adipose tissue, body composition, TRPV1 - transient receptor potential vanilloid type 1 receptor, PPAR gamma, HIV, hepatocyte growth factor (HGF)

#### **Editorial on the Research Topic**

#### **Peripheral Regulators of Obesity**

The incidence of obesity is increasing significantly worldwide and is considered one of the most significant public health challenges due to the associated disease states increasing morbidity and mortality and reducing quality of life. Despite the impact (both social and economic) of obesity little is known of the complex physiology that dictates the maintenance of the obese state and resistance to weight loss, and the link between obesity and complex diseases, including type 2 diabetes. Understanding the peripheral abnormalities that are present in obesity are essential to our development of treatments for these conditions. This special issue aimed to elucidate at least some of the complex interactions that occur particularly in the periphery. The special issue contained a mix of 4 review articles and 6 original manuscripts ranging from cell culture studies to clinical trials and large cross sectional observations.

The review by Christie et al. discusses TRPV1 structure and function, how this is linked to the endocannabinoid system, energy homeostasis, appetite control, and gastrointestinal vagal afferents. The review concluded with how TRPV1 activation may stimulate energy expenditure particularly via brown adipose tissue thermogenesis and its role in β-cell function in diabetes. However, despite TRPV1's potential roles in energy expenditure and blood glucose regulation, it has complex interactions with many signaling pathways. Thus further research to provide fundamental physiological understanding of TRPV1 to enable targeted treatment options for obesity and associated metabolic diseases is required.

Peroxisome proliferator-activated receptor γ (PPARγ) and its role in adipocytes via dynamic change of transcription was reviewed by Ma et al. PPARγ plays a key role in adipogenesis through ligand binding and selective interaction with transcriptional corepressors or coactivators. Of note, PPARγ plays a critical role in the determination of the brown and beige adipocytes. Ma et al. conclude via the discussion of the role of specific PPARγ phosphorylation sites and how stimulation of phosphorylation at different sites results in potentially different physiological outcomes.

The special issue continues with the review investigating insulin resistance induced by antiretroviral drugs in HIV-Patients by Pedro et al. This review discusses the increase in circulating proinflammatory cytokines and bacterial lipopolysaccharides resulting in downregulation of insulin signaling. Through discussion of what is known about the molecular signaling associated with antiretroviral induced insulin resistance as well as the link with gut microbiota and obesity and dyslipidemia, this review highlights the need for further research into this condition to enable the next generation of targeted medications to ensure the long term quality of life of HIV patients.

Edited and reviewed by:

*Konstantinos Tziomalos, Aristotle University of Thessaloniki, Greece*

#### \*Correspondence:

*Andrew J. McAinch andrew.mcainch@vu.edu.au*

#### Specialty section:

*This article was submitted to Obesity, a section of the journal Frontiers in Endocrinology*

Received: *12 March 2019* Accepted: *20 May 2019* Published: *04 June 2019*

#### Citation:

*McAinch AJ, Hryciw DH and Bajpeyi S (2019) Editorial: Peripheral Regulators of Obesity. Front. Endocrinol. 10:357. doi: 10.3389/fendo.2019.00357*

**4**

The theme of insulin resistance is also continued in the review by Oliveira et al. which discusses the role of hepatocyte growth factor (HGF) in insulin resistance and diabetes. This review discusses the structure and function of HGF and its role in the regulation of glucose metabolism in different cells and tissues. It then continues to provide insight into the involvement of HGF in inflammatory responses as well as cancer therapy and thus the complexity of targeting HGF in obesity or diabetes.

The reviews captured in this special issue provide an interesting snapshot of the current state of research into several signaling cascades that are associated with obesity and insulin resistance as well as the influence of other disease states on these signaling cascades in a number of tissues, including adipose tissue. The special issue then continues with a study by Merlin et al. which demonstrates that cultured white adipocytes derived from regions such as inguinal white adipose tissue (iWAT) with adequate sympathetic innervation respond efficiently to agonists, but only in combination with an additional priming stimulus such as rosiglitazone. This contrasts with in vivo iWAT depots in which chronic sympathetic tone is sufficient to maintain the capacity for brite (or beige) activation. This study emphasized the importance of activating both browning and thermogenic programs in cultured white adipocytes in order to reach maximum browning capacity.

Understanding the impact of not just obesity, but also how our body stores and distributes fat mass can have important implications on our health. Cui et al., determined that in Qingdao, China, being classified as overweight or having abdominal obesity as determined by waist circumference was associated with depression whereas waist-to-hip ratio was not. The impact of body composition not just body weight was also investigated by Colleluori et al. which observed a U-shaped distribution of fat mass when correlated to estradiol (Serum E2) levels in postmenopausal women which was different to the linear increase seen when only looking at body weight or body mass index (BMI). This study raises some speculations regarding E<sup>2</sup>

sensitivity and fat mass regulation in postmenopausal women. An inverse association with circulating SIRT1 and adiposity levels across the body weight range (underweight to obese) was also observed by Mariani et al. This inverse association was also associated with liver fat content, insulin and other markers of metabolic phenotype.

Even in the healthy weight range, according to BMI, Park et al. determined that young women classified as plump (around 55 kg) compared to slim (around 46 kg) young women showed more atherogenic features in LDL and HDL as well as elevated oxidation and glycation with loss of antioxidant ability and apoA-I. Despite these observations, interventions can improve the health outcomes, as determined in this special issue's final manuscript by Telles et al. In this study Telles et al. demonstrated an age effect, with women with central obesity between 30 and 45 years of age having a greater reduction in measures such as waist circumference and BMI following 12 weeks of yoga compared to the control group who just received nutritional advice. This reduction was not seen in the 46–59 year age range, which demonstrates that early interventions may have better health outcomes.

# AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 McAinch, Hryciw and Bajpeyi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Involvement of TRPV1 Channels in Energy Homeostasis

#### Stewart Christie<sup>1</sup> , Gary A. Wittert 1,2, Hui Li 1,2 and Amanda J. Page1,2 \*

<sup>1</sup> Vagal Afferent Research Group, Centre for Nutrition and Gastrointestinal Disease, Adelaide Medical School, University of Adelaide, Adelaide, SA, Australia, <sup>2</sup> Nutrition and Metabolism, South Australian Health and Medical Research Institute, Adelaide, SA, Australia

The ion channel TRPV1 is involved in a wide range of processes including nociception, thermosensation and, more recently discovered, energy homeostasis. Tightly controlling energy homeostasis is important to maintain a healthy body weight, or to aid in weight loss by expending more energy than energy intake. TRPV1 may be involved in energy homeostasis, both in the control of food intake and energy expenditure. In the periphery, it is possible that TRPV1 can impact on appetite through control of appetite hormone levels or via modulation of gastrointestinal vagal afferent signaling. Further, TRPV1 may increase energy expenditure via heat production. Dietary supplementation with TRPV1 agonists, such as capsaicin, has yielded conflicting results with some studies indicating a reduction in food intake and increase in energy expenditure, and other studies indicating the converse. Nonetheless, it is increasingly apparent that TRPV1 may be dysregulated in obesity and contributing to the development of this disease. The mechanisms behind this dysregulation are currently unknown but interactions with other systems, such as the endocannabinoid systems, could be altered and therefore play a role in this dysregulation. Further, TRPV1 channels appear to be involved in pancreatic insulin secretion. Therefore, given its plausible involvement in regulation of energy and glucose homeostasis and its dysregulation in obesity, TRPV1 may be a target for weight loss therapy and diabetes. However, further research is required too fully elucidate TRPV1s role in these processes. The review provides an overview of current knowledge in this field and potential areas for development.

#### Keywords: TRPV1, appetite regulation, metabolism, obesity, endovanilloid, endocannabinoid

# INTRODUCTION

Obesity has become the fifth leading cause of death, and the second leading cause of preventable death worldwide, closely following tobacco smoking (1, 2). There are multiple hormonal, neurotransmitter, and receptor systems involved in the regulation of energy balance. Pharmacological attempts to favorably modulate these systems to encourage weight loss have been somewhat effective, although not without adverse side effects. This has led to the search for more suitable targets. One such group of receptors/ion channels gaining attention for their possible role in energy homeostasis are the Transient Receptor Potential (TRP) channels.

#### Edited by:

Andrew J. McAinch, Victoria University, Australia

#### Reviewed by:

Denis P. Blondin, Université de Sherbrooke, Canada Thomas Alexander Lutz, Universität Zürich, Switzerland

\*Correspondence: Amanda J. Page Amanda.page@adelaide.edu.au

#### Specialty section:

This article was submitted to Obesity, a section of the journal Frontiers in Endocrinology

Received: 30 March 2018 Accepted: 04 July 2018 Published: 31 July 2018

#### Citation:

Christie S, Wittert GA, Li H and Page AJ (2018) Involvement of TRPV1 Channels in Energy Homeostasis. Front. Endocrinol. 9:420. doi: 10.3389/fendo.2018.00420

**6**

TRP channels are a superfamily of about 28 non-selective cation channels divided into 7 subfamilies including TRP vanilloid (TRPV), and TRP ankyrin (TRPA) (3). They were first identified in 1969 from an irregular electroretinogram in a mutant strain of the Drosophila fly (4). The electroretinogram presented a short increase in retinal potential which gave rise to the name "transient receptor potential" (5). Since their discovery, TRP channels have been identified as osmo- and mechano-sensitive (6). For example, TRPA1 is associated with pain sensations and inflammation (7), and TRPV1 is associated with pain and temperature regulation (8).

Endotherms use energy to create heat to maintain body temperature and in colder climates it has been shown that humans expend more energy for thermoregulation compared to warmer climates (9). Given the high energy costs of generating heat to maintain an optimal cellular environment thermoregulation can also play an important role in energy homeostasis. TRPV1 channels are involved in thermoregulation, making them a possible target for the modulation of energy expenditure. Further, it is becoming apparent that TRPV1 may be involved in the regulation of appetite via the modulation of appetite hormones and/or by acting on gastrointestinal vagal afferents. This is a process that may involve interaction with the endocannabinoid system considering that endocannabinoids such as anandamide (AEA), produced in the gastrointestinal tract are also endogenous TRPV1 agonists. In addition, there are suggestions that TRPV1 may be involved in the regulation of insulin secretion in the pancreas. Studies in obese individuals have suggested that TRPV1 may be dysfunctional or dysregulated due to loss of effect on energy homeostasis. For this reason TRPV1 may be a potential target for pharmacological manipulation to aid in weight loss with recent studies suggesting selective blockade or activation of specific functions of TRPV1. However, due to its complexity this may prove difficult. This review explores TRPV1 structure and modulation and will focus on its involvement in energy homeostasis, diabetes, and possible pharmacological manipulation.

#### TRPV1 CHANNELS

TPRV1, the first channel in the vanilloid family, is highly permeable to calcium and was discovered in 1997 by cloning dorsal root ganglia expressed genes in human embryonic kidney cells (10). It is expressed in a wide range of central and peripheral tissues. Centrally, TRPV1 is highly expressed in the brain stem, mid-brain, hypothalamus and limbic system (11). Peripherally it is expressed in many tissues including the vagal and spinal sensory nerves (12), stomach (13), and adipose tissue (14).

#### TRPV1 Structure

The TRPV1 channel consists of four identical subunits located in the plasma membrane with each subunit (**Figure 1**) consisting of an N-terminus, a transmembrane region, and a C-terminus (15, 16). The N-terminus contains an ankyrin repeating domain consisting of 6 ankyrin subunits (16) which in its tertiary structure forms six α-helices connected by finger loops (15). Sites on the N-terminus are capable of phosphorylation by protein kinases with the S116 phosphorylation site being one of functionality (17). A linker section connects the N-terminus to the transmembrane region via the pre-helical segment (pre-S1), and connects TPRV1 subunits together (15–18).

The transmembrane region of each TRPV1 subunit comprises 6 helical segments (S1–S6), where S1–S4 contribute to the voltage-sensing domain, and S5–S6 contribute to the poreforming domain (16). S1-S4 are connected to S5-S6 by a linker segment, and act as a foundation which allows the linker segment to move, contributing to pore opening and TRPV1 activation. The transmembrane region also contains binding sites for several ligands. For example, vanilloids (e.g., capsaicin) are capable of binding to S3 and S4, and protons (H+) are capable of binding to S5 and the S5–S6 linker (pore helix) (15).

Lastly, the C-terminus consists of a TRP domain (TRP-D) which interacts with pre-S1 suggesting a structural role (16). Following the TRP domain are several protein kinase A (PKA) and protein kinase C (PKC) phosphorylation sites, and sites for binding calmodulin and phosphatidylinositol-4,5-bisphosphate (PIP2) (15, 16).

#### TRPV1 Channel Activation or Modulation

TRPV1 is activated by a wide variety of different stimuli including heat, protons (pH < 5.9) (8, 19), capsaicin the irritant compound in hot chilies (10), allicin and diallyl sulfides from garlic (20, 21), peperine from black pepper (22), and gingerol from ginger (23). Spider and jellyfish venom-derived toxins are also TRPV1 agonists (24, 25).

Endogenous agonists are referred to as endovanilloids. To qualify as an endovanilloid the compound should be produced and released in sufficient amount to evoke a TRPV1-mediated response by direct binding and subsequent activation of the channel. Further, to permit regulation of the channel the signal should have a short half-life. Therefore, the mechanisms for synthesis and breakdown of the endovanilloid should be in close proximity to TRPV1. As the binding sites for endogenous ligands of TRPV1 are intracellular (26, 27) then the ligand could also be produced within the cell or there should be a mechanism to bring it into the cell. Three different classes of lipid are known to activate TRPV1 i.e., N-acyl-ethanolamines [NAEs, e.g., AEA (28)], some lipoxygenase products of arachidonic acid and N-acyl-dopamines (e.g., N-arachidonoyldopamine, N-oleoyldopamine) (29). Further, adipose tissue B lymphocytes (B1 cells) that regulate local inflammatory responses produce leukotrienes including leukotriene B4 which is also a TRPV1 agonist (30).

**Abbreviations:** AC, adenylyl cyclase; AEA, anandamide; AMPK, adenosine monophosphate kinase; BAT, brown adipose tissue; BGL, blood glucose level; BMP, bone morphogenic protein; CAMK, calmodulin dependent kinase; CB, cannabinoid; DMV, dorsal motor nucleus of the vagus; GI, gastrointestinal; GLP, glucagon like peptide; PGC, PPAR gamma coactivator; PI3K, phosphatidylinositol 4-5 bisphosphate 3-kinase; PIP2, phosphatidylinositol-4,5-bisphosphate; PKA, protein kinase A; PKC, protein kinase C; PLC, phospholipase C; PPAR, peroxisome proliferator activated receptor; PRDM, positive regulatory domain; SIRT, sirtuin; SNS, somatic nervous system; TRPV, transient receptor potential vanilloid; UCP, uncoupling protein; WAT, white adipose tissue.

Intracellularly, calmodulin, a calcium-binding messenger, mediates the negative feedback loop formed by calcium (31). Calcium binds and activates calmodulin allowing it to bind to the N-terminus or C-terminus of TRPV1 inhibiting TRPV1 activity (15). Other secondary messengers such as PKA, PKC, and PIP<sup>2</sup> are also capable of modulating TRPV1 activity. PKA can enhance or activate TRPV1 through phosphorylation of sites (S116 and T370) on the N-terminal (32) and may play a role in the capsaicin induced Ca2<sup>+</sup> dependent desensitization of TRPV1 activation, a phenomenon which has been extensively reviewed elsewhere (33). PKC directly activates TRPV1 through phosphorylation of the S2–S3 linker region (S502) and C-terminal sites (S800), and also potentiates the effect of other ligands such as protons (34, 35). PIP<sup>2</sup> is a negative regulator, inhibiting TRPV1 activity when bound to the C-terminal sites (TRP domain: K710) (36).

# Interactions Between TRPV1 and the Endocannabinoid System

#### The Endocannabinoid System

The endocannabinoid system consists of endocannabinoids, their receptors and the enzymes involved in endocannabinoid synthesis and degradation. This system is involved in many physiological processes including memory, mood, and relevant to this review promotion of food intake (37). Endocannabinoids are endogenous lipid messengers (e.g., AEA and 2-arachydonoylglycerol) which activate their receptors, cannabinoid receptor-1 (CB1) and cannabinoid receptor-2 (CB2) (38, 39).

These endogenous lipid messengers are synthesized on demand and degraded by cellular uptake and enzymatic hydrolysis [see review (40) and **Figure 2**]. Briefly, the first step in the synthesis of AEA and NAEs is the transacylation of membrane phosphatidyethanolamine-containing phospholipids to N-acylphosphatidyl-ethanolamines (NAPEs) (41, 42). There are a number of ways that NAPEs are metabolized to their corresponding NAE including catalyzed hydrolysis by the NAPE-hydrolysing enzyme phospholipase D (NAPE-PLD) (43). In contrast, diacylglycerol lipase (DAGL) is responsible for the formation of 2-AG (44). There is still some controversy on whether there is an endocannabinoid membrane transporter [See reviews (45, 46)]. Nonetheless, endocannabinoids can be cleared from the extracellular space. Further, there are intracellular proteins that can shuttle these lipids to specific intracellular locations (e.g., TRPV1 for AEA) (47, 48), the best characterized of these are the fatty acid-binding proteins (FABP e.g., FABP5 and 7) (49). The enzymes responsible for the breakdown of AEA and NAEs are fatty acid amide hydrolase (FAAH) and Nacylethanolamine-hydrolysing acid-amidase (NAAA). NAAA is predominantly located in the lungs where it is localized to the lysosomes of macrophages (50, 51). FAAH is more ubiquitous and FAAH-1 is located on the endoplasmic reticulum whereas FAAH-2 (not found in rodents) is located in the lipid rafts (52, 53). Monoacylglycerol lipase (MAGL) is the enzyme responsible for the majority of 2-AG hydrolysis in most tissues (54–56).

The receptors for endocannabinoids, CB1 and CB2 are members of the G-protein coupled receptor family, being predominantly coupled to the Gi/oα proteins that inhibit adenylyl cyclase thereby reducing cellular cAMP levels (57, 58). However, coupling to other effector proteins has also been reported, including activation of G<sup>q</sup> and G<sup>s</sup> proteins, inhibition of voltage-gated calcium channels, activation of inwardly rectifying potassium channels, β-arrestin recruitment and activation of mitogen-activated protein kinase (MAPK) signaling pathways (59). As a result of CB receptor signaling through multiple effector proteins the probability of biased signaling (liganddependent selectivity for specific signal transduction pathways) increases. Biased signaling is thought to occur when different ligands bind to the receptor causing different conformational changes to the receptor enabling the receptor to preferentially signal one pathway over the other (60, 61). This is attractive, in

FABP, fatty acid-binding protein; MAGL, monoacylglycerol lipase; NAE, N-acylethanolamines; NAPE, N-acylphosphatidylethanolamine; NAPE-PLD, NAPE-specific phospholipase D.

terms of development of pharmacotherapies for various diseases, as it suggests the possibility of being able to design a drug that will activate/inhibit a specific intracellular pathway.

The endocannabinoid system drives food intake via CB1 (62). Administration of CB1 agonists induces feeding in rodents (63) and humans (64), while blocking CB1 reduces food intake (65). Further, overactivity of the endocannabinoid system perpetuates the problems associated with obesity (62) and drugs targeting CB1 have been used therapeutically to manage obesity but withdrawn due to CNS side effects (66). New evidence indicates the endocannabinoid system can control food intake by a peripheral mechanism of action (66). Peripherally-restricted CB1 antagonists, with no direct central effects, reduce food intake and body weight in rodents (67, 68).

#### The Endocannabinoid System and TRPV1

Endocannabinoids, such as AEA, are also endogenous ligands for TRPV1 (69). Capsaicin, an agonist of TRPV1, has an anti-obesity effect in rodents (14) and reduces food intake in humans (70). Therefore, the effects of endocannabinoids on food intake will depend on the site of action (**Figure 3**). This is complicated further as effects may be mediated via cross talk between TRPV1 and CB1. It has been demonstrated that CB1 can enhance or inhibit TRPV1 channel activity depending on whether it activates the phospholipase C (PLC)-PKC or inhibits the adenylate cyclase (AC)-PKA pathways respectively (71) (**Figure 3**). This interaction appears to be dose-dependent. Moderate to high concentrations of AEA (1–10µM) have been shown to activate TRPV1 in a PKC dependent manner (34, 35). Conversely, low doses of AEA (3–30 nM) inhibit TRPV1 activity (72, 73), presumably through CB1 mediated inhibition of AC (74). Therefore, enzymatic synthesis and breakdown of endocannabinoids are potentially important determinants of TRPV1 activity in tissue, such as neuronal tissue, that coexpress TRPV1 and CB1 (71). A clearer understanding of the role endocannabinoids play in food intake regulation in health and obesity is required to determine the physiological relevance of these different interactions. In mice, the levels of AEA in the small intestinal mucosa and plasma were elevated in high fat diet-induced obese mice compared to controls (68) but still within the low dose range (3–30 nM) shown to inhibit TRPV1 (72, 73). Consistent with these observations food intake was reduced by treatment with a peripherally restricted CB1 antagonist (68). Similar observations were made in humans

with plasma anandamide levels elevated in obese compared to overweight or lean individuals, again to levels consistent with TRPV1 inhibition. Therefore, the physiological significance of TRPV1 activation observed at moderate to high concentrations of AEA remain to be determined.

### INVOLVEMENT OF TRPV1 IN ENERGY HOMEOSTASIS

Reports on TRPV1 mediated regulation of energy intake and expenditure are conflicting. Nonetheless, epidemiological data indicate that consumption of food containing capsaicin is associated with a lower prevalence of obesity (75, 76). Further, in a clinical trial capsinoid supplementation for 12 weeks decreased body weight in overweight individuals compared to the placebo control group (77). In a separate trial, capsinoid supplementation for only 4 weeks resulted in a trend toward a decrease in body weight (78). Under laboratory conditions, dietary capsaicin supplementation had no effect on body weight (14) in mice fed a standard laboratory diet. However, in high-fat diet-induced obese mice, dietary supplementation of capsaicin significantly reduced weight gain (14, 79, 80). Further, reduced weight gain was also observed in high fat diet mice after topical application of capsaicin (81). This appears to be consistent across species as a study in rabbits, fed a standard laboratory diet supplemented with cholesterol and corn oil, demonstrated that dietary capsaicin reduced weight gain (82). In contrast, it has been shown that a 6 week dietary capsaicin treatment had no effect on body weight in high fat diet mice (83). Weight gain in TRPV1-knockout (KO) mice has been reported to be reduced (84), increased (85) or unchanged (86, 87) compared to wild type mice. This variability may reflect the study design. For example, TRPV1 channels can be activated by the endogenous ligand AEA (69). The production of AEA is dependent on dietary fat and therefore even slight changes in diet will impact on research outcomes. A summary of the effect of TRPV1 on energy homeostasis in humans can be found in **Table 1**. The following sections integrate the data on energy intake and expenditure from human and animal studies in an attempt to draw some clear conclusions and directions for further study.

# Role of TRPV1 in Energy Intake

The effects of capsaicin supplementation on satiety and food intake are illustrated in **Table 1**. In human studies, dietary supplementation of a TRPV1 agonist such as capsaicin, or the less pungent sweet form capsiate, caused a short-term trend or significant decrease in energy intake along with an increase in satiety (88, 89, 91, 92, 97). These effects could at least in part be due to the effect of TRPV1 on appetite hormones and/or gastrointestinal vagal afferents. This will be discussed in detail below.

Conversely, other data from human (90, 92) and animal studies (82, 105–108) suggest that dietary supplementation of capsaicin has no effects on energy intake. This could be due to capsaicin mediated TRPV1 desensitization where food intake is initially reduced, due to capsaicin activation of the TRPV1 channel, but shortly returns to normal, due to a desensitization of the channel following the initial transient activation (14). In a Chinese adult cohort study, it has been shown that energy intake depends on the amount of chili consumed with individuals with chili consumption below 20 g per day and above 50 mg per day having reduced and increased energy intake respectively, compared to non-consumers (76). Therefore, it is possible that at low levels of consumption capsaicin activates TRPV1 leading to a reduction in food intake and at high levels it could be desensitizing TRPV1 leading to an increase in food intake. However, this is highly speculative and requires further investigation.

Dietary supplementation of capsaicin can also influence nutrient preference. It has been demonstrated that capsaicin ingestion reduced the desire for and subsequent intake of fatty foods (91, 97, 98), whilst also increasing the desire for and intake of carbohydrates (92, 97). Conversely, in other studies, capsaicin ingestion reduced the desire for and consumption of carbohydrates, and increased the desire for salt rich foods (91, 102). The sensory mechanisms responsible for the changes in food preferences remain to be determined.

#### TRPV1 and Appetite Hormones

There is evidence to indicate that TPRV1 interacts with appetite regulating hormones, most notably, ghrelin, leptin, and glucagon-like peptide-1 (GLP-1). Ghrelin is an orexigenic TABLE 1 | Effects of Capsaicin Supplementation on food intake and metabolism in Humans.


peptide mainly expressed in the stomach as an endogenous ligand for the growth hormone secretagogue receptor (GHSR) (109). It is involved in many processes including appetite regulation, secretion of gastric acid, gastrointestinal motility, and regulation of energy storage (110). It has been reported that TRPV1 activation reduced plasma ghrelin levels (93), which may account for the reduced food intake observed after capsaicin supplementation (88, 89, 91, 92, 97). However, this requires more intensive investigation. Within the stomach ghrelin maybe involved in the interaction between the endocannabinoid system and TRPV1. CB1 receptors co-expressed with ghrelin in specialized cells within the stomach wall (111). Ghrelin reduces gastric vagal afferent mechanosensitivity, in a manner dependent on nutritional status, via action at GHSR expressed on vagal afferents (112–115). Therefore, although the eating stimulatory effects of ghrelin are not thought to be mediated by vagal afferents (113), ghrelin acting on vagal afferents may impact on the amount of food consumed after the initiation of a meal. Inhibition of CB1 decreases gastric ghrelin secretion with subsequent, vagal afferent mediated, reductions in food intake (111). Therefore, part of the effect of endocannabinoids on vagal afferent activity maybe mediated indirectly via the activation of CB1 on ghrelin-producing cells. It is conceivable that the inhibitory effects of ghrelin are mediated via TRPV1 considering that, in the CNS, ghrelin effects on supraoptic magnocellular neurons are mediated via TRPV1 (116). Similar, interactions with the endocannabinoid system are observed centrally in areas associated with appetite regulation, including the hypothalamic arcuate and paraventricular nuclei (117). A comprehensive investigation of the interactions between the endocannabinoid system, ghrelin and TRPV1 is required to fully understand their role in appetite regulation.

Leptin is a satiety hormone produced and secreted in proportion to the amount of white adipose tissue (WAT). Data suggests that it is also secreted by gastric cells (118). There is evidence to suggest that TRPV1 and leptin may interact, since TRPV1 -/- mice exhibit increased basal leptin levels, even when normalized to WAT mass (85). Exogenous administration of leptin normally results in decreased food intake; however, this was not observed in TPRV1 -/- mice (85). Furthermore, there is evidence for direct interactions between leptin and TRPV1 in certain brain stem regions. For example, TRPV1 activation increased the frequency of miniature excitatory synaptic currents in leptin receptor containing neurons of gastric-related dorsal motor nucleus of the vagus (DMV) (119). These data suggest that TRPV1 may mediate the effects of leptin; however, further research is needed to substantiate these claims and to determine if leptin effects in the periphery are also mediated through TRPV1.

GLP-1 is a peptide hormone secreted by intestinal Lcells, pancreatic α-cells, and neurons in the brainstem and hypothalamus (120). Evidence suggests it is involved in appetite regulation, gastric emptying, gastrointestinal motility (121), insulin secretion, and glucagon inhibition (122). Capsaicin supplementation enhanced the increase in plasma GLP-1 levels observed after a meal (93) suggesting TRPV1 channel activation may play a role in GLP-1 secretion. This requires further investigation but has the potential to be a peripheral target for the treatment of obesity and/or diabetes.

#### TRPV1 and Gastrointestinal Vagal Afferents

Gastrointestinal vagal afferents are an important link between the gut and brain. They relay information on the arrival, amount and nutrient composition of a meal to the hindbrain where it is processed and gastrointestinal reflexes are coordinated with behavioral responses and sensations such as satiety and fullness (123–125). The role of gastrointestinal vagal afferents in the control of food intake has been extensively reviewed previously (126). Briefly, as food is ingested the vagal afferents innervating the stomach respond to mechanical stimulation as undigested food enters, fills and distends the stomach wall. There are two fundamental classes of mechanosensitive vagal afferent ending in the stomach according to location and response to mechanical stimulation (127, 128): mucosal receptors respond to fine tactile stimulation and tension receptors respond to distension and contraction of the stomach wall. Gastric mechanosensitive vagal afferents can be modulated by gut hormones and adipokines in a nutritional status dependent manner (114, 129, 130). As gastric emptying occurs, nutrients enter the small intestine and interact with nutrient receptors on the surface of specialized cells within the intestinal mucosa. This initiates an intracellular cascade that culminates in the release of gut hormones (126). These hormones can act in a paracrine fashion on vagal afferent endings innervating the small intestine and/or act as true hormones by coordinating activities within the gut or by entering the circulation and acting in the brain.

It has been demonstrated that TRPV1 is expressed in rat duodenal (131), mouse jejunal (132) and mouse gastric vagal afferents (13, 87). Activation of TRPV1, by oleoylethanolamide (OEA), caused depolarisation of nodose neurons and decreased short-term food intake (133). Further, OEA increased gastric vagal afferent tension receptor mechanosensitivity in lean but not high fat diet-induced obese mice (87). In standard laboratory diet fed TRPV1−/<sup>−</sup> mice, the response of gastric vagal afferent tension (but not mucosal) receptors to mechanical stimulation was reduced compared to TRPV1+/<sup>+</sup> mice (13, 87). This was associated with an increase in food intake in the standard laboratory diet fed TRPV1−/<sup>−</sup> mice (87). However, the increase in food intake could also be due to the involvement of TRPV1 in gut hormone release (93, 97) or its interaction with leptin in central regions, such as the DMV (119), as described in detail above. Nonetheless, this data suggests that TRPV1 is involved in gastric vagal afferent signaling.

In high fat diet-induced obese mice the response of gastric tension receptors to distension was dampened (87) an effect also observed in jejunal vagal afferents (134). Gastric tension receptor mechanosensitivity in high fat diet-fed TRPV1−/<sup>−</sup> mice was not significantly different compared to standard laboratory diet fed TRPV1−/<sup>−</sup> mice (87). This suggests that disrupted TRPV1 signaling plays a role in the dampened vagal afferent signaling observed in high fat diet-induced obesity, however, this requires further investigation. Interestingly, CB1 receptors are also expressed in vagal afferent neurons (135, 136) and therefore it is conceivable that there is an interaction between TRPV1 and CB1 in gastric vagal afferent signaling, however, this has yet to be confirmed.

#### Role of TRPV1 in Energy Expenditure

There is increasing evidence that capsaicin ingestion may have desirable metabolic outcomes such as increased metabolic rate and fat oxidation. It was reported that dietary capsaicin supplementation lowered the respiratory quotient indicating decreased carbohydrate oxidation and increased fat oxidation (90, 91, 96, 100). In contrast, there is data demonstrating that dietary capsaicin increased the respiratory quotient (103, 104). The differences in study design, which may account for the different outcomes, include method of ingestion (capsule vs. meal), active ingredient (capsinoid vs. capsaicin) and the population studied (habitual chili consumers, non-habitual, normal weight, overweight, fitness level). For example, the study by Lim et al. specifically used "runners" for their investigation (103). There is some evidence that capsaicin can elevate energy expenditure by action on the sympathetic nervous system (SNS) or adipose tissue; this is discussed below.

#### TRPV1 and the Sympathetic Nervous System

The SNS is involved in many processes and is probably best known for its involvement in the "flight or fight" response. Dietary supplementation of capsaicin increases postprandial SNS activity (70, 101, 102). Capsaicin excites TRPV1 containing afferent nerves, carrying a signal to the spinal cord (137). Efferent nerves are then excited by the central nervous system leading to elevated catecholamine (e.g., epinephrine, norepinephrine, and dopamine) release from the adrenal medulla (137–139). Catecholamines can bind β-adrenergic receptors increasing expenditure and thermogenic activity (104, 140). This suggests that TRPV1 may directly stimulate heat production. Further, these effects of TRPV1 on SNS activity are lost in obese subjects suggesting TRPV1 dysfunction in obesity (101).

#### TRPV1 and Adipose Tissue

Adipose tissue plays a key role in energy homeostasis (141). WAT generally stores excess energy as lipids, and oxidizes these stores when required, whereas, brown adipose tissue (BAT) is specialized for energy dissipation (142, 143). TRPV1 has been shown to be expressed in 3T3-L1 and HB2 adipocyte cell lines, brown adipocytes, BAT and WAT (144–147). Data indicate that TRPV1 may prevent the development of mature adipocytes from pre-adipocytes, and decrease their lipid content by increasing lipolysis (14). This may partially explain the decreased lipid accumulation during dietary supplementation of capsaicin. Further, it has been demonstrated that capsaicin induces browning in differentiating 3T3-L1 preadipocytes (145). Therefore, TRPV1 could be involved in the browning of WAT and the thermogenic activity of brown adipose tissue (BAT). The levels of TRPV1 mRNA in BAT and WAT are reduced in HFD-induced obesity and leptin receptor deficient mice (147) suggesting possible involvement in the development of obesity.

Browning is a process whereby WAT becomes thermogenic in nature, similar to BAT. The calcium influx from TRPV1 activation may mediate this process by activating the peroxisome proliferator-activated receptor gamma (PPARγ) and positive regulatory domain containing 16 (PRDM16) pathways (107). Calcium binds and activates calmodulin-dependent protein kinase II (CaMKII) leading to the subsequent activation of adenosine monophosphate activated protein kinase (AMPK) and sirtuin-1 (SIRT-1). SIRT-1 deacetylates PRDM and PPARγ causing browning events such as thermogenesis (**Figure 4**) (107).

TRPV1 activation may also promote BAT thermogenesis either through modulation of the SNS or via direct activation of BAT. However, research in this area is limited. The PPARγ and PRDM16 pathway, previously mentioned in WAT, has been shown to be activated by TRPV1, via SIRT1, in BAT (107, 108). Further, SIRT1 also activates peroxisome proliferatoractivated receptor gamma coactivator 1-α (PGC-1α). PGC-1α transcriptionally activates PPARα subsequently leading to the production of uncoupling protein-1 (UCP-1) (108). UCP-1 is a mitochondrial protein that uncouples the respiratory chain triggering a more efficient substrate oxidation and thus heat generation (148). Lastly, TRPV1 activates bone morphogenic

deacetylation of PRDM16 and PPARγ allowing their interaction and promotion

protein 8b in BAT, which also contributes to thermogenesis (**Figure 5**) (108).

Regulation of BAT by TRPV1 can also be via an indirect mechanism through modulation of the SNS. Activation of TRPV1 in the gastrointestinal tract, by capsaicin or its analogs, has been shown to enhance thermogenesis and activate UCP-1 in BAT in mice (149, 150) via a mechanism mediated via extrinsic nerves innervating the gastrointestinal tract (149). This is consistent with reports that TRPV1 ligands (capsaicin and acid) increase gastrointestinal afferent activity (151, 152) via TRPV1 (152). Further, it has been demonstrated that ingestion of capsinoids increases energy expenditure through activation of BAT in humans (153). Gastrointestinal vagal afferents have central endings in the nucleus tractus solitaries (NTS). The NTS has projections to BAT (154) where it regulates the sympathetic tone to BAT (155) and has been directly implicated in the control of thermogenesis (156, 157). Lipid activation of duodenal vagal afferents has been shown to increase BAT temperature via a cholecystokinin dependent mechanism (158). In contrast, it has been reported that vagal afferent activation decreases BAT sympathetic nerve activity and BAT thermogenesis in rats (159). Further, glucagon-like peptide-1 activation of gastrointestinal vagal afferents leads to a reduction in energy expenditure and BAT thermogenesis in mice (160). It is possible that different subtypes of gastrointestinal vagal afferent have different roles in the control of BAT thermogenesis, however, this requires further investigation along with the role of TRPV1 in this gut-brain-BAT pathway.

Capsaicin can also evoke a heat-loss response which could conceivably result in compensatory thermogenesis to maintain thermal homeostasis. Capsaicin evoked complex heat-loss responses have been shown in various mammals including the rat, mouse, guinea-pig, rabbit, dog, goat, and humans (161). In humans, cutaneous vasodilation and sweating in response to hot chili consumption is well recognized (161). In the rat it has

of WAT browning.

been demonstrated that capsaicin elicited cutaneous vasodilation resulting in a reduction in core body temperature (162). Simultaneously, capsaicin also enhanced heat production (162). In these experiments, it appeared that capsaicin independently activated pathways for heat production and heat loss and therefore the observed thermogenesis may not be a simple compensatory mechanism in response to heat loss, however, this requires further investigation.

#### INVOLVEMENT OF TRPV1 IN DIABETES

#### Type 1 Diabetes

Insulin is a hormone, secreted by β-cells of the pancreatic islets, which regulates blood glucose levels. Type 1 diabetes is an autoimmune disease involving T cell-targeted destruction of pancreatic β-cells. TRPV1 is expressed in sensory nerves innervating the pancreas. Chemical denervation of these TRPV1 containing pancreatic afferents, using high doses of capsaicin (approximately 150 mg/kg body weight), significantly reduced blood glucose levels and increased plasma insulin (163), suggesting that TRPV1 containing pancreatic afferents negatively regulate insulin secretion. Further, chemical destruction of TRPV1 expressing neurons in neonatal mice (164) was able to protect the mice from autoimmune diabetes (165). Chemical denervation of TRPV1 containing pancreatic afferents significantly reduced the levels of pre-type 1 diabetes immune markers such as CD4<sup>+</sup> and CD25<sup>+</sup> T-regulatory cells in pancreatic lymph tissue and reduced the infiltration of CD8- CD69 positive effector T-cells (165, 166); immune cells implicated in the destruction of pancreatic islets in type 1 diabetes. However, as the afferents are destroyed and the treatment is not selective for TRPV1, it is plausible that the observed effects have nothing to do with TRPV1. Pancreatic islets also include resident dendritic cells which are generally believed to express TRPV1 (167–169) although there is some controversy (170). Activation of TRPV1 channels on dentritic cells could activate cell function including antigen presentation to CD4<sup>+</sup> T cells. Further, TRPV1 channels have been shown to be expressed on rat pancreatic β-cells where they control the release of insulin leading to reduced blood glucose levels (171) and capsaicin has been shown to reduce blood glucose by increasing insulin levels in a streptozotocin-induced diabetic rat model (172). Taken together, these data suggest that TRPV1 may influence insulin secretion and type 1 diabetes acting via a number of different cell types within the pancreas.

# Type 2 Diabetes

Insulin resistance, closely linked to obesity (173), occurs when cells are less responsive to insulin. As a consequence there is reduced blood glucose uptake leading to increased blood glucose levels. Pancreatic β-cells normally respond to this by increasing output of insulin to meet the needs of the tissues. Development of type 2 diabetes stems from a failure of the β-cells to adequately compensate for insulin resistance (174). It has been demonstrated that postprandial insulin levels were lower after the consumption of a standardized meal seasoned with cayenne pepper (175). As the plasma glucose levels were not significantly different from the control group the authors suggested that glucose clearance occurred similarly with lower levels of insulin, implying increased insulin sensitivity after the consumption of the hot meal. Further, consumption of chili has been shown to decrease postprandial insulin levels in obese subjects (176). In support of these studies, TRPV1−/<sup>−</sup> mice have been shown to be more insulin resistant than wild type mice (85).

Type 2 diabetes is believed to be associated with inflammation (177, 178). It is believed that inflammation in the pancreas leads to an increase in the activity of TRPV1 which contributes to increasing levels of calcitonin gene-related peptide (CGRP) (179). CGRP is known to promote insulin resistance and obesity by decreasing insulin release from β-cells (180).

## TRPV1 AS A PHARMACOLOGICAL TARGET FOR OBESITY AND DIABETES

Behavioral interventions (e.g., diet and exercise) alone are seldom sufficient for the intervention of obesity and diabetes. Combining behavioral and pharmacological approaches is becoming increasingly more attractive. However, pharmacological interventions can have hard to access targets and/or adverse side effects. TRPV1 is present in the periphery making it an easily accessible target compared to drugs that target the central nervous system. However, TRPV1 interacts with other systems and shares pathways commonly used by other signaling molecules. Therefore, without a clear understanding of the interactions of TRPV1 with other systems, the targeting TRPV1 for the treatment of obesity and diabetes is unlikely to be successful, as evident from the numerous contradictory studies looking at the effect of capsaicin analogs on food intake and weight gain. Data suggest that manipulation of TRPV1 may be possible in such a way to reduce or eliminate any unwanted side effects. For example, three different TRPV1 ligands known to antagonize TRPV1 had different effects on thermoregulation (e.g., hyperthermia, hypothermia, or no effect) (181). In fact, TRPV1 can be manipulated in such a way, by action at different domains, to eliminate some functions of the TRPV1 channels without affecting others. For example, some antagonists block activation by capsaicin and high temperatures but not activation by low pH (182), and other antagonists block activation by capsaicin but not the activation by high temperature (183). However, this raises further questions on whether, for example, the observed effects are cell type specific. Again, this highlights the lack of fundamental knowledge on the role of TRPV1 in energy homeostasis and therefore the current challenges of targeting TRPV1 for the treatment of obesity.

# CONCLUSION

TRPV1 appears to be involved in energy homeostasis at a number of levels. In the periphery, TRPV1 activation or inhibition can have an impact of appetite and food intake through the control of appetite hormone levels or via the modulation of gastrointestinal vagal afferents, important for determining meal size and meal duration. In addition, TRPV1 plays a role in energy expenditure via heat production, either via direct thermogenesis or as a compensatory mechanism in response to TRPV1 induced heat-loss. Dietary supplementation with TRPV1 analogs, such as capsaicin, has yielded conflicting results with some studies demonstrating a decrease in food intake and increase in energy expenditure and others indicating the converse. This is probably reflective of the involvement of TRPV1 in a multitude of processes regulating food intake and energy expenditure. The story is complicated further by the interaction TRPV1 has with other systems involved in energy homeostasis, such as the endocannabinoid system. In addition, TRPV1 appears to be dysregulated in obesity, possibly due to alterations in the interaction with other systems. Therefore, although it is clear that TRPV1 plays a role in energy homeostasis without improved knowledge of the fundamental physiological

### REFERENCES


mechanisms involved and the interactions with other systems it is impossible to target this system for the treatment of obesity, the maintenance of weight loss and the metabolic diseases associated with obesity.

#### AUTHOR CONTRIBUTIONS

SC wrote the review under the supervision of GW and AP. HL edited the article and along with GW and AP provided assistance with the structure of the review.

### FUNDING

AP and GW were awarded a Diabetes Australia Research Trust grant.


amidase in macrophages revealed by immunochemical studies. Biochim Biophys Acta (2007) 1771:623–32. doi: 10.1016/j.bbalip.2007.03.005


vanilloid VR1 receptor-mediated response. Cell Mol Life Sci. (2003) 60:607– 16. doi: 10.1007/s000180300052


endocytosis of human monocyte-derived dendritic cells. FEBS Lett. (2013) 587:1440–5. doi: 10.1016/j.febslet.2013.03.027


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Christie, Wittert, Li and Page. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Deciphering the Roles of PPARγ in Adipocytes via Dynamic Change of Transcription Complex

Xinran Ma, Dongmei Wang, Wenjun Zhao and Lingyan Xu\*

*Shanghai Key Laboratory of Regulatory Biology, Institute of Biomedical Sciences and School of Life Sciences, East China Normal University, Shanghai, China*

Peroxisome proliferator-activated receptor γ (PPARγ), a ligand-dependent transcription factor highly expressed in adipocytes, is a master regulator of adipogenesis and lipid storage, a central player in thermogenesis and an active modulator of lipid metabolism and insulin sensitivity. As a nuclear receptor governing numerous target genes, its specific signaling transduction relies on elegant transcriptional and post-translational regulations. Notably, in response to different metabolic stimuli, PPARγ recruits various cofactors and forms distinct transcriptional complexes that change dynamically in components and epigenetic modification to ensure specific signal transduction. Clinically, PPARγ activation via its full agonists, thiazolidinediones, has been shown to improve insulin sensitivity and induce browning of white fat, while undesirably induce weight gain, visceral obesity and other adverse effects. Thus, deciphering the combinatorial interactions between PPARγ and its transcriptional partners and their preferential regulatory network in the processes of development, function and senescence of adipocytes would provide us the molecular basis for developing novel partial agonists that promote benefits of PPARγ signaling without detrimental side effects. In this review, we discuss the dynamic components and precise regulatory mechanisms of the PPARγ-cofactors complexes in adipocytes, as well as perspectives in treating metabolic diseases via specific PPARγ signaling.

#### Edited by:

*Andrew J. McAinch, Victoria University, Australia*

#### Reviewed by:

*Pawel K. Olszewski, University of Waikato, New Zealand Claire Joanne Stocker, University of Buckingham, United Kingdom*

> \*Correspondence: *Lingyan Xu lyxu@bio.ecnu.edu.cn*

#### Specialty section:

*This article was submitted to Obesity, a section of the journal Frontiers in Endocrinology*

Received: *04 April 2018* Accepted: *31 July 2018* Published: *21 August 2018*

#### Citation:

*Ma X, Wang D, Zhao W and Xu L (2018) Deciphering the Roles of PPAR*γ *in Adipocytes via Dynamic Change of Transcription Complex. Front. Endocrinol. 9:473. doi: 10.3389/fendo.2018.00473*

Keywords: PPAR gamma, adipocytes, transcription complex, obesity, metabolic diseases

# INTRODUCTION

Adipose tissue is critical in maintaining energy homeostasis and insulin sensitivity in vertebrate organisms by directing energy fluxes toward triglyceride synthesis for storage or fuel mobilization for utilization in response to different metabolic requirements. Genetic, dietary, pathological, or aging-related disruption of adipose tissue function is one of the underlying causes of the current pandemics of metabolic diseases including obesity and type 2 diabetes. Adipose tissue exerts its normal physiological function via a complicated and delicate network of transcription factors to orchestrate and fine-tune various molecular events in adipocytes development, functionality, and senescence, among which peroxisome proliferator activated receptor γ (PPARγ) holds special importance for its unique and indispensable role in adipogenesis, lipid metabolism and insulin sensitivity (1). Moreover, evidences from murine and Caenorhabditis elegans models indicate that PPARγ is also a key driver in promoting longevity (2, 3). As a ligand dependent transcription factor, PPARγ exerts pleiotropic functions through its vast sets of downstream transcriptional targets. Mechanistic studies reveal that, under different stimuli or physiopathological status, PPARγ forms distinct transcription complex with different set of cofactors and transcription partners to regulate specific transcriptional circuit. Though the upstream regulators of PPARγ transcription is well studied and documented, from an empirical standpoint, it would be intriguing to decipher the detailed molecular events during which PPARγ integrates and transduces diverse signals toward specific clusters of downstream targets. In this review, we aim to discuss our current understanding of the combinatorial interactions between PPARγ and its interaction partners and how their dynamic changes impact PPARγ downstream target gene transactivation under different metabolic scenario and during the aging process in adipose tissues, thus potentially provide us with more elaborate and precise strategies to screen novel therapeutic targets that target PPARγ and its cofactors in treating adipocyte dysfunctions and metabolic diseases.

### DISTINCT ROLES OF ADIPOCYTES IN ENERGY HOMEOSTASIS

Three distinct types of fat (white, brown, and beige adipose tissue) exist based on different anatomic locations and functions (4). White adipocytes feature a single large lipid droplet and store excess energy in forms of triglycerides. In sharp contrast to white adipocytes, brown adipocytes, characteristically contain multiple lipid droplets and packed with mitochondrial, are vital in defending body temperature through activation of uncoupling protein 1 (UCP1) and uncoupling the mitochondrial respiratory chain to produce heat, thus contribute greatly for energy expenditure and adaptive thermogenesis. Interspersed within white fat, the newly discovered beige adipocytes share functional similarities with both white and brown adipocytes. Briefly speaking, beige adipocytes are indistinguishable from white adipocytes morphologically under basal condition. While upon cold or adrenogenic signaling stimulation, beige adipocytes are activated (referred to as the "browning of white fat" effect) and take on a brown adipocyte-like look with multiple small lipid droplets and rich mitochondria content. Similar to brown fat, activated beige adipocytes have high UCP1 expression and enhanced thermogenic capacity and energy expenditure (5– 7). Importantly, upon activation, brown, and beige adipocytes uptake high levels of lipid and glucose for heat production, thus serve as a metabolic sink to clear excess nutrients in blood and contribute to insulin sensitivity and whole body lipid/glucose metabolism (8), in addition to the lowing body weight effect. Together, white, brown and beige adipocytes maintain the wholebody energy homeostasis.

Though rise from different precursors, white, brown and beige adipocytes all go through a finely tuned differentiation process (referred to as "adipogenesis") to mature and become fully functional (9). Then, mature adipocytes can respond to different metabolic stimuli to store or utilize energy, as well as crosstalk with other cell types via various adipokines, cytokines or lipid/glucose fluxes (10). Eventually, adipocytes undergo senescence and gradually lose their differentiation or thermogenic capability (11, 12). During various stages of the adipocyte lifespan, PPARγ is a well-established central player in orchestrating the numerous molecular events that ensure the normal physiological function of white, brown and beige adipocytes. PPARγ, a member of the nuclear receptor PPARs family, shares the common PPAR domain structures that feature a highly conserved DNA-binding domain (DBD) and a transactivation domain (AF1) at the N-terminal region, and a ligand-binding domain (LBD) and a ligand-dependent transactivation domain (AF2) at the C-terminal region. PPARγ has two isoforms, PPARγ1 and PPARγ2, the latter contains an additional NH2-terminal region composed of 30 amino acids due to different promoter usage and alternative splicing (13). PPARγ1 expresses ubiquitously while PPARγ2 expresses the highest in adipose tissues and is highly inducible in other tissues under high fat diet (HFD) (14). By forming distinct transcription complexes with different interaction partners or through epigenetic modifications, PPARγ exerts critical and pleiotropic functions in adipose tissues that include: (I) adipocyte differentiation and lipid storage, (II) acquisition of brown/beige adipocyte identity (III) maintenance of brown/beige adipocyte thermogenic capacity, (IV) brown/beige adipocyte functional decline in aging, and (V) regulation of diabetic gene program (15). The dynamics beween PPARγ and its transcription partners in these physiological processes are discussed below.

# PPARγ/Corepressors/Coactivators Dynamics Controls Adipocyte Differentiation

PPARγ is considered the master regulator for adipogenesis since ectopic expression of PPARγ alone in fibroblasts could successfully drive the adipogenesis program and no other factors could induce adipogenesis without the presence of PPARγ (16–18). In vivo studies reveal PPARγ is vital in controlling adipogenesis and lipid/glucose metabolism. For instance, aP2 and adiponectin-driven loss of PPARγ in adipose tissues consistently lead to impaired adipocyte differentiation and reduced fat weights or lipodystrophy, though different animal models show improved or worsen insulin sensitivity depend on the extent of PPARγ deficiency (19–21). In adult mice, PPARγ ablation in adipose tissues with tamoxifen-dependent Cre-ER(T2) recombination system leads to adipocyte death and subsequent renewal, suggesting that PPARγ is required for the survival of mature adipocyte (22, 23). In clinic, patients with heterozygous PPARγ mutation show partial lipodystrophy and insulin resistance (24). PPARγ is also a critical thrifty gene that governs hoarding gene programs for energy storage mostly in white adipocytes (25).

Adipocytes differentiation is a closely regulated process that rises in response to physiological needs, i.e., adipocytes turnover and excess nutrient influx. Though it may function differently in aging (26), during development and in adulthood, PPARγ controls the transcriptional activation of the adipogenesis process through ligand binding and selective interaction with transcription corepressors or coactivators (17). Corepressors bind to transcription factor/DNA complexes to recruit histone deacetylases (HDAC) and make the target DNA region less accessible for transcription. On the contrary, coactivators possess intrinsic histone acetyltransferase (HAT) activity or recruit other proteins with HAT activity to acetylate histones and loosen the chromatin in a limited region, allowing for increased basal transcription machinery accessibility (27). PPARγ constitutively forms a heterodimer with retinoid X receptor (RXR) and binds to a specific DNA sequence termed PPAR response elements (PPRE). In basal condition, unliganded PPARγ forms a complex with transcription corepressors such as silencing mediator of retinoid and thyroid hormone receptors (SMRT), the nuclear receptor corepressor (NCoR), and HDACs, which blocks its transactivation activity. Upon ligand binding, PPARγ undergoes a conformational change, dissociates corepressors and recruits coactivators i.e., CREB-binding protein (CBP), histone acetyltransferase p300 (p300) and PPAR-binding protein (PBP), thus initiates the adipogenic gene programs including Fatty Acid Binding Protein 4 (Fabp4), Cluster of differentiation 36 (Cd36), Adiponectin, Fatty Acid Synthase (Fasn), etc. (28, 29). It has been shown that coactivators deficiency in fibroblasts hinders adipocyte differentiation while adipocytes lack corepressors accumulate more lipids, which demonstrates the dynamic competition between activating vs. suppressive PPARγ-cofactors transcriptional complexes in adipogenesis (30, 31).

# PPARγ/PRDM16/EBF2/EHMT1 Complex Determines Brown/Beige Adipocyte Identity

It is demonstrated that brown adipocytes and skeletal muscle cells rise from the common Myf5+ Dermomyotomal precursors while beige adipocytes and white adipocytes share the common Pdgfra+ mesodermal stem cells (32). Besides its indispensable role in promoting differentiation of white, brown and beige adipocytes, another critical function of PPARγ resides in the determination of brown and beige adipocyte identity. To achieve this, PPARγ recruits PRDM16, EBF2, and EHMT1 to coordinate the transcriptional circuits toward the brown/beige lineage (33).

For brown/beige adipocytes to acquire their identity and become thermogenic poised adipocytes, PPARγ recruits PR (PRD1-BF1-RIZ1 homologous)-domain-containing 16 (PRDM16) to form a core transcription complex that determines the transition of brown adipocyte from skeletal muscle cell or beige adipocyte from white adipocyte. Mass spectrometry analysis indicates PPARγ as the only DNA binding transcription factor interacts with PRDM16 in a near stoichiometric manner (34). This is further proved when full length GST-PPARγ2 fusion protein binds to in vitro translated PRDM16. PRDM16 is a powerful inducer and maintainer of the thermogenic phenotype in both brown adipocyte development and the browning process (35, 36). Mechanistic studies show that PRDM16 binds to PPREs on the promoter and/or enhancer of brown fat-selective genes and highly stimulates the activity of a PPRE luciferase reporter. Besides, PRDM16 fails to promote adipogenesis in PPARγ-deficient fibroblasts (37), indicating it at least partially exerts function via PPARγ. In myocytes, in the presence of PPARγ agonists, PRDM16 induces multiple PPARγ target genes in adipogenesis like aP2 and adiponectin, as well as brown fat-selective gene program, including Ucp1 and Cidea. Later, it is revealed that PPARγ agonists could induce a white-to-brown fat conversion through stabilization of PRDM16 protein, suggesting the existence of a positive regulatory loop to strengthen the interaction of the PPARγ/PRDM16 complex to maintain the thermogenic capacity in beige adipocytes (38). In vivo studies echo the results from in vitro studies. Transgenic and tissue specific animal models show that increased expression of PRDM16 in rodent inguinal white adipose tissue (iWAT) and in mature adipocytes promotes beige adipocyte commitment and prevents HFD-provoked weight gain and systemic insulin resistance (39). Conversely, deletion of Prdm16 in adipocytes causes a profound loss of beige adipocytes in mice, leading to aggravated metabolic deregulation upon exposure to HFD (40).

Two other factors, Histone-lysine N-methyltransferase (EHMT1) and Early B-cell factor (EBF2), present in the PPARγ/PRDM16 complex and facilitate its function in brown/beige fat identity determination (41). EHMT1 is the only methyltransferase purified specifically with PRDM16 by mass spectrum. It induces the inhibitory H3K9me2 and H3K9me3 at PRDM16-resident myogenic gene promoter regions, thus enables the PPARγ/PRDM16/EHMT1 to promote the common Myf5<sup>+</sup> precursors toward brown adipocytes linage instead of muscle linage. EHMT1 also controls thermogenesis by stabilizing PRDM16 and extending its half-life. Similar to PRDM16, EHMT1 fat-specific deficient mice show obesity and insulin resistance (41). Moreover, it has been reported that activation of brown adipocytes in the adult human perirenal depot is highly correlated with PRDM16–EHMT1 complex expression (42).

On the other hand, it is demonstrated that PPARγ recruits EBF2 to its brown-selective binding site and coactivates the expression of brown fat-selective genes such as Ucp1, Pparα, and Prdm16 (43). Genome-wide ChIP-sequencing and motif analyses of the consensus binding site for EBF2 reveal that it is highly enriched in brown adipocytes compared to white and beige fat. Consistent with its function in coactivating PPARγ downstream brown fat-selective gene program, EBF2 deficient mice fail to thrive due to severe defects in brown fat development and themogenesis (43). Intriguingly, transcription regulator ZFP423 maintains white fat identify via inhibition of EBF2 and PRDM16 activity (44), further suggesting the critical role of PPARγ/PRDM16/EBF2 in brown adipocyte development.

Overall, the PPARγ/PRDM16/EBF2/EHMT1 complex determines the identity of brown/beige adipocytes. It has been shown that the loss of brown and beige fat may predispose the obese and aging population to metabolic dysfunction, thus future extensive study of this complex is fundamental for developing compounds that could drive precursor cells to undergo brown/beige adipocytes fate and potentially achieve enhanced thermogenesis and energy expenditure.

# PPARγ/PRDM16/PGC1α Modulates Brown/Beige Adipocyte Functionality

After the thermogenic endowment of brown/beige fat, the PPARγ/PRDM16 complex recruits another distinct set of cofactors to promote brown/beige fat function in adaptive thermogenesis and energy homeostasis, among which PPARγcoactivator PGC1α plays a central role. PGC1α is first cloned from murine brown fat cell cDNA library through yeast two-hybrid system using PPARγ partial protein as bait (45). As a coactivator, PGC1α is capable of binding to and coactivating many nuclear receptors and transcription factors to exhibit different functions, such as PPARα, β/δ for fatty acid β-oxidation, thyroid receptor and IRF4 for thermogenesis, HSF1 for heat shock responses, NRF1 and NRF2 for mitochondrial biogenesis (46–48). In brown fat, PGC1α promotes mitochondrial biogenesis and thermogenesis at least partially by coactivating PPARγ and enhancing PPARγ's transcription activity on the thermogenic gene program, including Cidea, Elovl3, and Ucp1 (49). Moreover, it has been found that PRDM16 directly interacts with PGC1α, which increases PGC1α expression and promotes PGC1α transactivation activity (50). These data suggest the existence of a PPARγ/PRDM16//PGC1α thermogenic transcription complex in the regulation of brown/beige functionality. This core complex in turn recruits other cofactors or undergoes various modifications to fine tune thermogenesis and energy homeostasis.

For example, MED1 is found to be enriched along with PPARγ/PRDM16 complex on the promoters of brown fatselective genes (51). As a component of mediator complex, MED1 plays a crucial role in regulating transcription in part through bridging the transcription factor-bound enhancer regions with the general transcriptional machinery and RNA Pol II at gene promoters. In the brown/beige adipocytes context, MED1 physically binds to PRDM16 and is then recruited to super enhancers at brown fat-selective genes. Consistently, MED1 binding at PRDM16 target sites reduces in PRDM16 deficiency BAT, causing a fundamental change in chromatin architecture at key brown fat-selective genes (52).

The SRCs family, also known as p160 proteins, includes SRC-1/NCoA-1, SRC-2/GRIP1/TIF2/NCoA2, and SRC-3/p/CIP/AIB1. They are transcriptional coactivators that interact with nuclear receptors and enhance their transactivation in a ligand-dependent manner (27, 53). In the metabolic context, they work as cofactors for PPARγ/PGC1α to influence brown/beige function and energy homeostasis. For Instance, SRC-1−/<sup>−</sup> mice showed partially impaired PPARγ function in the brown fat and are prone to obesity due to reduced energy expenditure and fatty acid oxidation upon HFD feeding (54, 55). Mechanistic study reveals that SRC-1 stabilizes PPARγ/PGC1α interaction and thus increases PGC1α-mediated adaptive thermogenesis (54). On the contrary, SRC-2−/<sup>−</sup> mice are protected against obesity. In white fat, PPARγ preferentially recruits SRC-2 as its coactivator to transactivate downstream fat uptake and storage pathways while in the brown fat SRC-2 competes with SRC-1 for PPARγ binding and disrupts SRC-1-induced interaction between PPARγ and PGC1α (56, 57). Meanwhile, in SRC-3−/<sup>−</sup>

mice, HFD feeding causes a depot-selective decrease in PPARγ2 level, resulting in body weight reduction, fat mass decrease and fat redistribution in knockout mice (58, 59). Besides, SRC-3 induces PGC1α acetylation and consequently inhibits its activity in brown fat (60). SRC-3 also cooperates with SRC-2 to attenuate PPARγ phosphorylation at S114, which in turn increase PPARγ transcriptional activity and adipogenesis (61). Collectively, these data summarize the crucial roles of SRC family members that cooperate or antagonize with each other to regulate the interaction of PPARγ/PGC1α complex and its transactivation activity on the thermogenic network.

The stability of PPARγ/PRDM16 complex also impacts the induction of brown fat gene programs and thus the brown/beige fat function. It is well demonstrated that SirT1-dependent deacetylation of PPARγ Lys268 and Lys293 facilitate the close interaction between PPARγ and PRDM16 and is essential for the selective induction of brown fat-selective genes and repression of visceral white fat-selective genes (62). Silencing SIRT1 in 3T3-L1 preadipocytes leads to their hyperplasia and increased expression of white fat and inflammatory markers with a parallel decrease in brown fat markers (63). Specific genetic ablation of SIRT1 in WAT leads to obesity, increased inflammatory infiltration, and insulin resistance similar to that observed in HFD induced obesity, further suggesting SIRT1 deficiency leads to whitening effects of mature adipocytes (64). Conversely, SirT1 pharmacologic activators, i.e., resveratrol is shown to be effective in inducing the browning of white fat, weight loss, and improved insulin sensitivity in mice, though possibly through multiple mechanisms in addition to SirT1 activation (65). On the other hand, a few negative regulators of browning work through disrupting PPARγ/PRDM16 interactions. For example, TLE3 is a white-selective cofactor that competes with PRDM16 for PPARγ interaction to specify lipid storage or thermogenic gene programs (66). In addition to transcriptional factors and cofactors, PexRAP, a peroxisomal lipid synthetic enzyme interacts with PPARγ and PRDM16 and disrupts PRDM16-mediated gene expression (67). Last but not least, it has to be noted that nuclear factor I-A (NFIA) is recently reported to co-localize with PPARγ at the brown fat specific enhancers and facilitate PPARγ chromatin accessibility for induction of brown fat gene programs through mechanisms independent of the PPARγ/PRDM16 complex (68).

Collectively, these findings extend our understanding on how the PPARγ/PRDM16/PGC1α thermogenic transcriptional complex ensures delicate regulation on brown/beige fat functions by fine tuning its various components and modification status. The fact that the thermogenic transcriptional complex is required for brown/beige adipocytes to be fully functional supports the notion of targeting it for future potential drug discoveries to treat obesity and metabolic diseases, while inhibition of the negative regulators of this complex provides alternative strategies of inducing browning for energy expenditure.

# PPARγ/Cofactors Complex in Aging

Life expectancy increases with the advancement in modern medicine and the improvement in life quality, bringing forward an aging society. It has been well accepted that energy homeostasis is tightly controlled by the central nervous system, most prominently in the hypothalamus and brainstem areas that receive and integrate peripheral signals such as insulin and leptin, and in turn regulate food intake and energy metabolism in peripheral organs including adipose tissues, livers, and muscle (69). Thus it is not surprising that metabolic dysfunction occurs in line with aging due to the gradual deterioration in brain function. In both rodents and humans, advanced aging features significant weight loss due to anhedonia, anorexia, lipodystrophy, and muscle atrophy caused by the deregulations in brain (70). Intriguingly, during aging, peripheral organs such as adipose tissues undergo programmed functional loss and senescence as well. It is well documented that during aging, adipose tissues undergo a gradual exhaustion in progenitor pool, impaired proliferation and differentiation capability in preadipocytes and programmed functional loss in mature adipocytes (71, 72), resulting in disrupted lipid and glucose metabolism and impaired energy homeostasis. The aging trajectory begins around middle age after the peak fertility is passed. Compared to the phase of advanced aging, this phase often features an increase in body weight and fat mass, impaired metabolic fitness, and deregulated lipid profiles, during which the senescence and dysfunction of adipose tissues play a profound role. Indeed, midlife weight gain is one of the major risk factors for metabolic syndrome such as Type 2 diabetes, cardiovascular diseases, hypertension, hyperlipidemia, hepatic steatosis, and certain types of cancer, which poses a serious burden on public health management nowadays (73, 74). Thus, how aging affects adipocyte function and its intrinsic mechanism attract great attentions in the recent years (75).

As the master regulator in adipocytes, with the triumph of specific cre/loxP recombination system, the in vivo roles of PPARγ and its cofactors in fat biology during development or in young adult animals have been extensively studied. However, the existing aP2-Cre/loxP or adiponectin-Cre/loxP system deletes gene in fat tissues in early stages of life, the results of which are hard to extrapolate to the aging process. It has been shown that the PPARγ agonist rosiglitazone (TZD) treatment rescues shortened lifespan in C. elegans and induces a transcriptional signature that overlaps with longevity-associated genes (3). Besides, hypomorphic PPARγ mice and PPARγ2 heterozygous mice both show reduced lifespan (2), indicating an important role of PPARγ in aging. Considering the metabolic derangement in adipose tissues during aging and the critical role of PPARγ in adipocytes, it merits further investigation on the dynamics of PPARγ and its cofactors specifically in the aging process and how they impact the functional decline in adipose tissues.

Using the adenoviral delivery strategy, Ma et al. reported the surprising finding that loss of PPARγ specifically in subcutaneous fat in aging mice (12-month-old) is sufficient to increase body weight and insulin resistance by accelerating the decay of browning effects of white fat and disrupting energy homeostasis. Detailed analysis reveals that in aging adipose tissues, PPARγ preferentially binds to the promoters of browning vs. whitening gene program, thus maintains browning capability of subcutaneous fat of aging mice rather than lipid storage,

as shown by gene program expression array and in vivo ChIP analysis (**Figure 1**). Meanwhile, although the total PPARγ levels do not change, PPARγ Ser273 phosphorylation increases during aging, which may potentially affect PPARγ affinity with its cofactors and rewire its metabolic circuitry toward brown gene program (26). Additionally, in rodent and human adipose tissues, SRC-1 expression levels decrease during aging, which may disrupt PPARγ/SRC-1 complex and its function in insulin sensitivity and adipogenic activity (76). These works offer novel insights on the unique combinatorial organization of the PPARγ/cofactors complex and the resulting target gene sets under the aging scenario, which might be largely different than previously reported PPARγ function in young animal models. From a therapeutic point of view, these results emphasize the need of personalized treatment when targeting PPARγ to treat metabolic disorders in patients of different ages. It would also be of importance to further investigate the potential benefits of targeting PPARγ in preventing age-associated metabolic diseases and promoting longevity.

### PPARγ Phosphorylation in Diabetic/Adipogenic/Osteogenic Gene Program

In addition to recruitment of different cofactors, the posttranslational regulations of PPARγ add another layer of regulation for the PPARγ/cofactors complex to selectively transactivate different gene circuits. In clinic, the full PPARγ agonists TZDs are very effective in glycemic control but have multiple adverse effects such as obesity, water retention, and increased risk of cardiovascular diseases and bone fractions. The opportunity for improvement emerges when Spiegelman's lab identifies the distinguish role of PPARγ Ser273 phosphorylation in lipid metabolism and insulin sensitivity (77–79). Briefly, in HFD induced obesity, protein kinase cyclin-dependent kinase 5 (Cdk5) is activated and in turn phosphorylates PPARγ at Ser273. This form of post-translational modified PPARγ selectively transactivates target genes involved in insulin resistance, without significant changes in adipogenic or osteogenic gene program (77). Consistently, Ma et al. report increased PPARγ Ser273 phosphorylation in iWAT of 12-month-old mice, which may contribute to impaired insulin sensitivity during aging (26).

PPARγ Ser273 phosphorylation is regulated by its interacting cofactors, which specifies a regulatory role of PPARγ/cofactors complex on the diabetic gene program. For example, PPARγ corepressor NcoR functions as an adaptor to enhance Cdk5 activity on interacting and phosphorylating PPARγ Ser273. Consistently, when fed a HFD, compared to wild type mice, fat-specific NCoR-deficient mice are prone to obesity yet have enhanced insulin sensitivity (80). Upon PPARγ Ser273 phosphorylation, phosphorylated PPARγ recruits thyroid hormone receptor-associated protein 3 (thrap3) to facilitate the expression of the diabetic gene programs (78). These data establish a delicate control of diabetic genes by PPARγ/cofactor complex through PPARγ Ser273 phosphorylation, as well as the potential of targeting PPARγ Ser273 phosphorylation to improve insulin sensitivity while eliminating the disadvantage of side effects. Of therapeutic significance, ERK have been shown to phosphorylate PPARγ Ser273 and induce insulin resistance in the absence of Cdk5, rendering ERK as promising anti-diabetic targets (79). Moreover, novel PPARγ partial ligands such as SR1664, F12016, and antibiotic ionomycin are designed or screened to inhibit PPARγ Ser273 phosphorylation, which show effective glycemic control with minimal side effects (81–83).

Aside from its numerous functions in adipocytes, PPARγ also plays vital role in controlling the balance between bone marrow adipogenesis and bone formation. For instance, PPARγ phosphorylation at Ser112 by mitogen-activated protein kinases (MAPKs) inhibits adipogenesis while promoting osteogenesis (84). A non-phosphorylatable Ser112 PPARγ mutation drives adipogenic differentiation and inhibited osteoblastogenesis in vitro. Consistently, mice carrying this homozygous PPARγ mutation (PPARγ-S112A mice) show increased bone marrow adipose tissue with reduced bone volume. In another study, PPARγ phosphorylation at different sites exhibits osteoblastic (pS112) or osteoclastic (pS273) activity thus regulates bone formation and absorption in a delicate balance. These collective data, along with the role of PPARγ Ser273 phosphorylation in promoting diabetic gene program, show the advantage of using PPARγ agonists against selective phosphorylation sites. For instance, SR10171, a PPARγ S273 phosphorylation blocker, increases both insulin sensitivity and osteoclastogenesis (85) while PPARγ partial agonist Telmisartan, which decreases PPARγ S273 without altering pS112 phosphorylation, promotes browning effects and improves insulin sensitivity without altering bone mass or bone biomechanical properties in mice (86). Thus, it would be desirable to develop pharmacologic agents targeting specific PPARγ phosphorylation sites, especially ones that dephosphorylate S273 while maintaining pS112 phosphorylation, to maximize their beneficial activities.

# PERSPECTIVES

In clinic, PPARγ Pro12Ala polymorphism is identified as one of the common risk polymorphisms for Type 2 diabetes in multiple GWAS studies (87). Indeed, the thiazolidinedione class of drugs, the PPARγ full agonists, has a long history of use in treating Type 2 diabetes mellitus for its insulin sensitizing effects, though multiple side effects exist including obesity, water retention, increased risk of cardiovascular diseases and bone fractions. Besides, recent studies indicate PPARγ play an important role in antiretroviral treatment related adipocyte dysfunction and lipodystrophy in HIV-infected patients. These studies potentiate the importance of developing PPARγ partial agonists to selectively activate a specific set of its numerous target genes in order to avoiding possible side effects. To date, various PPARγ partial agonists are in different stages of drug development yet none has progressed into clinic due to insufficient efficacy and potential adverse effects, which merits a close examination to gain more mechanistic insights of PPARγ.

Of note, when developing PPARγ agonists, it is necessary to put the interaction between PPARγ and its cofactors into the equation since after ligand binding, PPARγ recruits transcription partners and cofactors to form a transcription complex to achieve selective target gene transactivation. In this review, we highlight a comprehensive review of the dynamics in transcription partners, coregulators, and post-translational modifications of PPARγ in adipose tissues during various stages of the adipocyte life cycle and the resulting regulatory function on fat development, thermogenesis and adipocyte senescence (**Figure 2**). PPARγ dissociates corepressors (e.g., NCoR, HDAC, SMRT) and binds with coactivators including CPB, p300, and PBP to initiate adipocyte differentiation in white, brown and beige adipocytes. In brown and beige adipocytes, PPARγ complexes with PRDM16/EBF2/EHMT1 for brown/beige cell linage determination while their unique functions in adaptive thermogenesis and energy homeostasis are maintained and modulated by the PPARγ/PRDM16//PGC1α complex as well as their interaction with various cofactors i.e., MED1, SRC1/2/3, and TLE3. Of therapeutic relevance, under aging scenario, PPARγ may complex with different cofactors due to different post-translational modifications or cofactor expression levels and thus exert distinct function as compared to its function in young animals. Last but not least, modulation of phosphorylation status of PPARγ selectively activates the diabetic or adipogenic circuits in adipocytes, which serves as an attractive strategy for therapeutic development. In recent years, the triumph of high-throughput sequencing technology has offered us the whole picture of the binding profiles of PPARγ on the genome-wide scale in white, brown and beige adipocytes under different metabolic conditions, which enable us to gain a deeper understanding of the common and distinct target gene sets regulated by PPARγ in different fat depots, under different diet regimes or ages, or in response to different environmental or drug stimulus. However, challenge remains as how PPARγ achieves differential regulatory patterns in different scenarios. To achieve this, multiple ChIP sequence analyses using various PPARγ cofactors are needed to link the dynamics of PPARγ and

fat senescence (D) and Diabetic gene program (E).

its transcription partners with specific transactivation circuits and functionality.

During gene transcription, upon transcription factor binding, cofactors add an extra layer of transcriptional regulation on target genes and endow enhanced regulatory complexity. To combat the high prevalence of metabolic diseases worldwide, the needs for novel PPARγ agonists that feature maximum beneficial effects i.e., efficient glycemic control or enhanced "browning" effects while avoid undesirable side effects are heightened. With the crystal structure of PPARγ fully resolved and the accumulating data on its conformational changes upon full or partial agonists binding, a large body of works focuses on finding potential novel PPARγ agonists via computational predictions and virtual screening strategies. However, despite binding to the PPARγ ligand binding domain with high affinity as predicted, these compounds often face the problem of low transactivation activity in vitro, which hinders their progression into clinic. In this regard, the importance of cofactors during PPARγ transactivation events is underappreciated. It would be more informative to resolve the structure changes of PPARγ in the presence of various cofactors i.e., PRDM16 and PGC1α and take the dynamics between PPARγ and its cofactors into consideration when designing new PPARγ agonists in the future. For example, strategies could be developed to screen for agonists that could strengthen the interaction between PPARγ and specific coactivators or block PPARγ phosphorylation at specific sites.

In previous reports, specific PPARγ deletion in adipose tissues in young adult animals results in lipodystrophy or massive adipocyte death, indicting an indispensable role of PPARγ in adipose tissue development and survival. Unexpectedly, PPARγ deletion in a temporal-specific manner in inguinal fat of aging animals causes significant increases in body weight and fat mass, which is in sharp contrast to young control animals (26). Further analyses show PPARγ has unique spectrums of interaction partners and target genes in aging, which is in consistent with previous reports of PPARγ as a potential driver for increased longevity (2, 3, 26). These findings shed first light on the potential of using PPARγ agonists in aged population to promote healthspan and lifespan. As a transcription factor with pleiotropic functions, targeting PPARγ could have multiple benefits against various aspects of aging. Firstly, aging features metabolic derangement, toward which PPARγ agonists show high efficacy in glycemic control, thus promoting metabolic fitness during aging. Secondly, caloric restriction (CR) is a wellestablished nutritional intervention that increases longevity and promotes metabolic health in various species from C. elegance to primates (88). It is reported that PPARγ expression levels are dramatically induced by CR treatment in metabolic organs, especially in inguinal fat for browning of white fat, suggesting that PPARγ activation might mimic CR effects at least in fat tissues (89). Thirdly, recent reports show compounds classically

#### REFERENCES


functioning through glycemic control (i.e., metformin) or immunosuppression (i.e., rapamycin) are effective in promoting longevity (90, 91). Considering the dual function of PPARγ in glucose metabolism and inflammation control, it would be interesting to test its agonists in longevity.

Overall, although PPARγ has been recognized as a promising target in preventing and treating metabolic diseases, challenges remain in finding the suitable agonists with high therapeutic efficacy and low side effects. Understanding how PPARγ dynamically complexes with its cofactors to transactivating specific downstream target gene sets in different metabolic conditions might provide a novel angle in search of new therapeutic compounds against various aspects of obese- and aging-associated metabolic diseases.

#### AUTHOR CONTRIBUTIONS

XM and LX conceived, drafted and revised the review. DW and WZ drafted the review.

#### FUNDING

This project was supported by funding from National Natural Science Foundation of China (31770840 to XM), Shanghai Pujiang Program (17PJ1402600 to XM and 17PJ1402700 to LX), the China Postdoctoral Science Foundation (2017M611499 to DW), and the Program for Professor of Special Appointment (Eastern Scholar) at Shanghai Institutions of Higher Learning to XM.


and a subcutaneous to visceral fat switch. Cell (2014) 156:304–16. doi: 10.1016/j.cell.2013.12.021


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Ma, Wang, Zhao and Xu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Insulin Resistance in HIV-Patients: Causes and Consequences

Marcelo N. Pedro1†, Guilherme Z. Rocha1†, Dioze Guadagnini <sup>1</sup> , Andrey Santos <sup>1</sup> , Daniela O. Magro<sup>2</sup> , Heloisa B. Assalin<sup>1</sup> , Alexandre G. Oliveira1,3, Rogerio de Jesus Pedro<sup>1</sup> and Mario J. A. Saad<sup>1</sup> \*

<sup>1</sup> Department of Internal Medicine, Faculty of Medical Sciences, State University of Campinas-UNICAMP, Campinas, Brazil, <sup>2</sup> Department of Surgery, Faculty of Medical Sciences, State University of Campinas-UNICAMP, Campinas, Brazil, <sup>3</sup> Biosciences Institute, São Paulo State University (UNESP), Rio Claro, Brazil

Here we review how immune activation and insulin resistance contribute to the metabolic alterations observed in HIV-infected patients, and how these alterations increase the risk of developing CVD. The introduction and evolution of antiretroviral drugs over the past 25 years has completely changed the clinical prognosis of HIV-infected patients. The deaths of these individuals are now related to atherosclerotic CVDs, rather than from the viral infection itself. However, HIV infection, cART, and intestinal microbiota are associated with immune activation and insulin resistance, which can lead to the development of a variety of diseases and disorders, especially with regards to CVDs. The increase in LPS and proinflammatory cytokines circulating levels and intracellular mechanisms activate serine kinases, resulting in insulin receptor substrate-1 (IRS-1) serine phosphorylation and consequently a down regulation in insulin signaling. While lifestyle modifications and pharmaceutical interventions can be employed to treat these altered metabolic functions, the mechanisms involved in the development of these chronic complications remain largely unresolved. The elucidation and understanding of these mechanisms will give rise to new classes of drugs that will further improve the quality of life of HIV-infected patients, over the age of 50.

Keywords: insulin resistance, diabetes, HIV, CVD, cART, LPS

#### INTRODUCTION

The introduction and evolution of antiretroviral drugs over the past 25 years has completely changed the clinical prognosis of HIV-infected patients (1). These drugs have transformed the disease into a chronic condition, and increased life expectancy, which is similar to the general uninfected population (2). Nowadays, the deaths of HIV-infected individuals, who appropriately follow their therapy regimen, are related to non-communicable and HIV-related chronic diseases, mainly atherosclerotic cardiovascular disease (CVD) (3–6). Some of the mechanisms responsible for this increased cardiovascular risk, in HIV-infected patients, involve HIV infection and inflammation, dyslipidemia, insulin resistance, as well as metabolic, and body composition changes induced by antiretroviral therapy (7–11). Moreover, non-HIV related risk factors, such as aging, can also contribute to the development of these metabolic alterations and risk factors (12–15).

Here we review how immune activation and insulin resistance contribute to the metabolic alterations observed in HIV-infected patients, and how these alterations increase the risk of developing CVD. In section Sources of Immune Activation and Insulin Resistance in HIV Patients, we discuss how HIV infection and inflammation, combination antiretroviral therapy (cART), and

#### Edited by:

Andrew J. McAinch, Victoria University, Australia

#### Reviewed by:

Caterina Conte, Università Vita-Salute San Raffaele, Italy Kathleen Grace Mountjoy, University of Auckland, New Zealand

> \*Correspondence: Mario J. A. Saad msaad@fcm.unicamp.br

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Obesity, a section of the journal Frontiers in Endocrinology

Received: 28 May 2018 Accepted: 16 August 2018 Published: 05 September 2018

#### Citation:

Pedro MN, Rocha GZ, Guadagnini D, Santos A, Magro DO, Assalin HB, Oliveira AG, Pedro RdJ and Saad MJA (2018) Insulin Resistance in HIV-Patients: Causes and Consequences. Front. Endocrinol. 9:514. doi: 10.3389/fendo.2018.00514

**30**

gut microbiota contribute to immune activation and insulin resistance. In section Consequences of Immune Activation and Insulin Resistance in HIV-Infected Patients, we review the clinical consequences of immune activation and insulin resistance, as well as how these processes are involved in the development of age-related metabolic diseases in HIV patients.

### SOURCES OF IMMUNE ACTIVATION AND INSULIN RESISTANCE IN HIV PATIENTS

The immune activation of HIV-infected patients, whether on cART or not, is usually accompanied by insulin resistance (16). In this section, the role of HIV infection and inflammation, cART, and gut microbiota in immune activation and molecular mechanism of insulin resistance are discussed.

#### Effect of HIV Infection and Inflammation on Insulin Resistance

It is generally accepted that there is a correlation between innate immune system activation and insulin resistance, which contributes to glucose metabolism dysregulation and dyslipidemia (8). Immune activation results in chronic inflammation, that varies in severity, and has been observed in untreated HIV patients and patients undergoing cART (17). However, untreated HIV-patients display an enhanced inflammatory state, which is characterized by high levels of proinflammatory cytokines, like tumor necrosis factor alpha (TNF-α), and interleukins (IL-6 and IL-1β), and is associated with a procoagulant state (12). Under these conditions, the insulin resistance is probably severe and could occur in the liver, muscle, and adipose tissue. In fact, severe insulin resistance in the adipose tissue (as observed in HIV untreated patients), may prevents adipose mass gain as described in mice (18–20) (**Figure 1A**).

In patients undergoing antiretroviral drug therapy, there is a decrease in proinflammatory cytokines, which do not completely return to normal, thus indicating that some level of inflammation persists (21). A variety of factors, such as: virus production, cytomegalovirus infection, regulatory T-cell loss, and/or lymphoid structure damage could contribute to this persistent inflammation (22, 23). There is still insulin resistance, but it is mild or moderate. As previously demonstrated in animal models of obesity (24, 25), less severe insulin resistance in adipose tissue allows normal or increased glucose uptake and lipid conversion in this tissue, favoring weight gain and contributing to explain the increase in visceral adipose tissue (VAT) in these patients (26, 27) (**Figure 1A**).

In HIV-treated patients, the activation of the innate immune system and insulin resistance is similar to what has been described in obesity and type 2 diabetes mellitus (DM2) (28, 29). The innate immune system and insulin signaling are integrated and toll-like receptors (TLRs), inducible nitric oxide synthase (iNOS), protein kinase R (PKR), c-Jun N-terminal kinases (JNK), and NF-κB are connected to the insulin receptor (IR) and its downstream signaling pathway IRS/PI3K/Akt. Upon activation of the innate immune system, proteins involved

FIGURE 1 | (A) Tissue specific insulin resistance. In HIV untreated patients there is severe insulin resistance with increased LPS and cytokines that involves liver, hypothalamus, muscle, vessels and adipose tissue. After cART treatment the insulin resistance is mild/moderate, with reduced LPS and cytokines that probably spares adipose tissue. (B) Molecular mechanism of LPS-induced insulin resistance. The increase in LPS circulating levels in HIV patients will induce an increase in circulating inflammatory cytokines, and these increases will activate TLR4, IL-6, and TNFalpha receptors, which will induce ER stress, mitochondrial dysfunction, activation of inflammasome and an increase in intracellular lipid accumulation. Then, there will be activation of PKR, JNK, and IKKβ/NF-κB pathways in liver, muscle, adipose tissue, macrophages and also in other tissues. The activation of these serine kinases (PKR, JNK, and IKKβ) will induce serine phosphorylation of the IRS1/2 and consequently a down regulation in insulin signaling.

in insulin signaling pathways become posttranscriptionally modified, resulting in reduced insulin action (**Figure 1B**) (30, 31).

It is important to mention that bacterial lipopolysaccharide (LPS) from the Gram negative intestinal bacteria is continuously produced in the gut (secondary to death of Gram negative bacteria) and translocated to the circulation (32). This translocation depends on many factors including immune system, integrity of epithelia barrier, diet, and many other environmental factors. The increase in circulating LPS, through its own receptor -TLR4- induce the release of inflammatory cytokines that can contribute to insulin resistance (33, 34).

Previous data showed that there is an increase in LPS circulating levels in HIV patients, whether on treatment or not (35), which can induce at the same time TLR4 activation and endoplasmic reticulum (ER) stress (19, 36, 37). Then, there will be activation of PKR, JNK, and IKKβ/NF-κB pathways in liver, muscle, adipose tissue, macrophages, and also in other tissues. The activation of these serine kinases (PKR, JNK, and IKKβ) will induce serine phosphorylation of the insulin receptor substrate 1 and 2, and consequently a down regulation in insulin signaling (38, 39). The activation of NF-κB pathways in liver, adipose tissue, and macrophages will induce the production of proinflammatory cytokines (i.e., TNF-α, IL-1β, and IL-6), creating an inflammatory vicious cycle, which is even worst with the increased adiposity (40–43). Certainly, this aggravates inflammation and insulin resistance.

In addition to increase in circulating LPS and in proinflammatory cytokine TNF-α, IL-1β, and IL-6, the JNK and NF-κB pathways can also be activated by intracellular mechanisms that involve oxidative and endoplasmic reticulum (ER) stresses, activation of inflammasome and an increase in intracellular lipid accumulation (39, 44, 45) (**Figure 1B**). Additionally, augmented iNOS activity and the nitrosylation of insulin pathway proteins have been shown to promote insulin resistance (46, 47). In summary, an increase in LPS and proinflammatory cytokines and in intracellular mechanisms will activate serine kinases, resulting in insulin receptor substrate-1 (IRS-1) serine phosphorylation and insulin signal transduction inhibition (30, 31, 48).

# Effect of Viral Suppression on Insulin Resistance

Protease inhibitors (PI) or nucleoside analog reverse transcriptase inhibitors (NRTI) have been shown to induce insulin resistance, dyslipidemia, and lipodystrophy, and consequently increase cardiovascular risk (17, 49–51). These drugs increase the nuclear localization of SREBP-1 (sterol regulatory element-binding protein 1), which is a transcription factor that regulates the expression of genes associated with lipid synthesis (52). In the liver, these antiviral drugs can increase the levels of free intracellular cholesterol and lipids (53), which can affect aging and the immune system response. In the muscle and adipose tissue, these drugs can induce ER stress and reduce glucose transporter 4 (GLUT4) translocation to the plasma membrane (54, 55). The NRTIs also inhibit mitochondrial DNApolymerase, respiratory chain function, and ATP production, ultimately leading to adipocyte death (56–60).

HIV patients undergoing cART exhibit a partial reversal of immune activation and inflammation. Additionally, cART reduces opportunistic infections and cardiovascular risk factors, which is likely a result of some reduction in inflammation (61), although residual markers of inflammation and coagulation remains elevated in ART-treated HIV-infected patients (62). In treated patients the dyslipidemia correlates better with Creactive protein and IL-6 levels, rather than with CD4 count or HIV viral load, suggesting that immune activation has a central role in the development of dyslipidemia (63). While cART improves some of the observed alterations, it does not reverse the immune activation or chronic inflammation completely. In fact, patients undergoing cART still present a proinflammatory and prothrombotic state, accompanied by changes in the number and size of low-density lipoprotein (LDL) and high-density lipoprotein (HDL) particles, which increases the risk of these patients developing cardiovascular complications (62, 64–67).

# Effect of Microbiota Modulation on Insulin Resistance

While the progression from HIV infection to AIDS is primarily modulated by T cell activation and systemic inflammation, there is evidence that the gastrointestinal mucosa immune system also participates in this process (68). The human gut microbiota is mainly composed of four phyla: Firmicutes, Bacteroidetes, Actinobacteria, and Proteobacteria. The general population predominantly harbors Bacteroidetes, followed by Firmicutes; however, the composition of the gut microbiota is influenced by diet, age, geography, drugs, and cultural behaviors (69–72). Over the past 10 years, it is becoming clear that microbiota populations are modulated and may have a causal effect in more prevalent chronic conditions such as: obesity, diabetes, hypertension, and CVD (73).

Interestingly, HIV infection can also modulate the levels of bacteria of the gut microbiota. In fact, there is a decrease in the levels of the phylum Bacteroidetes, but some genders of this phylum, as Prevotella, increases when analyzed in treated and untreated patients (74–76). Such a change in the gut microbiota could result in increased tryptophan catabolism, chronic inflammation, and increased cardiovascular risk (77, 78). Additionally, an increase in Prevotella could augment circulating trimethylamine (TMA) levels, which is transformed into trimethylamine oxide (TMAO), and can have a role in the development of atherosclerosis (79).

Previous data showed that increased levels of choline and TMAO are associated with cardiovascular diseases (80). It is well known that ingested choline is transformed by gut microbiota in TMA, which enters portal circulation and in liver is converted in TMAO (80). It is interesting that fasting TMAO levels are independent predictor of atherosclerotic disease and high-risk mortality in coronary artery disease patients (81, 82). The mechanisms by which TMAO induces or accelerates atherosclerosis is not completely understood, it may involve macrophage activation and increase in foam cells and also modulation of platelet aggregation and adhesion (83). Moreover, besides promoting atherosclerosis lesion development, TMAO also aggravate pressure-overload heart failure in mice (84).

Recently, our group demonstrated that, in HIV patients, a close correlation exists between increased circulating LPS levels, a marker for intestinal permeability, and insulin resistance (35). The increased translocation of LPS and elevated serum levels induce the activation of the innate and adaptive immune systems (35, 85, 86). As discussed previously, in macrophages and most tissues, LPS binds to and activates TLR4, which initiates a complex cascade of signaling events, resulting in the downstream activation of the JNK and NF-kB pathways, and consequently

insulin resistance and systemic inflammation. Additionally, LPS has also been shown to increase adipose tissue, which augments body weight gain (35, 87, 88). Moreover, in HIV-infected patients, elevated levels of LPS have been linked to endothelial dysfunction and adverse metabolic outcomes (35, 88–91).

It is important to mention that the effect of cART introduction in the modulation of gut microbiota is not completely understood. However, very recently Ji et al. showed that this modulation after cART was differentially correlated with the immune status, especially in patients with CD4 + T cell counts>300/mm<sup>3</sup> (92). In these patients it was shown that the alpha diversity was correlated with CD4 + T cell counts, but the specific role of cART in increasing microbial diversity is still controversial (78, 92, 93). This correlation may explain the conflicting results in previous studies investigating alpha diversity in intestinal microbiota in HIV patients (93–96), indicating that this diversity is consequence of the immune status of the subjects. The immunological profile and cART seem to contribute together to alter the gut microbiota.

# CONSEQUENCES OF IMMUNE ACTIVATION AND INSULIN RESISTANCE IN HIV-INFECTED PATIENTS

Chronic immune activation and insulin resistance can contribute to obesity, dyslipidemia, CVDs, and non-alcoholic fat liver disease (NAFLD) as well as neurocognitive disorders, metabolic disorders, bone abnormalities, and non-HIV associated cancers (12, 97–99). While the evolution of these complications depends on genetic and environmental factors, each condition has the potential of aggravating another (**Figure 2**).

# Obesity and Lipodystrophy

Before the new generation of antiretroviral therapies, HIV was often associated with lipodystrophy, which is a marker for metabolic alterations and includes a broad spectrum of clinical alterations (100, 101). Previous studies showed that HIV infection severity was associated with an increased prevalence of lipodystrophy, which is secondary to HIV-infected macrophages infiltration and enhanced local inflammation in the adipose tissue (102, 103). In the past, the development of lipodystrophy was partially related to drugs (i.e., stavudine and zidovudine) included in the treatment regimen, but it is also influenced by age, CD4 levels, viral load, therapy duration, and race (especially caucasians). Remarkably, the new classes of cART and inhibitors (fusion, integrase, and entry) do not alter the metabolic parameters of fat distribution (1, 50).

Obesity and visceral adiposity are commonly observed in HIV-treated patients and are the result of factors associated with both traditional treatments and cART. As with most obese people, the increase in adipose tissue is associated with inflammatory and metabolic responses. Since many HIV patients have low muscle mass, excess adipose tissue may be present, even when the BMI is within the normal range. In fact, a recent study showed that when considering BMI, 60–70% of HIV-infected

patients are considered overweight or obese (104–108). In most patients, there is an increase in visceral adipose tissue (VAT), which is usually indicative of a more deleterious metabolic profile (109–111).

In obese populations, metabolically healthy obesity is characterized by less VAT and reduced inflammation (112– 115). These same more benign metabolic conditions have also been documented in some HIV infected patients considered overweight or obese. The reasons why some obese individuals, with or without HIV, have a more aggressive metabolic profile and associated risk factors are not completely understood, but it might involve adipocyte size and/or number, recruited inflammatory cells, hypoxia, and/or adiponectin levels (116– 118).

The extrapolation of data from the obese population without HIV to those with HIV must be interpreted with care, since a large portion of the data related to adipose tissue dysfunction includes HIV patients with lipodystrophy. Although this alteration in fat distribution still occurs in treated HIVinfected patients, the prevalence of obesity is increasing, and in some patients, an association between obesity and lipodystrophy has been observed (119).

#### Dyslipidemia

During the 80's and early 90's, before the introduction of antiretroviral therapy, dyslipidemia was evident in more severe HIV cases, and was characterized by high triglyceride (TG) levels and low levels of HDL-cholesterol (HDL-C) and LDLcholesterol (LDL-C). Although the exact mechanisms that account for the development of this kind of dyslipidemia are not fully understood, there is data suggesting that it may be induced by insulin resistance resulting from HIV infection and inflammation (120–122). This observed pattern of dyslipidemia is not only observed in HIV patients, but is also detected in other infections and inflammations, and can become atherogenic if it persists (123, 124). In this regard, an increase in TNF-α level can impair the clearance of TGs and reduce the antilipolytic effect of insulin, thus stimulating lipolysis in HIV-patients with lipodystrophy.

Besides activation of the innate immune system and insulin resistance, which contributes to glucose metabolism dysregulation and dyslipidemia, other mechanisms may also contribute to explain this pattern of dyslipidemia. One such mechanism involves the adenosine-triphosphate binding cassette transporter A1 (ABCA1), a transmembrane transporter present in macrophages, which interacts with the HIVproduced accessory protein Nef. Under physiological conditions, ABCA1 shuttles cholesterol from macrophages (in peripheral tissues) to HDL, which was previously shown to be to reduce cardiovascular risk (87, 125). However, in HIVinfected patients, Nef downregulates ABCA1 expression and reduces the efflux of cholesterol to HDL. As a result, lipid accumulates inside the macrophage, and is transformed into foam cells, which is associated with atherosclerosis (125).

Another mechanism, that can contribute to the understanding of why this pattern of dyslipidemia is commonly observed in HIV-patients, is through the inhibition of an intracellular peroxisome protein, proliferatoractivated receptor gamma (PPAR-γ) (126). This protein is critical for adipocyte differentiation, and is inhibited by the HIV viral protein, vpr. The inhibition of PPARγ blocks adipocyte differentiation, and leads to fatty acid accumulation and lipotoxicity. Moreover, HIV replication is also associated with an increase in fatty acid synthase (FAS) activity, which impacts fatty acid synthesis (127). These data suggest that in HIV-infected patients, not undergoing cART, there is an increase in fatty acid production that can contribute to the appearance of dyslipidemia and insulin resistance.

On the other side, it has been shown that cART not only suppresses HIV infection and reduces inflammation, but it also changes the dyslipidemia pattern, which is characterized by an increase in TGs and LDL-C, a reduction in HDL-C and maintenance of insulin resistance (128, 129). In fact, a recent meta-analysis study showed that cART patients have a higher risk of developing hypercholesterolemia and display higher TG levels than non-treated HIV patients (27). The previously described modulation of ABCA1 by Nef, which reduces the efflux of HDL, is also reversed by cART (130–132).

#### Cardiovascular Diseases and NAFLD

As previously mentioned, there is still an increased risk of CVD in the HIV-infected population, despite cART and the control of risk factors (1, 51). Even with the development of new antiretroviral drugs, the chronic immune activation and insulin resistance remain and contribute to this greater risk (12, 51). Also, recent data has shown that HIV-infected patients can present left ventricular systolic and diastolic dysfunction and myocardial fibrosis (133, 134).

Abnormal liver enzymes are also common in HIV-infected patients, despite the absence of alcohol consumption or viral hepatitis. These abnormalities have been associated with an increased prevalence of NAFLD and non-alcoholic steatohepatitis (NASH) (135). The actual prevalence of NAFLD and NASH in HIV-infected patients is not known because the methods used to define these alterations vary among studies (136). Recent data has shown that in HIV-infected patients, treated with cART, the prevalence of NAFLD is around 40%, and that these patients have a higher risk of developing NASH or cirrhosis than obese patients without HIV (137– 141). These hepatic alterations are secondary to multiple factors, and immune activation, insulin resistance, cART, and aging are certainly involved in the process. It is also important to mention that NAFLD is a risk factor for diabetes and CVDs (142).

# Neurocognitive Disorders, Metabolic Disorders, Bone Abnormalities, and Non-HIV Associated Cancers

The HIV-infected population can develop behavioral abnormalities, motor dysfunction, and dementia (12). The clinical presentation can vary from mild neurocognitive disorders to severe HIV-associated dementia (143, 144). The prevalence of these abnormalities is ∼50% (145), however, since the introduction of modern cART, the prevalence of severe forms of neurocognitive disorders has been dramatically reduced (146).

With regards to glucose metabolism, it is clear that current cART is much less metabolically toxic than previous therapies. However, HIV and cART are independently associated with glucose intolerance and diabetes (49). Again, while these abnormalities are secondary to multiple factors, glucose intolerance and diabetes are known to exacerbate the risk of these patients developing CVD.

Recent data has shown that HIV-infected patients have fracture rates three times higher than the control population (147, 148). In fact, decreased bone mineral density, osteopenia, and osteoporosis have been observed in these patients, and are probably related to immune activation and systemic inflammation, cART, low vitamin D, and/or aging.

The risk of some non HIV-associated cancers is 50% higher in HIV-infected patients than in non-infected patients (149, 150). For example, HIV infection is associated with a higher incidence of virus-related cancers such as: Kaposi sarcoma, lymphomas, and anal and liver cancer, which is most likely secondary to the poor immunological control of oncogenic viruses (12).

Overall, recent data showed that the mortality of HIV-infected people decreased in those below 65 years old, but increased after this age (112, 115), which is likely due to the increased risk of developing CVDs and dying from acute coronary syndrome. Furthermore, there is also an increase in the occurrence of coronary artery disease in young HIV-infected patients, when compared with the uninfected control population (113, 114, 151).

#### CONCLUSION

The introduction of antiretroviral drugs has changed the clinical prognosis of HIV-infected patients and the deaths of these individuals are now related to atherosclerotic CVDs, rather than from the viral infection itself. However, HIV infection, cART, and intestinal microbiota are associated with immune activation and insulin resistance, which can lead to the development of a variety of diseases and disorders, especially with regards to CVDs. While lifestyle modifications and pharmaceutical interventions can be employed to treat these altered metabolic functions, the mechanisms involved in the development of these chronic complications remain largely unresolved. The elucidation and understanding of these mechanisms will give rise to new classes of drugs that will further improve the quality of life of HIV-infected patients, over the age of 50.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

MP and GR contributed to discussions and wrote, edited, and reviewed the article. DG, DM, HA, AO, and RP edited and reviewed the article. AS prepared to figures of the article, edited, and reviewed the article. MS contributed to discussions and wrote and reviewed the article. All authors approved the final version.

#### FUNDING

We also acknowledge the financial support CEPID/Fapesp 201307607-8, OCRC (Obesity and Comorbidities Research Center), INCT (National Institute of Science and Technology for Diabetes and Obesity), 465693/2014-8 and CAPES/CNPq (Conselho Nacional de Desenvolvimento Científico e Tecnológico).


among HIV-infected patients. J Clin Endocrinol Metab. (2015) 100:2873–82. doi: 10.1210/jc.2015-1461


age-related changes to monocyte activation equivalent to 12 and 4 years of aging, respectively. J Acquir Immune Defic Syndr. (2015) 69:11–7. doi: 10.1097/QAI.0000000000000559


atrophy and fat deposition in a cohort of HIV-infected men and women. Clin Infect Dis. (2005) 40:1837–45. doi: 10.1086/ 430379


fatty liver disease on HIV-infected individuals. AIDS (2015) 29:1927–35. doi: 10.1097/QAD.0000000000000799


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Pedro, Rocha, Guadagnini, Santos, Magro, Assalin, Oliveira, Pedro and Saad. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Role of Hepatocyte Growth Factor (HGF) in Insulin Resistance and Diabetes

Alexandre G. Oliveira1,2 \* † , Tiago G. Araújo1,3†, Bruno de Melo Carvalho1,4 , Guilherme Z. Rocha<sup>1</sup> , Andrey Santos <sup>1</sup> and Mario J. A. Saad<sup>1</sup> \*

<sup>1</sup> Department of Internal Medicine, State University of Campinas, Campinas, Brazil, <sup>2</sup> Department of Physical Education, Institute of Biosciences, São Paulo State University (UNESP), Rio Claro, Brazil, <sup>3</sup> Department of Physiology and Pharmacology, Federal University of Pernambuco, Recife, Brazil, <sup>4</sup> Institute of Biological Sciences, University of Pernambuco, Recife, Brazil

In obesity, insulin resistance (IR) and diabetes, there are proteins and hormones that may lead to the discovery of promising biomarkers and treatments for these metabolic disorders. For example, these molecules may impair the insulin signaling pathway or provide protection against IR. Thus, identifying proteins that are upregulated in IR states is relevant to the diagnosis and treatment of the associated disorders. It is becoming clear that hepatocyte growth factor (HGF) is an important component of the pathophysiology of IR, with increased levels in most common IR conditions, including obesity. HGF has a role in the metabolic flux of glucose in different insulin sensitive cell types; plays a key role in β-cell homeostasis; and is capable of modulating the inflammatory response. In this review, we discuss how, and to what extent HGF contributes to IR and diabetes pathophysiology, as well as its role in cancer which is more prevalent in obesity and diabetes. Based on the current literature and knowledge, it is clear that HGF plays a central role in these metabolic disorders. Thus, HGF levels could be employed as a biomarker for disease status/progression, and HGF/c-Met signaling pathway modulators could effectively regulate IR and treat diabetes.

Keywords: hepatocyte growth factor, HGF, obesity, insulin resistance, diabetes, inflammation, beta cells

# INTRODUCTION

Obesity is a rapidly growing worldwide epidemic. It is estimated that more than 1 billion adults worldwide are obese (BMI ≥ 30 Kg/m<sup>2</sup> ) or overweight (BMI between 25 and 30 kg/m<sup>2</sup> ) (1–3). It is projected that by 2025, 40% of the population in the US, 30% in England and 20% in Brazil will be affected by obesity (1–3). The WHO has declared that obesity is a key factor in the development of type 2 diabetes (1). Along these lines, there are a plethora of studies that have investigated the roles of and relationships among nutrition, physical activity and genetic susceptibility as determinants of the current prevalence of obesity. Additionally, it is well established that there is a direct relationship between obesity, the onset of insulin resistance (IR) and diabetes.

Insulin resistance is manifested by reducing the ability of insulin to activate the insulin signaling pathway (4, 5). At the molecular level, IR is characterized by diverse alterations in various intracellular signaling pathways. In fact, it has been shown that insulin signaling is impaired in the liver, muscle, adipose tissue, hypothalamus, and others tissues, in IR states (6, 7). While there are several biological events that can lead to the impairment of the insulin signaling pathway, chronic

#### Edited by:

Andrew J. McAinch, Victoria University, Australia

#### Reviewed by:

Negar Naderpoor, Monash University, Australia Kerry Loomes, University of Auckland, New Zealand

#### \*Correspondence:

Alexandre G. Oliveira alexandre.gabarra@unesp.br Mario J. A. Saad msaad@fcm.unicamp.br

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Obesity, a section of the journal Frontiers in Endocrinology

Received: 05 March 2018 Accepted: 10 August 2018 Published: 30 August 2018

#### Citation:

Oliveira AG, Araújo TG, Carvalho BM, Rocha GZ, Santos A and Saad MJA (2018) The Role of Hepatocyte Growth Factor (HGF) in Insulin Resistance and Diabetes. Front. Endocrinol. 9:503. doi: 10.3389/fendo.2018.00503

**40**

inflammation is perhaps the best described. Inflammation promotes serine phosphorylation of insulin receptor substrate (IRS), through the action of serine kinases such as c-Jun Nterminal kinase (JNK) and inhibitor of kappa B kinase (IKKβ) (6, 8, 9).

Diabetes is another major global health problem, from which the social and economic consequences are devastating. For example, this disease is directly or indirectly responsible for approximately 4 million deaths per year, which corresponds to 9% of the total world mortality (10). In 2013, the International Diabetes Federation (IDF) reported that diabetes affected 382 million people worldwide, and this number is expected to grow to 592 million, by 2035 (10, 11).

Obesity, IR and type 2 diabetes are closely related to subclinical and chronic inflammation, which is characterized by abnormal cytokine production, including an increase in acute phase molecules and other mediators, as well as inflammatory signaling pathway activation (6, 12, 13).

Within this inflammatory context, the liver may have an important contribution, because during overeating, which is common in obesity, it expresses and secretes several inflammatory proteins (14, 15). The liver may also play an important role in IR (16). Likewise, in obesity-induced IR, the suppression of hepatic glucose production becomes impaired, contributing to the development of hyperglycemia. Furthermore, the liver participates in metabolic regulation, through the expression and secretion of organokines, which are also referred to as hepatokines. Interestingly, these proteins influence metabolic processes through paracrine and endocrine signaling, rather than by autocrine signaling (17). Hepatokines include several growth factors such as fibroblast growth factor 21 (18), insulin-like growth factors, angiopoietin-related growth factor, and HGF. The characteristics and roles of the latter will be the focus of this review.

HGF synthesis and secretion are up-regulated in IR states, and is thought to be an important component of the pathophysiology of IR (19–30). Interestingly, the receptor of HGF, c-Met, is structurally related to the insulin receptor (INSR) (31). Furthermore, HGF signaling participates in the modulation of the metabolic flux of glucose in different insulin sensitive cell types such as β-cells, enterocytes, adipocytes, hepatocytes and myocytes (25, 31–36). HGF signaling also plays a key role in β-cell homeostasis (33, 37–47). Lastly, HGF has been shown to participate in the modulation of the inflammatory response (48–51). This review discusses the significance of each of these five HGF features in the pathophysiology of IR and diabetes and presents evidence for HGF in the etiology of cancer.

# STRUCTURE AND MOLECULAR BIOLOGY OF HGF AND ITS RECEPTOR (C-MET)

Originally, HGF was identified as a liver-regenerative circulating factor, since levels of this protein increased following liver injury or hepatectomy (47, 52, 53). Currently, HGF is recognized as a molecule with diverse activities, and has been demonstrated to have mitogenic, morphogenic, and motogenic effects in different tissues (**Figure 1**) (54).

The HGF locus is harbored on chromosome 7q21.1, covering approximately 70 kb and consists of 18 exons and 17 introns (55). The translation of a single 6 kb transcript results in a prepro-polypeptide of 728 amino acids (56). In the liver, HGF is synthesized and secreted as an inactive single chain precursor (proHGF, 92 kDa). The active form of HGF is produced by proteolytic cleavage of proHGF, between Arg<sup>494</sup> and Val<sup>495</sup> . Active HGF consists of a heavy chain (62 kDa) and a light chain (32–36 kDa), which are covalently linked by a disulfide bridge (57, 58). In vitro studies have shown that blood coagulation factor XIIa, urokinase, plasminogen activator and HGF activator can all activate HGF. Among these proteases, HGF activator (34 kDa) is the most efficient at converting proHGF into mature HGF (59). Both the liver and, to a lesser extent, gastrointestinal tissue synthesize HGF activator (60, 61). Following translation, this protease circulates as an inactive precursor, activator proHGF (96 kDa), and is activated upon thrombin cleavage (62). These studies highlight the important role HGF activator plays in regulating HGF activity. Moreover, a HGF activator inhibitor (HAI-1) was recently isolated, and acts as a physiological HGF activator inhibitor at the cell surface (53, 63). However, its role needs to be better elucidated.

Biological activities (i.e., proliferation/survival/motogenesis) mediated by HGF binding to the c-Met receptor require the active form of HGF. This receptor is encoded for by the c-met protooncogene (p190MET) (33). Additionally, the c-Met receptor is a dimer composed of an alpha-chain (50 kDa), and a beta-chain (145 kDa), which are linked together by a disulfide bond. The alpha chain is highly glycosylated and exposed on the cell surface, while the beta-chain is a transmembrane protein containing the kinase domain, the tyrosine autophosphorylation site and the

multifunctional binding site (53, 64). The c-Met receptor is more closely related to the INSR, in terms of the overall structure of the protein and the sequence of the kinase domain, than any other member of the receptor tyrosine kinase (RTK) subclasses (65).

Upon HGF binding, c-Met activation and subsequent signal transduction rely on the presence of adenosine triphosphate (ATP) and Mg2<sup>+</sup> ions, which promote receptor dimerization and trans-phosphorylation of the catalytic domain. Additionally, HGF binding to c-Met and subsequent phosphorylation of intracellular transducers such as: PI3K (p85 subunit), Gap, Ras, PLC-gamma, Src related tyrosine kinase, Grb-2, Gab-1, IRS-1, IRS-2 (**Figure 1**) (33, 53, 56, 66) has been shown to promote HGF functions in the cell.

For example, Fafalios et al. (31) demonstrated that c-Met activation in the liver induces the formation of a c-Met-INSR complex, which stimulates the recruitment of insulin receptor substrates (IRS-1 and IRS-2) and thereby amplifies insulin signaling (**Figure 2**). Furthermore, this study showed that activation of this complex improves hepatic glucose metabolism, by increasing glucose uptake and decreasing hepatic glucose production (31). These findings reinforce the notion that HGFinduced signaling, at least in the liver, imparts protection against IR. These effects need to be further elucidated in both the liver and other metabolic tissues in an IR state.

# THE ROLE OF HGF IN THE MODULATION OF GLUCOSE FLUX IN DIFFERENT INSULIN SENSITIVE CELL TYPES

Several studies have demonstrated that HGF has regulatory effects on glucose transport and metabolism. One such study

signaling pathway in hepatocytes. Adapted from Fafalios et al. (31).

showed that HGF overexpression, in murine β-cells, increases the transcription of GLUT-2 and glucokinase, leading to increased glucose uptake and metabolism in these cells (32). On the other hand, the ablation of HGF/c-Met signaling, in adult murine β-cells, led to a decrease in GLUT-2 expression, followed by glucose intolerance and an impairment in glucose-stimulated insulin secretion (33). Additionally, HGF was shown to interfere with the function of the Na+/glucose cotransporter and GLUT-5, in rat intestinal cells (34). Furthermore, HGF promotes the translocation of GLUT4 to the cell membrane and increases glucose uptake through the activation of PI3K, in 3T3-L1 adipocytes (25). With regards to regulating glucose transport and metabolism in the liver, it has been demonstrated that HGF is able to stimulate hepatic glucose uptake and suppress hepatic glucose output (31). In fact, an in vitro study demonstrated that HGF substantially increases glucose transport and metabolism in myotubes, through a mechanism mediated by the PI3K/Akt pathway, resulting in increased GLUT-1 and GLUT-4 plasma membrane expression (35). More recently, the same group presented further evidence of HGF participation in skeletal muscle glucose homeostasis (36). In this study, a transgenic model, capable of increasing skeletal muscle HGF levels by threefold, without altering plasma HGF levels was employed. When these transgenic mice were fed a high-fat diet there was an observed improvement in glucose tolerance, and an increase in Akt phosphorylation levels in the gastrocnemius muscles (36). Thus, reinforcing the notion that HGF plays an important role in obesity-mediated IR in muscle.

# REGULATION OF PANCREATIC ISLET β-CELL HOMEOSTASIS BY HGF

There is a great interest in identifying endogenous regulatory factors that control both β-cell mass expansion and insulin secretion. In β-cell cultures, Otonkoski et al. (40) investigated the mitogenic and morphogenic activity of various growth factors, and were the first to observe the insulinotropic activity of HGF. Later, Hayek et al. (42) showed that adding HGF to culture medium induced β-cell proliferation. Furthermore, it was observed that mRNAs, encoding for HGF and c-Met receptor, are highly expressed in pancreatic β-cells during early development, and remain at low levels during puberty and throughout adult life (41). These studies have raised interest in identifying HGF as a potential element in the development of hyperplasia in β-cells, and subsequent hyperinsulinemia.

In 2000, Garcia-Ocana et al. developed transgenic mice that overexpressed HGF, with the intention of studying the role of this protein on the pancreatic islets, in vivo. The authors concluded that in vivo HGF overexpression increased β-cell proliferation, the number of islets, β-cell mass, as well as insulin production (47). Together these effects resulted in animals that displayed moderate hypoglycemia and hyperinsulinemia (**Figure 3A** summarizes main results of this study). In 2005, two studies used a c-Met receptor knockout mouse model to investigate the physiological consequence of HGF inhibition in pancreatic β-cells, and the results were discordant in some

aspects (33, 43). For example, Roccisana et al. (33) showed that c-Met receptor knockout animals presented a reduction in insulin secretion and a decrease in glucose tolerance, which was accompanied by a reduction in GLUT-2 expression in the βcells. However, there were no observed changes in total β-cell mass, proliferation or islet morphology (33). The main findings of this research is illustrated in **Figure 3B**. On the other hand, Dai et al. (43) demonstrated that the size of the islets in this animal model was decreased, which was accompanied by reduced insulin content in the pancreas, lower insulin levels in the serum and mild hyperglycemia.

Furthermore, another study in mice demonstrated that the absence of c-Met, in the pancreas, led to a substantial increase in pancreatic β-cell death and a reduction in cell mass, which resulted in the manifestation of hypoinsulinemia. Consequently, these mice had a tendency to develop hyperglycemia in response to diabetogenic stimuli (46). Thus, these data indicate that HGF/c-Met-induced β-cell survival could be employed as a potential therapy for diabetes.

Another study from the Garcia-Ocaña group, demonstrated that HGF signaling is required for β-cell regeneration, following β-cell ablation (45). In this study, mice that underwent partial pancreatectomy (Ppx) and received a daily dose of HGF showed an increase in the rate of β-cell proliferation, when compared to animals that did not receive exogenous HGF. Conversely, c-Met knockout mice subjected to Ppx, had reduced β-cell mass, glucose intolerance, and decreased insulin secretion, when compared to wild type post-Ppx mice (45). More recently, a different mouse model was used to study the HGF-mediated regenerative effect (44). IRS2 KO mice are known to have an increased incidence of β-cell failure, increased apoptosis and reduced proliferation, which contributes to the development of diabetes in these mice (44, 67, 68). In order to investigate the effect of HGF on the observed β-cell failure, the IRS2 KO mice were cross-bred with transgenic mice overexpressing HGF in the β-cells. HGF overexpression was able to compensate for the negative effects related to the absence of IRS2, by normalizing β-cell mass and improving glucose homeostasis in the context of IR (44). Taken together, these findings indicate that HGF could be employed as a regenerative factor in the treatment of diabetes.

As mentioned above, the expression of c-Met in the β-cells results in enhanced HGF sensitivity, which can influence cell growth, survival and insulin secretion. Furthermore, HGF has both paracrine and endocrine properties, which are constantly regulating pathways involved in β-cell homeostasis. Following this line of reasoning, it was shown in hyperglycemia/diabetes, induced by stressful situations, that HGF has a protective role in physiological status of β-cells (37–39). Hence, HGF appears to be involved in the compensatory response of β-cells in two common conditions characterized by IR and associated with the pathophysiology of diabetes: (1) obesity, (2) pregnancy (37, 39). The latter has been shown to be directly associated with the development of gestational diabetes mellitus (GDM) (38).

In obesity, the most common insulin resistant condition, there is a compensatory pancreatic β-cell response against this hormonal resistance, which increases insulin secretion and maintains homeostatic glucose metabolism for a certain period of time. This compensatory β-cell response induces hyperinsulinemia, and is the first step in the development of diabetes (39). Since HGF stimulates insulin secretion and increases β-cell mass, both in vitro and in vivo, and due to the fact that increased circulating levels of this hormone are associated with obesity, HGF might link insulin resistance and β-cell hyperplasia. In fact, our group demonstrated through a dose-dependent, longitudinal approach that: (a) circulating HGF levels strongly correlate with β-cell mass increase in the IR state; (b) the β-cell mass increases in a HGF dose-dependent manner; (c) pharmacological blockage of the HGF signaling pathway abolishes the β-cell mass compensatory response; and (d) circulating HGF levels increase before the onset of the IRinduced β-cell compensatory response, thus suggesting a causal role in this specific event. Moreover, c-Met inhibition negatively affects the already impaired insulin signaling in the liver of dietinduced obese rats (37, 39). These HGF-mediated effects on βcell proliferation and expansion indicate that this hormone has a crucial role in increased metabolic demand, which is commonly observed in obesity-induced IR.

Another adaptive β-cell mechanism, in which HGF participates, occurs during pregnancy. In this situation, the β-cell adapts to an increased demand for insulin and undergoes structural and functional changes, such as β-cell expansion and enhanced insulin secretion. However, any perturbations or disturbances in this adaptive process can lead to the development of GDM (69). Interestingly, while searching for potential therapeutic targets, for the treatment of GDM, it was found that HGF expression is upregulated in rat islet endothelium at gestational day 15 (70), which also happens to be when maximal β-cell proliferation is detected (38). Based on this evidence and the other roles of HGF in β-cell homeostasis, it was hypothesized that maternal βcell adaptation is mediated by HGF. In fact, it was recently demonstrated that in the absence of c-Met, during pregnancy, there was a substantial impairment in β-cell proliferation and survival, resulting in β-cell mass reduction. In pregnant mice, this sequence of events culminates with hypoinsulinemia, consequent hyperglycemia and glucose intolerance, all of which are features of GDM (38). These observations implicate the HGF/c-Met signaling pathway in the pathogenesis of GDM, since ablation and inhibition of HGF or c-Met results in the disease. In this regard, targeting the HGF/cMet signaling pathway might represent a potential strategy for the treatment of GDM.

### ROLE OF ADIPOSE TISSUE DERIVED HGF

Obesity is associated with numerous metabolic changes, such as hyperinsulinemia and IR (71). The IR that occurs in adipose tissue may contribute to systemic IR and a variety of hepatic alterations (72, 73). Furthermore, secretion of adipokines, hormones, cytokines, chemokines, growth factors, acute phase proteins and HGF from the white adipose tissue (WAT) further complicates this condition (74). It is important to mention that even though HGF is considered a hepatokine, the synthesis and secretion of HGF has been demonstrated in murine 3T3- L1 adipocytes (20) and human adipose tissue (21, 22). Thus, it is quite plausible that the HGF synthesized in human adipose tissue plays an important role in obesity (23). Furthermore, due the importance of adipose tissue in IR, adipose-derived HGF expression might also have an influence on the obesity associated IR etiology. In fact, a recent study examined HGF secretion in different types of adipose tissue and observed that the HGF secretion pattern of the perivascular adipose tissue differs from the subcutaneous and visceral adipocytes (26). Surprisingly, this research also showed that, in both in vitro and in vivo models, perivascular fat cells have a much higher capacity for secreting adipokines, especially HGF, than other fat cell types. Additionally, it was shown that subjects with greater perivascular fat mass secreted more HGF (26), bringing to light the pathophysiological relevance of this hormone.

In humans, it has been demonstrated that levels of circulating HGF are elevated in obesity (27), metabolic syndrome (23), and diabetes mellitus (75). In obese individuals, circulating HGF levels were increased by more than 3-fold, when compared to lean individuals (27). HGF levels have also been correlated with anthropometric measures of obesity such as waist circumference, body mass index and body fat mass (24, 30). Furthermore, an association between weight loss induced by bariatric surgery and reduced HGF levels was previously observed, in obese patients (22, 28).

In patients with IR, a significant association between HGF and plasma insulin was observed suggesting a pathological link between them (23). A more thorough, 10-year prospective study proposed that elevated circulating levels of HGF were significantly associated with the development of IR (19). Furthermore, another prospective study, with over 12 years of follow-up, which included multi-ethnical individuals, also demonstrated that higher levels of serum HGF were positively associated with IR (76). Based on these data and the similarities between HGF and insulin, it is tempting to speculate that "HGF resistance" exists, which may contribute to glucose metabolism dysregulation. Finally, another study demonstrated that there is a significant correlation between circulating HGF levels and the Visceral Adiposity Index (VAI), which is a genderspecific mathematical index based on simple anthropometric and metabolic parameters as a presumed surrogate marker of adipose tissue function and distribution (29).

In summary, adipose tissue may be an important source of HGF, which may comprise an important component in adipose tissue dynamics. Thus, further emphasizing the role of HGF in the pathophysiological processes of obesity/IR/diabetes.

### THE INVOLVEMENT OF HGF IN THE MODULATION OF THE INFLAMMATORY RESPONSE

Over the past 10 years, evidence has indicated that there is a correlation between IR and inflammation (6, 8, 77–79). In fact, subclinical chronic inflammation has been shown to be a major mechanism for the development of IR in peripheral tissues. This mechanism results in a chronic upregulation of proinflammatory cytokines (i.e., TNF-α and IL-6), which are known to exert deleterious effects on the insulin signaling pathway in adipocytes, hepatocytes, and myocytes (6, 8, 80). Furthermore, the actions of macrophages have been studied extensively, since they have been shown to accumulate in the liver and WAT, and promote IR and liver disease (81–83).

Studies have shown that HGF is involved in the modulation of the inflammatory response. Notably, Coudriet et al. demonstrated that HGF could decrease the acute phase of the inflammatory response by attenuating the production of IL-6, and increasing IL-10 levels, in lipopolysaccharide (LPS) stimulated bone marrow derived macrophages (48). Similarly, Flaquer et al., using db/db mice to study diabetic nephropathy, demonstrated that when the HGF gene was delivered to these animals, there was a reduction in the circulating levels of IL-6 and MCP-1, an increase in the number of M2 macrophages and an improvement in glomeruli inflammation (49).

Two studies from the Morishita group (50, 51), demonstrated novel HGF-induced anti-inflammatory mechanisms. The first study utilized a transgenic mouse-model that overexpressed HGF (HGF-Tg), which inhibited LPS-induced oxidative stress and vascular tissue inflammation. These HGF-mediated protective effects against the LPS-induced reactive oxygen species (ROS) production and inflammation were achieved via epidermal growth factor receptor (EGFR) degradation and angiotensin II (Ang II) signaling inhibition (50). The second study was divided into two stages: an in vitro study and an in vivo study. The in vitro stage of the study demonstrated that HGF disrupts NF-κB signaling in RAW264 macrophages, as well as in co-culture with 3T3-L1 adipocytes, leading to an overall reduction in the expression of MCP-1, TNF-α, and IL-6 (51). The second stage employed ApoE KO mice, which exhibit a chronic inflammation phenotype, including adipose tissue macrophage infiltration, adipocyte hypertrophy and fatty liver. After crossing the ApoE KO mice with HGF-Tg, the chronic inflammation, presented by the ApoE KO mice, was significantly reduced. The authors also detected an increase in serum adiponectin levels in the ApoE KO/HGF-Tg mice (51). Taken together, these data suggest that HGF suppresses the production of pro-inflammatory cytokines and, conversely, increases the secretion of adiponectin, thus breaking the macrophageadipocyte inflammatory cycle (**Figure 4**). Furthermore, it has been recently shown that administering an antibody against HGF to wild type mice fed a high-fat diet, resulted in a more pronounced IR (84). A similar result was previously observed with pharmacological blockage of HGF signaling (37). Together, these results reinforce the already presented protective role of HGF on glucose metabolism in obesity. The observed results from HGF-Tg mice fed a high fat diet (HFD) also highlight the different anti-inflammatory effects of HGF in adipose tissue including the reduction in the levels of chemoattractants (MCP-1 and CXCL2), inflammatory cytokines (TNF-α and iNOS), as well as, pan macrophage marker (F4/80) (84). Thus, HGF represents a potential therapeutic target for inhibiting the inflammatory response in adipose tissue and improving IR.

# HGF SIGNALING IN CANCER

Activation of the HGF/c-Met signaling pathway is a hallmark of cancer cells and both HGF and c-Met have emerged as therapeutic targets. HGF, produced by some cancer cells, stimulates c-Met, through activation of the autocrine signaling system. Alternatively, HGF present in the tumor microenvironment activates c-Met receptors displayed on the surface of cancer cells, through the activation of the paracrine signaling system. The origin of HGF present in the tumor microenvironment is unknown and HGF could be synthesized by a variety of tissues throughout the body, such as the liver itself or by the adipose tissue. Recently, adipose tissue has been shown to be related to the increased prevalence and aggressiveness of some types of cancers (85– 87). HGF is known to promote proliferation, migration, invasion and survival of cancer cells, and to confer resistance to therapy. Thus HGF is a possible link between obesity and cancer.

While researchers are currently investigating the relationship between HGF and cancer, some properties of HGF in the tumor microenvironment should be considered. Tumor stromal cells (macrophages, inflammatory cells, endothelial cells, and cancer-associated fibroblasts) secrete a variety of growth factors, chemokines and cytokines promoting a pro-inflammatory microenvironment for wound healing (88, 89). In fact, tumors could be considered wounds that do not heal (90). HGF is an important component of the fibroblast secretome (91), and has been shown to stimulate cancer cell invasiveness (88, 89), and promote the epithelial-mesenchymal transition (EMT), cell scattering and migration (52, 92–94).

As evidenced by the embryonic lethality in mice deficient in HGF or c-Met, it is thought that HGF/c-Met signaling stimulates the EMT, and plays a major role in embryogenesis, organogenesis, tissue repair and wound healing. As a growth factor, HGF promotes the growth and survival of cancer cells, increases tumor aggressiveness, stimulates metastasis and is associated with resistance to therapy (95, 96).

Recently, established colon cancer cell lines and primary colon tumors have been shown to produce large amounts of HGF (97). High levels of HGF were also observed in the serum and tumor tissue of stage II and III colon cancer patients, and especially in patients with lymph node and liver metastasis (98, 99). Moreover, increased HGF levels have been correlated with lymph node metastasis and relapse in individuals with breast cancer (100, 101), multiple myeloma (102), and myeloid leukemia (103).

Sensitivity to anticancer treatment is largely influenced by stromal cells, and HGF derived from these cells is a notable factor that provides resistance to target drugs (104). In general, cancer cells do not produce HGF, themselves. Rather, cancer-associated fibroblasts are primarily responsible for HGF production, and as a result activate c-Met paracrinally. Thus, HGF might be an important stromal cell component of the tumor microenvironment (105–108).

Unfortunately, conventional cancer treatment (radiotherapy or chemotherapy), is not able to distinguish a malignant cell from a normal one. Targeted therapies have the ability to circumvent this apparent lack of specificity. In fact, drugs that can block specific cancer cell dependent pathways such as: EGFR, BRAF, HER2, and HGF/c-Met signaling are being developed, and are predicted to have fewer side effects. Previous studies have shown that cells that undergo EMT are more resistant to cell death (109, 110), thus it is possible that the HGF-induced EMT is responsible for the increased resistance to therapy. In fact, in lung cancer cells with c-Met amplification, HGF has been shown to be responsible for the observed resistance to anti-MET treatment (111, 112). Thus, HGF production by the stromal cells inhibits the response to MET kinase inhibitors, as well as other signaling pathways (111). Additionally, EMT is also responsible for the acquisition of a stem cell phenotype, which is metabolically altered and markedly resistant to therapies. Moreover, HGF induces Wnt signaling in colon cancer cells, resulting in a cancer stem cell phenotype in vitro and in vivo (113). Lastly, activation of HGF/c-Met signaling also contributes to the cancer stem cell phenotype in other types of cancer such as: gliomas (114, 115), colon cancer (107), head and neck cancer (116), prostate cancer (117) and pancreatic cancer (118).

The data related to the HGF/c-Met signaling pathway and cancer suggests that the pathway inhibition may stunt the growth and progression of cancers. Thus, inhibition of HGF synthesis or activity in stromal cells may be an effective approach for future targeted cancer therapy.

#### CONCLUSION AND PROSPECTS

It has been nearly three decades since human HGF cDNA was successfully cloned (52). Since that time to the present day, the biological functions of the HGF/c-Met axis have been extensively investigated. The results presented from the concerted efforts of numerous laboratories have provided compelling evidence for the essential physiologically relevant functions of HGF, as well as its therapeutic potential.

The liver-originated hormone, HGF, considered a hepatokine, can potentially interfere with energy metabolism (37, 39). Similar to what was observed with adipokines and myokines, HGF has uncertain mixed functions, which can improve the metabolic profile of type 2 diabetes or induce IR (14, 119). Future investigations into the physiological role of liver-derived factors will likely yield potential biomarkers and/or lead to the discovery of novel therapies against type 2 diabetes and other metabolic complications.

Although the great amount of evidence has suggested that increased levels of HGF are associated with the manifestation of IR, this review has highlighted some of the beneficial effects stemming from HGF activity observed in obesity. Additionally, due to the fact that HGF markedly increases glucose metabolism and transport in myocytes and adipocytes, future research needs to focus on elucidating the details of how HGF contributes to the protective mechanism against the development of IR, in the liver, and possibly other tissues, such as skeletal muscle and WAT. Further support, for HGF being utilized as a therapeutic target in the treatment of IR, comes from its role in β-cell hyperplasia, and subsequent hyperinsulinemia.

Beyond emphasizing the proposed utility of HGF as potential therapeutic target in IR, this review also discussed how maternal β-cell adaptation, during pregnancy, depends on the HGF/c-Met signaling pathway, and that impairment of this pathway can lead to the onset of GDM. Regarding obesity-induced IR, which is strongly associated with an inflammatory state, HGF also emerges as a positive factor. This is because HGF has the ability to blunt inflammation in adipose tissue, by increasing M2 macrophage polarization, and consequently augmenting the expression of IL-10, both of which are positively associated with insulin sensitivity (see **Figure 4**). Additionally, the HGF/c-Met pathway plays a key role in the onset and progression of human cancers, and pathway activation is associated with a poor prognosis. While HGF may serve as the link between obesity and cancer, it should be noted that given the association of HGF levels with tumor progression, metastasis and less favorable response to cancer treatment, employing HGF for the modulation of IR or treating diabetes may be significantly limited by its cancerpromoting properties.

Future research in the area should focus on understanding the regulatory mechanisms, and developing inhibitors and activators for the modulation of endogenous HGF levels. In fact, innovative mechanisms for controlling HGF expression and/or activity could generate unique therapies for the treatment of IR and diabetes.

#### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

# FUNDING

This work was supported by grants from Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP), Coordenação

#### REFERENCES


de Aperfeiçoamento de Pessoal de Nível Superior/PNPD (CAPES/PNPD), and Conselho Nacional de Pesquisa (CNPq) (Instituto Nacional de Ciência e Tecnologia-Obesidade e Diabetes).

to elevated serum HGF in obesity. Am J Physiol Endocrinol Metab. (2006) 291:E843–8. doi: 10.1152/ajpendo.00174.2006


maternal beta-cell adaptation and gestational diabetes mellitus. Diabetes (2012) 61:1143–52. doi: 10.2337/db11-1154


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Oliveira, Araújo, Carvalho, Rocha, Santos and Saad. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Jon Merlin1 , Masaaki Sato1 , Ling Yeong Chia1 , Richard Fahey1 , Mohsen Pakzad2 , Cameron J. Nowell1 , Roger J. Summers1 , Tore Bengtsson3 , Bronwyn A. Evans1 and Dana S. Hutchinson1,4\**

*1Drug Discovery Biology, Monash Institute of Pharmaceutical Sciences, Monash University, Parkville, VIC, Australia, 2Department of Toxicology and Pharmacology, Faculty of Pharmacy, Pharmaceutical Sciences Research Center, Tehran University of Medical Sciences, Tehran, Iran, 3Department of Molecular Biosciences, The Wenner-Gren Institute, Stockholm University, Stockholm, Sweden, 4Department of Pharmacology, Monash University, Clayton, VIC, Australia*

#### *Edited by:*

*Sudip Bajpeyi, The University of Texas at El Paso, United States*

#### *Reviewed by:*

*Daniel John Fazakerley, University of Sydney, Australia Sarah M. Turpin-Nolan, Max-Planck-Institut für Stoffwechselforschung, Germany*

*\*Correspondence: Dana S. Hutchinson dana.hutchinson@monash.edu*

#### *Specialty section:*

*This article was submitted to Obesity, a section of the journal Frontiers in Endocrinology*

> *Received: 31 January 2018 Accepted: 02 May 2018 Published: 30 May 2018*

#### *Citation:*

*Merlin J, Sato M, Chia LY, Fahey R, Pakzad M, Nowell CJ, Summers RJ, Bengtsson T, Evans BA and Hutchinson DS (2018) Rosiglitazone and a β3-Adrenoceptor Agonist Are Both Required for Functional Browning of White Adipocytes in Culture. Front. Endocrinol. 9:249. doi: 10.3389/fendo.2018.00249*

The recruitment of brite (or beige) adipocytes has been advocated as a means to combat obesity, due to their ability to phenotypically resemble brown adipocytes (BA). Lineage studies indicate that brite adipocytes are formed by differentiation of precursor cells or by direct conversion of existing white adipocytes, depending on the adipose depot examined. We have systematically compared the gene expression profile and a functional output (oxygen consumption) in mouse adipocytes cultured from two contrasting depots, namely interscapular brown adipose tissue, and inguinal white adipose tissue (iWAT), following treatment with a known browning agent, the peroxisome proliferator-activated receptor (PPARγ) activator rosiglitazone. Prototypical BA readily express uncoupling protein (UCP)1, and upstream regulators including the β3-adrenoceptor and transcription factors involved in energy homeostasis. Adipocytes from inguinal WAT display maximal UCP1 expression and mitochondrial uncoupling only when treated with a combination of the PPARγ activator rosiglitazone and a β3-adrenoceptor agonist. In conclusion, brite adipocytes are fully activated only when a browning agent (rosiglitazone) and a thermogenic agent (β3-adrenoceptor agonist) are added in combination. The presence of rosiglitazone throughout the 7-day culture period partially masks the effects of β3-adrenoceptor signaling in inguinal white adipocyte cultures, whereas including rosiglitazone only for the first 3 days promotes robust β3-adrenoceptor expression and provides an improved window for detection of β3-adrenoceptor responses.

Keywords: adipocyte, beta adrenergic receptors, uncoupling protein 1, adrenoceptor, seahorse xf96 analysis, rosiglitazone, CL316243

# INTRODUCTION

The remarkable prevalence of obesity worldwide has sparked considerable interest in therapeutic strategies that are effective and safe in promoting weight loss. In conjunction with lifestyle modification, such strategies range from surgical intervention to reduce stomach capacity, through to pharmaceutical interventions aimed primarily at modulating neural pathways that affect food/

caloric intake. There are no pharmacological interventions that primarily act on adipose tissue, despite its obvious role in obesity. There are two types of adipose tissue with distinct functions: white adipose tissue (WAT) that stores chemical energy in the form of triacylglycerol, and brown adipose tissue (BAT) that releases chemical energy in the form of heat (thermogenesis).

Classical BAT depots have been studied extensively in adult rodents. They are highly innervated and are activated by centers in the brain responsive to cold exposure, leading to the release of norepinephrine (NE) from sympathetic nerves. Upon binding of NE to BAT β3-adrenoceptors (β3-ARs), increased levels of intracellular cyclic AMP (cAMP) promote lipolysis, and this breakdown of triglycerides leads to release of free fatty acids that upregulate and activate uncoupling protein 1 (UCP1). Activated UCP1 uncouples mitochondrial respiration leading to heat generation, thus β3-AR signaling increases respiration and nonshivering thermogenesis, with prototypical BAT adipocytes being remarkably rich in mitochondria (1).

β3-AR signaling in response to NE or synthetic receptor agonists occurs almost exclusively in peripheral tissues, including adipose tissue/adipocytes, bladder, and gastrointestinal tract. While there is detectable expression in the rodent brain (2), and central administration of β3-AR agonists directly into hypothalamic/third ventricle regions can reduce food intake and weight (3, 4) and increase c-fos immunoreactivity (5), it is unlikely that peripheral administration of β3-AR agonists such as CL316243 and mirabegron would have direct central effects, since they do not readily cross the blood–brain barrier (6, 7). Any effects that peripheral administration of β3-AR agonists has centrally are thought to occur indirectly through fatty acids liberated from lipolysis in peripheral adipose tissues.

In addition to prototypical white or brown adipocytes (BA), beige or brite adipocytes have been described (8, 9). These cells reside in WAT depots but can be "browned" by various stimuli, most notably cold exposure or activation of β-AR signaling, and by the peroxisome proliferator-activated receptor (PPARγ) agonist rosiglitazone. Activation of brite/beige adipocytes leads to an increase in mitochondrial uncoupling similar to that occurring in BAT. Two studies indicate that brite or beige adipocytes contribute significantly to whole body energy expenditure: mouse models that have increased beige/brite adipocytes in WAT are protected from diet-induced obesity (10), and browning of WAT contributes to non-shivering adaptive thermogenesis in the absence of classical BA (11). Adult humans possess both brown and beige/brite adipose tissue. Humans were thought to lose BAT after infancy, but phenotypically beige/brite adipocytes have been isolated from human supraclavicular fat depots and neck biopsies (9, 12, 13).

There is considerable interest in identifying additional agents that promote browning of adipose tissue, as increased expenditure of energy as heat would be of therapeutic utility in obesity and type 2 diabetes. To date, agents and processes with browning potential fall into a number of classes, namely (i) cold exposure, activation of the sympathetic nervous system (SNS), and β3-AR agonists; (ii) G protein-coupled receptor (GPCR), ion channel, and signaling pathway modulators; (iii) exercise and associated factors; (iv) growth factors and cytokines; (v) nutritional and dietary factors; and (vi) PPAR agonists. The evidence for these as browning agents has been reviewed elsewhere (14).

Many studies of browning agents have employed a combination of cultured adipocytes and whole animal experiments. Clearly *in vivo* experiments have the advantage that test compounds are acting on cell populations as they exist in whole animals, thereby providing valid information on predicted clinical efficacy. It is important to understand the mechanism of action, however, particularly in relation to the precise cells targeted by browning agents. In the whole animal, such agents could be acting directly on adipocytes, but it is equally possible that they are targeting the central nervous system (15), or indirectly the sympathetic nervous system (16).

Cultured adipocytes thus offer a system for characterizing the direct effect of browning agents, and also have advantages in facilitating high-throughput screening of compounds. The ideal model system would be cultured human adipocytes with the potential to undergo browning, however, there have been difficulties in using human cultures or cell lines—(i) even in human subjects with highly inducible BAT (17), beige/brite adipocytes are localized to the neck and supraclavicular regions, and presumably arise from specialized cells within these adipose depots, and (ii) human primary cultures or immortalized lines such as SGBS cells require strongly adipogenic media in order to differentiate, including, for example, rosiglitazone, dexamethasone, 3-isobutyl-1-methylxanthine (IBMX), cortisol, transferrin, triiodothyronine, and insulin (9, 18). In particular, the inclusion of rosiglitazone and IBMX (to increase cAMP), is highly likely to promote browning in conjunction with differentiation.

The primary cultured adipocytes most often utilized to study browning are differentiated from the stromal vascular fraction (SVF) of mouse inguinal WAT (iWAT) depots. Again these cultures generally include rosiglitazone at least for the first 2–4 days of culture (19–23), thus the mature adipocytes are likely to have undergone browning as well as differentiation. The aim of our study was to systematically clarify the effect of rosiglitazone on cultured adipocytes in the presence or absence of recognized browning agents targeting the β3-AR. We have examined adipocytes isolated from FVB/N mouse interscapular brown and inguinal white depots and cultured in a minimal medium consisting of DMEM supplemented with 10% newborn calf serum, 4.5 g/l glucose, and 2.4 nM insulin (8). We tested the effect of 1 µM rosiglitazone added to the culture medium for the entire 7 days or for the first 3 days only. The adipocyte cultures were treated for a further 24 h with CL316243 (in the absence of rosiglitazone), as a recognized browning agent. We find that BA cultures differentiate well even in the absence of rosiglitazone, whereas inguinal white adipocytes (iWA) require rosiglitazone for at least the first 3 days of culture. Substantial browning occurs only after 7-day rosiglitazone treatment in iWA, though induction of UCP1 and the thermogenic gene Cpt1b can be induced by CL316243 after 3 days of rosiglitazone. The highest levels of UCP1 mRNA occur following 7-day rosiglitazone combined with CL316243 treatment, and the vast majority of BA and iWA cells become positive for UCP1 immunostaining under these conditions.

# MATERIALS AND METHODS

#### Ethical Statement

All experiments were conducted with ethical permission from the Monash University Animal Ethics Committee, ethics approval numbers MIPS.2015.14 and VCP.2009.22, which complied with the National Health and Medical Research Council of Australia (NHMRC) guidelines for use of animals in scientific research.

# Adipocyte Culture

Adipocyte isolation and culturing was performed as described previously (24). Inbred FVB/N mice (3–4 weeks of age, either sex) were bred at the Monash University Parkville animal facility. Mice were killed by CO2 inhalation and BAT isolated from the interscapular, cervical, and axillary depots, while WAT was isolated from the subcutaneous depots. Pooled tissue pieces were finely minced in DMEM and transferred to a digestion solution [0.2% (wt/vol) collagenase type II, 0.1 M HEPES (pH 7.4), 123 mM NaCl, 5 mM KCl, 1 mM CaCl2, 4.5 mM glucose, 1.5% (wt/vol) BSA]. Tissues were digested for 30 min at 37°C with continuous mixing. Cells were filtered through a 250 µm nylon mesh filter into sterile tubes and kept on ice for 15 min whereupon the mature adipocytes float to the top. The top layer of the suspension was removed and the remaining cell suspension filtered through a 25-µm nylon mesh filter and centrifuged (700 × *g*, 10 min). The pellet containing preadipocytes was resuspended in DMEM and centrifuged (700 × *g*, 10 min). The pellet was suspended in culture medium (6 ml/animal for WAT, 4.8 ml/animal for BAT) and plated in either Seahorse cell culture plates (100 µl/well), 8-well chamber slides (200 µl/well), or 6-well plates (2 ml/well). The culture medium consisted of DMEM containing 4.5 g/l glucose, 10% (vol/vol) newborn calf serum, 2.4 nM insulin, 25 µg/ml sodium ascorbate, 10 mM HEPES, 4 mM l-glutamine, 50 U/ml penicillin, and 50 µg/ml streptomycin, and supplemented where indicated with 1 µM rosiglitazone. Adipocytes were grown at 37°C in 8% CO2. Cells were washed in pre-warmed DMEM and medium renewed on day 1, then every second day. All experiments were conducted on day 7. For experiments treated for 7 days with 1 µM rosiglitazone (7-day Rosi), rosiglitazone was included in control media from day 1 to 7 (the day of experiment). In some experiments as indicated, 1 µM rosiglitazone was only included in the culture media from day 1 to 3, whereupon the cells were cultured in the absence of rosiglitazone until use on day 7.

# Reverse Transcription-qPCR

Cells were serum starved on day 6 in DMEM/Nutrient Mix F-12 (1:1) with 4 mM l-glutamine, 0.5% BSA, 2.4 nM insulin, 10 mM Hepes, 50 IU/ml penicillin, 50 µg/ml streptomycin, and 50 µg/ml sodium ascorbate, with rosiglitazone (1 µM) as indicated. Media was replaced with DMEM containing 4.5 g/l glucose, 0.5% BSA, 25 µg/ml sodium ascorbate, 10 mM HEPES, 4 mM l-glutamine, 50 U/ml penicillin, and 50 µg/ml streptomycin, and supplemented with norepinephrine (1 µM), or CL316243 (1 µM) as indicated. Media was aspirated, the cells washed in warmed PBS, and plates rapidly frozen at −80°C until use. Total RNA was extracted using RNeasy Plus Mini Kits (QIAGEN), as per the manufacturer's instructions (samples in **Figure 5** were extracted using TriReagent (Sigma-Aldrich), according to the manufacturer's instructions). RNA samples were DNAse treated using DNA*-free* DNA Removal Kit (Invitrogen), according to manufacturer's instructions. Where tissues were used, interscapular BAT or inguinal WAT were excised from similarly aged (3- to 4-week-old FVB/N mice of either sex) and housed mice (room temperature 22°C), and tissues rapidly frozen (−80°C). RNA from tissues was extracted using TriReagent according to manufacturer's instructions (Sigma-Aldrich).

For preparation of cDNA, 0.5 µg of RNA was reversetranscribed using iScript Reverse Transcription Supermix for RT-qPCR (Bio-Rad), in a total volume of 10 µl. For each independent sample, qPCR was performed in duplicate using TaqMan Gene Expression assays (Life Technologies) for Prdm16, Hoxc9, Ppara, Ppargc1a, Pparg, Ppargc1b, Fabp4, Adipoq, Slc27a1, Fasn, Pck1, Acaca, Fabp3, Acox1, Sirt3, Cpt1b, Cox4i1, Prdx3 Acadl, Vdac1, Pdk4, Slc2a1, Slc2a4, Hk2, Pfkm, Gapdh, Pgk1, Ucp1, Adrb3, and the reference gene Actb. The cDNA was diluted to the equivalent of 2.5 ng/µl of starting RNA and 4 µl added to 6 µl reaction mix comprising 1× TaqMan Gene Expression Assay and 1× TaqMan Fast Advanced Master Mix dispensed in 96-well plate, as per manufacturer's instructions. After initial denaturation at 95°C for 30 s, fluorescence was detected over 40 cycles (95°C for 5 s, 60°C for 30 s). Cq values were automatically calculated by the Realplex analysis module. qPCR presented in **Figure 5** was performed on a CFX Connect™ real-time PCR detection system (BioRad) and samples were initially denatured at 50°C for 2 min, 95°C for 10 min, and fluorescence detected over 40 cycles (95°C for 15 s, 60°C for 1 min), and Cq values automatically calculated by the BioRad analysis module. All data are expressed as expression of the gene of interest relative to Actb, calculated as ( ) 2 1\* ,000 −∆Cq . Multiplication of all values by 1,000 does not change the relative expression levels and was done for two reasons; (i) it facilitates viewing of the data because even values for poorly expressed genes are greater than 1.0, and (ii) it positions expression values in the same range as RPKM or FPKM values obtained in RNA-Seq studies, as the average value for Actb in numerous cell and tissue types is 1,000. All statistics for gene expression were performed on ΔCq values, as these data are normally distributed. For the purpose of statistics, where genes were not detected within 40 cycles (and were, therefore, not detectably expressed), an over-assumption Cq value of 40 was used.

# Custom PCR Array

We designed a custom mouse qPCR array in 384-well plates that comprised 2 sets of 192 genes, including Actb as a reference gene, a genomic DNA control (Lonza) and 190 genes representing a broad cross-section of targets downstream of 25 different transcription factors (**Table 1**). Each array plate was used to analyze expression in a control and rosiglitazone-treated adipocyte culture from BA or iWA. cDNA was prepared as described above, then 4 µl equivalent to 10 ng of starting RNA was added to each well containing 1× SYBR Green PCR Master Mix (Applied Biosystems) and Lonza-dispensed primer sets in a total volume of 10 µl. qPCR reactions were carried out at the Australian Genome Research Facility (Parkville, VIC, USA) on a 7900HT Real-Time PCR System (Applied Biosystems).

#### Table 1 | Categorization of genes assessed by the StellARray custom array system.


#### Ingenuity Pathway Analysis (IPA)

Patterns of gene expression were examined using IPA software. For each of the adipocyte datasets, we uploaded fold-change between rosiglitazone treated and control cultures, along with the final expression value in rosiglitazone-treated cultures and *P* values for significant differential expression. The datasets were filtered for fold-changes of at least 2 in either direction, expression values normalized to Actb (as described above) of at least 1.0 in at least one culture condition, and *P*-values <0.05 (Student's *t*-test), leaving 171 genes. IPA core analysis was used to determine enrichment of molecular and cellular functions or upstream regulators predicted from observed patterns of up or down-regulation among the expressed genes. IPA uses Fisher's Exact Test to test the hypothesis that patterns of gene expression related to functions or to upstream regulators are not due to chance. This analysis provides *P* values that signify the degree of enrichment of differentially expressed genes, and activation *z*-scores that take into account the direction of change compared to effects predicted from the IPA knowledge base (25). A *z*-score of at least 2.0 is considered significant.

#### External Data

We reanalyzed publically available RNA sequencing data obtained from Hao et al. (26) (upstream analysis shown in **Figure 6**). In this published study 8-week-old male C57BL/6J mice were housed at thermoneutrality (28–30°C) on a 12:12 h light–dark cycle and fed standard chow diet *ad libitum*. After 8 days, mice were either retained at thermoneutrality or transferred to 4°C for 2 or 4 days. Mice were killed by cervical dislocation for isolation of total RNA from interscapular BAT and inguinal WAT. Data consisting of tag counts per million were generated by digital gene expression profiling, mapped to NCBI RefSeq mRNAs, and deposited in Gene Expression Omnibus (accession number GSE63031). We uploaded the 0- and 2-day cold exposure datasets to IPA to compare upstream regulators between BAT and iWAT, as described in the legend to **Figure 6**. The mouse experiments described by Hao et al. (26) were approved by the Norwegian or Danish Animal Research Authority.

#### Immunocytochemistry for Detection of UCP-1

Cells were seeded in 8-well culture chamber slides (BD Biosciences, Franklin Lakes, BJ). On day 7, cells were fixed with 4% formaldehyde in PBS for 15 min, and quenched with 150 mM Tris pH 8.0 for 10 min. Cells were permeablized with 0.1% Triton X-100 for 10 min, blocked for 1 h at room temperature (5% BSA in PBS), and incubated with UCP1 (Abcam) primary antibody solution (diluted 1:1,000 in 1% BSA in PBS) overnight at 4°C. Alexa Fluor 488-conjugated goat anti-rabbit IgG (1:1,000 dilution in 1% BSA in PBS) was added and incubated at room temperature for 2 h. Images were observed on a Leica DMLB epifluorescence microscope. Images were acquired using a DC350F camera with IM500 software (Leica Microsystems AB; Kista, Sweden). Quantification was performed using Fiji (ImageJ version 1.50 g) (27), using a script developed by C.J. Nowell. Cells identified by DAPI staining in images were judged to be negative/positive for the protein of interest (UCP-1) in a blinded manner and counted using ImageJ software, performed by J. Merlin. No distinction was made in the relative intensities of staining within images.

#### Measurement of Oxygen Consumption Rates

Oxygen consumption rates (OCR) were measured using the Seahorse xF96 (Seahorse Bioscience). On day 7, adipocytes were washed twice in XF assay media (Seahorse Bioscience) supplemented with 25 mM glucose, 0.5 mM sodium pyruvate, 2 mM l-glutamine and 1% fatty free BSA, and 160 µl added/ well. OCR was measured as described in detail (28) with some modifications. 6 baseline rate measurements were made using a 2 min mix, 5 min measure cycle. After 6 basal measurements, oligomycin A (5 µM) was added for 6 measurements, followed by a combination of antimycin A (1 µM) and rotenone (0.1 µM) for 6 measurements. OCR rates immediately prior to oligomycin A injection at measurement number 6 were used as the basal rates and defined as 100%.

#### Statistical Analysis

All data are expressed as mean ± SEM of *n*. Statistical significance was determined by Student's *t*-test, multiple comparisons one-way ANOVA with Tukey's test, or multiple comparisons Kruskal–Wallis test (for non-parametric analyses), as indicated in Figure legends. *P* < 0.05 was considered significant.

#### RESULTS

### Rosiglitazone Treatment of Cultures From Inguinal WAT Increases Differentiation and the Expression of Genes Related to Thermogenesis

7-day rosiglitazone treatment (1 µM) promoted expression of genes associated with both maturation of white adipocytes, and thermogenic re-programming (**Figure 1**; **Table 2**). While BA cultures differentiate in the absence of rosiglitazone, displaying robust expression of the mature adipocyte marker Fabp4 (**Figure 1B**), iWA cultures express lower levels of Fabp4 (36% relative to BA), indicating that a much lower proportion of cells undergo differentiation under control conditions (in DMEM containing 4.5 g/l glucose, 10% (vol/vol) newborn calf serum, and 2.4 nM insulin). However, following 7d of rosiglitazone treatment, Fabp4 mRNA levels are similar to that in iWAT tissues, and also to BA tissue and cultures (**Figure 1**; **Table 2**). Likewise, the genes Acaca (acetyl CoA carboxylase), Fasn (fatty acid synthase), Slc27a1 (FATP1), Pck1 (phosphoenolpyruvate carboxykinase), and Adipoq (adiponectin), all characteristic of mature adipocytes, are expressed in control and treated BAs, whereas all of these genes are low in control iWA, and significantly increased by rosiglitazone (**Figure 1B**). Conversely, the white adipocyte marker Hoxc9 is expressed to some extent in both control and rosiglitazone-treated iWA cultures but is negligible in BA under any conditions (**Figure 1**).

The BA transcriptional regulator Prdm16 is expressed in both control and rosiglitazone-treated BA cultures, and is substantially increased in iWA cultures treated with rosiglitazone (**Figure 1A**), suggesting browning of iWA under 7-day rosiglitazone treatment as we have shown previously (29). Pparg, encoding the rosiglitazone-targeted nuclear receptor PPARγ, a major regulator of adipocyte differentiation, is expressed in control white adipocyte cultures but also significantly increased by rosiglitazone in treated iWA. These findings indicate that control iWA cultures contain few fully differentiated adipocytes, but a significant population of Hoxc9-positive preadipocytes (**Figure 1A**). Upon rosiglitazone treatment, levels of Fabp4, Adipoq, Fasn, Slc27a1, and Acaca expression in iWA equal or exceed those found in treated BA cultures, indicating a high degree of differentiation.

**Figure 1D** shows genes involved in fatty acid metabolism and mitochondrial function. In particular, Cpt1b (carnitine palmitoyltransferase 1B) is the rate-limiting step for fatty acid oxidation and consequent stimulation of mitochondrial respiration, while Fabp3 is also required for efficient fatty acid oxidation (30). Cpt1b mRNA is 65 times higher in control BA than in iWA cultures, and in BA, undergoes 6.4-fold induction in the presence of rosiglitazone. iWA display a striking 455-fold induction, reaching levels similar to those in BA. Similarly, Fabp3 is induced by 630-fold in iWA cultures following rosiglitazone treatment. Additional genes related to thermogenesis (Acadl, Vdac1, Acox1, Sirt3, Pdk4, Prdx3, Cox4i1) display significantly higher expression in rosiglitazone-treated iWA than in control cultures. Ppara (PPARα), which promotes expression of genes essential for many aspects of fatty acid metabolism, and Ppargc1a (PGC-1α), a master regulator of adipocyte browning and mitochondrial biogenesis, are induced by 77- and 83-fold, respectively, in adipocytes from iWAT following rosiglitazone treatment (**Figure 1A**).

Analysis of genes related to glucose handling (**Figure 1C**) showed significant increases in the insulin responsive GLUT transporter Slc2a4 (GLUT4), and glucose metabolism enzymes Hk2, Pfkm, and Gapdh, in iWA cultures treated with rosiglitazone, further indicating the differentiation of iWA cultures toward an insulin-sensitive mature adipocyte population. Rosiglitazone treatment also increased the expression of Slc2a4, Gapdh, and Pgk1 in BAs, likewise suggesting changes in insulinmediated glucose handling, but also a significant increase in Slc2a1 (GLUT1), which has been shown to be the major GLUT involved in β3-AR-mediated glucose uptake in BA (31).

In parallel with individual qPCR assays, we obtained a comprehensive view of rosiglitazone-induced changes in cultured mouse adipocytes by analyzing mRNA levels of 160 expressed genes quantified using a custom qPCR array (categorized in **Table 1**). Products of these genes participate in a broad cross-section of molecular and cellular functions, and represent targets of diverse upstream transcriptional regulators. We assessed the expression of 4 genes from both the qPCR array and TaqMan assays (Cpt1b, Fabp3, Ppargc1a, and Sirt3). Because expression is normalized to the reference gene Actb, the two methods give closely corresponding values for a given sample, showing less inter-assay variation than the inter-sample differences seen using either method. Genes displaying greater than twofold change of expression in either direction (*P* < 0.05) were subject to enrichment analysis using IPA software (Qiagen). The significantly enriched molecular and cellular functions are similar between BA and iWA (**Figure 2A**). The top score is for molecular transport (which encompasses a wide array of cellular processes), but the remaining significant functions are related largely to energy homeostasis, fatty acid metabolism, and ATP generation. **Figure 2B** shows the upregulated genes expected to contribute to these functions in rosiglitazone-treated iWAs.

As shown in **Figure 2A**, there are four functions that show an apparent difference between BA and iWA. Increases in the "transport of lipid," "metabolism of nucleic acid component," and "transport of carbohydrate" functions are significant in only one of iWA or BA, but in each of these three cases the number of genes showing a direction of expression change consistent with increased function is small. For example, significantly increased transport of carbohydrate is predicted in BA, based on enhanced expression of 10 genes positively linked to this function (Hk2, Ppargc1a, Ppargc1b, Ptgs2, Slc2a1, Slc2a4, Sod1, Tnf, Trib3, and UCP1). In iWA, only seven of these genes display increased expression, so in these cells transport of carbohydrate is no longer predicted to be significantly affected. This finding is not borne out by the functional phenotype of iWA, as rosiglitazone treatment

Table 2 | Comparison of gene expression in interscapular brown (BA) and inguinal white adipocytes (iWA) cultured for 7 days in the presence or absence of rosiglitazone (Rosi, 1 µM, 7 days), and in freshly isolated brown adipose tissue (BAT) and iWAT from mice housed at 22°C.


*a Data for Ucp1 and Adrb3 expression in control and rosiglitazone-treated adipocytes obtained from Merlin et al. (29). All data are expressed relative to Actb expression\*1,000* ± *SEM of three independent experiments (as per Figure 1).*

significantly increases glucose uptake in response to norepinephrine and CL316243 (29). This reflects our observation that the limiting step in agonist-stimulated glucose uptake is the abundance of the β3-AR, which is markedly enhanced when iWAs are cultured in the presence of rosiglitazone [**Table 2**; (29)]. Interestingly, the function "synthesis of reactive oxygen species" is predicted to be decreased in iWA (*z*-score of −2.2, based on 23 differentially expressed genes), while this function is not significantly affected by rosiglitazone treatment in BA (*z*-score of +0.27, 16 genes). Genes displaying changes in expression expected to enhance or inhibit synthesis of reactive oxygen species (ROS) are shown in **Figure 2C** (iWA) and **Figure 2D** (BA). The overlapping genes all display the same direction of change in expression in iWA and BA, but in iWA cultures there are 15 genes for which up- or down-regulation is expected to reduce ROS. The induction of genes encoding anti-oxidant enzymes, including superoxide dismutase SOD1 and SOD2 in iWAT also occurs *in vivo* following rosiglitazone treatment (32).

# Brite Adipocytes Display Increased UCP1 Protein and Uncoupling Capacity in Response to 24 h Treatment With the **β**3-AR Agonist CL316243

As chronic β-AR treatment of BA upregulates UCP-1 mRNA and protein levels (1), we treated BA and iWA cultures in the presence/

absence of 7-d rosiglitazone (1 µM) with norepinephrine (1 µM, 24 h) or CL316243 (1 µM, 24 h). Norepinephrine and CL316243 caused a small increase in UCP1 mRNA expression (**Figure 3B**) and immunoreactivity (**Figures 3A,C**) in control BA, but this response was significantly amplified in rosiglitazone-treated BA cultures. The effect of combined rosiglitazone and β-AR agonist treatment was even more evident in iWA cultures. These data are consistent with the presence or absence of β3-ARs in these cultures (29). There were significant changes in the proportion of UCP1-positive cells between all control cultures and those treated with rosiglitazone alone (**Figures 3A,C**). Following an additional 24 h in the presence of CL316243, there were also significant differences in UCP1-positive cells between control and rosiglitazone-treated iWA cultures (**Figure 3C**).

In order to investigate whether the increase in UCP1 mRNA/ protein levels following rosiglitazone, norepinephrine, or CL316243 treatment affected mitochondrial uncoupling in white adipocytes, we measured OCR in the presence of the ATP synthase inhibitor oligomycin (5 µM), and in the presence of rotenone (0.1 µM) and antimycin A (1 µM), that define non-mitochondrial sources of OCR. In control iWA cells, oligomycin inhibited OCR by ~50% (**Figure 4**). In iWA cells treated 7-day with rosiglitazone (1 µM), oligomycin inhibited OCR by only 34%. The insensitivity to oligomycin was enhanced when rosiglitazone treated cells were also treated with either NE (1 µM, 2 and 24 h) or CL316243 (1 µM, 2 and 24 h). This suggests an increase in uncoupled respiration. This was further assessed by treating the cells with a combination of rotenone (0.1 µM) and antimycin A (1 µM), that define non-mitochondrial sources of OCR (~36% of OCR was due to non-mitochondrial sources; **Figure 4**). After correction for non-mitochondrial OCR (indicated in dotted arrows in **Figures 4C,D**, and illustrated as % OCR due to proton leak in **Figures 4E,F**), we can see that mitochondrial uncoupling (proton leak) accounts for 14% of OCR in control iWA cells, 27% in rosiglitazone-treated cells, and this level rises to 40–50% in rosiglitazone-treated iWA cells further treated with NE or CL316243. These results suggest that rosiglitazone-induced brite adipocytes possess a greater mitochondrial uncoupling capacity, and that β3-AR treatment of brite adipocytes significantly increases this capacity.

# The Effects of Rosiglitazone on Adipocyte Differentiation, and the Thermogenic Potential of iWA, Are Separable Following 3-Day Treatment

Our studies indicate that 7-d rosiglitazone treatment promotes both differentiation and browning of iWA cultures. It is common practice to induce adipocyte differentiation using an adipogenic cocktail, generally containing IBMX, dexamethasone, insulin, high glucose, and often 1 µM rosiglitazone, particularly in studies involving human adipocytes. However, the presence of rosiglitazone, and perhaps IBMX that increases cellular cAMP and thus mimics the effect of β-AR agonists, are confounding factors in any attempt to attribute browning capacity to agents being tested using cultures derived from the stromal vascular fraction (14). Several published studies have included rosiglitazone in mouse adipocyte cultures only for the first 2–4 days before removing rosiglitazone

Figure 2 | Molecular and cellular functions of genes differentially expressed in control or rosiglitazone-treated adipocytes. Ingenuity Pathway Analysis (IPA) software was used to compare downstream functions consistent with observed patterns of target gene up- or down-regulation in adipocyte cultures treated with rosiglitazone relative to vehicle (A). Functions are ranked according to the sum of activation *z*-scores across brown adipocyte (BA) and inguinal white adipocytes (iWA) cultures, irrespective of positive or negative direction. *Z*-scores ≥|2| describe statistically significant matches between known functions and observed patterns of gene regulation, and also whether functions are predicted to be up- or down-regulated downstream of the observed changes (activated shown in orange, inhibited shown in blue) (25). Network of genes upregulated in iWAT that contribute to key non-redundant cellular functions (oxidation of fatty acid, concentration of ATP and energy homeostasis) (B). Differentially expressed genes identified by IPA that contribute to synthesis of reactive oxygen species (ROS) in iWA (C) and BA (D). In both panels, coloring of genes red or pink denotes upregulation and green denotes downregulation. In panel (C), blue lines signify genes showing a change in expression consistent with reduced ROS synthesis, while yellow lines signify genes that are not consistent with this function. The *z*-score for iWA is −2.2, below the significance threshold of |2|, whereas that for BA is +0.27 and not significant. iWA cultures treated with rosiglitazone display fourfold increases in expression of the key anti-oxidant genes Sod1 and Sod2 relative to control cultures.

from the media for the remainder of the differentiation protocol, instead of including rosiglitazone for the entire differentiation period [e.g., Ref. (19, 21–23, 33)]. We therefore examined changes in gene expression in both BA and iWA cultures treated with rosiglitazone for the entire 7 days compared to the first 3 days only (**Figure 5**). In iWA cultures, 3-d rosiglitazone promoted increases in Adrb3 mRNA that were 50% of those seen after the full 7-d treatment. 1–3 day rosiglitazone did not significantly alter expression of UCP1, Fabp3, Fabp4, or Cpt1b, in contrast to the substantial increases seen in 7d-treated cultures. We also tested the effect of CL316243 (1 µM, 24 h) on cultures differentiated with rosiglitazone for 3 or 7 days. CL316243 increased the expression of Ucp1 in 1–3 day cultures, whereas it had no effect on control cultures, indicating that 3 days does provide a suitable window to test for browning capacity. After 7 days in the presence of rosiglitazone, CL316243 treatment for 24 h did increase UCP1 mRNA further (although not significant in this cohort of adipocytes due to high variation observed), and also seen in **Figure 3**, however, there was no increase in Fabp3 or Cpt1b expression associated with 24 h CL316243. Thus longer exposure to rosiglitazone essentially masks some of the effects of CL316243.

# Modulation of Gene Expression in BA and iWA Cultured From the SVF Compared to *In Vivo* Adipose Depots

Our gene expression data and previous functional studies (29) indicate that although BA and iWA cultures are derived from adipocyte precursors in the SVF, the mature adipocytes retain distinct properties according to their site of origin. It has been shown recently that *in vivo* treatment of mice with rosiglitazone induces a population of mature UCP1-positive inguinal adipocytes that have a distinct profile compared to populations induced by *in vivo* treatment with CL316243 (34). This differs from our findings based on adipocytes derived from the SVF, as in our cultures prior differentiation with rosiglitazone was required in order for cells to induce expression of the β3-AR and thereby become responsive to CL316243. This suggests that the mature adipocytes residing in iWAT depots differ from those derived from differentiation of the SVF. We compared the regulation of gene expression by rosiglitazone in our cultures with that demonstrated by transcriptome sequencing of RNA from BAT and iWAT subjected to sympathetic stimulation due to cold exposure of mice for 2 days (26). We also compared the expression of key adipocyte and thermogenic genes in our cultures with that in freshly isolated BAT and iWAT (**Table 2**). These tissues were isolated from mice subject to mild thermal stress by housing at 22°C, rather than at thermoneutrality (35, 36).

We used IPA to compare upstream regulators associated with all BA and iWA genes differentially expressed in the presence of rosiglitazone (twofold change in either direction, *P* < 0.05). As would be expected for cells treated with rosiglitazone, high-scoring upstream regulators across all cultures were closely associated with the PPARγ/PGC-1α network previously implicated in activating thermogenesis (**Figure 6A**). We also performed IPA upstream regulator analysis on the data from RNA sequencing of adipose depots from cold-exposed mice

Figure 4 | Effect of combined rosiglitazone and β-adrenoceptor agonist treatment on oligomycin-insensitive oxygen consumption in inguinal white adipocytes (iWA). Control or rosiglitazone (1 µM, 7 days) treated iWA were treated with (A,C) 1 µM norepinephrine (NE; 2 or 24 h) or (B,D) 1 µM CL316243 (CL; 2 or 24 h) prior to measurement of oxygen consumption rates (OCR). Cells were treated in the Seahorse XF96 with the ATP-synthase inhibitor oligomycin (5 µM) or a combination of the mitochondrial inhibitors 0.1 µM rotenone (Rote) and 1 µM antimycin A (Anti A). Data are mean ± SEM of 12–38 independent experiments performed in duplicate. Data in (A,B) are representative traces of two experiments performed in duplicate. Arrows indicate the addition of oligomycin (5 µM) or the combination of rotenone (0.1 µM) and antimycin A (1 µM), with basal OCR set to 100% before the addition of oligomycin at rate 6 to account for variations in the raw data between adipocyte cultures made on different days. (E,F) The relative changes in OCR between oligomycin (defining OCR due to ATP synthase) and rotenone/antimycin A (defining OCR due to non-mitochondrial sources) are expressed as % OCR due to proton leak [calculated from the results presented in (C,D) indicated with the dotted arrows]. Data are non-parametrically distributed and, therefore, statistically analyzed by non-parametrically analysis. \**P* < 0.05, \*\**P* < 0.01, \*\*\**P* < 0.001 indicate statistical significance from control cells, # *P* < 0.05 indicates statistical significance from rosiglitazone-treated cells, <sup>α</sup> *P* < 0.05 indicates statistical significance from adrenergic treatment alone (multiple comparisons Kruskal–Wallis test, one-way ANOVA).

[(26), **Figure 6B**]. In BAT and iWAT, seven of the highest scoring upstream regulators from the *in vivo* study corresponded to those seen in our rosiglitazone-treated cultures. This highlights the previously characterized thermogenic pathway involving activation of PPARγ by endogenous fatty acids released due to sympathetic stimulation of lipolysis (37). When we compared expression of key adipocyte and thermogenic genes in control and rosiglitazone-treated cultures with that in native AT depots, many showed similar levels in rosiglitazone-treated cultures and in BAT or iWAT. There are, however, striking differences in levels of UCP1, Acot11, and Pck1 mRNA, which each display over 10-fold higher expression in native BAT and iWAT than in the corresponding rosiglitazone-treated cultures (**Table 2**; **Figure 6**).

#### DISCUSSION

The presence of active BAT in humans (38–41) has reignited research into ways to promote increased thermogenesis in humans as a strategy to combat obesity and its related disorders such as type 2 diabetes. This is in part due to the overwhelming

Figure 5 | Effect of different lengths of time of rosiglitazone on gene expression in white and brown adipocytes (BA). Control or rosiglitazone [1 µM, 3 days (3d), or 7 days (7d)] treated brown or inguinal white adipocytes (iWA) were treated with 1 µM CL316243 (24 h) prior to measurement of several genes, including Ucp1 (A,B), Fabp4 (C,D), Adrb3 (E,F), Fabp3 (G,H), or Cpt1b (I,J). Data represent mean ± SEM of 6 (BA) or 7–8 (iWA) independent experiments, performed in duplicate, relative to β-actin (Actb) expression. \**P* < 0.05, \*\**P* < 0.01, \*\*\**P* < 0.001 indicate statistical significance between all treatments and the control cultures. # *P* < 0.05, ##*P* < 0.01, ###*P* < 0.001 indicate statistical significance between all treatments and the Rosi (3d) treated cultures. ∞*P* < 0.05, ∞∞*P* < 0.01, ∞∞∞*P* < 0.001 between all treatments and the Rosi (7 days) treated cultures. Data analyzed by Tukey's multiple comparisons of one-way ANOVA performed on the ΔCt values, which are normally distributed. One sample (iWA treated with rosiglitazone for 3 days) was excluded from all analysis due to poor integrity of its RNA. A single measurement of *Adrb3* for iWA treated with CL316243, and of Fabp4 for BA treated with Rosi 7 days, was excluded from the analysis due to failure of the qPCR reaction.

evidence that activation of β3-ARs leading to increased UCP1 expression and function in BAT can reverse obesity, and prevent the development of insulin resistance and diabetes in rodents (42–45). However, translation of these responses observed in rodents has not led to effective treatments in humans (46, 47). This may in part be due to human tissue comprising adipocytes that are more akin to mouse beige or brite adipocytes rather than classical BAs, based on gene expression profiling (9, 48). Thus research into beige/brite adipocytes, and how they can be induced by different stimuli [reviewed in Ref. (14)] may offer greater therapeutic potential. It is critical to this research that model systems comprising cultured adipocytes respond in the same way to browning agents as endogenous cells *in vivo*.

We have characterized FVB/N mouse adipocytes induced by rosiglitazone in culture, utilizing a method for *in vitro* differentiation (8) that in control cultures lacks potential brite-inducing agents commonly used in adipocyte differentiation cocktails (such as triiodothyronine, IBMX, and rosiglitazone itself). We show that BA undergo differentiation in control cultures, but still display increased expression of UCP1 mRNA and protein in the presence of rosiglitazone. iWA cultures undergo both differentiation and browning due to inclusion of rosiglitazone. In BA cultures treated with rosiglitazone, 80% of cells become UCP1-positive, while this figure is 45% in iWA cultures (**Figure 3B**). These findings align with our observation that the key transcriptional regulators Prdm16 and PPARα are expressed at somewhat lower levels in iWA. Importantly, PPARγ and PGC-1α show comparable robust expression in rosiglitazone-treated BA and iWA (**Figure 1**).

## Cultured Adipocytes From iWAT Depots Require Priming by Rosiglitazone for Induction of UCP1 and Thermogenic Genes by the **β**3-AR Agonist CL316243

We showed previously that the majority of BA cells express β3-AR protein in control and rosiglitazone-treated cultures, whereas in iWA cultures the proportion of β3-AR-positive cells increase markedly in the presence of rosiglitazone (29), concomitant with

Figure 6 | Comparison analysis [ingenuity pathway analysis (IPA)] of transcriptional regulators governing gene expression in rosiglitazone-treated adipocytes and 2-day cold-exposed mice. IPA software was used to determine upstream regulators consistent with observed patterns of target gene up- or down-regulation in adipocyte cultures treated with rosiglitazone relative to vehicle-treated control cultures (A). Transcriptional regulators are ranked according to the sum of activation *z*-scores across brown adipocyte and inguinal white adipocytes (iWA) cultures, irrespective of positive or negative direction. *Z*-scores ≥|2| describe statistically significant matches between known regulators and observed patterns of up- and down-regulated genes, and also predict the activation state of each putative regulator (activated shown in orange, inhibited shown in blue) (25). As an example, the top transcriptional regulators in rosiglitazone-treated iWAs are PGC-1α (*z* = + 4.88, *P* value = 1.83E-46), PPARγ (*z* = + 4.35, *P* value = 5.89E-44), and PPARα (*z* = + 3.38, *P* value = 1.19E-32). These are predicted to be activated based on *z*-score, and are themselves upregulated in the treated cultures (83-, 3-, and 77-fold, respectively). *P* values (Fisher's Exact Test) describe the significance of overlap between the observed differentially expressed genes and all genes associated with a particular upstream regulator in the Ingenuity database, curated from published literature. As an additional comparison, we re-analyzed literature-derived RNA sequencing data for adipose depots from cold-exposed mice (26) (B). 8-week-old mice were housed at 28–30°C for 8 days, then one group maintained at thermoneutrality and a second group housed at 4°C for 2 days. Fold-change values were calculated for all genes expressed in iWAT and brown adipose tissue (BAT) of mice at 4°C compared to control mice at 28–30°C. These fold changes were analyzed using IPA and upstream regulators compared by *z*-score between BAT and iWAT. (C) Shows the predicted upstream regulators across the entire set of 171 genes found to be expressed in adipocyte cultures, ranked by log(*P*-value of overlap), demonstrating that the dataset is not biased toward the regulators predicted to govern differential expression in adipocyte cultures.

increased levels of β3-AR mRNA (**Figure 5**). This in turn facilitates the response to β3-AR agonists (29). This may represent a special case, however, it is equally likely that receptors targeted by other agents are similarly induced during adipocyte differentiation and/ or browning. In mice, the β3-AR is expressed in native BAT and iWAT at levels similar to those in rosiglitazone-treated cultures (**Table 2**), consistent with widespread findings that *in vivo* treatment with CL316243 can induce browning in iWAT depots (9, 35, 49). The utility of cultured adipocytes as a screening platform is clearly dependent on the presence of the target receptors.

Wu et al. (9) suggested that the inherent capacity for brite adipogenesis *in vivo* is independent of external factors, such as innervation, blood flow, oxygen supply, and nutrients. While differentiation of cultured adipocytes from the SVF does occur in a cell-autonomous manner, properties of the precursor cells appear to be specified by the environment from which they were derived, with a significant contribution from SNS innervation. *In vivo*, BAs possess a well-established ability to adapt to chronic β-AR activation; prolonged exposure to cold increases UCP1 expression in BAT (50, 51), which is inhibited by sympathetic denervation of BAT (37, 52). This upregulation in UCP1 occurs *via* β3-ARs and can be inhibited by β3-AR selective antagonists (53, 54). The requirement for intact sympathetic innervation has also been demonstrated in iWAT. For example, 3-week-old mice have high levels of UCP1 expression in iWAT, but this is markedly reduced by 8 weeks of age (35). At 8 weeks, mature adipocytes within iWAT re-express UCP1 in response to CL316243 stimulation, but this effect is abolished in mice undergoing prior surgical denervation of iWAT depots at 3 weeks. Thus sympathetic tone due to the mild cold-stress associated with housing mice at room temperature (55) maintains browning capacity in iWAT. We observed high UCP1 expression in native iWAT from animals housed at 22°C, reflecting this thermal stress (**Table 2**). Conversely, UCP1 mRNA is undetectable in the iWAT of mice housed at thermoneutrality [28–30°C (56)].

Importantly, reports of CL316243-induced browning have used mice housed at room temperature rather than at thermoneutrality (9, 35, 49). Previous studies have demonstrated an *in vitro* effect of CL316243 on UCP1 expression and function only in white adipocytes cultured in the presence of rosiglitazone or adipogenic cocktails (9, 49). Likewise, we found that while 24 h β3-AR stimulation (NE or CL316243) increased the proportion of UCP1-expressing cells in both control and treated BA cultures, neither agent alone induced significant browning of iWA cultures (**Figure 3**). These findings may recapitulate the requirement for intact SNS innervation of WAT seen *in vivo*, as sympathetic tone and consequent β-AR activation would be expected to drive lipolysis and subsequent activation of PPARγ by endogenous fatty acids. From our overall profiling of gene expression using the custom qPCR array, we examined predicted upstream regulators (IPA) based on genes differentially expressed following 7d rosiglitazone in cultured adipocytes, and these regulators were compared to those predicted in BAT and iWAT of mice subject to cold exposure for 2 days (26). In both cases, the top upstream regulators included PGC-1α, PPARγ, PPARα, PPARδ, CEBPα, ERRα (estrogen-related receptor α), and KLF15 (**Figure 6**), consistent with a common mechanism of induction that is dependent on PPARγ.

# Functional Demonstration of iWA Browning

The therapeutic potential of adipocyte browning is contingent upon the dissipation of energy through mitochondrial uncoupling becoming dominant over energy storage as triglycerides. We previously have shown that increased UCP1 function in brite adipocytes (defined as the percentage of OCR due to proton leak) is increased in brite adipocytes as compared to white adipocytes (29), which correlates with an increase in the expression of several genes involved in energy production (**Figure 1**). To assess whether combined treatment of iWA cells with rosiglitazone and a β-AR agonist (2 or 24 h treatment with either NE or CL316243) could further increase UCP1 function, we analyzed our OCR data, as described by Collins and colleagues (57), by defining the mitochondrial capacity that is not coupled to ATP synthesis. Rosiglitazone treatment of iWA induces a significant increase in this non-coupled spare capacity at the level of mitochondria, which is further increased in iWA also treated with either NE or CL316243 (**Figure 4**). These results indicate that the combination of both a browning agent (rosiglitazone) and a β3-AR agonist are required for maximal effects on mitochondrial function in brite adipocytes. This demonstrates that inguinal-derived brite adipocytes are geared toward mitochondrial uncoupling (and energy expenditure), consistent with a previous report that brite adipocytes are thermogenically competent, in that their mitochondria are functional and able to uncouple *via* UCP1 (58).

UCP1 is activated upon release of fatty acids generated following adrenergic stimulation of lipolysis, however, a recent study has also demonstrated that UCP1 activity is sensitized in the presence of ROS due to sulfenylation at Cys253 (59). *In vivo*, increased ROS due to deletion of Sod2 (encoding Mn-superoxide dismutase) leads to enhanced expression of key fatty acid oxidation genes in iWAT, and elevated mitochondrial oxygen consumption (60). Furthermore, UCP1 plays a central role in protecting adipocytes from mitochondrial dysfunction in the presence of high ROS and calcium overload (61). Thus, UCP1 is both activated by and protects against the effects of increased mitochondrial ROS. Our gene expression profiling and IPA indicates that rosiglitazone treatment may have dual protective actions in iWA, by increasing expression of UCP1 and at the same time inducing Sod1, Sod2, and other genes implicated in reduced ROS synthesis (**Figure 2C**).

# Culture Conditions That Promote Adipogenesis but Not Browning

Another issue raised by our study is whether it is possible to separate the effects of rosiglitazone on adipocyte differentiation from those on thermogenic programming in iWA. Certainly 7d rosiglitazone treatment promotes induction of UCP1, with further substantial increases following an additional 24 h in the presence of CL316243 (**Figures 3** and **5**). In contrast, Fabp3 and Cpt1b are increased by 7d rosiglitazone but there is no potentiation of expression by CL316243, indicating that responses to the β3-AR agonist are in part masked by prolonged rosiglitazone treatment. The magnitude of responses was smaller when rosiglitazone was included in the cultures only for the first 3 days of the 7-day culture period, however, this protocol did provide an improved window for observing the browning effect of 24 h CL316243 over and above rosiglitazone alone, with significant increases in expression of UCP1 (96-fold). The capacity to respond to CL316243 in these 3d cultures may be dictated solely by rosiglitazone-induced expression of the β3-AR, or it may reflect other aspects of the differentiation process. We cannot say at this time whether the 3d iWA most closely resemble the brite adipocytes seen *in vivo* (34), though profiling both 3d and 7d cultures would be worthwhile as it would facilitate a better understanding of the relationship between these adipocytes and those induced by PPARγ activation in mice.

# CONCLUSION

While there is a wealth of data on the gene expression of brite adipocytes (8, 9, 56, 62–66), these studies have generally aimed to identify differential genetic markers between brown, white, and brite adipocytes. We have instead sought to determine the similarities of brite adipocytes to conventional BAs in metabolic responses and profiles of thermogenic gene expression. We have demonstrated that the induction of brite adipocytes involves the upregulation of metabolic genes associated with thermogenesis, along with expression of β3-ARs, and this allows for adrenergically mediated activation of UCP1. Furthermore, brite adipocytes display the adaptive capacity observed in BAs, capable of responding to sustained adrenergic stimuli. Our findings indicate that cultured white adipocytes derived from regions such as iWAT with adequate sympathetic innervation respond efficiently to agonists, but only in combination with an additional priming stimulus such as rosiglitazone. This contrasts with *in vivo* iWAT depots in which chronic sympathetic tone is sufficient to maintain the capacity for brite activation. Our results emphasize the importance of activating both browning and thermogenic programs in cultured white adipocytes in order to reach maximum browning capacity. This understanding will translate into better model systems for the screening of beige or brite capacity to combat obesity and type 2 diabetes.

# ETHICS STATEMENT

All experiments were conducted with ethical permission from the Monash University Animal Ethics Committee, ethics approval numbers MIPS.2015.14 and VCP.2009.22, which complied with the National Health and Medical Research Council of Australia (NHMRC) guidelines for use of animals in scientific research.

# REFERENCES


# AUTHOR CONTRIBUTIONS

JM designed and performed most of the experiments with assistance from MS and CN (confocal microscopy). RF, MP, and LC (qPCR), DH (adipocyte cultures, Seahorse studies), and BE (IPA analysis). JM, BE, and DH wrote the manuscript with help and suggestions from RS and TB. DH, BE, and TB conceived the research. All authors discussed the results and commented on the manuscript.

# FUNDING

DH was supported by a National Health and Medical Research Council of Australia (NHMRC) Career Development Fellowship (545952), a Monash University Research Accelerator Program, and supported by a Bioplatforms Australia/NHMRC grant. RS is supported by a NHMRC Program Grant (519461, 1055134). MS was supported by a NHMRC International Training Fellowship (606763). JM was supported by an Australian Postgraduate Award. MP was supported by an Iranian Ministry of Health Fellowship. TB is supported by Diabetesfonden, Vinnova, and Vetenskapsrådet (Sweden).


the supraclavicular region of adult humans. *Cell Metab* (2013) 17(5):798–805. doi:10.1016/j.cmet.2013.04.011


**Conflict of Interest Statement:** TB owns stocks in the following pharmaceutical companies: Sigrid Therapeutics AB, Atrogi AB, and Glucox Biotechnology AB. DH owns stocks in Glucox Biotechnology AB. DH and RS are consultants for Atrogi AB.

*Copyright © 2018 Merlin, Sato, Chia, Fahey, Pakzad, Nowell, Summers, Bengtsson, Evans and Hutchinson. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Association Between Different Indicators of Obesity and Depression in Adults in Qingdao, China: A Cross-Sectional Study

Jing Cui 1,2†, Xiufen Sun3†, Xiaojing Li 1,2†, Ma Ke<sup>4</sup> , Jianping Sun1,2, Nafeesa Yasmeen<sup>5</sup> , Jamal Muhammad Khan<sup>6</sup> , Hualei Xin1,2, Shouyong Xue<sup>7</sup> \* and Zulqarnain Baloch<sup>8</sup> \*

*<sup>1</sup> Qingdao Municipal Center for Disease Control and Prevention, Qingdao, China, <sup>2</sup> Qingdao Institute of Preventive Medicine, Qingdao, China, <sup>3</sup> Qingdao Shi'nan Municipal Center for Disease Control and Prevention, Qingdao, China, <sup>4</sup> College of Traditional Chinese Medicine, Shandong University of Traditional Chinese Medicine, Jinan, China, <sup>5</sup> Institute of Microbiology, Agriculture University Faisalabad Pakistan, Bahawalpur, Pakistan, <sup>6</sup> Department of Patho-biology, The Islamia University of Bahawalpur, University College of Veterinary and Animal Sciences, Bahawalpur, Pakistan, <sup>7</sup> Qingdao Shi'bei Municipal Center for Disease Control and Prevention, Qingdao, China, <sup>8</sup> College of Veterinary Medicine, South China Agricultural University, Guangzhou, China*

#### Edited by:

*Andrew J. McAinch, Victoria University, Australia*

#### Reviewed by:

*Scott B. Teasdale, University of New South Wales, Australia Guoqin Xu, Guangzhou Sport University, China*

#### \*Correspondence:

*Shouyong Xue sbjkxue@163.com Zulqarnain Baloch znbalooch@yahoo.com*

*†These authors have contributed equally to this work*

#### Specialty section:

*This article was submitted to Obesity, a section of the journal Frontiers in Endocrinology*

Received: *26 May 2018* Accepted: *29 August 2018* Published: *10 October 2018*

#### Citation:

*Cui J, Sun X, Li X, Ke M, Sun J, Yasmeen N, Khan JM, Xin H, Xue S and Baloch Z (2018) Association Between Different Indicators of Obesity and Depression in Adults in Qingdao, China: A Cross-Sectional Study. Front. Endocrinol. 9:549. doi: 10.3389/fendo.2018.00549* Background: This study was designed to investigate the perceived relationship between body weight and depression risk in a Chinese population in Qingdao, China.

Methods: A population-based cross-sectional survey was performed with 4,573 participants (between 35 and 74 years) from the year 2009 to 2012 in Qingdao, China. We applied the Zung self-rating depression scale to ascertain the level of depression in participants. The associations between different indicators of obesity [body mass index (BMI), waist circumference (WC), and waist-to-hip ratio (WHR)] and depression were assessed by logistic regression based on the Chinese criteria of obesity. Sensitivity analysis was done based on the Asian and WHO criteria of obesity.

Results: The Zung scores for the 243 participants (5.2%) were over 45 and they were entitled as depression. Furthermore, multivariable logistic analyses revealed that being overweight [odds ratios (OR): 1.48, 95% confidence intervals [95% CI]: 1.08–2.03] and having abdominal obesity (WC category in Chinese criteria) (OR: 1.47, 95% CI: 1.08-2.00) were often associated with a higher risk for depression compared to normal weight subjects. Sensitivity analysis revealed that abdominal obesity (Asian criterion) (OR: 1.41, 95% CI: 1.03-1.91) was a significant risk factor for depression. Similarly, being overweight (WHO criterion) (OR: 1.39, 95% CI: 1.03-1.87) was an obvious risk factor for depression.

Conclusion: Being overweight and having abdominal obesity (WC category) were found to be linked with a higher risk of depression. However, abdominal obesity (WHR category) was not associated with depression.

#### Keywords: depression, body mass index, waist circumference, waist-to-hip ratio, cross-sectional study

**Abbreviations:** BMI, Body mass index; CI, Confidence intervalsl; OR, Odds ratio; WC, waist circumference; WHO, World Health Organisation.

#### INTRODUCTION

Significant progress has indeed been made in improving public health in the world; however, many serious health issues still need to be resolved such as depression and obesity. The global prevalence of being overweight and obese was estimated to be 39% and 13% in adults aged 18 years and over in 2017, respectively (1). In 2018, the World Health Organization (WHO) estimates that depression is seen to affect approximately 300 million people worldwide (2). In China, the prevalence of obesity has been continuously increasing during the last few decades (3–5). Additionally, depression is also frequently diagnosed at public and private clinics. According to health officials, approximately 26 million people are struggling with symptoms of mild depression annually in China (6).

Several population-based studies conducted in Western countries demonstrated that the risk for depression was positively associated with being overweight and obesity (7–12). In another study, a U-shaped association between BMI and depression has been reported (10). However, estimates obtained from the Chinese population are found to be controversial (13–15). Several research activities have reported a negative correlation between the risk for depression and obesity among middle-aged and the elderly Chinese (13, 14). In contrast, Li et al. observed a significant positive association between obesity and increased depression scores in the Chinese elderly (aged ≥ 65 years) (15). Nonetheless, mechanisms underlying the association between obesity and depression are still unclear. Some studies have shown that the brain–reward regions for depression and obesity, such as the dopaminergic pathways, can enhance positive mood. Yet, they also cause increased intake of "comfort" food, which may lead to obesity (16, 17). As a matter of fact, the perceptions regarding obesity are different between the Western world and the Chinese. While obesity is a stigma in Western countries (18), it is considered as a symbol of wealth in China (13–15). Regardless of these differences, it may be due to population diversity, different body weight criteria, and depression standards in China.

The association between obesity and depression has been normally studied in adolescents and the elderly population in China (19–21); yet there is a need to explore various indicators of obesity in detail. Hence, this large population-based cross-sectional survey was carried out to assess the potential relationship between depression and different body weight levels among the general population of Qingdao, China.

#### METHODS

#### Ethical Considerations

All the participants voluntarily signed the consent prior to their participation. The security, anonymity, and the privacy of participants were strictly respected. This study was approved by the Ethics Committee of the Qingdao Municipal Center for Disease Control and Prevention.

#### Study Population

This cross-sectional population-based survey was conducted in the eastern city of Qingdao, Mainland China during 2009 to 2012. Inclusion and exclusion criteria have already been described in our previous study (22). Briefly, a total of 6,100 Chinese adults (between 35-74 years old) were selected with a stratified, random cluster sampling procedure and invited to participate in the survey. A total of 5,110 individuals agreed to take part in our survey, with a response rate of 83.8%. From the 5,110 individuals, 454 participants were excluded due to an insufficient Zung score or body mass index (BMI) or waist circumference (WC) or waist-to-hip ratio (WHR) baseline information. Therefore, results from the current study were based on data from 4,656 participants, accounting for 76.3% of the initially invited individuals.

#### Questionnaires

A standard questionnaire was designed to collect basic information which included age, gender, marital status, educational background, occupation, smoking, and alcohol consumption. The marital status was divided into married/cohabiting and unmarried (single, divorced, or widowed). Educational attainment was categorized into three groups: illiterate/ elementary school, junior high school, and senior high school or higher. Occupational physical activity (PA) was categorized into light (housewife, retired, or unemployed), moderate (teacher, doctor, or nurse), and heavy (worker, farmer, or fisherman), according to occupation. Smoking was classified into current smokers (smoking every day and occasional smoking in the past years) and noncurrent smokers (including ex-smokers and nonsmokers). Alcohol consumption was categorized into regular drinkers (drinking every day in the past years) and noncurrent drinkers (including ex-drinkers, rare drinkers, and nondrinkers). Personal monthly income was categorized into ≤999 Chinese Yuan (CNY), 1,000–2,999 CNY, and ≥3,000 CNY.

The plasma glucose level was determined by the glucose oxidase method. All subjects (without prediagnosed diabetes) underwent a standard 2-h 75-g oral glucose tolerance test. Diabetes was diagnosed on the basis of the WHO/International Diabetes Federation criteria (23). Subjects with fasting plasma glucose (FPG) level of ≥7.0 mmol/l and/or 2-h postload plasma glucose (2-hPG) level of ≥11.0 mmol/l were diagnosed with diabetes. The individual was taken as nondiabetic if FPG was <7.0 mmol/l and/or 2-hPG <11.1 mmol/l.

Blood pressure was measured using a mercury sphygmomanometer. Measurements were taken three times in 5-min intervals. The mean of the three readings was used for data analysis. Hypertension was defined by a mean systolic blood pressure of ≥140 mmHg and/or a mean diastolic blood pressure of ≥90 mmHg and/or an established diagnosis of hypertension at the baseline (24).

#### Anthropometric Measurement

We measured the participant's height and weight while wearing light clothes and no shoes. The BMI was calculated as the weight in kilograms divided by the height in squared meters (kg/m<sup>2</sup> ). Waist circumference (WC) was measured at the midpoint between the lower rib margin and the iliac crest. Hip circumference (HC) was measured at the maximal horizontal girth between the waist and the thigh. The WHR was calculated by dividing WC (in cm) by HC (cm).

#### Assessment of Depression

Depression was assessed using the Zung Depression Rating Scale (ZDRS) (25, 26). The ZDRS questionnaire contains 20 questions divided into 10 positively and 10 negatively phrased questions. Each question was scored as 1 through 4, and the total scores ranged from 20 to 80 based on the participant's feelings (sadness, indifference, or happiness) toward family, work, and living conditions, respectively. Participants were divided into two major groups: normal (20–44) and depressed (≥45) (25).

#### Criteria for Obesity

Overweight and obesity were defined in accordance with the published criteria (26, 27). According to the Chinese criteria, subjects were divided into four BMI categorical groups as: underweight (BMI <18.5 kg/m<sup>2</sup> ), normal weight (18.5–23.9 kg/m<sup>2</sup> ), overweight (24.0–27.9 kg/m<sup>2</sup> ), and obese (≥28.0 kg/m<sup>2</sup> ). According to Asian criteria, subjects were subdivided as: underweight (BMI <18.5 kg/m<sup>2</sup> ), normal weight (18.5–22.9 kg/m<sup>2</sup> ), overweight (23.0–24.9 kg/m<sup>2</sup> ), and obese (≥25.0 kg/m<sup>2</sup> ) (27) and according to the WHO criteria as: underweight (BMI < 18.5 kg/m<sup>2</sup> ), normal weight (18.5–24.9 kg/m<sup>2</sup> ), overweight (25.0–29.9 kg/m<sup>2</sup> ), and obese (≥30.0 kg/m<sup>2</sup> ) (28).

Based on the value of WC, subjects were further subdivided into two groups as follows: normal (<85.0 cm for males and <80.0 cm for females); abdominal obesity, (≥85.0 cm for males and ≥80.0 cm for females) according to Chinese criteria (26). Alternatively, subjects were subdivided into: normal (<90.0 cm for males and <80.0 cm for females) and abdominal obesity, (≥90.0 cm for males and ≥ 80.0 cm for females) according to the Asian guidelines (27). Finally, according to the WHO guidelines, subjects were divided into normal (<94.0 cm for males and < 80.0 cm for females) and abdominal obesity (≥94.0 cm for males and ≥80.0 cm for females) (28).

Subjects were divided into two categories based on their WHR cutoffs points (Chinese guidelines): normal (<0.9 for males and <0.85 for females) and abdominal obesity (≥0.90 for males and ≥ 0.85 for females) (26); or the Asian WHR cutoff points: normal (<0.95 for males and <0.80 for females) and abdominal obesity (≥0.95 for males and ≥0.80 for females) (29) or the WHO WHR cutoff points: normal (<1.00 for males and <0.85 for females) and abdominal obesity (≥1.00 for males and ≥0.85 for females) (28).

### Statistical Analysis

Statistical analyses were performed using IBM SPSS Statistics 17.0. Continuous variables and categorical variables were presented as the mean ± standard deviation and the number (percentages), respectively. Using data from the 2010 census in Qingdao, the age-standardized prevalence of depression was calculated according to different obesity criterion for the age group of 35–74 years. Logistic regression was used to determine the possible link between the risk for depression and different indicators of obesity (BMI, WC, and WHR). Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated by logistic regression to obtain the association between depression and each variable. Sensitivity analysis was done based on the Asian and WHO criteria of obesity. A P-value <0.05 was considered to be statistically significant.

# RESULTS

The demographic information, social-economic status, lifestyle information, and anthropometric measurement have been shown in **Table 1**. In this study, a total of 4,656 participants were enrolled; among them, 1,804 were males (38.6%) and 2,852 were females (61.4%). The mean age was 52.3 ± 10.7 years (rang 35–74 years).

TABLE 1 | Characteristics of the population included in the study.


*BMI, body mass index; WC, waist circumference; HC, hip circumference; WHR, waistto-hip ratio; PA, physical activity; CNY, Chinese Yuan. ? obesity and abdominal obesity is defined according to Chinese criteria. n, number.*

According to the BMI category, the age-standardized prevalence of depression was higher in underweight (10.5%) and overweight participants (6.0%) compared with the normal weight (4.3%); however, the difference was not statistically significant. Based on the WC category, a slightly higher agestandardized prevalence of depression was observed in the abdominal obesity participants (5.7%) compared with the normal WC participants (4.3%), and the difference was statistically significant. The corresponding age-standardized prevalence of depression was 5.5% in normal WHR participants and 4.7% in the abdominal obesity participants, respectively, and the difference was nonsignificant. The age-standardized prevalence of depression according to BMI, WC, and WHR has shown in **Figure 1**.

In this study, we determined the role of potential risk factors in depression using multivariable regression analysis according to the Chinese criterion (**Table 2**). Overweight indicated higher risk of depression (OR: 1.49, 95%CI: 1.08-2.05) compared with normal weight. We observed a significant relationship between abdominal obesity and depression risk according to WC.

Next, we performed a sensitivity analysis according to the Asian and WHO criteria (model 1 and model 2, respectively; **Table 3**). In model 1, sensitivity analysis based on Asian criterion revealed that overweight and obesity (BMI category) were not significantly associated with depression, while abdominal obesity (according to WC) was significantly associated with depression similar to that of Chinese criterion. In model 2, based on the WHO criterion, abdominal obesity (according to WC) was not significantly associated with depression, while overweight participants in the BMI category were significantly associated with depression.

#### DISCUSSION

In this study, we investigated the association between depression and body weight among adult people in Qingdao, China. Our results showed that the prevalence of depression was higher among the overweight participants (according to BMI category in China) and abdominal obesity participants (according to WC category in China) compared with other groups. Additionally, we defined prevalence of depression by using a Zung score equal to or greater than 45. In Qingdao, the overall prevalence of depression rate of 5.2% (5.5% in males and 5.0% in females) was lower than the documented rates in South Korea (overall 5.7%, 3.9% in males and 7.0% in females) (30). Additionally, a crosssectional study of 512 891 Chinese adults, aged 30–79 years, had also revealed a lower pooled prevalence of depression compared with the current study (2.4%) (31). This difference might be due to social and family factors. Residents of the eastern coastal areas could not convincingly adapt to rapid industrialization and urbanization; thus it lead to depression. Secondly, accompanied with China entering the aging society, the "empty nest" phenomenon is the leading cause of unaccompanied, unattended, and psychosocial problems (such as depression) in middle-aged and elderly adults (32, 33). Depression, in turn, affects the middle-aged and the elderly population's social and physical

activities, which strongly influenced the wellbeing of elderly adults (34).

The prevalence of depression is higher in underweight than normal individuals. However, there was an insignificant positive association between underweight and depression in general adults in Mainland China. Our results are inconsistent with the estimates reported in Japan, Korea, and Taiwan (13, 30, 35, 36). These trends can be observed in different age populations, and underweight elderly adults are far more likely to be depressed (13, 30, 35, 36). Yet, the common population tends to have a distortion of body weight, which entitles thinness as a beauty symbol owing to social standard (37). Our observations also support this fact.



*Model 1, model 2 and Model 3, the risk factors listed in the table were entered at the same time.*\**P* < *0.05 for factors associated with depression.*

The prevalence of depression was higher in overweight and abdominal obesity (WC category) participants compared with the normal participants. In contrast, we did not observe a similar association in obese and abdominal obesity (WHR category) participants. This trend can be explained by the Chinese cultural heritage that usually associates overweight with a higher economic status, since those who can afford to eat more could attain more body weight and vice versa. In China, prevalence of obesity was substantially higher in rich people (38, 39). The overweight carries a social stigma, such as body image, self-esteem, and social life, which can help a bit for depression (40).

In this study, our analysis revealed a significant positive association between being overweight and depression, while a nonsignificant association between being obese and depression. However, the association of BMI with depression is still controversial. A nonsignificant positive relationship between TABLE 3 | Sensitivity analyses for depression according to the Asian and WHO criteria.


*Model 1 is according to Asia criterion. Model 2 is according to WHO criterion. All models were adjusted for gender, age, hypertension, diabetes, resident districts, marital status, educational attainment, occupational physical activity, smoking status, alcohol drinking status, personal monthly income. BMI category, WC category, and WHR category entered models separately.* \**P* < *0.05 for factors associated with depression.*

being overweight and depression has been reported in developed countries such as Canada (41). Additionally, Palinkas et al. observed a nonsignificant inverse association between being overweight and depression in women (42). Moreover, a nonsignificant negative association between depression and obesity has also been reported previously (12, 43). In contrast, several studies reported a significant positive association between depression and obesity (11, 12, 20, 21, 44, 45). Young people regarded thinness as a beauty symbol. However, middle-aged and elderly people are still more likely to view obesity as a symbol of wealth and happiness in the traditional Chinese culture. Meanwhile, in this study, participants are mainly middle-aged and the elderly people aged 35 years or over, which is the possible reason for the current observations.

Next, we examined the effect of abdominal obesity on depression. We observed a significant positive association between abdominal obesity and depression. These results are in agreement with those of Takeuchi et al. (46) and Vogelzangs et al. (47). In contrast, Herva et al. (48), Gil et al. (49), and Zavala GA (12) reported a nonsignificant positive association between depression and abdominal obesity. The absence of significant associations between abdominal obesity (WHR category) and depression were inconsistent with the results reported by Ahlberg et al. (50).

Here, for the first time, we reported a significant association between being overweight, abdominal obesity (WC category), and depression in the general population of Qingdao China, but not in obese or abdominal obesity (WHR category) participants, which might be due to the following reasons. First, cultural factors are known to influence the association between depression and body weight. A higher socioeconomic status is generally associated with more concern for body image and abnormal body weight, which can possibly be a stress factor that leads to depression (51, 52). Second, diet is an important factor in the association between depression and body weight status. Emotions of individuals can also affect food intake. Depression was associated with emotional eating such as negative emotional overeaters (53). Moreover, a negative emotion, such as depression, is further inclined to uncontrolled eating and/or overeating (54). However, this will ultimately lead to a high body weight gain. Finally, we cannot underestimate the roles of neurological mechanisms (16, 17, 55) and/or genetic susceptibility (56), which may also lead to depression and an abnormal weight status.

Previous studies had reported conflicting results in China, partly because different studies had used different criteria to define obesity. Therefore, in this study, we performed sensitivity analyses according to the Asian and WHO criteria of obesity. Sensitivity analyses showed a significant positive association between depression and abdominal obesity (WC category in Asian criteria), and being overweight (BMI category in WHO criteria) after adjustment for gender, age, hypertension, diabetes, resident districts, marital status, educational attainment, occupational PA, smoking status, alcohol-drinking status, and personal monthly income. These results were slightly different from the results produced using the Chinese criteria. Taken together, we suggest that the association between depression and weight status is influenced by the criteria adopted. Additionally, the reciprocal and complex associations of being overweight, abdominal obesity, and depression are strongly intertwined, and these phenomena are likely to be true for the earlier-mentioned results. Hence, further large-scale research is needed to reveal the association between obesity and depression among the whole Chinese population.

This study suffered from a few limitations. First, the present study was a cross-sectional study that does not reflect the underlying mechanisms between depression and weight status. Therefore, a follow-up study is imperative. Second, this study was performed on a relatively small sample size of the adult community in Qingdao, China.

# CONCLUSION

We observed that overweight and abdominal obesity (WC category) participants were at a higher risk of depression according to Chinese criterion. However, abdominal obesity according to WHR was not associated with depression.

#### REFERENCES


Furthermore, Asian and WHO criteria of obesity might influence the association between depression and body weight status. On all accounts, controlling being overweight and having abdominal obesity had a protective impact on depression. Future research will involve a larger multicenter study in China to further investigate the relationship between body weight and depression.

# ETHICS AND CONSENT TO PARTICIPATE

All the participants voluntarily signed the informed consent before their participation and the consent of ethics, including three urban districts (Shinan, Shibei, and Sifang) and three rural districts (Huangdao, Jiaonan, and Jimo), was obtained from the ethics committee in Qingdao Municipal Center for Disease Control and Prevention. This study was approved by the local ethics committee at Qingdao Municipal Health and Family planning commission.

# AVAILABILITY OF DATA AND MATERIALS

The aggregate data supporting findings contained within this manuscript will be shared upon request submitted by the corresponding author. Identifying patient data will not be shared.

# AUTHOR CONTRIBUTIONS

ZB, JC, and JS were the primary authors and leading investigators. NY, JC, MK, ZB, and HX carried out the experiments, analyzed experimental results. ZB, JC, and JS wrote the manuscript. All authors read and approved the final manuscript.

# FUNDING

This study was funded by the Bayer Healthcare in China and World Diabetes Foundation [WDF07-308].

#### ACKNOWLEDGMENTS

We are grateful to the participants, primary care doctors, and nurses who participated in this survey.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Cui, Sun, Li, Ke, Sun, Yasmeen, Khan, Xin, Xue and Baloch. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Fat Mass Follows a U-shaped Distribution Based on estradiol levels in Postmenopausal Women

*Georgia Colleluori1,2,3, Rui Chen1,2, Nicola Napoli3 , Lina E. Aguirre4 , Clifford Qualls5 , Dennis T. Villareal1,2 and Reina Armamento-Villareal1,2\**

*1Division of Endocrinology and Metabolism, Department of Internal Medicine, Baylor College of Medicine, Houston, TX, United States, 2Center for Translational Research on Inflammatory Diseases, Michael E. DeBakey VA Medical Center, Houston, TX, United States, 3Division of Endocrinology, University Campus Biomedico of Rome, Rome, Italy, 4Division of Endocrinology and Metabolism, Department of Internal Medicine, New Mexico VA Health Care System, Albuquerque, NM, United States, 5Division of Mathematics and Statistics, University of New Mexico School of Medicine, Albuquerque, NM, United States*

Objective: Estradiol (E2) regulates adipose tissue resulting in increased fat mass (FM) with declining E2. However, increased visceral fat and hyperestrogenemia are features of obese individuals. It is possible that adipocytes in obese individuals are less sensitive to E2 resulting in higher FM. Our objective is to identify the range of serum E2 for which postmenopausal women have the lowest FM and best body composition.

#### *Edited by:*

*Andrew J. McAinch, Victoria University, Australia*

#### *Reviewed by:*

*David Scott, Monash University, Australia Fernando Lizcano, Universidad de La Sabana, Colombia*

*\*Correspondence:*

*Reina Armamento-Villareal reina.villareal@bcm.edu*

#### *Specialty section:*

*This article was submitted to Obesity, a section of the journal Frontiers in Endocrinology*

> *Received: 15 January 2018 Accepted: 25 May 2018 Published: 02 July 2018*

#### *Citation:*

*Colleluori G, Chen R, Napoli N, Aguirre LE, Qualls C, Villareal DT and Armamento-Villareal R (2018) Fat Mass Follows a U-Shaped Distribution Based on Estradiol Levels in Postmenopausal Women. Front. Endocrinol. 9:315. doi: 10.3389/fendo.2018.00315*

Methods: Cross-sectional data from 252 community-dwelling postmenopausal women, 42–90 years old. Subjects were stratified into categories of E2 (pg/ml): (1) ≤10.5; (2) 10.6–13.9; (3) 14.0–17.4; and (4) ≥17.5. Body composition by dual-energy X-ray absorptiometry. Serum E2 by radioimmunoassay. Between-group comparisons by analysis of covariance.

results: E2 linearly increased with increasing body weight and body mass index (*r* = 0.15 and *p* = 0.01 for both), but not with total FM (kg) or % FM (*r* = 0.07, *p* = 0.34 and *r* = −0.04, *p* = 0.56, respectively). However, total FM (kg) followed a U-shaped distribution and was significantly lower in group 3 (27.6 ± 10.6), compared with groups 1: (34.6 ± 12.5), 2: (34.0 ± 12.4), and 4: (37.0 ± 10.6), *p* = 0.005. % FM was also lowest in group 3. While fat-free mass (FFM, kg) increased with increasing E2 (*p* < 0.001), % FFM was highest in group 3.

conclusion: In our population of postmenopausal women, FM followed a U-shaped distribution according to E2 levels. E2 between 14.0 and 17.4 pg/ml is associated with the best body composition, i.e., lowest total and % FM and highest % FFM. Given the role of E2 in regulating body fat, high FM at the high end of the E2 spectrum may suggest reduced E2 sensitivity in adipocytes among obese postmenopausal women.

clinical Trials: ClinicalTrials.gov identifier: NCT00146107.

Keywords: estradiol, obesity, adipocyte, adipose tissue, body composition, estrogen receptor, estrogen receptor alpha, estrogen receptor beta

# INTRODUCTION

Over one-third of US adults are obese (1). Considering the 10-fold rise in the worldwide incidence of obesity among children and adolescents during the past four decades, the number of obese adults could increase further in the coming years (2). High visceral fat mass (FM), common in obese individuals, is a risk factor for the development of cardiovascular disease, type 2 diabetes, and certain types of cancer, reason for which obesity is considered a major public health burden of the twenty-first century (1, 3). Research on the mediators of body fat deposition and distribution is thus crucial. Estrogen has been recognized as a regulator of FM (4–7), and in sufficient amount favors a "pear-shaped" fat distribution among women through promotion of both visceral lipolysis and subcutaneous adipogenesis (4, 8). Thus, the decreased lipid utilization and visceral FM accumulation in postmenopausal women is attributed to a severe reduction in estrogen levels (4–7). In rat adipocytes, estrogen depletion resulted in increased lipoprotein lipase and lipid deposition, while estradiol (E2) treatment reversed this process (9).

Although high E2 levels could be theorized as protective against developing obesity, it is interesting to note that obesity is classically characterized by relative hyperestrogenemia. It is in fact expected that increased aromatase activity in the expanded adipose tissue volume in obese individuals would result in high E2 levels (10). However, because these individuals remain obese, this observation may suggest impaired regulation of FM by E2 or reduced E2 sensitivity.

We thus hypothesize that total body fat varies according to E2 levels with individuals in the lower and higher end of the E2 spectrum having higher body fat. To the best of our knowledge, there are no reports describing the optimum E2 levels associated, primarily, with the lowest body fat and, secondarily, with the best body composition profile in postmenopausal women. The aim of this study is to determine the association of body fat and fatfree mass (FFM) with circulating E2 and to identify the range of serum E2 levels associated with the lowest body fat and the best body composition profile in postmenopausal women.

# MATERIALS AND METHODS

#### Study Population

This is a secondary analysis of cross-sectional data collected from otherwise healthy community-dwelling postmenopausal women participating in a study investigating the relationship between polymorphisms in the different CYP450 enzymes that metabolize estrogen and the effect on bone health (11, 12) or those undergoing a lifestyle intervention trial (13). These protocols were approved by the Washington University School of Medicine institutional review board. Both studies were conducted in accordance with the guidelines in the Declaration of Helsinki for the ethical treatment of human subjects. Participants were recruited using advertisement or direct mailing. All of the participants were provided written informed consent. Exclusion criteria include the use of medications that affect gonadal hormone levels such as estrogens and androgens, medications that affect bone metabolism such as glucocorticoids, anticonvulsants, bisphosphonates, selective estrogen receptor modulators (e.g., raloxifene and tamoxifen), aromatase inhibitors, as were medical conditions that affect bone metabolism such as hyperparathyroidism, chronic renal failure, Paget's diseases of the bone, chronic liver disease, and osteomalacia. All women in this study were at least 1 year from their last menstrual period or had a bilateral oophorectomy. Subjects were stratified into different groups based on serum E2 (pg/ml) as follows: (1) ≤10.5; (2) 10.6–13.9; (3) 14.0–17.4; and (4) ≥17.5.

#### Body Mass Index (BMI)

Subjects were requested to empty their pockets, remove shoes, heavy clothing, and heavy jewelries (e.g., watches, necklaces, and bracelets) to measure their body weight using a standard weighing scale. A stadiometer was used to measure height. BMI was calculated as body weight in kilograms divided by the square of the height in meters (kilograms/meters2 ).

#### Body Composition

Because body composition was not part of the original protocol in one study and was added on after several women were enrolled in the study (11, 14), we only have the body composition data in 147 women. Total body mass, FFM, FM, and truncal fat were measured using whole body dual-energy X-ray absorptiometry (Hologic Delphi 4500/w; Hologic, Waltham, MA, USA; Enhanced Whole Body 11.2 software version, Hologic) as previously described (15). The coefficients of variation for these measurements in our laboratory are <2% (15). The percentages of total FM were obtained from the estimated readings given by the machine for the different regions of interest. % FFM was calculated as the FFM divided by the whole body mass multiplied by 100 [(FFM/whole body mass) × 100].

## Biochemical Data

Serum E2 levels were measured by ultrasensitive radioimmunoassay (RIA, Diagnostic System Laboratory, Webster, TX, USA) (11, 14). The coefficient of variability for this assay in our laboratory is less than 10% (11, 12, 16).

#### Statistical Analysis

Between-group differences in body composition, BMI, age, and E2 were assessed by analysis of covariance. The main outcome investigated, total FM, was normally distributed in our population (*p* = 0.41) according to the Shapiro–Wilk test. We performed power calculation for the definition of the groups to optimize the trade-off between-group sample sizes versus variances; this resulted in 4 groups with a minimum of 32 participants per group with 80% power and 0.05 alpha to detect a difference of at least 8 kg of FM between the nadir (group 3) and the end groups (groups 1 and 4). Differences in body composition and BMI were analyzed adjusting for age. Least significant difference procedure to discriminate among the means was performed as *post hoc* analysis. Data were managed using Excel 2010 (Microsoft, Redmond, WA, USA) and analyzed using Statgraphics X64 (Statgraphics Technologies, The Plains, VA, USA). Results are expressed as the mean ± SD in the text and tables and mean ± SE in the figures. A *p* < 0.05 was considered statistically significant.

#### RESULTS

Data from 252 women were included in our analysis (**Table 1**): mean age (66.1 ± 7.6 years old), weight (79.3 ± 18.5 kg), BMI (29.8 ± 6.5 kg/m2 ), and E2 (14.1 ± 5.6 pg/ml). Subjects' characteristics according to E2 categories are reported in **Table 1**.

Body mass index did not significantly differ between those with and without body composition measurement (29.3 ± 5.9 vs 27.8 ± 5.5 kg/m2 , respectively, *p* = 0.07). Similarly, E2 levels were not significantly different between the two groups (14.4 ± 6.0 vs 14.1 ± 5.5 pg/ml for those with and without body composition assessment, respectively). However, age and body weight were significantly higher in participants who had body composition (66.6 ± 6.5 vs 63.7 ± 8.6 years old, *p* = 0.002; 78.0 ± 18.0 vs 73.2 ± 16.8 kg, *p* = 0.03, respectively).

Body mass index and body weight were the highest in group 4 (E2 ≥ 17.5 pg/ml, **Table 1**) and were both positively correlated to circulating E2 (*r* = 0.15 and *p* = 0.01 for both variables). On the other hand, total FM (kg) and % FM did not follow a linear relationship with circulating E2 (*r* = 0.07, *p* = 0.34 and *r* = −0.04, *p* = 0.56, respectively). However, analysis according to the different categories of E2 levels showed that both FM (kg) (**Figure 1A**) and % FM (**Figure 1B**) follow a U-shaped trend according to E2 levels. FFM (kg) increased with increasing E2 (*r* = 0.22, *p* = 0.007, **Figure 2A**), while % FFM did not follow a linear relationship with E2 (*r* = −0.001, *p* = 0.98). % FFM in fact follows a bell-shaped curve (**Figure 2B**) with women in group 3 having the highest % FFM. Truncal FM (kg) was the lowest in group 3 (E2: 14.0–17.4 pg/ml), while % truncal FM and % truncal lean mass did not differ significantly among the groups (**Table 2**).

#### DISCUSSION

Our study showed that postmenopausal women with circulating E2 between 14.0 and 17.4 pg/ml measured by RIA have the best body composition profile with the lowest total and % FM, and the highest % FFM. Furthermore, FM in postmenopausal women followed a U-shaped distribution according to E2 levels, with individuals with E2 levels <14.0 and >17.4 pg/ml having higher body fat.

According to the US Centers for Disease Control and Prevention, obesity and its related disorders are a leading cause of mortality worldwide (1). Hormones play a key role in obesity etiology and progression, though obesity is known to be multifactorial in origin, with causes including genetic, lifestyle, and environmental factors. Gender-dependent adipose tissue distribution and body composition changes after menopause are among the clinical observations linking estrogen to adipocyte development (5). During menopausal transition, women experience a shift from a gynoid to an android fat distribution (5, 7), and become three times more likely to develop obesity and metabolic syndrome than premenopausal women (17). Consistently, visceral FM reduction and improved cardiometabolic profile were reported in a meta-analysis including over 100 clinical trials (18) of women undergoing postmenopausal hormone replacement. The role of E2 in regulating fat distribution is well illustrated in male-to-female transgender patients having a change from an android to a gynoid fat habitus after 12 months of E2 therapy (19).

Interestingly, obesity is often accompanied by a state of relative hyperestrogenemia because of the high aromatase expression/ activity in the adipocytes of obese subjects (10). Considering the positive role exerted by estrogens on adipose mass and distribution, the coexistence of high-visceral fat content and elevated estrogen levels in obese individuals may appear inconsistent. This scenario raises the possibility that adipocytes in obese individuals, who have high E2 levels, are not as sensitive to the lipolytic effects of estrogens. This in turn may suggest a different mechanism for the increased body fat at both ends of the estrogen spectrum, from estrogen deficiency on one end to estrogen resistance on the other end. Another explanation for the association between higher body fat at higher end of estrogen level spectrum is the type of obesity developed from a multifactorial cause, independent of estrogen levels. Obesity, together with its associated high aromatase activity, then subsequently leads to the development of higher estrogen levels and possibly to E2 resistance.

To date, results from studies exploring the correlation between estrogens and body fat or BMI are contradictory (20–25). Despite numerous studies investigating the impact of estrogen adequacy or lack thereof on FM (4), to the best of our knowledge, there is no information on the optimum circulating E2 in relation to body fat and FFM in postmenopausal women. Our study showed the existence of a narrow range of serum E2 (14.0–17.4 pg/ml) for which postmenopausal women have the lowest total and % body fat, and highest % FFM. This identifies the E2 levels associated with the best body composition profile in this population.


*Data are reported as mean* ± *SD adjusted for age.*

*\*p* < *0.05 for comparison with group 4.*

γ *All groups differ significantly from each other (p* < *0.05).*

total FM [kg–%, (A,B)]. Data are reported as mean ±SE. \*Comparison with group 3 (E2: 14.0–17.4), *p* < 0.05.

Furthermore, FM followed a U-shaped curve according to E2 levels, with those with E2 levels below 14.0 and above 17.4 pg/ml having higher FM. Considering our results, it is possible that the persistently high-body fat in obese women leads to downregulation of ER-alpha (ERα) in adipocytes and reduced E2 sensitivity despite higher E2 levels.

Estrogen is also the main sex hormone regulating FM in men (26). In fact, estrogen resistance was first described in a man with an inactivating mutation of the ERα (23). Besides the peculiar skeletal profile, the subject was obese (23). Both ERα and ER-beta (ERβ) are expressed in adipose tissue; however, the former seems to be more relevant to its regulation (4). ERα knock-out male and female mice had higher adiposity compared with wild-type

mice, while body fat in ERβKO mice was similar to wild-type mice (4). A study conducted on Swedish women reported lower ERα expression in the subcutaneous adipose tissue of obese (*N* = 17) when compared with non-obese (*N* = 16), suggesting a different sensitivity to estrogens depending on BMI (27). On the other hand, ERα expression in subcutaneous adipose tissue was found to increase after weight loss in 23 obese women (28). Since body composition and estrogen levels were not measured in these studies, one can only speculate, but not confirm, that increased body fat and higher estrogen levels were present in these patients. Given our data, it is possible that obese subjects in that study could have either high- or low-circulating E2.

Comparison with group 4 (E2: 10.5–13.9), *p* < 0.05.

δ


Table 2 | Body composition of the trunk according to estradiol (E2) categories.

*Data are reported as mean* ± *SD adjusted for age.*

*\*p* < *0.05 for comparison with group 3.*

δ *p* < *0.05 for comparison with group 4.*

We speculate that a possible sequence of events following the drop in estrogen levels during menopause which include the physiologic increase in body fat, which in turn leads to increased adipose tissue production of estrogen in an effort to compensate for the expanded FM. However, the persistent hyperestrogenemia could lead to ER downregulation and thus reduced adipocyte response to estrogen.

After the age of 40, the prevalence of obesity significantly increases in American women (29) and E2 deficiency is postulated as a critical triggering factor (5). E2 deficiency promotes metabolic dysfunction predisposing to metabolic syndrome, type 2 diabetes, and cardiovascular events (7). Considering the regulation of adipose tissues by estrogen and its implications on the cardiometabolic health (7), the identification of optimum E2 values associated with the best body composition profile would be of clinical relevance. In our group of postmenopausal women, serum E2 values between 14.0 and 17.4 pg/ml corresponded to the best body composition profile; aside from the lowest absolute and percent FM, % FFM was also relatively higher in group 3. FFM increased with increasing E2 levels possibly because of the beneficial effect exerted by E2 on both bone mineral density (30) and muscle homeostasis as suggested by a very recent publication (31).

Our study has limitations. We used of RIA for E2 measurement. Liquid chromatography/mass spectrometry (LC/MS) is considered the gold standard for sex steroid assay because of its high sensitivity and ability to detect very low E2 concentrations (32). However, RIA has been used in prior studies with good reproducibility (30). Nevertheless, according to a study by Hsing et al. exploring the reproducibility of both techniques, steroid hormone levels measured by RIA and MS were highly correlated despite the consistently higher E2 assayed by RIA (32). More importantly, they concluded that the intra- and interassay variability of both techniques were equally reliable (32). It is thus possible that the E2 cutoffs we identified will be shifted downwards if our E2 assay was done by LC/MS instead of RIA. Other limitations include the small sample size and the lack of adjustment for confounders, such as additional hormones, diet, and exercise (data not available in our study), which could affect body composition. In addition, the cross-sectional study design does not allow us to establish a causal relationship between the outcomes investigated and E2 levels.

In conclusion, to the best of our knowledge, this is the first study investigating the optimum E2 levels associated with the lowest FM and the best body composition in postmenopausal women. Since FM is comparably high in people with low- and high-serum E2, our observation raises questions regarding E2 sensitivity and the ability of E2 to regulate FM and distribution in women with high E2 levels, i.e., E2 ≥ 17.5 pg/ml. Prospective investigation using LC/MS to measure E2 with a bigger sample size is needed to confirm our observation and to explore the possibility of E2 resistance among individuals with relative hyperestrogenemia.

# ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the Washington University School of Medicine institutional review board with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Washington University School of Medicine institutional review board.

#### AUTHOR CONTRIBUTIONS

RA-V and DV designed the studies. GC, RC, NN, LA, DV, and RA-V conducted the study and collected the data. GC, CQ, and RA-V analyzed and interpreted the data. GC, RC, NN, LA, CQ, DV, and RV drafted the manuscript. Revising manuscript content and approving the final version: all take responsibility for the manuscript content, the integrity of the data analysis, and approval of the final version of the manuscript.

#### ACKNOWLEDGMENTS

This study was supported by the resources at the Washington University School of Medicine St Louise, MO, USA; the Michael E. DeBakey VA Medical Center, Houston, TX, USA; and the Center for Translational Research in Inflammatory Diseases (CTRID) at the Michael E. DeBakey VA Medical Center, Houston, TX, USA. The authors thank participants for their contribution.

#### FUNDING

This work was supported by National Institutes of Health grants R03 R049401 (to RA-V), K12 HD01459 (Building Interdisciplinary Research Careers in Women's Health), P30- DK020579, and grant RO1-AG025501.

# REFERENCES


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Colleluori, Chen, Napoli, Aguirre, Qualls, Villareal and Armamento-Villareal. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Inverse Association of Circulating SIRT1 and Adiposity: A Study on Underweight, Normal Weight, and Obese Patients

Stefania Mariani <sup>1</sup> \*, Maria R. di Giorgio<sup>1</sup> , Paolo Martini <sup>2</sup> , Agnese Persichetti <sup>3</sup> , Giuseppe Barbaro<sup>1</sup> , Sabrina Basciani <sup>1</sup> , Savina Contini <sup>1</sup> , Eleonora Poggiogalle<sup>1</sup> , Antonio Sarnicola<sup>2</sup> , Alfredo Genco<sup>4</sup> , Carla Lubrano<sup>1</sup> , Aldo Rosano<sup>5</sup> , Lorenzo M. Donini <sup>1</sup> , Andrea Lenzi <sup>1</sup> and Lucio Gnessi <sup>1</sup>

*<sup>1</sup> Section of Medical Physiopathology, Food Science and Endocrinology, Department of Experimental Medicine, Sapienza University of Rome, Rome, Italy, <sup>2</sup> Italian Hospital Group, Center for the Treatment of Eating Disorders and Obesity "Villa Pia", Guidonia, Italy, <sup>3</sup> Department of Molecular Medicine, Sapienza University of Rome, Rome, Italy, <sup>4</sup> Department of Surgical Sciences, Policlinico Umberto I, Sapienza University of Rome, Rome, Italy, <sup>5</sup> Roman Academy of Public Health, Rome, Italy*

#### Edited by:

*Beverly Sara Muhlhausler, University of Adelaide, Australia*

#### Reviewed by:

*Frida Renstrom, Lund University, Sweden Kathleen Grace Mountjoy, University of Auckland, New Zealand*

> \*Correspondence: *Stefania Mariani stefaniamariani@yahoo.com*

#### Specialty section:

*This article was submitted to Obesity, a section of the journal Frontiers in Endocrinology*

Received: *11 January 2018* Accepted: *20 July 2018* Published: *07 August 2018*

#### Citation:

*Mariani S, di Giorgio MR, Martini P, Persichetti A, Barbaro G, Basciani S, Contini S, Poggiogalle E, Sarnicola A, Genco A, Lubrano C, Rosano A, Donini LM, Lenzi A and Gnessi L (2018) Inverse Association of Circulating SIRT1 and Adiposity: A Study on Underweight, Normal Weight, and Obese Patients. Front. Endocrinol. 9:449. doi: 10.3389/fendo.2018.00449* Context: Sirtuins (SIRTs) are NAD+-dependent deacetylases, cellular sensors to detect energy availability, and modulate metabolic processes. SIRT1, the most studied family member, influences a number of tissues including adipose tissue. Expression and activity of SIRT1 reduce with weight gain and increase in conditions of starvation.

Objective: To focus on SIRT1 plasma concentrations in different conditions of adiposity and to correlate SIRT1 with fat content and distribution, energy homeostasis and inflammation in under-weight, normal-weight, and obese individuals.

Materials and Methods: 21 patients with anorexia nervosa, 26 normal-weight and 75 patients with obesity were evaluated. Body fat composition by dual-energy X-ray absorptiometry, ultrasound liver adiposity, echocardiographic epicardial fat thickness (EFT), inflammatory (ESR, CRP, and fibrinogen), and metabolic (FPG, insulin, LDL- and HDL-cholesterol, triglycerides) parameters, calculated basal metabolic rate (BMR) and plasma SIRT1 (ELISA) were measured.

Results: SIRT1 was significantly higher in anorexic patients compared to normal-weight and obese patients (3.27 ± 2.98, 2.27 ± 1.13, and 1.36 ± 1.31 ng/ml, respectively). Linear regression models for each predictor variable adjusted for age and sex showed that SIRT1 concentration was inversely and significantly correlated with EFT, fat mass %, liver fat content, BMR, weight, BMI, WC, LDL-cholesterol, insulin, ESR. Stepwise multiple regression analysis revealed that age and EFT were the best independent correlates of SIRT1 (β = −0.026 ± 0.011, *p* = 0.025, and β = −0.516 ± 0.083, *p* < 0.001, respectively).

Conclusions: Plasma SIRT1 shows a continuous pattern that inversely follows the whole spectrum of adiposity. SIRT1 significantly associates with EFT, a strong index of visceral fat phenotype, better than other indexes of adiposity studied here.

Keywords: sirtuins, circulating SIRT1, adiposity, anorexia, obesity, body fat mass

# INTRODUCTION

Sirtuins (SIRTs) are nutrient sensing, metabolic regulators, and chromatin silencers (1). SIRT1, the most-studied SIRT, is best known for mediating lifespan extension by consistently improving health during aging. Results mainly derived from animal studies show that SIRT1 protects against or delays the onset of metabolic diseases, neurodegeneration, cardiovascular diseases, and some types of cancers (1, 2). SIRT1 modifies the acetylation status of many different targets in cytoplasm, mitochondria, and nucleus, and it carries out its protective roles by activating key transcription factors, improving lipid metabolism, reducing inflammation, and acting as a tumor suppressor by preserving genomic integrity. SIRT1 plays also an essential role in adaptive metabolic and endocrine responses (3). Several metabolic disorders such as liver steatosis, diabetes, and obesity associate with defects in SIRT1 pathways. Obesity is associated with low NAD(+)/SIRT pathway expression in subcutaneous adipose tissue of BMI-discordant monozygotic twins, highlighting a strong relationship of reduced SIRTs expression with inflammation, insulin resistance, and impaired mitochondrial homeostasis (4). Visceral adiposity negatively correlates with SIRT1 expression (5, 6). Accordingly, we have previously shown an inverse association between plasma SIRT1 and ectopic fat distribution in patients affected by obesity (7, 8) in particular with epicardial and liver fat depots, both typical examples of visceral fat with particularly detrimental effect because of localized and systemic toxic effects (9, 10). Although the main source of circulating SIRT1 is not known (11–13), these results indicate that the negative metabolic effects of obesity could be related, at least in part, to the reduced levels of SIRT1 in the blood. Moreover, what regulates circulating SIRT1 vs. tissue SIRT1 is still unknown.

Conversely, SIRTs tissue enzymatic activity increases in conditions of nutrient depletion and starvation. SIRT1 expression rises in cultured cells, and in multiple tissues of mice after overnight or 24 h fasting (14, 15). An increased expression of SIRT1 is seen after long periods of calorie restriction (CR) in mice (16). Analogously, in man, 30 days and 7 weeks of CR cause a rise of tissue and plasma levels of SIRT1, respectively (17, 18). Indeed, SIRT1 has been identified as a novel factor responsible for some beneficial effects of CR, and previous studies showed that weight loss induces an increase in tissue and circulating SIRT1 levels in obese patients (19, 20).

Therefore, SIRT1 may act differently in states of nutritional excess compared with states of nutritional deprivation.

Circulating SIRT1 has not been studied yet in underweight individuals or in subjects who restrict eating. Thus, in relation to its opposite behavior in condition of hyper- or hyponutrition, we evaluated SIRT1 blood concentration, body fat composition, markers of energy homeostasis, inflammation and some metabolic parameters in underweight, normal-weight, and obese individuals, i.e., in subjects with defect or excess of body fat mass. The aim of the study was to investigate the plasma SIRT1 concentration across the whole spectrum of adiposity, and its relationship with fat distribution and metabolic, inflammatory and energy settings.

# SUBJECTS AND METHODS

Study participants were recruited among subjects referring to the High Specialization Center for the Care of Obesity (CASCO), Department of Experimental Medicine, "Sapienza" University of Rome, and from the Italian Hospital Group, "Villa Pia," Guidonia, Italy, from January 2015 to February 2017. The study was approved by the ethical committee of the Sapienza University of Rome, Policlinico Umberto I, and was concordant with Helsinki Declaration. Each patient gave a written informed consent before admission to the study.

Over the 2 year recruitment period, a total of 50 patients with anorexia nervosa (AN), 400 obese individuals and 150 normalweight consecutive subjects were screened. After screening, 122 patients were included. 21 underweight patients with AN based on the diagnostic criteria of the DSM-5 (3 males, 18 females, age range 16–68 year, BMI range 10.63–20.23 Kg/m<sup>2</sup> ); 26 normalweight control individuals (7 males, 19 females, age range 20–59 year, BMI range 20.22–24.83 kg/m<sup>2</sup> ); 75 patients affected by obesity (19 males, 56 females, age range 18–65 year, BMI range 31.36–59.0 kg/m<sup>2</sup> ). The subjects were excluded either on the basis of the criteria reported below or declined to participate. A portion of the obese and normal-weight patients were included in two previous studies (7, 8).

The exclusion criteria were: uncontrolled hypertension, heart diseases, lung diseases, type 1 diabetes, uncontrolled type 2 diabetes, corticosteroids for systemic use, any medication potentially affecting body weight or body composition, cirrhosis and other chronic liver diseases, acromegaly, hypothyroidism, acute illness, current or past presence of hepatitis B surface antigen and antibody to hepatitis C virus, excessive alcohol intake (≥140 g/week for men or 70 g/week for women).

All patients underwent complete medical examination and anthropometric measurements [body weight (kg), height (m), waist circumference (WC) at the level of umbilicus (cm)]. Body weight was measured by Tanita BWB-800A digital medical scale (Tanita Corporation, Arlington Heights, IL, USA). BMI was calculated by the formula weight (kg)/height(m)<sup>2</sup> .

Fasting plasma glucose (FPG, mg/dl) and insulin (mU/L), total cholesterol (TC, mg/dl), high-density lipoprotein (HDL) cholesterol (mg/dl), low-density lipoprotein (LDL)-cholesterol (mg/dl), triglycerides (TG, mg/dl), erythrocyte sedimentation rate (ESR, mm/h), C-reactive protein (CRP, µg/L), fibrinogen (g/L), and SIRT1 (ng/ml) were assessed after a 12-h overnight fast. Plasma samples for SIRT1 analyses were frozen at −80◦C until measurement. Because intermittent fasting might influence the circulating levels of SIRT1, a special attention was paid to withdrawing the blood at the same 12 h time interval from the last meal for all patients. Dual energy X-ray absorptiometry (DXA) body composition, echocardiographic epicardial fat thickness measurements (mm) and liver adiposity by ultrasound were also recorded.

The estimated BMR value was calculated using the Harris & Benedict equation and expressed in kcal/day. Following the equations for men and women:

$$\text{Men} = 66.4 + 13.75 \times \text{(Wt)} + 5 \times \text{(Ht)} - 6.8 \times \text{(Age)}$$

$$\text{Women} = 655 + 9.6 \times \text{(Wt)} + 1.85 \times \text{(Ht)} - 4.7 \times \text{(Age)}$$

#### SIRT1 Assay

SIRT1 was determined by a monoclonal antibody-based ELISA method using a commercially available human SIRT1 ELISA kit (MyBioSource, Cod. GDMBS705558) with an inter- and intraassay coefficient of variation of 10 and 8%, respectively, and a detection limit of 0.039 ng/mL.

Microtiter plates were coated with equal amount of primary mouse anti-human SIRT1 monoclonal IgG. 100 µL standard and plasma samples were pipetted in each well and the protocol was followed by using secondary avidin conjugated horseradish peroxidase. The formation of horseradish peroxidase was measured at 405 nm using ELISA reader (Quanta Biotech, UK). Seven different concentrations of purified SIRT1 (0.15, 0.312, 0.625, 1.25, 2.5, 5.0, and 10 ng/mL) were used to plot a standard curve. A calibration curve was added to each plate used.

# Body Composition Evaluation by DXA Analysis

DXA was performed by one single experienced technician using a DXA scan (Hologic Inc., Bedford, MA, USA, QDR 4500 W). The coefficient of variation for fat mass (FM) was <1.5%. Body composition was measured in the whole body and, with the use of specific anatomic landmarks determined by a standard software (Hologic Inc., S/N 47168 VER. 11.2), in the trunk, which included neck, chest, abdominal, and pelvic areas. The upper perimeter was the inferior edge of the chin and the lower borders intersect the middle of the femoral necks without touching the brim of the pelvis. Scans were performed according to the manufacturer's instructions.

### Determination of Liver Adiposity

The determination of liver fat content was based on liver-kidney contrast measured with ultrasonography by one single trained radiologist with extensive experience in abdominal ultrasound examinations. The analysis was carried out using a Esaote Medica apparatus equipped with a convex 3.5 MHz probe (Esaote MyLab40, Esaote Europe B.V., The Netherlands). The severity of liver adiposity was based according to the brightness of the liver estimated as a numerical value: 0 = absent; 1 = mild lipid accumulation; and 2 = moderate/severe lipid accumulation.

# Echocardiographic Epicardial Fat Thickness Measurements

Epicardial Fat Thickness (EFT) was measured through a validated echocardiographic procedure (21). Participants underwent highresolution M-B-mode transthoracic echocardiography using a 2.5-MHz probe, and spectral Doppler exam of the common carotid artery using a 7.5- MHz probe (Esaote MyLab40, Esaote Europe B.V., The Netherlands). The EFT was identified as the echo-free space between the outer wall of the myocardium and the visceral layer of the pericardium, and its thickness was measured perpendicularly on the free wall of the right ventricle (RV) at end-systole in three cardiac cycles. The average value of three cardiac cycles from each echocardiographic view was considered. All echocardiograms were recorded by the same experienced operator who was blinded to the other study data.

#### Statistical Analysis

Variables were expressed as mean ± SD. Differences between groups were analyzed using Student's T-test. A matrix correlation among variables was calculated. Each variable, in relation to SIRT1, was tested by the use of regression analyses, taking into account sex and age for their potential confounding effect. Violations of normality of the regression models were tested through the Shapiro-Wilk test. In the stepwise regression analysis, we included significant (p < 0.05) predictors from linear regression along with variables deemed important, a priori, on clinical grounds. To avoid colinearity, the correlation between variables was assessed and the more clinically relevant variable of a pair of highly correlated variables was included. To arrive to a parsimonious model, covariates were selected with a stepwise regression procedure using backward elimination. The parameters selected were age, sex, waist circumference, EFT, liver steatosis, HDL-cholesterol, ESR, and basal metabolic rate. All p-values presented were two-tailed, and values <0.05 were considered statistically significant. Data were analyzed with the use of STATISTICA software, version 6.1 (Stat Soft, Inc., Tulsa, Oklahoma).

# RESULTS

The characteristics of the study population, stratified according to the patients BMI, are summarized in **Table 1**. Statistical significances presented for participant characteristics are all obtained from unadjusted analysis. The mean BMI was 16.22 ± 2.44 kg/m<sup>2</sup> , 23.39 ± 1.24 kg/m<sup>2</sup> , and 40.95 ± 6.83 kg/m<sup>2</sup> in anorexic patients, normal weight and obesity group, respectively. WC was constantly ≥80 cm in females and ≥94 cm in males affected by obesity. The differences in weight, BMI, WC, EFT, total-FM%, trunk-FM% were statistically significant (p < 0.001) across the groups. BMR was significantly higher in obese patients compared to underweight (p < 0.0001) and normal-weight (p < 0.0001) patients and between underweight and normalweight patients as well (p < 0.05).

# Circulating SIRT1 Levels

Underweight patients showed the highest values of SIRT1 followed by normal-weight and obese individuals. The differences in SIRT1 levels were statistically significant between obese subjects and both normal-weight (p = 0.002) and underweight patients (p < 0.0001).

# Fat Amount and Distribution

The characteristics of the adiposity of the patients are summarized in **Table 1**. EFT, total FM % and truncal FM % were significantly reduced in underweight patients compared to both normal-weight subjects and patients affected by obesity (p < 0.001).

Both underweight and obese patients had an abnormally high accumulation of liver fat evaluated by ultrasonography compared to normal-weight. However, the degree of liver steatosis was significantly lower in underweight patients (mild degree) compared to obese patients (moderate/severe degree) (p < 0.0001).

#### Metabolic and Inflammatory Parameters

There were important metabolic differences between the categories of patients (**Table 1**). FPG was lower in underweight patients compared to normal-weight and obese patients (p < 0.0001). As expected, the highest basal insulin was found in the obesity group. The differences in insulin levels between underweight and normal-weight subjects (p = 0.013), and between normal-weight and obese patients (p = 0.002) were statistically significant.

LDL-cholesterol levels were comparable in underweight and normal-weight patients, while obese individuals showed higher values of both total and LDL-C (p < 0.05). Indeed,

TABLE 1 | Demographic, anthropometric and clinical characteristics of the patients.


*SIRT1, sirtuin1; BMI, body mass index; WC, waist circumference; EFT, epicardial fat thickness; FPG, fasting plasma glucose; HDL-C, high-density lipoprotein-cholesterol; LDL-C, low-density lipoprotein-cholesterol; ESR, erythrocyte sedimentation rate; CRP, Creactive protein; BMR, basal metabolic rate. Values are expressed as means* ± *SD.* \**The severity of liver adiposity was based according to the brightness of the liver estimated as a numerical value: 0* = *absent; 1* = *mild lipid accumulation; and 2* = *moderate/severe lipid accumulation. For each variable, missing values were* < *2%. Information on missing values is therefore not provided in the table.*

HDL-C was higher in anorexic patients compared to normalweight (p < 0.0001) and obese (p < 0.0001), while there were not differences between normal-weight and obese patients (p = 0.534).

Analogously, the triglycerides concentrations did not differ between underweight and normal-weight subjects (p = 0.45), but were significantly higher in patient affected by obesity compared to normal-weight individuals (p = 0.03). All the markers of inflammation followed a clear pattern with a statistical significant increase from underweight, to normal-weight, to obese patients.

#### Regression Analysis

**Table 2** shows the regression analysis results for each predictor variable in relation to SIRT1 adjusted for age and sex. SIRT1 was inversely associated with EFT, total FM%, liver steatosis, body weight, BMI, and WC. Concerning the metabolic variables, SIRT1 was negatively associated with LDL-cholesterol, insulin, and BMR. Finally, SIRT1 was inversely correlated with ESR.

There was no significant association between SIRT1 and triglycerides, HDL-cholesterol, fasting glycaemia, trunk FM%, fibrinogen and CRP.

Given that metabolic and inflammatory markers are influenced by degree of adiposity, we ran an additional set of analyses that included adjustment for WC, beyond age and sex, to assess whether the associations observed for SIRT1 were independent from adiposity. We found that the association between SIRT1 and either inflammatory (ESR, CRP, fibrinogen) or metabolic (FPG, insulin, HDL-cholesterol, LDL-cholesterol, triglycerides) parameters was abolished once adjusted for WC, suggesting that the major drive for the variation of circulating SIRT1 levels is the adiposity per se (data not shown). WC was


*BMI, body mass index; WC, waist circumference; EFT, epicardial fat thickness; FPG, fasting plasma glucose; HDL-C, high-density lipoprotein-cholesterol; LDL-C, low-density lipoprotein-cholesterol; ESR, erythrocyte sedimentation rate; CRP, C-reactive protein; BMR, basal metabolic rate.*

adjusted for because WC is a reliable representative of adiposity and SIRT1 expression parallels visceral fat.

#### Backward Stepwise Regression Analysis

Multivariate stepwise regression analysis was used to identify factors that influence circulating SIRT1 across AN, obese and normal-weight groups. We considered only a sub-set of the variables initially tested in linear regressions for the backward stepwise analysis (age, sex, WC, EFT, liver steatosis, HDLcholesterol, ESR and BMR), depending on both preliminary statistics and clinical appraisal.

The results from the analysis provide the set of independent variables that best explain the variance in plasma SIRT1 levels in the current sample, although the results are limited by the small sample size. In the study population, age and EFT were the sole determinants of circulating SIRT1 with a β-coefficient of −0.026 (p = 0.025) and −0.516 (p = < 0.001), respectively, and a R 2 value of 0.3698 (**Table 3**).

#### DISCUSSION

In this study, we compared the circulating levels of SIRT1 in condition of deficiency, normal content or excess body fat in underweight, normal-weight, and obese patients. We found a significant negative correlation between plasma SIRT1 and adipose tissue, with the highest levels observed in participants with extremely reduced fat content (**Figure 1**). This observation is novel and opens new questions dealing with the regulation of SIRT1 production and its function in relation to adipose tissue.

Several studies have provided insights into the mechanisms underlying endocrine, metabolic, and adaptive changes in states of chronic starvation (22).

SIRT1 concentrations were negatively correlated with insulinemia and LDL-cholesterol. This is likely explained by the compelling evidence that SIRT1 overexpression offers substantial benefits on serum cholesterol and insulin levels and increased resistance to high-fat diet induced glucose intolerance and insulin resistance (23–25). No association was found between SIRT1 levels and triglycerides, although contradictory results have been obtained with resveratrol, a potent SIRT1 activator, that was found either to reduce (26) or to have no effects (27) on plasma triglycerides.

TABLE 3 | Stepwise multiple regression analysis results to identify predictor variables associated with circulating SIRT1 (ng/ml).


*EFT, epicardial fat thickness. Variables included in the starting model for circulating SIRT1 were: age, sex, waist circumference, EFT, liver steatosis, high-density lipoprotein cholesterol, erythrocyte sedimentation rate and basal metabolic rate. We left in the model all variables that met the 0.15 significance level for entry into the stepwise model. R-squared value* = *0.3698.*

SIRT1 is an essential negative inflammatory regulator in high-fat diet or alcohol induced fatty liver diseases, mainly through deacetylating NF-κB and down-modulating NF-κB transcriptional activity, thereby reducing macrophage infiltration and pro-inflammatory cytokines production in the liver as well as in the adipose tissue (28). Indeed, in line with other studies (29, 30), we found that underweight patients were less inflamed compared to normal-weight and obese patients and ESR was inversely associated to the pattern of SIRT1 and proportional to fat mass. This coincides with the assumption that obese patients, generally, show a pro-inflammatory phenotype and express less SIRT1 than lean subjects.

It is worth to be mentioned that additional adjusted regression models for WC, a reliable predictor of visceral adiposity, abolished the association between SIRT1 and inflammatory and metabolic parameters, indicating that fat content is the most relevant determinant of SIRT1 circulating levels in this study.

In the stepwise regression analysis, epicardial fat, out of all the markers of adiposity included in the study whose expression is linearly associated with SIRT1, is the variable most strongly associated with SIRT1. The strict association between EFT and blood SIRT1 was not unexpected, being already seen previously (8). But, then again, human studies have shown that SIRT1 is expressed in visceral adipose tissue and reduced by obesity (31) and echocardiographic measurement

circulating levels of SIRT1 and weight, fat abundance and distribution, inflammatory, and metabolic parameters in underweight, normal weight, and obese individuals.

of epicardial fat can provide a more specific and sensitive measurement of intraabdominal visceral fat (32). Therefore, a possible explanation for the preferred association between EFT and SIRT1 might depend on the robust representativeness of epicardial fat as visceral fat as opposed to other measures of adiposity used and SIRT1. In line with previous studies (8), partly based on the same study sample, the inverse relationship between SIRT1 and EFT adds a new potential mechanism to the evidence supporting the role of epicardial adipose tissue in the development of atherosclerosis and its complications, inflammation, and metabolic syndrome in obese patients.

Furthermore, it is relevant the negative association seen between age and circulating SIRT1 levels. This observation confirms what previously reported in a comprehensive study aimed at identifying the pattern of serum SIRT1 activity according to age (33).

Individuals with AN have lower resting energy expenditure than normal-weight controls (34) and CR is a powerful stimulus for SIRT1 activation (18), likely an adaptive mechanism to preserve energy for vital functions. Accordingly, in SIRT1 gain-of-function transgenic mice, SIRT1 behaves as a "thrifty gene" that protects against metabolic diseases by instructing the organism to limit energy consumption and expenditure (23). Although our data are purely associative in nature, they seem to confirm the hypothesis that SIRT1 levels have the tendency to match with energy saving since the higher the SIRT1 values the lower the BMR values. Further studies to reveal the relationship between SIRT1 and BMR are warranted.

Although dysregulations of peripheral adipokines, gutsecreted peptides and central neurotransmitters involved in appetite modulation have been detected in patients with AN (35), the significance of these derangements for the development, course and prognosis of eating disorders is still not clear. Actually, there are no conclusive data as to whether alterations of feeding regulatory substances precede the appearance of an eating disorder or are the consequence of the nutritional aberrations occurring in the disorder. It has been suggested, although not definitively proved, that those alterations, even when secondary to malnutrition and/or to aberrant eating behaviors, might contribute to the genesis and the maintenance of some symptomatic aspects of AN, thus affecting the course and the prognosis of the disease. Whether the high levels of circulating SIRT1 in AN individuals is a consequence of the feeding behavior of these patients and whether they may modulate eating-related or non-eatingrelated psychopathological aspects of AN deserve to be deeply investigated. Interestingly, hypothalamic SIRT1 stimulates food intake and weight gain (36), raising the hypothesis that forms of AN might associate with SIRT resistance. These considerations may be an interesting starting point to study whether SIRT resistance might play a role in the pathogenesis of AN.

SIRT1 is found in a wide range of tissues and organs, highly expressed in liver and adipose tissue and regulated by nutritional status. In general CR stimulates SIRT expression (17, 18, 37) while high calorie diet reduces it. Thus, SIRT1 tissue expression and activity is influenced by the availability of energy suggesting that SIRT1 could have a role in the regulation of normal energy balance. Accordingly, plasma SIRT1 levels and fat mass are inversely regulated, with SIRT1 concentrations being increased in a catabolic condition and decreased in conditions of extreme BMIs.

Remarkably, CR dependent changes occur in a highly tissuespecific manner, as demonstrated by comparing circadian gene expression in the liver vs. epidermal and skeletal muscle stem cells (38). De novo oscillating genes under CR show an enrichment in SIRT1 targets in the liver due to enhanced SIRT1 activity (39). Therefore, we hypothesize that the increased circulating SIRT1 levels recorded in severely underweight patients may reflect the reorganization of metabolic pathway linked to SIRT1 in the liver of calorie restricted anorexic individuals.

The measurement of the circulating SIRT1 in severely underweight patients may provide new pathogenetic hypothesis for some of the features of AN.

Limitations of our study are the relatively small number of study subgroups and the use of calculated BMR values. Moreover, males and females were not analyzed separately because of the scarcity of males in our sample. We recognize that our results and conclusions are based on observational data and that the associations between SIRT1 levels and the variables measured do not establish causative roles. The strength of our study is the separation and comparison of different weight subgroups.

In conclusion, circulating SIRT1 inversely parallels the entire spectrum of fat phenotype, basal metabolic rate, inflammatory status, and eating behavior from anorexia to obesity through normal weight.

### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the guidelines, Ethical Committee of the University of Rome La Sapienza with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the ethical committee of the Sapienza University of Rome, Policlinico Umberto I.

# AUTHOR CONTRIBUTIONS

SM and LG conceived the project, developed the overall research plan, and wrote the manuscript. PM, AP, and AS selected the patients and collected the patients' data. MdG made the ultrasound determination of liver adiposity. GB made the echocardiographic epicardial fat thickness measurements. SB and SC made the circulating SIRT1 assay. SM, EP, and AR made the statistical analysis. SM, LG, AG, CL, LD, and AL interpreted the data and critically revised the manuscript for important intellectual content. All authors read and approved the final manuscript.

# FUNDING

Sapienza University of Rome, Ricerca di Ateneo 2016.

#### REFERENCES


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Mariani, di Giorgio, Martini, Persichetti, Barbaro, Basciani, Contini, Poggiogalle, Sarnicola, Genco, Lubrano, Rosano, Donini, Lenzi and Gnessi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Slim Body Weight Is Highly Associated With Enhanced Lipoprotein Functionality, Higher HDL-C, and Large HDL Particle Size in Young Women

Ki-Hoon Park 1,2,3†, Dhananjay Yadav 1,2,3†, Suk-Jeong Kim1,2,3, Jae-Ryong Kim<sup>4</sup> and Kyung-Hyun Cho1,2,3 \*

*<sup>1</sup> Department of Medical Biotechnology, Yeungnam University, Gyeongsan, South Korea, <sup>2</sup> Research Institute of Protein Sensor, Yeungnam University, Gyeongsan, South Korea, <sup>3</sup> LipoLab, Gyeongsan, South Korea, <sup>4</sup> Department of Biochemistry and Molecular Biology, Smart-Aging Convergence Research Center, College of Medicine, Yeungnam University, Daegu, South Korea*

#### Edited by:

*Sudip Bajpeyi, The University of Texas at El Paso, United States*

#### Reviewed by:

*Andrew James Murphy, Baker Heart and Diabetes Institute, Australia Brie Sorrenson, University of Auckland, New Zealand Andras G. Lacko, University of North Texas Health Science Center, United States C. Roger White, University of Alabama at Birmingham, United States*

\*Correspondence:

*Kyung-Hyun Cho chok@yu.ac.kr*

*†Co-first authors.*

#### Specialty section:

*This article was submitted to Obesity, a section of the journal Frontiers in Endocrinology*

Received: *09 March 2018* Accepted: *29 June 2018* Published: *19 July 2018*

#### Citation:

*Park K-H, Yadav D, Kim S-J, Kim J-R and Cho K-H (2018) Slim Body Weight Is Highly Associated With Enhanced Lipoprotein Functionality, Higher HDL-C, and Large HDL Particle Size in Young Women. Front. Endocrinol. 9:406. doi: 10.3389/fendo.2018.00406* There has been no information about the correlations between body weight distribution and lipoprotein metabolism in terms of high-density lipoproteins-cholesterol (HDL-C) and cholesteryl ester transfer protein (CETP). In this study, we analyzed the quantity and quality of HDL correlations in young women (21.5 ± 1.2-years-old) with a slim (*n* = 21, 46.2 ± 3.8 kg) or plump (*n* = 30, 54.6 ± 4.4 kg) body weight. Body weight was inversely correlated with the percentage of HDL-C in total cholesterol (TC). The plump group showed 40% higher body fat (26 ± 3 %) and 86% more visceral fat mass (VFM, 1.3 ± 0.3 kg) than the slim group, which showed 18 ± 2% body fat and 0.7 ± 0.2 kg of VFM. Additionally, the plump group showed 20% higher TC, 58% higher triglyceride (TG), and 12% lower HDL-C levels in serum. The slim group showed 34% higher apoA-I but 15% lower CETP content in serum compared to the plump group. The slim group showed a 13% increase in particle size and 1.9-fold increase in particle number with enhanced cholesterol efflux activity. Although the plump group was within a normal body mass index (BMI) range, its lipid profile and lipoprotein properties were distinctly different from those of the slim group in terms of CETP mass and activity, HDL functionality, and HDL particle size.

Keywords: body weight, blood pressure, lipoproteins, apoA-I, HDL-cholesterol

# INTRODUCTION

Maintenance of a slim body shape is greatly desired for health and beauty purposes, especially in young women. However, there has been no information about the correlations between body weight distribution and lipoprotein metabolism. It is well known that serum high-density lipoprotein-cholesterol (HDL-C) level is inversely correlated with the incidence of coronary artery disease (1). HDL plays important roles in promoting anti-atherosclerotic, anti-diabetic, and anti-thrombotic activities (2) in serum and interacts with many antioxidant enzymes such as paraoxonase (3, 4). However, HDL can be altered under various health conditions (5), such as changed dietary patterns (6), pathogen infection (7), and environmental stress (8). Therefore, HDL-C may be a good biomarker for the diagnosis of many diseases and disease progression by monitoring changes in its antioxidant and anti-inflammation abilities (9). However, it is well known that dysfunctional HDL is frequently found in patients with metabolic diseases such as obesity (10).

Although overweight status and obesity are associated with low HDL-C levels, there has been no report characterizing HDL particles from overweight subject. In the context of metabolic syndrome, obese persons (body mass index, BMI >30) with insulin resistance were shown to have low HDL-C levels, specifically 41% in women (<45 mg/dL), and 31% in men (<35 mg/dL) (11). Reduction of the HDL-C level is linked with increased very low-density lipoprotein (VLDL) production via elevation of cholesteryl ester transfer protein (CETP) activity. Further, obese subjects often show increased CETP activity (12, 13).

Apolipoprotein A-I (apoA-I), the major protein of HDL, exerts antioxidant and anti-inflammatory activities in lipid-free and lipid-bound states along with cholesterol efflux activity (14) in macrophages and insulin secretion activity in pancreatic beta cells. During the progression of obesity or insulin resistance in early metabolic syndrome, tumor necrosis factor-α (TNF-α) is up-regulated while apoA-I production is down-regulated. On the other hand, apoA-I gene expression can be down-regulated by increased production of adipokines and inflammatory cytokines (15, 16).

Progression of overweight status is associated with various complications such as hypertension, hyperglycemia, and coronary artery disease mediated via several mechanisms, including reduced HDL-C and elevated total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), and triglyceride (TG) levels (17). Although a high quantity of serum lipids has been well established as a cause of obesity and hypertension in adolescents, there has been no report on the quality of lipids and lipoproteins according to body weight. Furthermore, there has been no report comparing lipoprotein quality between two groups (slim and plump) showing a slight difference in body weight within a normal range, whereas lipid and lipoprotein properties of obese patients have been relatively well investigated.

In the current study, we compared lipoprotein biomarkers between slim (around 46 kg) and plump (around 55 kg) young women groups, although both are considered as having a normal body weight. Between the plump and slim type, lipid and lipoprotein qualities were compared to identify new biomarkers contributing to a slim body shape in young women.

#### MATERIALS AND METHODS

#### Subjects

We randomly recruited healthy female volunteers (n = 51) enrolled at Yeungnam University in 2015. Heavy alcohol consumers (>30 g EtOH/day) and those who had consumed any prescribed drugs to treat hyperlipidemia, diabetes mellitus, or hypertension were excluded. All subjects had unremarkable medical records without illicit drug use or past history of systemic diseases. The study was approved by the Institutional Review Board at Yeungnam University (Gyeongsan, South Korea) endorsed the protocol (IRB #7002016-A-2016-021) and the participants signed an informed consent form prior to research commencement.

### Anthropometric Analysis

Blood pressure was measured each morning at 2-week intervals by an Omron HEM-1000 (Kyoto, Japan). Height, body weight, BMI, total body fat (%), total body fat mass (kg), and visceral fat mass (VFM) (kg) were measured individually at the same time of day at 2-week intervals using an X-scan plus II body composition analyzer (Jawon Medical, Gyeongsan, Korea).

#### Vascular Stiffness Analysis

Systolic and diastolic blood pressures were measured in triplicate using a semi-automated non-invasive oscillometric sphygmomanometer using the SphygmoCor system (Actor Medical, Sydney, Australia) following a 5 min rest period (18). Pulse wave was analyzed with SphygmoCor software. Augmentation pressure is the gain in aortic pressure above its first systolic push, which is thought to be determined by pressure wave reflection but may be influenced by ventricular ejection aspects. However, the augmentation index, the ratio of augmentation pressure to aortic or central pulse pressure, is widely used as a measure of pressure wave reflection (19, 20). Measurements were performed by a technician trained in the technique and blinded to the characteristics of each subject.

#### Plasma Analysis

Blood was obtained from all subjects following overnight fasting. Blood was collected using a vacutainer (BD Biosciences, Franklin Lakes, NJ, USA) containing EDTA (final concentration, 1 mM). Plasma was isolated by low-speed centrifugation (3,000 rpm) and stored at −80◦C until analysis. To analyze plasma, TC, TG, HDL-C, glucose, aspartate aminotransferase (AST), and alanine aminotransferase (ALT) levels were measured using commercially available kits (Cleantech TS-S; Wako Pure Chemical, Osaka, Japan).

#### Characterization of Lipoproteins

Very low-density lipoprotein (VLDL, d <1.019 g/mL), LDL (1.019 <d <1.063), HDL<sup>2</sup> (1.063 < d <1.125), and HDL<sup>3</sup> (1.125 < d <1.225) were isolated from pooled plasma of each group via sequential ultracentrifugation (21), and the density was adjusted by addition of NaCl and NaBr in accordance with standard protocols. Samples were centrifuged for 22 h at 10◦C at 100,000 g using a Himac CP-100NX (Hitachi, Tokyo, Japan) using P50AT4-0124 rotor in our laboratory. Protein concentrations of lipoproteins were determined via Lowry protein assay, as modified by Markwell et al. (22). Expression levels of apoA-I (28 kDa) and apo-B (550 kDa) were determined by SDS-PAGE.

To assess the degree of lipoprotein oxidation, the concentration of oxidized species in lipoproteins was determined by the thiobarbituric acid-reacting substance (TBARS) assay method using malondialdehyde as a standard (23). To compare the extent of glycation between groups, advanced glycation end products (AGEs) in lipoproteins were determined from reading fluorometric intensities at 370 nm (excitation) and 440 nm (emission), as our previous report (24), using a spectrofluorometer LS55 (Perkin Elmer, Shelton, CT, USA) with the WinLab software package (version 4.0).

#### Cholesteryl Ester Transfer Protein Assay

A rHDL-containing apoA-I and cholesteryl oleate were synthesized in accordance with the method described by Cho (25) and Cho et al. (26) using trace amounts of [3H]-cholesteryl oleate (TRK886, 3.5 µCi/mg of apoA-I; GE Healthcare). The rHDL was immobilized using CNBr-activated Sepharose 4B resin (Amersham Biosciences) for easy separation after the reaction in accordance with the manufacturer's instructions. CE transfer reaction was performed in 300-µL reaction mixtures containing human serum (20 µL) or HDL<sup>3</sup> (20 µL, 2 mg/mL) as a CETP source, [3H]-rHDL-agarose (20 µL, 0.25 mg/mL) as a CE-donor, and human LDL (20 µL, 0.25 mg/mL) as a CE-acceptor. After incubation of 4 h at 37◦C, the reaction was halted via brief centrifugation (10,000 g) for 3 min at 4◦C. The supernatant containing CE-acceptor (150 µL) was then subjected to scintillation counting, and percentage transfer of [3H]-CE from [3H]-rHDL to LDL was calculated.

# Ferric Reducing Ability of Plasma Assay and Paraoxonase Activity

The ferric reducing ability of plasma (FRAP) was determined using the method described by Benzie and Strain (27). The antioxidant activities of individual HDL fractions were estimated by measuring increases in absorbance produced by generated ferrous ions. Paraoxonase-1 (PON-1) activity was then determined by measuring the initial velocity of pnitrophenol production at 37◦C, as determined by measuring the absorbance at 405 nm (microplate reader, Bio-Rad model 680; Bio-Rad, Hercules, CA, USA), as described previously (28) with slight modifications (29). Prior to the measurement, HDL was extensively dialyzed against PBS to remove EDTA.

#### Phagocytosis of LDL Into Macrophages

THP-1 cells, a human monocytic cell line, were obtained from the American Type Culture Collection (ATCC, #TIB-202TM, Manassas, VA, USA) and maintained in RPMI-1640 medium (Hyclone, Logan, UT) supplemented with 10% fetal bovine serum until needed. Cells that had undergone no more than 20 passages were incubated in medium containing phorbol 12-myristate 13 acetate (PMA, 150 nM) in 24-well plates for 48 h at 37◦C in a humidified incubator (5% CO2, 95% air) in order to induce differentiation into macrophages. Differentiated and adherent macrophages were then rinsed with warm PBS, followed by incubation with 450 µL of fresh RPMI-1640 medium containing 0.1% FBS and 50 µg of each LDL (1 mg of protein/mL in PBS) for 48 h at 37◦C in a humidified incubator. After incubation, cells were washed with PBS three times and then fixed in 4% paraformaldehyde for 10 min. Next, fixed cells were stained with oil-red O staining solution (0.67%) and washed with distilled water. THP-1 macrophage-derived foam cells were then observed and photographed using a Nikon Eclipse TE2000 microscope (Tokyo, Japan) at 400× magnification. Cell media (0.2 mL) were then analyzed by TBARS assay to evaluate changes in levels of oxidized species using malondialdehyde (MDA) as a standard.

# Anti-atherogenic Activity of HDL<sup>3</sup>

Differentiated and adherent macrophages were then rinsed with warm PBS and incubated with 400 µL of fresh RPMI-1640 medium containing 0.1% fetal bovine serum, 50 µg of oxLDL (1 mg of protein/mL in PBS), and 30 µg of HDL<sup>3</sup> (2 mg of protein/mL in PBS) from each group for 48 h at 37◦C in a humidified incubator. After incubation, cells were stained with oil-red O solution (0.67%) to visualize the amount of lipid species in cells. THP-1 macrophage-derived foam cells were then observed and photographed using a Nikon Eclipse TE2000 microscope (Tokyo, Japan) at 400× magnification. The stained area was quantified via computer-assisted morphometry using Image Proplus software (version 4.5.1.22, Media Cybernetics, Bethesda, MD).

#### Cholesterol Efflux

THP-1 cells were incubated in medium containing phorbol 12-myristate 13-acetate (PMA, 150 nM) in a plate for 48 h at 37◦C in a humidified incubator in order to induce differentiation into macrophages. Macrophages were treated with radiolabeled cholesterol (0.1 µCi of [3H]-cholesterol) in RPMI 1640 medium (Hyclone, Logan, UT) containing 1% fetal bovine serum (Hyclone, Logan, UT) per well (0.5 mL) for 48 h. Media containing the isotope was saved and replaced with fresh media containing 0.3 mM 8-(4-chlorophenylthio) cyclic adenosine monophosphate (cAMP, Cat# C3912, Sigma-Aldrich, St. Louis, MO) to up-regulate the cellular cholesterol pump (ATP-binding cassette transporter 1, ABCA1) for 18 h. After removal of media containing cAMP, 28 µg of HDL<sup>3</sup> was added and the sample incubated with serum-free media (0.5 mL) for 24 h. Subsequently, cell media (0.5 mL) in individual wells were collected into a 1.7 mL tube. Cells were then rinsed with PBS three times and dissolved in 0.2 mL of RIPA buffer (50 mM Tris-HCl [pH 8.0], 150 mM NaCl, 5 mM EDTA [pH 8.0], 1% NP-40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate) to lyse cells. An aliquot of the cell lysate (0.1 mL) was mixed with a scintillation cocktail (3 mL) to quantify the amount of isotope indicating uptake of cholesterol into cells. After scintillation counting of [3H]-cholesterol in cells and media, the amount of effluxed cholesterol from cells was calculated using the following formula as previously reported (30):

% Cholesterol efflux = (media counts × dilution factor) media count × dilution factor + cell lysis count <sup>×</sup> dilution factor <sup>×</sup> <sup>100</sup> % Net efflux = % cholesterol efflux with HDL3 − % blank efflux without HDL3

#### ELISA and Western Blotting

The quantity of apoA-I in serum was determined in each group by ELISA using a commercially available kit (Quantikine ELISA, DAPA10, R&D systems, Minneapolis, MN). In order to quantify serum CETP, each well of a polystyrene microplate (#3590; Corning Inc., Corning, NY, USA) was coated with anti-human CETP rabbit antibody (ab19012; Abcam, Cambridge, UK) at a concentration of 0.25µg/mL, followed by incubation overnight at 4◦C. Equally, diluted serum sample was incubated for 2 h at room temperature. After extensive washing, antihuman CETP mouse antibody (ab2726; Abcam, 1µg/mL) was added and the sample incubated for 2 h at room temperature. To develop the color reaction, anti-mouse IgG antibody (ab6728; Abcam, 0.5µg/mL) conjugated with horseradish peroxidase was added. For color development, 3,3′ ,5,5′ tetramethylbenzidine (TMB) substrate solution (Cat. No. 555214; BD Biosciences, Flanklin Lakes, NJ, USA) was added and quantified using a Victor X4 microplate reader (Perkin Elmer, Waltham, MA).

Apolipoprotein/lipoprotein compositions were compared via sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) using identical protein loading quantities (3 µg of total protein per lane) from individual HDL3, and expression levels of apolipoprotein were analyzed via immunodetection. Antihuman apoA-I antibody (ab7613), anti-paraoxonase antibody (ab24261), and anti-IL-6 antibody (ab6672) were purchased from Abcam (Cambridge, UK). Relative band intensity (BI) was compared via band scanning with Chemi-Doc <sup>R</sup> XRS + (Bio-Rad, Hercules, CA) using Image Lab software (Version 5.2).

#### Electron Microscopy

Transmitted electron microscopy (TEM) was performed with a Hitachi electron microscope (model H-7600; Ibaraki, Japan) operated at 80 kV as in our previous report (31). HDL<sup>2</sup> were negatively stained with 1% sodium phosphotungstate (pH 7.4) with a final apolipoprotein concentration of 0.3 mg/mL in TBS.

#### Zebrafish and Embryos

Wildtype zebrafish and embryos were maintained according to standard protocols. Zebrafish maintenance and experimental procedures were approved by the Committee of Animal Care and Use of Yeungnam University (Gyeongsan, Korea). Zebrafish and embryos were maintained in a system cage (3 L volume, acrylic tank) and 6-well plates, respectively, at 28◦C during treatment under a 14:10 h light:dark cycle.

# Microinjection of Zebrafish Embryos

To compare antioxidant and anti-inflammatory activities of HDL between slim and plump, HDL<sup>3</sup> from each group was injected into zebrafish embryos, as in our previous report (32). Embryos at 1-day post-fertilization (dpf) were individually microinjected using a pneumatic picopump (PV820; World Precision Instruments, Sarasota, FL, USA) equipped with a magnetic manipulator (MM33; Kantec, Bensenville, IL, USA) and a pulled microcapillary pipette-using device (PC-10; Narishigen, Tokyo, Japan). To minimize bias, injections were performed at the same position on each yolk. Filter-sterilized solution of each LDL or HDL<sup>3</sup> was injected into flasks of embryos. Following injection, live embryos were observed under a stereomicroscope (Motic SMZ 168; Hong Kong) and photographed using a Motic cam2300 CCD camera.

# Imaging of Reactive Oxygen Species (ROS)

Injection of HDL<sup>3</sup> from slim and plump group into the zebrafish embryo in the presence of ox LDL was done and changes in reactive oxygen species (ROS) levels in larvae were imaged by dihydroethidium (DHE; cat # 37291, BioChemika) staining, as previously described (33). Images were obtained by fluorescence microscopy (Ex = 588 nm and Em = 605 nm) on a Nikon Eclipse TE2000 instrument (Tokyo, Japan). To avoid bias, red fluorescence was measured in the trunk area away from the injection site. Quantification of the stained area was carried out via computer-assisted morphometry using Image Proplus software (version 4.5.1.22, Media Cybernetics, Bethesda, MD).

#### Data Analysis

All data are expressed as the mean ± SD from at least three independent experiments with duplicate samples. Data comparisons were assessed by Student's t-test using the SPSS program (version 14.0; SPSS, Inc., Chicago, IL, USA). In the human study, data in the same group were evaluated via one-way analysis of variance (ANOVA) using SPSS (version 14.0; Chicago, IL, USA), and differences between the means were assessed using Duncan's multiple-range test. Statistical significance was defined as p < 0.05.

# RESULTS

# BMI and Serum HDL-C

All participants were very similar in age (21.5 ±1.2 years old) and height (161.5 ± 4.2 cm). Participants were of normal body weight and BMI but were divided into a plump group (around 55 kg of BW) and slim group (around 46 kg of BW).

The plump group showed a significantly higher BMI, body fat percentage, and VFM than the slim group, as shown in **Table 1**. The systolic blood pressure and diastolic blood pressure were not found to be significant between the slim and plump group.

# Serum Lipid Level and Body Shape

The plump group showed 1.2- and 1.6-fold higher serum TC and TG levels, respectively, than the slim group, although all TC and TG levels were within their normal ranges (**Table 2**). However, the plump group showed 12% lower HDL-C and 1.7 fold higher TG/HDL-C than the slim group. The slim group showed 1.3-fold higher % HDL-C and serum apoA-I content than the plump group. Serum uric acid level of the plump group was slightly higher than that of the slim group. However, there was no significant difference in serum glucose, uric acid, or AST/ALT level between the groups.

# CETP Mass and Activity

Interestingly, the plump group showed significantly higher CETP mass than the slim group (**Table 2**). Furthermore, there was a significant difference in serum CETP activity (p = 0.0202)

#### TABLE 1 | Anthropometric data.


\*\*\**p* < *0.001 vs. plump group.*

between the slim and plump groups although around 10% of activity was different (**Table 2**).

#### Correlation of HDL-C and Body Shape

As shown in Supplementary Figure 1A, height generally increased with elevation of % HDL-C. Although there was no statistical difference, height and % HDL-C showed positive correlations. However, % HDL-C was inversely correlated with body weight, as shown in Supplementary Figure 1B. Correlations between BMI and % HDL-C (**Figure 1A**) and between % HDL-C and % body fat mass (**Figure 1B**) also showed clear inverse relationships. Visceral fat mass was inversely correlated with % HDL-C (p < 0.0001, Supplementary Figure 2A). However, visceral fat was positively correlated with CETP mass (Supplementary Figure 2B).

#### Correlation of HDL-C and Blood Pressure

Both systolic and diastolic blood pressures were negatively correlated with % HDL-C as shown in Supplementary Figure 3. Diastolic blood pressure and % HDL-C showed steeper slopes than systolic blood pressure, suggesting that % HDL-C was more closely associated with lowered diastolic blood pressure (p = 0.0031). The plump group showed 5–6 mmHg higher diastolic and systolic blood pressures than the slim group.

The plump group showed a remarkably high central augmentation pressure and index with an arterial stiffness reference age of 30 ± 13.9-years-old, whereas the slim group showed a reference age of 20.8 ± 2.3-years-old (**Table 1**).

#### Expression Level of apoA-I in HDL

To compare the expression level of apoA-I in HDL, the same amount of total proteins per individual HDL was loaded on



\**p* < *0.05;* \*\**p* < *0.01;* \*\*\**p* < *0.001 vs. plump group.*

each lane of a gel. As shown in **Figure 2**, the slim group showed 2.2- and 1.3-fold higher expression levels of apoA-I in HDL<sup>2</sup> and HDL3, respectively, compared to the plump group as determined by SDS-PAGE. Based on Western blotting with an apoA-I antibody, the slim group showed 1.5- and 1.6-fold higher apoA-I levels in HDL<sup>2</sup> and HDL3, respectively, compared to the plump group.

# HDL<sup>2</sup> Particle Size and Cholesterol Efflux

The slim group showed a 15% larger HDL particle size and 1.9-fold higher number of particles than the plump group as shown in **Figure 3A** and a representative photo of **Figure 3** from TEM analysis. HDL<sup>2</sup> from the slim group showed significantly enhanced cholesterol efflux activity from macrophages (around 24 ± 2% efflux), whereas the plump group showed 21 ± 1% efflux (p = 0.018) during 24 h of incubation of each HDL as shown in **Figure 3B**.

#### Antioxidant Ability

Serum PON activity of the slim group was 1.3-fold higher than that of the plump group up to 60 min of incubation, as shown in **Figure 4A**. Serum FRAP ability was not significantly different between the groups, and the two groups showed saturated reduction potential at the same incubation time, as shown in **Figure 4B**. The slim group showed higher HDL-associated antioxidant enzyme activity, although their antioxidant potentials in serum were similar. This result suggests that HDL from the slim group showed enhanced functionality.

The slim group also showed higher HDL-associated antioxidant activity than the plump group (**Figure 5**). Regarding FRAP ability, the slim group showed 22% higher reduction potential than the plump group during 20 min of incubation (**Figure 5A**). Additionally, the slim group showed 1.7-fold higher PON activity than the plump group during 60 min of incubation (**Figure 5B**).

# Compositions of Lipoproteins

As shown in **Figure 6A**, the plump group showed 1.9- and 1.3 fold higher protein contents in VLDL and LDL, respectively, but 20% lower protein content in the HDL<sup>2</sup> fraction. The lower protein content in HDL<sup>2</sup> from the plump group was strongly associated with reduced serum apoA-I content (**Table 2**). However, the plump group showed 10–13% higher glycation extent in all lipoproteins, especially VLDL and LDL (**Figure 6B**). At the same protein amount, the plump group showed 9% higher cholesterol content in VLDL and LDL (**Figure 6C**). However, the plump group showed 27% less cholesterol content in HDL3, whereas there was no difference in HDL2- C. Interestingly, the plump group showed elevated TG contents in all lipoprotein fractions, especially 1.3- and 1.2-fold higher TG contents in VLDL and HDL3, respectively. HDL<sup>3</sup> from the plump group showed higher TG and lower cholesterol contents than the slim group (**Figure 6D**), suggesting functionality of lipoproteins may be impaired at the beginning of overweight status.

# Modification of Lipoproteins

LDL from the plump group showed faster electromobility, as shown in **Figure 7A**, due to putative modifications such as oxidation and glycation. The plump group showed greater migration distance (1.3 ± 0.2 cm) than the slim group (0.9 ± 0.1 cm). TBARS assay was performed to determine oxidized products, showed all lipoproteins from the plump group showed higher MDA contents (**Figure 7B**). Especially, the plump group showed 45% more MDA content in HDL<sup>2</sup> compared to the slim group.

Cupric ion treatment revealed that the plump group was 1.3 fold more sensitive to oxidation than the slim group (**Figure 7C**) during 150 min of incubation. After incubation, Cu2+-treated LDL from the plump group showed faster electromobility with

FIGURE 3 | Cholesterol efflux activity and a representative photo of negatively-stained HDL2 and HDL3 from slim and plump groups (electron microscopy). All micrographs are shown at a magnification of 40,000×. The scale bar corresponds to 100 nm. (A) Shows measured particle size of HDL and particle number in the designated area. (B) Shows cholesterol efflux activity from macrophages.

fainter band intensity as shown in the photo of **Figure 7C**, suggesting that LDL from the plump group was more susceptible to oxidation.

#### Uptake of LDL Into Macrophages

After 48 h of treatment with each LDL, the plump group showed 2-fold higher oil-red O-stained area than the slim group

(Supplementary Figure 4), suggesting that LDL from the plump group was taken up 2-fold more into macrophages. This result supports the earlier findings that LDL from the plump group was more glycated (**Figure 6B**) and oxidized (**Figure 7**), as modified LDL was more phagocytosed by macrophages to accelerate foam cell production.

#### Embryo Survivability

Microinjection of LDL from the plump group caused the most severe embryo death up to 69% survival, whereas LDL from the slim group resulted in 75% embryo survival at 48 h postinjection, as shown in Supplementary Figure 5B. This difference in embryo survival can be attributed to more oxidized in LDL from the plump group (**Figure 7**). DHE staining revealed that plump LDL-injected embryos showed the strongest red intensity, suggesting the highest production of reactive oxygen species (Supplementary Figures 5A,C).

Injection of oxLDL caused the lowest embryo survival (around 44%), whereas PBS injection resulted in 84% survival, as shown in the graph of Supplementary Figure 6. In the presence of oxLDL, co-injection of HDL<sup>3</sup> from the slim group enhanced survival (around 70%), whereas HDL<sup>3</sup> from the plump group resulted in lower survivability (around 59%). oxLDL alone-injected embryos showed the strongest red fluorescence, which was 3-fold higher than PBS-injected embryos. However, HDL<sup>3</sup> from the slim and plump groups caused 25 and 16% reduction of red fluorescence, respectively, suggesting that HDL from the slim group was more effective in preventing embryonic death and ROS production mediated by oxLDL. Taken together, the plump group showed more dysfunctional HDL with smaller particle size, reduced antioxidant ability, greater atherogenic properties, such as higher CETP activity and mass with more inflammatory properties in human cells and zebrafish embryos.

# DISCUSSION

To our knowledge, this is the first report demonstrating differences in lipoproteins and lipid levels between young women with slim and plump body weights. It has been well established that overweight and obese adolescents are worldwide epidemic issue (34) due to the prevalence of hypertension, cardiovascular disease, and metabolic syndrome, etc. However, there are few drugs capable of treating obesity long-term other than orlistat (Xenical) and lorcaserin (Belviq), which have several side effects such as physical and emotional disorders (35).

Elevated BP is the most frequent comorbidity identified in overweight adolescents (36). Our current results clearly show that % HDL-C was inversely correlated with body weight, body fat mass, and blood pressure (Supplementary Figures 1–3). However, there were no correlations with the amount of HDL-C (mg/dL), suggesting that % HDL-C in TC is a reliable biomarker to predict body fat mass and blood pressure (**Table 2**). There is little information about lipoprotein profiles and obesity in young women. Furthermore, there has been no report about the lipid and lipoprotein profiles of young women with a moderately plump body weight. Until now, there have been limitations in interpreting the physiological roles of HDL in health and beauty since most studies have focused on the quantity of HDL-C in order to explain its beneficial activities.

Recently, the importance of HDL functionality has been well investigated to prevent CVD and metabolic syndrome via protection of LDL from modification (37). In our study, the slim group showed enhanced antioxidant activity in both serum and HDL (**Figures 4**, **5**) with less oxidation of LDL (**Figure 7**). Although the plump group showed lower protein content in HDL2, its proteins were more glycated and modified (**Figure 6**). Further, although the two groups showed a difference of 8.4 ± 2 kg in body weight and the plump group had a normal BMI, their lipid profiles and lipoprotein properties were distinctly different in terms of CETP mass and HDL functionality.

Elevated levels of CETP activity was found to be a major determinant of the atherogenic dyslipidemia, and atherosclerotic cardiovascular disease (38). Our study have also reported an elevated level of CETP mass and %CE-transfer in the plump group which was found to be significant when compared with the slim group. CETP is that enzyme accountable for moving cholesterol esters and triglycerides between, LDL, VLDL, and HDL. Lower level of CETP nurture HDL formation. Considering higher HDL levels are associated with lower risk of atherosclerosis, the elevated level of CETP is thought to play in promoting the disease by lowering in HDL-C (39). CETP mass activity was directly correlated with body weight and VFM (Supplementary Figure 2B), whereas apoA-I mass was inversely correlated with VFM (**Tables 1**, **2**). In a Japanese study, obese children showed increased plasma CETP activity as well as lower apoA-I content and higher serum TG levels (40). Another study on obese Japanese adults (BMI 33.1 kg/m<sup>2</sup> ) reported increased serum TG and reduced HDL-C levels via elevation of CETP activity (12). Further, obese women with BMIs around 35.1 ± 0.7 kg/m<sup>2</sup> showed 1.22-fold higher CETP content in serum compared to lean women (22.4 ± 0.3 kg/m<sup>2</sup> ) along with 1.7 fold higher serum TG and 23% lower HDL-C levels (41). These reports are in good agreement with our current results in terms of lowered HDL-C and increased TG serum levels via elevation of CETP mass, although subjects in the current study were not obese. The higher percentage of HDL-C in the slim group was associated with lower body fat and VFM contents. In both groups, the percentages of HDL-C and apoA-I contents were inversely correlated with blood pressure. VLDL and LDL from the plump group showed 13 and 30% more glycated end products and TG.

Injection of LDL from the slim group into zebrafish embryos resulted in 75% survival, whereas the overweight group showed 70% survival, indicating that LDL from the plump group had more inflammatory properties. The lowest embryo survivability (around 44%) was seen after injecting oxLDL while PBS injection lead to 84% survivability. The co-injection of HDL<sup>3</sup> and oxLDL in the slim group enhanced the survivability up to 70% in the zebrafish embryos, although the oxLDL and HDL<sup>3</sup> from the plump group resulted a lower survivability of 59%. The study observed that the HDL<sup>3</sup> from the slim and plump groups elicited 25 and 16% reduction in the red fluorescence that suggested a protective role of HDL extracted from slim group in the embryonic death and ROS production.

Further, the slim group showed 22% higher antioxidant activity in serum and HDL as well as 26% higher apoA-I content in HDL than the plump group. ApoA-I content in HDL from the slim group was 2-fold higher while LDL from the slim group was less oxidized compared to the plump group. LDL from the plump group showed 2-fold greater uptake into macrophages, whereas HDL from the plump group did not prevent ox-LDL phagocytosis by macrophages.

Although obese patients are often associated with low serum HDL-C and antioxidant activities, there has been no information about HDL particle number and size. The current results show that changes in obese status were associated with smaller HDL particle size and number as well as lowered apoA-I content. In addition to HDL, LDL showed a smaller particle size with increased susceptibility to oxidation. It is well known that PON-1 hydrolyzes lipid peroxides in order to prevent oxidation of LDL. Reduced HDL-associated PON-1 activity in the plump group (**Figure 5**) can explain the increase in LDL oxidation (**Figure 7**) and phagocytosis by macrophages (Supplementary Figure 4). Similar to our current results, obese patients displaying obstructive sleep apnea showed reduced serum PON-1 activities with elevated oxLDL levels, inflammation, and endothelia dysfunction (42). Impairment of antioxidant ability is well correlated with reduction of apoA-I content in HDL (43). It has been reported that obesity and metabolic syndrome are frequently associated with reduced apoA-I and elevated apoC-III contents in HDL along with decreased particle size (44). Plasma apoA-I level has been shown to be inversely associated with obesity in the Framingham Offspring Study conducted on 4,260 young adult men and women (45). Although apoA-I level was lower in the plump group in the current study, there was no difference in apoC-III level since the plump group was not obese.

In older men, diet-induced weight loss has been shown to be associated with increased HDL-C and apoA-I levels as well as reduced body fat mass and CETP activity. Among body fat mass areas, intra-abdominal fat (IAF) and abdominal subcutaneous fat (SQF) were shown to be inversely correlated with apoA-I content, suggesting that body fat mass and distribution are intimately associated with HDL functionality (46). Furthermore, older men showed an increased LDL particle size, whereas HDL particle size was not reported.

Obese subjects show smaller LDL and HDL particle sizes (47). In a Turkish study, obese children showed lower serum HDL-C and higher TG levels than the control despite similar TC. They also showed lower PON activity, increased BMIs, and reduced HDL-C content (48), which are in good agreement with the current results. To prevent obesity progression, HDL should be functionally enhanced as well as increased in particle size via reduction of CETP activity. A cohort study studying the hearts of young overweight subjects (2,017 participants, aged 12–15-yearsold) reported higher serum C-reactive protein, serum amyloid A, and CETP activities. Interestingly, the overweight group showed 1.3-fold higher HDL2-associated CETP activity, whereas there was no difference in HDL3-associated CETP activity (49), similar to our current results.

#### CONCLUSIONS

In conclusion, despite having normal BMIs, the plump group showed more atherogenic features in LDL and HDL as well as elevated oxidation and glycation with loss of antioxidant ability and apoA-I. Elevation in the CETP might be a key player in HDL metabolism during obesity progression, accumulating in visceral fat to impair the lipoprotein levels, antioxidant activity, and lipoprotein functionality. It is necessary to monitor the quality of HDL in terms of composition, structure and functional properties in order to increase the particle size and antioxidant activity.

#### REFERENCES


#### ETHICS STATEMENT

Informed consent was obtained from all participants prior to enrollment in the study, and the Institutional Review Board at Yeungnam University (Gyeongsan, South Korea) approved the protocol (IRB #7002016-A-2015-003).

#### AUTHOR CONTRIBUTIONS

K-HP, DY, S-JK performed the experiments. J-RK analyzed the data. K-HC wrote the manuscript and supervised the whole project.

#### ACKNOWLEDGMENTS

This work was supported by a grant from the Medical Research Center Program (2015R1A5A2009124) through the National Research Foundation (NRF), funded by the Ministry of Science, ICT and Future Planning of Korea.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo. 2018.00406/full#supplementary-material


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Park, Yadav, Kim, Kim and Cho. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Twelve Weeks of Yoga or Nutritional Advice for Centrally Obese Adult Females

Shirley Telles\*, Sachin K. Sharma, Niranjan Kala, Sushma Pal, Ram K. Gupta and Acharya Balkrishna

*Patanjali Research Foundation, Haridwar, India*

Background: Central obesity is associated with a higher risk of disease. Previously yoga reduced the BMI and waist circumference (WC) in persons with obesity. Additional anthropometric measures and indices predict the risk of developing diseases associated with central obesity. Hence the present study aimed to assess the effects of 12 weeks of yoga or nutritional advice on these measures. The secondary aim was to determine the changes in quality of life (QoL) given the importance of psychological factors in obesity.

Material and Methods: Twenty-six adult females with central obesity in a yoga group (YOG) were compared with 26 adult females in a nutritional advice group (NAG). Yoga was practiced for 75 min/day, 3 days/week and included postures, breathing practices and guided relaxation. The NAG had one 45 min presentation/week on nutrition. Assessments were at baseline and 12 weeks. Data were analyzed with repeated measures ANOVA and *post-hoc* comparisons. Age-wise comparisons were with *t*-tests.

#### Edited by:

*Deanne Helena Hryciw, Griffith University, Australia*

#### Reviewed by:

*Olivia Holland, Griffith University, Australia Lannie O'Keefe, Victoria University, Australia*

> \*Correspondence: *Shirley Telles shirleytelles@gmail.com*

#### Specialty section:

*This article was submitted to Obesity, a section of the journal Frontiers in Endocrinology*

Received: *28 May 2018* Accepted: *30 July 2018* Published: *17 August 2018*

#### Citation:

*Telles S, Sharma SK, Kala N, Pal S, Gupta RK and Balkrishna A (2018) Twelve Weeks of Yoga or Nutritional Advice for Centrally Obese Adult Females. Front. Endocrinol. 9:466. doi: 10.3389/fendo.2018.00466* Results: At baseline and 12 weeks NAG had higher triglycerides and VLDL than YOG. Other comparisons are within the two groups. After 12 weeks NAG showed a significant decrease in WC, hip circumference (HC), abdominal volume index (AVI), body roundness index (BRI), a significant increase in total cholesterol and LDL cholesterol. YOG had a significant decrease in WC, sagittal abdominal diameter, HC, BMI, WC/HC, a body shape index, conicity index, AVI, BRI, HDL cholesterol, and improved QoL. With age-wise analyses, in the 30–45 years age range the YOG showed most of the changes mentioned above whereas NAG showed no changes. In contrast for the 46–59 years age range most of the changes in the two groups were comparable.

Conclusions: Yoga and nutritional advice with a diet plan can reduce anthropometric measures associated with diseases related to central obesity, with more changes in the YOG. This was greater for the 30–45 year age range, where the NAG showed no change; while changes were comparable for the two groups in the 46–59 year age range. Hence yoga may be especially useful for adult females with central obesity between 30 and 45 years of age.

#### Trial registration: (CTRI/2018/05/014077).

Keywords: central obesity, anthropometry, anthropometric indices, risk of disease, yoga, nutritional advice

# BACKGROUND

The Indian Council of Medical Research-India Diabetes (ICMR-INDIAB) study is an ongoing cross sectional national study of the prevalence of obesity, diabetes, and hypertension across the Indian sub-continent (1). In Phase 1 of the study 16,000 individuals aged 20 years and above were sampled from the whole of India. Generalized, abdominal, and combined obesity (i.e., generalized as well as abdominal) were determined based on the BMI and waist circumference using the World Health Organization Asia Pacific Guidelines (2, 3). The ICMR-INDIAB study reported a higher prevalence of isolated abdominal obesity compared to generalized obesity across India. This finding was given importance considering the association independent of ethnicity, between central obesity and several chronic noncommunicable diseases such as cardiovascular disease, type 2 diabetes mellitus and certain cancers (1).

Lifestyle changes which include increased physical activity, a healthy diet and a positive attitude have been recommended for the management of obesity (4). For people who are obese and physically less active, physical activity may seem challenging (5). This would prevent persons who are obese from initiating and adhering to increased physical activity. The practice of yoga has been adopted for weight control because of its increasing popularity and relative safety while supervised (6, 7). A systematic meta-analysis assessed the effects of yoga for weight related outcomes in overweight and obese persons assessed in randomized controlled trials (RCTs) (7). Thirty trials with a total of 2,173 participants from bibliographic databases such as MEDLINE, Scopus, and Cochrane Library were screened from their inception to March 2016. The risk of bias was assessed using the Cochrane risk of bias tool which demonstrated methodological drawbacks in the 30 trials. Nonetheless the authors of the meta-analysis concluded that yoga can be considered a safe and effective intervention to reduce the BMI in overweight and obese individuals.

A single study demonstrated a decrease in central obesity in 60 female participants after 12 weeks of yoga (8). This inference was based on the waist circumference. There are several other anthropometric measures and derived anthropometric indices which have been shown to predict the risk of developing diseases associated with central obesity (9).

The aim of this study was to determine if in adult females with central obesity a 12 week program of yoga or of nutritional advice could (i) alter anthropometric measurements associated with a risk of developing diseases associated with central obesity and (ii) positively influence the lipid profile and quality of life.

#### MATERIALS AND METHODS

#### Participants

Fifty two healthy Asian Indian adult females with central obesity with ages between 30 and 59 years (group average age ± SD; 43.98 ± 6.89 years) were recruited for the trial. The trial profile is provided in **Figure 1**. The sample size was determined from a previous study on centrally obese females (8). A required sample size (n = 20) was obtained using Cohen's formula for the effect size of 0.40 calculated from the mean and SD values of waist circumference which were changed significantly after 12 weeks of yoga, alpha at 0.05, powered at 0.90 using G power software (10). Participants were recruited through advertisements in local newspapers and flyers distributed in nearby residential areas and hospitals. Participation in the study was voluntary with no remuneration. The inclusion criteria were (i) waist circumference ≥ 80 cm (11), (ii) BMI ≥ 25 kg/m<sup>2</sup> (12), (iii) ages between 30 and 59 years, and (iv) willingness to take part in the study. The exclusion criteria were (i) obesity secondary to hormonal imbalance, medication such as steroids or secondary to any other medical condition, (ii) any physical or psychological disability which would have prevented the participants from taking part in the yoga program or attending the nutritional advice session, (iii) involvement in any other dietary or exercise program during the 12 months prior to, at the time of or during the study, and (iv) any co-morbidities associated with obesity such as cardiovascular disease, type-2 diabetes mellitus, or hypertension. No participant was excluded from the trial for the above mentioned reasons. The participants' written signed informed consent was taken. The baseline characteristics of the participants are given in **Table 1**.

## Study Design

The present single blind comparative controlled trial was carried out between April and August 2016 where assessors were blinded to the group to which the participants belonged. The yoga group practiced yoga for 75 min/day, for 3 consecutive days in a week, over a 12 week period. Along with this they were given a diet plan for 1,900–2,000 Kcal/day developed based on the guidelines from the National Institute of Nutrition, India (13). The nutritional advice group was given lectures on nutrition (one 45 min lecture/week) and the same diet plan for 1,900–2,000 Kcal/day developed based on the guidelines from the National Institute of Nutrition, India (13). Adherence to yoga was based on the attendance in the yoga class noted by the yoga teacher. The compliance to the diet plan was based on the 24 h diet recall questionnaire which was administered at the start and end of the 12 week period. This did not cover the dietary intake during the rest of the time which is a limitation. Both groups were assessed for levels of physical activity using the International Physical Activity Questionnaire—Short Form administered at the beginning and end of 12 weeks. The study had the approval of the ethics committee of the institute, which was formed and based on the guidelines of the Indian Council of Medical Research and is in accordance with the Helsinki Declaration (Approval number: YRD/016/022). The trial is registered with the Clinical Trials Registry of India (CTRI/2018/05/014077). The present study is part of a larger nationwide trial of which the data are still being analyzed, comparing yoga with nutritional advice in different regions of India.

At the start of the trial an attempt was made to convince participants to be randomly allocated to yoga or nutritional advice groups. However participants had time and practical constraints. Hence based on their convenience they were assigned to either intervention. However several participants of the nutritional advice group mentioned that they were interested to learn yoga later on.

TABLE 1 | The baseline characteristics of the participants.


#### Assessments

The following assessments were carried out by individuals who were blinded to the group to which the participants belonged.

### Anthropometry

#### Waist Circumference (WC)

Participants were lightly clothed and asked to stand upright with their feet 25–30 cm apart and weight evenly distributed on both feet. The tape measure which was used for assessments (Gülick Anthropometric tape Model J00305, Lafayette Instrument, U.S.A.) was fitted around the abdomen without compressing soft tissue. The waist circumference was measured to the nearest 0.1 cm in a horizontal plane midway between the inferior costal margin and the iliac crest.

#### Sagittal Abdominal Diameter (SAD)

The participant was asked to lie supine on their back. The caliper used to measure the supine sagittal abdominal diameter has two arms attached to a vertical scale [Holtain-Kahn Abdominal Caliper 50 cm (98.609XL), U.K.]. The standard method was followed by which the lower arm of the caliper was placed under the participant (14). After a normal exhalation, the upper arm of the caliper was lowered to reach the mid-point between the inferior costal margin and the iliac crest. The lower arm was repositioned if required. The reading on the vertical scale between the upper and lower arms of the caliper gave the supine sagittal abdominal diameter in cm.

#### Hip Circumference (HC)

The hip circumference was measured around the pelvis at the point of maximal protrusion of the buttocks. The ratio of the waist circumference to the hip circumference was derived and is a ratio of the fat stored centrally inside the abdomen (waist circumference) and fat stored peripherally (hip circumference).

#### Derived Anthropometric Indices

Seven anthropometric indices were derived from direct measurements using standardized formulae (9, 15–18). These indices are as follows.

#### Body Mass Index (BMI)

The body mass index (BMI) was calculated as the body weight (in kg), in light clothing and without footwear, divided by the height (in meters squared). The accuracy of the weighing machine (Model DS 215 N, Essae-Teraoka Pvt. Ltd, Bengaluru, India) was up to 0.05 kg. The height was measured to the nearest 0.1 cm. BMI = Weight (kg)/ Height <sup>2</sup> (m).

#### Waist-Hip Ratio (WHR)

WHR = WC (cm)/HC (cm).

#### A Body Shape Index (ABSI) (9)

ABSI = WC (m)/[BMI <sup>2</sup>/<sup>3</sup> (kg/m<sup>2</sup> ) Height ½ (m)] where WC and height are expressed in m and BMI in kg/m<sup>2</sup> .

#### Conicity Index (CI) (9)

CI = 0.109 <sup>−</sup><sup>1</sup> WC (m) [Weight (kg) / Height (m)] <sup>−</sup>1/<sup>2</sup> where WC is measured in cm, weight in kg and height in m.

#### Abdominal Volume Index (AVI) (9)

AVI = [2WC<sup>2</sup> (cm) + 0.7 (WC - HC) <sup>2</sup> (cm)]/1000 where WC and HC (hip circumference) are expressed in cm.

#### Visceral Adiposity Index (VAI) (15)

VAI female = {WC (cm)/[36.58 +1.89BMI (kg/m<sup>2</sup> )]} [TG (mmol/L)/0.81] [1.52/HDL (mmol/L)]

where WC is expressed in cm, BMI in kg/m<sup>2</sup> , Triglycerides in mmol/L, and HDL in mmol/L.

#### Body Roundness Index (BRI) (9)

BRI = 364.2 – 365.5 [1 –π <sup>−</sup><sup>2</sup> WC <sup>2</sup> (m) Height <sup>−</sup><sup>2</sup> (m)] <sup>1</sup>/<sup>2</sup> where WC and height are expressed in m.

#### Biochemical Measures

Antecubital venous blood samples were collected under sterile conditions. Total cholesterol, triglycerides, high density, and low density lipoprotein cholesterol were estimated using appropriate enzymes followed by spectrophotometry.

#### 24 h Diet Recall (Energy Intake/Day)

A structured interview was carried out to recall the food and fluids which the participants had consumed in the 24 h prior to the study. The method was as follows: participants were asked (i) to recall and list the foods they had consumed during the 24 h preceding the assessment, (ii) the method of preparation (e.g., raw, cooked, boiled, or baked), (iii) the size of utensils used, the interviewer had four types of utensils (i.e., a bowl, cup, glass, and spoon), each of which had four sizes. Participants were asked to indicate the size used by them. The volume of each of the utensils of different sizes had already been determined (19). The total energy intake/day as well as the amount of macronutrients was calculated based on norms for Indian foods (13, 19, 20).

### Level of Physical Activity and Energy Expenditure/Day

Physical activity was assessed using the International Physical Activity Questionnaire Short Form (IPAQ) (21). Data obtained at the beginning and end of the 12 week period for both groups were extracted using the guidelines published by the IPAQ for data processing (22). From these data the basal metabolic rate of each participant was determined and energy expenditure/day was calculated using the Harris-Benedict equation (23).

### Quality of Life

The Moorehead-Ardelt Quality of Life Questionnaire was used to assess six aspects of the quality of life (QoL) (24). These were general self-esteem, physical activity, social contacts, satisfaction concerning work, pleasure related to sexuality, and focus on eating behavior, with scores ranging from −0.5 to +0.5. The sum of these 6 scores provided a total QoL score. Each score was classified into 5 categories (very poor: −3.0 to −2.1; poor: −2.0 to −1.1; fair: −1.0 to +1.0; good: 1.1 to 2.0: and very good: 2.1 to 3.0).

#### Interventions

#### Nutritional Advice

Participants received a 45 min presentation on nutrition (1 presentation/week) for 12 weeks. The person who gave the presentation had a minimum of 12 years of education including 2 years of training in science. The presenter used slides made by the research institute conducting the trial. The 12 topics of the presentations were: (i) five basic food groups, (ii) vegetarian diet, (iii) proteins, (iv) fats, (v) carbohydrates, (vi) dietary fiber, (vii) vitamins, (viii) minerals, (ix) probiotics, (x) iron deficiency, (xi) calcium, and (xii) antioxidants.

#### Yoga

The yoga intervention consisted of (i) a universal prayer (3 min)<sup>1</sup> , (ii) yoga postures (asanas, 42 min), (iii) voluntarily regulated breathing techniques (pranayamas, 24 min), and (iv) guided relaxation with meditation (6 min). Yoga classes were conducted on three successive days in a week between (i) 05:30 h and 06:45 h, during the 12 weeks. Attendance was noted in each yoga class by the yoga instructor. The instructor was also asked to note any adverse event during the classes. The yoga instructor had been teaching yoga for 5 years. All the participants in the yoga group attended at least 18 out of a total of 36 yoga classes. Details of the yoga intervention are given in **Table 2**.

#### Diet Plan

Participants of both groups were given a diet plan for a balanced diet of 1900–2000 Kcal/day. The diet plan included fruits, vegetables, lentils, complex carbohydrates, and dairy products. Hence it was a lacto-vegetarian diet. The ratio of carbohydrates, protein and fats in the diet was based on the guidelines for a balanced diet for an Indian population published by the National Institute of Nutrition, India (13), which states that a balanced Indian diet should have 10–15% of total calories from proteins, 20–30% of calories from fats and 50–60% of calories from carbohydrates. The total energy and nutritive values of the foods were determined from a database of Indian foods (20).

**<sup>1</sup>**May all be prosperous and happy May all be free from illness May all see what is uplifting May no one suffer Peace, peace, peace

#### TABLE 2 | Details of the yoga intervention.



*Those practices which are commonly used are not detailed.*

# Data Analysis

#### Group as a Whole

The data obtained at the beginning and end of the 12 week period for the two groups were compared with a repeated measures analysis of variance (ANOVA) followed by multiple post-hoc comparisons which were Bonferroni adjusted. The ANOVA had one Within subjects factor i.e., States, with two levels (pre and post) and one Between subjects factor i.e., Groups which were (i)

#### TABLE 3 | Anthropometric variables.


*Values are group mean*± *SD.*

\**p* < *0.05;* \*\**p* < *0.01;* \*\*\**p* < *0.001, post-hoc analyses with Bonferroni adjustment, post values compared with pre.*

#*p* < *0.05, post-hoc analyses with Bonferroni adjustment, pre values compared with pre.*

@*p* < *0.05, post-hoc analyses with Bonferroni adjustment, post values compared with post.*

*BMI, Body mass index; SAD, Sagittal abdominal diameter.*

#### TABLE 4 | Lipid profile.


*Values are group mean* ± *SD.*

\**p* < *0.05;* \*\**p* < *0.01, post-hoc analyses with Bonferroni adjustment, post values compared with pre.*

@@*p* < *0.01, post-hoc analyses with Bonferroni adjustment, post values compared with post.*

#*p* < *0.05, post-hoc analyses with Bonferroni adjustment, pre values compared with pre.*

TABLE 5 | Estimated energy intake/day based on 24 h diet recall questionnaire.


*Values are group mean* ± *SD.*

nutritional advice group with a diet plan of 1900–2000 Kcal/day and (ii) yoga group who had a similar diet plan of 1900–2000 Kcal/day, as the nutritional advice group.

#### Analysis Based on Age

Both nutritional advice and yoga groups were subdivided based on age as (i) participants between 30 and 45 years and (ii) those who were 46 years and above.

For (i) and (ii) there were separate between group comparisons of values at baseline and after 12 weeks using t-tests for unpaired data. Within the groups baseline data and date at 12 weeks were compared with t-tests for paired data.

# RESULTS

None of the participants reported any adverse events during the trial. At baseline the visceral adiposity index (VAI), levels of triglycerides and of VLDL differed significantly between the nutritional advice and yoga groups. The group mean values ± TABLE 6 | Estimated energy expenditure/day based on (i) International Physical Activity Questionnaire—Short Form and (ii) Harris-Benedict equation to determine the basal metabolic rate.


*Values are group mean* ± *SD.*

#### TABLE 7 | Quality of life.


*Values are group mean* ± *SD.*

\**p* < *0.05, post-hoc analyses with Bonferroni adjustment, post values compared with pre.*

SD for the (i) anthropometric measures and anthropometric indices, (ii) lipid profile, (iii) energy intake/day and energy expenditure/day, and (iv) quality of life are given in **Tables 3**–**7**.

#### Repeated Measures Analysis of Variance (RM-ANOVA)

The ANOVA values for the Within-Subjects factor (States), Between-Subjects factor (Groups) and interaction between the two are given below. The details of the ANOVA are available in the **Supplementary Material**.

#### Within-Subjects Factor (States)

There were significant differences between post and pre states for waist circumference (F = 36.92, p < 0.001), sagittal abdominal diameter (F = 9.66, p < 0.01), hip circumference (F = 34.41, p < 0.001), BMI (F = 14.88, p < 0.001), a body shape index (F = 20.77, p < 0.01), conicity index (F = 13.09, p < 0.001), abdominal volume index (F = 31.05, p < 0.001), body roundness index (F = 28.78, p < 0.001), total cholesterol (F = 5.71, p < 0.05), general self-esteem (F = 6.05, p < 0.05), and total quality of life (F = 7.80, p < 0.01). For all comparisons mentioned above the condition of sphericity was met and value of epsilon (Huynh-Feldt, Greenhouse-Geisser) was 1. The degrees of freedom for all variables were 1,50.

#### Between-Subjects Factors (Groups)

There were significant differences between the groups in a body shape index (F = 4.10, p < 0.05), triglycerides (F = 8.23, p < 0.01), and VLDL (F = 9.94, p < 0.01). The condition of sphericity was met for the comparisons and value of epsilon (Huynh-Feldt, Greenhouse- Geisser) was 1. The degrees of freedom for all variables were 1,50.

#### Interaction Between States and Groups

Interaction between States and Groups was significant for waist circumference (F = 5.11, p < 0.05), total cholesterol (F = 4.69, p < 0.05), LDL cholesterol (F = 5.77, p < 0.05), and HDL cholesterol (F = 5.44, p < 0.05). A significant interaction between States and Groups suggests that the effects of the two factors are interdependent. For all comparisons the condition of sphericity was met and value of epsilon (Huynh-Feldt, Greenhouse-Geisser) was 1. The degrees of freedom for all variables were 1, 50 (States) x 50 (Groups).

#### Post-hoc Analyses

There were two post-hoc comparisons: (i) Between groups [at baseline (pre) and at 12 weeks (post)] and (ii) Within groups, comparing values obtained at 12 weeks (post) with those at baseline (pre).

#### Post-hoc Between Groups Comparisons

At baseline (pre) the visceral adiposity index (VAI, p < 0.05), VLDL (p < 0.05) and triglyceride levels (p < 0.05) were significantly higher in the nutritional advice group compared to the yoga group. After 12 weeks (post) a body shape index (p < 0.01), the triglyceride levels (p < 0.01), and VLDL (p < 0.01) levels were higher in the nutritional advice group compared to the yoga group.

TABLE 8 | Anthropometric variables and derived indices for two age range i.e., 30-45 and 46-59 year.


*Values are group mean* ± *SD.*

\**p* < *0.05;* \*\**p* < *0.01;* \*\*\**p* < *0.001, paired t-test, post values compared with pre.*

@*p* < *0.05, Unpaired t-test, post values compared with post.*

#*p* < *0.05, Unpaired t-test, pre values compared with pre.*

*BMI, Body mass index; SAD, Sagittal abdominal diameter.*

#### Post-hoc Comparisons Within a Group (Post-Pre)

The nutritional advice group showed a significant decrease in the waist circumference (p < 0.05), hip circumference (p < 0.01), abdominal volume index (p < 0.01), body roundness index (p < 0.01), a significant increase in total cholesterol (p < 0.01), and LDL cholesterol (p < 0.05).

The yoga group had a significant decrease in waist circumference (p < 0.001), sagittal abdominal diameter (p < 0.01), hip circumference (p < 0.001), BMI (p < 0.001), waist/hip ratio (p < 0.05), a body shape index (p < 0.05), conicity index (p < 0.01), abdominal volume index (p < 0.001), body roundness index (p < 0.001), HDL cholesterol (p < 0.05), improved quality of life, i.e., focus on eating behavior (p < 0.05) and total quality of life (p < 0.05).

#### Age Wise Analysis (t-Tests)

The participants of both groups were divided as two groups based on age, viz., 30–45 years and 46–59 years. For each age group, comparisons were made between nutritional advice and yoga groups using t-tests for unpaired data and within each group using t-tests for paired data. The age wise group mean values ± SD for the (i) anthropometric measures and anthropometric indices, (ii) lipid profile, (iii) energy intake/day and energy expenditure/day, and (iv) quality of life are given in **Tables 8**–**12**.

#### Age Group 30–45 Years

#### Between Groups

At baseline the triglyceride levels and the focus on eating behavior (an aspect of the quality of life) were lower in the yoga group (p < 0.05, both cases) compared to the nutritional advice group. There were no differences at 12 weeks between groups.

#### Within Groups

There were no significant changes in the nutritional advice group. In the yoga group there was a decrease in waist circumference, sagittal abdominal diameter, hip circumference, BMI, a body shape index, conicity index, abdominal volume index, and body roundness index (p < 0.05 in all cases).

#### TABLE 9 | Lipid profile for two age range i.e., 30–45 and 46–59 year.


*Values are group mean* ± *SD.*

#*p* < *0.05, Unpaired t-test, pre values compared with pre.*

TABLE 10 | Estimated energy intake/day based on 24 h diet recall questionnaire for two different age range i.e., 30–45 and 46–59 year.


*Values are group mean* ± *SD.*

# Age Group 46–59 Years

#### Between Groups

At baseline the waist circumference and satisfaction concerning work (of the quality of life scale) were higher in the nutritional advice group compared to the yoga group (p < 0.05). At 12 weeks the hip circumference and satisfaction concerning work remained higher in the nutritional advice group compared to the yoga group (p < 0.05 both cases).

#### Within Groups

The nutritional advice group showed a decrease in waist circumference, sagittal abdominal diameter, hip circumference, waist-hip ratio, a body shape index, conicity index, abdominal volume index, and body roundness index (p < 0.05 in all cases). The yoga group had a significant decrease in waist circumference, sagittal abdominal diameter, hip circumference, BMI, a body shape index, abdominal volume index, body roundness index, and total quality of life (p < 0.05 in all cases). The significant results have been summarized in the **Supplementary Material**.

#### DISCUSSION

Following 12 weeks of nutritional advice there was a significant decrease in waist circumference, hip circumference, abdominal TABLE 11 | Estimated energy expenditure/day based on (i) International Physical Activity Questionnaire—Short Form and (ii) Harris-Benedict equation to determine the basal metabolic rate for two different age range i.e., 30-45 and 46-59 years.


TABLE 12 | Quality of life for two different age range i.e., 30–45 and 46–59 years.


*Values are group mean*±*SD.*

\**p* < *0.05, Paired t-test, post values compared with pre.*

@*p* < *0.05, Unpaired t-test, post values compared with post.*

#*p* < *0.05, Unpaired t-test, pre values compared with pre.*

volume index, and increase in total cholesterol and LDL cholesterol. The yoga group at the end of 12 weeks showed a decrease in waist circumference, sagittal abdominal diameter, hip circumference, BMI, waist-hip ratio, a body shape index, conicity index, abdominal volume index, body roundness index, HDL cholesterol, and better total quality of life. When both groups were considered as two age ranges (i.e., 30–45 and 46–59 years), the results were different. For the 30–45 years age group the nutritional advice group showed no change after 12 weeks whereas the yoga group showed most of the changes mentioned above for the group as a whole. In contrast to this for the 46–59 year age group, the nutritional advice and yoga groups showed comparable benefits with reduction in most-anthropometric measures and indices at 12 weeks. Hence yoga may be especially useful for adult females between 30 and 45 years of age.

The waist circumference, hip circumference, abdominal volume index, and body roundness decreased in both groups irrespective of age. The waist circumference correlates with increased risk of cardiovascular disease (25), the abdominal volume index has been correlated with impaired glucose tolerance and higher incidence of type 2 DM and metabolic syndrome (26). Body roundness index was associated with higher occurrence of non-alcoholic fatty liver disease (27). Hence the risk of these conditions could be considered to be lower in both groups at 12 weeks. The other changes in the yoga group also suggest a reduction in the risk of developing obesity associated diseases such as carotid artery stiffness (28) and insulin resistance (29), based on a decrease in SAD, cardiovascular disease (based on reduction in waisthip ratio and conicity index (30), and cardiometabolic disease (based on reduced a body shape index and BMI) (31). Hence these risk factors reduced in the yoga group irrespective of age.

The serum lipid profile was assessed using quantitative methods. The yoga group showed a significant decrease in HDL cholesterol. This reduction in HDL cholesterol has been seen in two other studies, in which obese participants received yoga for 6 days and 15 days (32, 33). In both studies participants consumed a plant based lacto vegetarian diet comparable to that of the present study. In another study, 12 weeks of yoga practice resulted in a significant decrease in total cholesterol, triglycerides and LDL levels with a non-significant increase in HDL levels (34). The participants' diet was not described. It has been observed that those diets which are most effective in reducing the risk of atherosclerosis are usually associated with the greatest decrease in protective HDL cholesterol levels (35– 37). These diets are typically plant based high fiber low fat diets. However this reduction in protective HDL cholesterol levels following such diets need not necessarily be harmful as separate studies have shown that even if HDL levels decrease, the anti-inflammatory efficacy of HDL cholesterol may be enhanced despite reduction in absolute levels of HDL (38, 39).

In the nutritional advice group there was a significant increase in total cholesterol, and LDL levels. The explanation for this increase is not clear as the nutritional advice group was given the same dietary instructions as the yoga group. Mental stress levels have a positive correlation with total cholesterol, triglycerides and LDL cholesterol levels (40–42). Practicing yoga is one of the methods for stress reduction. Even though the stress levels were not measured in the present study it may be speculated that in the absence of yoga intervention the nutritional advice group continued to experience higher stress levels which contributed to the increase in total cholesterol, and LDL cholesterol levels. Though this is a speculation it is supported by the results of the participants' response to the six items of the Moorehead-Ardelt Quality of Life Questionnaire.

The yoga group showed significantly higher scores in focus on eating behavior (an aspect of quality of life) and total quality of life after 12 weeks. Previously the Moorehead-Ardelt Quality of Life Questionnaire has been used to compare the quality of life in obese persons who were experienced in yoga compared to those without any yoga experience and demonstrated a better quality of life in the group with prior yoga experience (43). These findings are of importance as psychological wellbeing is important for the long term successful management of obesity (44).

The centrally obese participants of the present trial showed no differences in their energy intake or energy expenditure after 12 weeks irrespective of the group to which they belonged. Energy intake was derived from the 24 h diet recall questionnaire which does not give an accurate idea of the diet during the 12 week period. Hence though the present results suggest that the energy intake in a day did not differ significantly with the energy expenditure in a day, between groups after 12 weeks it must be emphasized that both energy intake/day and energy expenditure/day were assessed by qualitative methods which lack the accuracy and objectivity of quantitative assessments.

The main limitations of the present findings are the study design and small sample size. Both the yoga and nutritional advice groups were given their intervention based on convenience, though the nutritional advice group did express an interest to learn yoga at some stage after the trial. The ideal design would have been a randomized controlled trial but after recruitment it was clear that though the participants were motivated to learn yoga, for personal reasons such as time constraints they were unable to state that they could complete 12 weeks of yoga practice successfully. This point demonstrates the practical difficulties a person may have in learning and practicing any intervention. Apart from this, though the supine sagittal abdominal diameter through anthropometry is an acceptable method to measure visceral adipose tissue, magnetic resonance imaging (MRI), dual-energy x-ray absorptiometry (DEXA), and computed tomography (CT) scans would be more accurate (30). To distinguish between types of adipose tissue in central obesity these methods are essential. The other limitations include the lack of quantitative measures to assess energy expenditure/day and energy intake/day. Also the present sample included females alone, all of whom were generational vegetarians. Hence generalizing the findings cannot be done.

# CONCLUSIONS

Yoga and nutritional advice with a diet plan can reduce anthropometric measures associated with diseases related to central obesity, with more changes in the yoga group. This difference was greater for the 30–45 years age range, where the nutritional advice group showed no change; while changes were comparable for the two groups in the 46–59 year age range. Hence yoga may be especially useful for adult females with central obesity between 30 and 45 years of age.

# AUTHOR CONTRIBUTION

ST, SS, and AB designed the study. NK and SP performed data collection and analyses. ST and SS wrote the manuscript. NK, SP, and RG prepared the manuscript. ST, SS, NK, SP, RG, and AB proofread the manuscript.

# FUNDING

The study was funded by Patanjali Research Foundation (Trust), Haridwar, India.

# ACKNOWLEDGMENTS

The authors gratefully acknowledge the help of Dr. Jaideep Arya, Chief Central Coordinator of Patanjali Yog Samiti, Haridwar, Uttarakhand, India. The meticulous work of slide preparation for the nutritional advice group and the arrangement of data by Deepshikha Tyagi, Babita Vishwakarma, Kumar Gandharva, Savita Agnihotri, Alok Singh, Sadhna Verma, Deepak Pal, and Ankit Gupta was an important and indispensable contribution.

#### REFERENCES


# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo. 2018.00466/full#supplementary-material

the Korean population: the Korean National Health and Nutrition Examination Survey 2008-2011. BMC Endocr Disord. (2015) 15:79. doi: 10.1186/s12902-015-0075-5


and obese adults. Med Sci Monit. (2014) 20:894–904. doi: 10.12659/MSM.8 89805


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer OH and handling Editor declared their shared affiliation.

Copyright © 2018 Telles, Sharma, Kala, Pal, Gupta and Balkrishna. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.