# CROSS TALK BETWEEN THE IMMUNE SYSTEM AND METABOLISM

EDITED BY : Jixin Zhong, Wen Kong and Jie Chen PUBLISHED IN : Frontiers in Endocrinology

#### Frontiers eBook Copyright Statement

The copyright in the text of individual articles in this eBook is the property of their respective authors or their respective institutions or funders. The copyright in graphics and images within each article may be subject to copyright of other parties. In both cases this is subject to a license granted to Frontiers. The compilation of articles constituting this eBook is the property of Frontiers.

Each article within this eBook, and the eBook itself, are published under the most recent version of the Creative Commons CC-BY licence. The version current at the date of publication of this eBook is CC-BY 4.0. If the CC-BY licence is updated, the licence granted by Frontiers is automatically updated to the new version.

When exercising any right under the CC-BY licence, Frontiers must be attributed as the original publisher of the article or eBook, as applicable.

Authors have the responsibility of ensuring that any graphics or other materials which are the property of others may be included in the CC-BY licence, but this should be checked before relying on the CC-BY licence to reproduce those materials. Any copyright notices relating to those materials must be complied with.

Copyright and source acknowledgement notices may not be removed and must be displayed in any copy, derivative work or partial copy which includes the elements in question.

All copyright, and all rights therein, are protected by national and international copyright laws. The above represents a summary only. For further information please read Frontiers' Conditions for Website Use and Copyright Statement, and the applicable CC-BY licence.

ISSN 1664-8714 ISBN 978-2-88966-170-1 DOI 10.3389/978-2-88966-170-1

#### 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

# CROSS TALK BETWEEN THE IMMUNE SYSTEM AND METABOLISM

Topic Editors: Jixin Zhong, Case Western Reserve University, United States Wen Kong, Huazhong University of Science and Technology, China Jie Chen, Xiamen University, China

Citation: Zhong, J., Kong, W., Chen, J., eds. (2020). Cross Talk Between the Immune System and Metabolism. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88966-170-1

# Table of Contents


Anup Bhusal, Md Habibur Rahman, In-Kyu Lee and Kyoungho Suk


Mildren Porchas-Quijada, Zyanya Reyes-Castillo, José Francisco Muñoz-Valle, Sergio Durán-Barragán, Virginia Aguilera-Cervantes, Antonio López-Espinoza, Mónica Vázquez-Del Mercado, Mónica Navarro-Meza and Patricia López-Uriarte


Wei Wei, Shi Zhao, Sha-li Fu, Lan Yi, Hong Mao, Qin Tan, Pan Xu and Guo-liang Yang

*109 Characteristics of the Urinary Microbiome From Patients With Gout: A Prospective Study*

Yaogui Ning, Guomei Yang, Yangchun Chen, Xue Zhao, Hongyan Qian, Yuan Liu, Shiju Chen and Guixiu Shi

# Editorial: Cross Talk Between the Immune System and Metabolism

Jie Chen<sup>1</sup> , Wen Kong<sup>2</sup> and Jixin Zhong3,4 \*

*<sup>1</sup> Jiangxi Provincial People's Hospital Affiliated to Nanchang University, Nanchang, China, <sup>2</sup> Department of Endocrinology, Wuhan Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China, <sup>3</sup> Department of Rheumatology and Immunology, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China, <sup>4</sup> Department of Medicine, Case Western Reserve University, Cleveland, OH, United States*

Keywords: immune system, metabolism, metabolic disease, chronic inflammation, metabolic dysregulation

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

#### **Cross Talk Between the Immune System and Metabolism**

Metabolic dysregulation leads to a number of diseases including diabetes, obesity, hypertension, gout and rheumatoid arthritis (RA). These diseases have been associated with chronic inflammatory processes. To date, increasing evidence indicates that immune system plays an important role in the process of metabolic diseases. For example, the activation of type 1 immunity was identified in obesity-associated metabolic dysfunction (1). Metabolism has been shown to play a critical role in RA, a chronic autoimmune disease with involvement of a series of pro-inflammatory and immune-regulatory cytokines and mediators (2, 3). Innate immunity is also of key importance in the pathogenesis of gout (4). Although the association between the immune system and metabolic diseases have been well-recognized, the understanding of immune-related mechanisms of these diseases are not fully understood. This special issue exhibits a number of original research articles and review papers on the topic of cross talk between the immune system and metabolism.

As a central immune organ, the thymus provides a place for naive T cell differentiation, development and maturation. With thymic senescence, the epithelial network shrinks and is replaced by adipose tissue, leading to a decline of its immune function (5). Meanwhile, beige adipose tissue plays a key role in metabolism. By investigating the beige-specific and beige-indicative markers and metabolic profile (OCR/ECAR ratio), Banfai et al. reported in this issue that thymic adipose tissue emerging with senescence was actually beige adipose tissue, which builds a bridge between immune organ and metabolic tissue. T cells develop within the thymus and play a central role in adaptive immunity. Sirtuins are nicotine adenine dinucleotide (NAD+)-dependent enzymes involved in the cell metabolism (6). A review by Jonathan L et al. summarized some recent progresses in the role of sirtuins in regulating adaptive immunity (Warren and MacIver).

Alarmins play vital roles in innate and adaptive immune responses and participate in a wide range of pathophysiological processes such as inflammation and oncogenesis (7). Guo et al. investigated how sodium butyrate exerts its anti-inflammatory activity by inhibiting an "alarmin," HMGB1, and thus exhibits an anti-diabetic effect in type 1 diabetes. Shang et al. examined the role of another alarmin, IL-33, in an animal model of human gout, MSU-induced inflammation. A review by Tu and Yang summarizes the potential mechanisms of IL-33/ST2 axis in the metabolic disorders.

Hormones have regulatory effects on immunologic processes. In this special issue, a research by Porchas-Quijada et al. evaluated the relationship of anti-ghrelin autoantibodies with clinical, body-composition, and metabolic parameters in RA patients. Their findings support the previously reported functions of these natural autoantibodies as carriers and modulators of the stability and physiological function of natural hormones Porchas-Quijada et al.. In addition, a mini review by

#### Edited and reviewed by:

*Cunming Duan, University of Michigan, United States*

> \*Correspondence: *Jixin Zhong zhongjixin620@163.com*

#### Specialty section:

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

Received: *03 July 2020* Accepted: *20 July 2020* Published: *24 September 2020*

#### Citation:

*Chen J, Kong W and Zhong J (2020) Editorial: Cross Talk Between the Immune System and Metabolism. Front. Endocrinol. 11:588. doi: 10.3389/fendo.2020.00588*

**4**

Montesinos and Pellizas summarized the cellular and molecular mechanisms involved in thyroid hormones effects on innate immunity. In recent years, studies into the microbiomes have revealed their close relationship with human diseases (8). A study by Ning et al. investigated the alterations of urinary microbiome in gout patients, indicating the new prospects for microbiome in the diagnosis and treatment of gout.

Metabolic disorder can lead to serious complications. For instance, diabetic cardiomyopathy and encephalopathy are common severe complications of diabetes that cause mortality and morbidity in diabetic patients. Using mouse models of diabetes, Bhusal et al. investigated the role of LCN2 in the pathogenesis of diabetic encephalopathy, which help explain the pathogenic mechanisms that cause this complication. Ying et al. found that Phloretin prevented diabetic cardiomyopathy, possibly by suppressing the interaction between Nrf2 and Keap 1. Their work indicates a suppressive effect of Phloretin in high glucose-induced injury of cardiomyocytes. Intensive antidiabetic therapy in diabetic patients may cause hypoglycemia, which has been found to be associated with an increased risk for adverse cardiovascular outcomes and all-cause mortality (9). In

#### REFERENCES


this special issue, Wei et al. investigated the association between hypoglycemia as assessed by continuous glucose monitoring and the major adverse cardiovascular events or all-cause mortality.

Collectively, the original research and review articles in this special issue cover a series of important aspects in the field of interaction between the immune system and metabolism, which may provide new insights into the diagnosis and treatment of metabolic diseases and their complications.

#### 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 National Natural Science Foundation of China (81974254 and 81670431), Jiangxi Provincial Natural Science Foundation (20202ACBL206011), and Outstanding Innovation Team of Jiangxi Provincial People's Hospital (no. 19-008).


**Conflict of Interest:** 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 © 2020 Chen, Kong and Zhong. 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.

# Sodium Butyrate Ameliorates Streptozotocin-Induced Type 1 Diabetes in Mice by Inhibiting the HMGB1 Expression

Yu Guo1†, Zheng Xiao1†, Yanan Wang<sup>1</sup> , Weihua Yao<sup>1</sup> , Shun Liao<sup>1</sup> , Bo Yu<sup>1</sup> , Jianqiang Zhang<sup>1</sup> , Yanxiang Zhang1,2, Bing Zheng1,2 \*, Boxu Ren1,2 \* and Quan Gong1,2 \*

*<sup>1</sup> Department of Immunology, School of Medicine, Yangtze University, Jingzhou, China, <sup>2</sup> Clinical Molecular Immunology Center, School of Medicine, Yangtze University, Jingzhou, China*

#### Edited by:

*Jie Chen, Xiamen University, China*

#### Reviewed by:

*Hsien-Hui Chung, National Cheng Kung University, Taiwan Sonia Liao, University of Electronic Science and Technology of China, China*

#### \*Correspondence:

*Bing Zheng hxzheng@yangtzeu.edu.cn Boxu Ren boxuren188@163.com Quan Gong gongquan1998@163.com*

*†These authors have contributed equally to this work*

#### Specialty section:

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

Received: *04 August 2018* Accepted: *04 October 2018* Published: *25 October 2018*

#### Citation:

*Guo Y, Xiao Z, Wang Y, Yao W, Liao S, Yu B, Zhang J, Zhang Y, Zheng B, Ren B and Gong Q (2018) Sodium Butyrate Ameliorates Streptozotocin-Induced Type 1 Diabetes in Mice by Inhibiting the HMGB1 Expression. Front. Endocrinol. 9:630. doi: 10.3389/fendo.2018.00630*

Type 1 diabetes (T1D) is an autoimmune disease characterized by the immune cell-mediated progressive destruction of pancreatic β-cells. High-mobility group box 1 protein (HMGB1) has been recognized as a potential immune mediator to enhance the development of T1D. So we speculated that HMGB1 inhibitors could have anti-diabetic effect. Sodium butyrate is a short fatty acid derivative possessing anti-inflammatory activity by inhibiting HMGB1. In the current study, we evaluated the effects of sodium butyrate in streptozotocin (STZ)-induced T1D mice model. Diabetes was induced by multiple low-dose injections of STZ (40 mg/kg/day for 5 consecutive days), and then sodium butyrate (500 mg/kg/day) was administered by intraperitoneal injection for 7 consecutive days after STZ treatment. Blood glucose, incidence of diabetes, body weight, pancreatic histopathology, the amounts of CD4+T cell subsets, IL-1β level in serum and pancreatic expressions levels of HMGB1, and NF-κB p65 protein were analyzed. The results showed that sodium butyrate treatment decreased blood glucose and serum IL-1β, improved the islet morphology and decreased inflammatory cell infiltration, restored the unbalanced Th1/Th2 ratio, and down-regulated Th17 to normal level. In addition, sodium butyrate treatment can inhibit the pancreatic HMGB1 and NF-κB p65 protein expression. Therefore, we proposed that sodium butyrate should ameliorate STZ-induced T1D by down-regulating NF-κB mediated inflammatory signal pathway through inhibiting HMGB1.

Keywords: sodium butyrate, HMGB1, Th1/Th2, type 1 diabetes, streptozotocin

#### INTRODUCTION

Diabetes, as a non-communicable disease, has become the major cause of mortality and disease burden in the world. In recent decades, the incidence of diabetes has increased continuously that the estimated morbidity of diabetes in China was 10.9% in 2013 according to the latest published national survey (1). Type 1 diabetes (T1D) is characterized by the chronic hyperglycemia resulting from an immunologic disorder in which the autoreactive immune cells attack insulin-producing pancreatic β-cells (2). T1D is also known as juvenile-onset diabetes because it usually occurs in children and young adults (3). So far, the most suitable treatment for T1D is still insulin.

**6**

However, the usage of insulin is restricted because of the inevitable chronical cardiovascular complications caused by unmanageable blood glucose and destruction of β-cells (4). Recently, alloislet transplantation has been becoming an appealing method for T1D treatment is, but therapeutic approaches are limited due to deficiency of donor (5). In recent onset of T1D, the insulitis is usually present that is characterized by immune cell inflammatory infiltration within pancreatic islets (6, 7), the process is thought to be important for autoimmune diabetes progression (8). Consequently, it is reasonable to prevent or treat T1D with some anti-inflammatory agents.

High-mobility group box 1 protein (HMGB1), a highly conserved chromosomal protein, can be passively released from damaged cells or secreted from immune cells. It was recognized as an innate signal to mediate the autoimmune initiation and progression of the systemic lupus erythematosus and rheumatoid arthritis (9, 10). HMGB1 was also recognized to be involved in the development of both type 1 and 2 diabetes (11–13). In the type 1 diabetes model, streptozotocin (STZ) is a widely used diabetogenic agent (14), Previous studies have shown that HMGB1 is activated and the expression is increased in STZ-induced diabetic mice (15, 16). In addition, Extracellular HMGB1 was also discovered as proinflammatory cytokine (17, 18). Comparing with early-acting role of TNF-α and IL-1, HMGB1 was identified as a late-acting cytokine to influence the progression of sepsis (19), so we speculated that anti-HMGB1 therapeutics would become an effective approach to treat inflammation-related autoimmune disease. For example, administration of anti-HMGB1 antibody reduced the diabetes incidence and delayed the onset of diabetes in NOD mice (11). Sodium butyrate, a short fatty acid derivative, is present in human diet such as butter and cheese and it is also notably produced in the large intestine through fermentation of dietary fiber (20). It can act as a direct HMGB1 antagonist, and showed the effects to attenuate myocardial ischemia/reperfusion injury (21), to protect against acute lung injury (ALI) induced by severe burn (22), and to reduce pancreas injury in severe acute pancreatitis (23), through modulating the expression of HMGB1.

Given the role of HMGB1 in T1D initiation and progression (11, 15, 16). And sodium butyrate, as a specific HMGB1 antagonist, has shown anti-inflammatory effect in various animal models, so whether sodium butyrate have some protective effect on the T1D development through inhibiting the HMGB1? In current study, we reported the potential beneficial effects of sodium butyrate in STZ-induced type 1 diabetes mouse model and its underlying molecular mechanisms.

## MATERIALS AND METHODS

#### Animals

Male BALB/c mice (6–8 weeks, 25 ± 2 g) were purchased from Wuhan Centers for Disease Prevention & Control and the mice were bred and maintained in a pathogen-free facility, where kept the room temperature about 25◦C, humidity about 50%, a standard 12 h dark/light cycles. The studies were in accordance with protocols approved by the Institutional Animal Care and Use Committee (IACUC) at the Yangtze University. The diabetes was induced by treating the male BALB/c mice multiple low doses of STZ (Sigma-Aldrich, Shanghai, China). Namely, the mice were received intraperitoneal injection of STZ at a dose of 40 mg/kg/day (dissolved in 0.1 mol/L citrate buffer, pH 4.5) for 5 consecutive days. Non-diabetic mice were received with an equal volume of vehicle. To observe the diabetic status of the mice, the non-fasting glucose from tail blood sampling were monitored by a glucose meter (OneTouch, LifeScan). Diabetes onset were diagnosed when blood glucose level >16.7 mmol (300 mg/dl) on 2 consecutive tests (24).

## Drug Treatment

Sodium butyrate (Sigma-Aldrich, Shanghai, China) were dissolved in 0.9% sodium chloride solution and administered by intraperitoneal injection of 500 mg/kg/day at day 6 after STZ injection, the treatment were followed for 7 consecutive days.

#### Serum Collection

The mice were sacrificed at the end of experiment, after anesthetized with diethyl ether, the blood were collected using retro-orbital venous plexus puncture and then stayed at room temperature for 30 min, separated the serum through centrifugation at 12,000× g for 15 min at 4◦C. The sera were kept at −70◦C for ELISA.

#### ELISA for Cytokine Assay

The amount of IL-1β in serum was determined using a commercial kit (MultiSciences, Hangzhou, China) according to the manufacturers' instructions.

#### Histological and Morphological Analyses

The mice pancreases were removed and fixed in 4% formaldehyde at room temperature for 24 h, then the fixed tissues were infiltrated with paraffin, three series of 4µm thick sections were prepared and subsequently subjected to standard hematoxylin and eosin staining to assess the pancreatic histopathologic changes.

## Flow Cytometry Analysis of CD4+T Cell Subsets

CD4+T cell subsets from the spleen and pancreatic lymph nodes (PLNs) were determined by flow cytometry. Briefly, the lymphocyte were isolated from fresh spleen and PLNs by mechanical dissociation, then centrifuged and adjusted the supernatant cell number to 2 × 10<sup>6</sup> , erythrocytes were lysed using red cell lysis buffer (Tiangen, Beijing, China), washed twice with RPMI-1640 (containing 10% FBS), added 0.5 µl PMA and BFA (MultiSciences, Hangzhou, China) and incubated for 5 h, followed by incubation with PE-labeled IL-4, PE-labeled IL-17A, FITC-labeled CD4, and APC-labeled IFN-γ (BD Pharmingen, Shanghai, China) at 4◦C for 30 min. The cells were then subjected to flow cytometry analysis.

#### Western Blot Analysis

Pancreas tissues were homogenized in the RIPA lysis buffer (MultiSciences, Hangzhou, China) containing various inhibitors. The lysates were separated by 10% SDS-PAGE and then electrotransferred onto polyvinylidene difluoride (PVDF)

membranes. The membranes were incubated with primary antibody for the protein of interest or anti-β-Actin, the membranes were washed with Tris-buffered saline with Tween and incubated with HRP-conjugated secondary antibody. Immunoreactivity was detected using an enhanced chemiluminescence reagent (MultiSciences, Hangzhou, China).

## Statistical Analysis

Results are shown as mean ± standard deviation (SD). Graphical presentation and statistical analyses were carried out with GraphPad Prism software. The Student's t-test were used for comparison of the mean for two groups (plasma glucose and body weight).The difference of diabetes onset between groups were determined using the log-rank (Mantel–Cox) test. Comparisons between groups for cytokine secretion, the amount of CD4+T cell subsets and western blot were performed by oneway ANOVA. P < 0.05 was considered statistically significant.

## RESULTS

## Administration of Sodium Butyrate Decreases Plasma Glucose and Delays the Onset of Diabetes

To determine the effect of sodium butyrate on diabetes, the mice were intraperitoneal injected with 500 mg/kg sodium butyrate after the STZ injection. Vehicle-treated mice developed hyperglycaemia within 7 day after the last STZ injection, whereas the mice administrated with sodium butyrate exhibited lower non-fasting serum glucose levels compared with vehicle group (**Figure 1A**). Although sodium butyrate can't block the progression of diabetes, sodium butyrate treatment significantly postponed the development of diabetes (**Figure 1B**). The incidence of diabetes (non-fasting blood glucose level > 16.7 mmol) was first observed in vehicle group at day 12 compared with sodium butyrate group at day 21. In addition, we have also monitored the effect of sodium butyrate on the body weight and food intake of diabetic mice, but there was no significant difference between vehicle and sodium butyrate group (**Figure 1C** and **Supplementary Figure 1**).

## Pancreatic Histopathologic Changes Are Improved by Sodium Butyrate

Histological examination of mice pancreases were performed to evaluate the effect of sodium butyrate on STZ-induced mice. As shown in **Figure 2**, the healthy mice had intact islet morphology. However, islet boundary became a little vague and cell number inside islet decreased in diabetic mice. Moreover, heavy inflammatory cell infiltration at one side was evident. Here, although sodium butyrate treatment could not prevent the inflammatory cells infiltration, morphology of islet was improved when compared with vehicle group.

group, and sodium butyrate treated mice (H&E staining, magnification 1000×).

## The Ratio of Th1/Th2 and Th17 Cells Are Regulated by Sodium Butyrate

It has been reported that the pathogenesis of some inflammatory diseases are associated with the imbalance of Th1/Th2 and Th17 cells (25–27). We determined whether the anti-diabetic effect of sodium butyrate are related with the changes of Th1/Th2 in the spleen and Th17 cells in PLNs. As shown in **Figure 3**, vehicle treated diabetic mice showed a higher percentage of Th1 and lower percentage of Th2 compared with non-diabetic mice. More importantly, the ratio of Th1/Th2 and Th17 cells in diabetic mice are higher than non-diabetic mice. Whereas, sodium butyrate regulated the increased ratio to a relative low level that was closed to health non-diabetic mice. These results demonstrated that sodium butyrate should exhibit the anti-diabetic effect through modulating the unbalanced Th1/Th2 and decreased Th17 cells to the normal level.

## Proinflammatory Cytokine IL-1β Are Down-Regulated in Sodium Butyrate Treated Diabetic Mice

Considering the role of proinflammatory cytokines on the pathogenesis of diabetes, population with detectable levels of circulating IL-1β cytokines have increased risk to develop diabetes (28–30). We speculate that sodium butyrate could relieve diabetes reflected by decreased level of IL-1β. As expected, diabetic mice showed increased levels of IL-1β, but sodium butyrate treatment significant decreased the level (**Figure 4**). The data suggested that sodium butyrate could inhibit proinflammatory response of diabetes.

## Sodium Butyrate Inhibits the Pancreatic HMGB1 and NF-κB p65 Protein Expression

Pancreatic HMGB1 and NF-κB p65 protein expression were analyzed by western blot. Results showed that both HMGB1 and NF-κB p65 protein expression were up-regulated in diabetic mice compared with healthy non-diabetic mice, in contrast, the protein expression was markedly down-regulated in the mice treated with sodium butyrate compared with diabetic mice (**Figure 5**).

## DISCUSSION

Current strategies for T1D treatment include lifelong insulin delivery, maintaining normal glycemic level, eating healthy foods, and keeping to a healthy weight. As is known to us, people with diabetes have a higher risk to develop one or more complications, so it is important to control and maintain the blood glucose within the normal level.

HMGB1, a highly conserved non-histone nuclear protein, was proven to be involved in the pathogenesis of inflammatory and autoimmune disease (31), and it can serve as endogenous alarmin to alert the innate immune system to promote host defense or tissue repair. The role of HMGB1 in autoimmune disease was first confirmed in rheumatoid arthritis (RA) (32). Extranuclear HMGB1expression was increased in the synovia of patients and animal models with rheumatoid arthritis (RA), and blockade of HMGB1 expression in experimental animal models can attenuate the RA (32, 33). In addition, HMGB1 was also involved in pathogenesis of systemic lupus erythermatosus (SLE), the patients with SLE show increased level of HMGB1 in the epidermis, and the increased plasma levels of HMGB1 correlated closely with disease activity (34). T1D is also an autoimmune disease characterized by destruction of the insulin secreting β-cells, previous study has shown that HMGB1 seems to be involved in T1D pathogenesis and it can act as a potent innate alarmin to mediate the initiation and progression during T1D development in NOD mice, a model of spontaneous T1D (11). When NOD mice were treated with HMGB1 neutralizing antibody, the insulitis progression was significantly inhibited and the diabetes incidence was also decreased (11). HMGB1 was also thought to be involved in the pathogenesis of type 2 diabetes (T2D) (35). It has been reported that serum HMGB1 level was increased in patients with T2D, and in vitro study showed that high glucose can activate HMGB1 expression in mesangial cells (35). A recent investigation about the effect of

FIGURE 3 | Sodium butyrate treatment regulates CD4<sup>+</sup> T cell subsets (A). Splenic and PLNs' lymphocyte originated from sodium butyrate treated mice (*n* = 5), vehicle group (*n* = 5) and healthy mice (*n* = 3) were harvested for CD4, IL-4, IFN-γ, and IL-17A staining. The percentages of Th1, Th2, and Th17 cells were determined by flow cytometric analysis (B). The data are listed in the table as means ± SD of three independent experiment. The ratio of splenic Th1/Th2 and Th17 cell in PLNs was significantly higher than healthy mice, whereas sodium butyrate treatment restored unbalanced Th1/Th2 and regulated Th17 cells to normal level. \**p* < 0.05, \*\**p* < 0.01 vs. healthy group, #*p*<0.05, ##*p*<0.01 vs. vehicle group.

HMGB1 on high concentration glucose induced mesothelial cells (MCs) injury also demonstrated that high glucose promoted HMGB1 translocation and secretion from the nucleus of MCs (36). Moreover, it has also been reported that high glucose can induce retinal pericytes, vascular smooth muscle cells and human aortic endothelial cells to secret HMGB1 (37–39). So it is undoubted that HMGB1 expression could be activated under high glucose induction.

Based on the potential role of HMGB1 in the pathogenesis of diabetes, It was presumed that HMGB1 inhibitors would probably affect the diabetes onset. Sodium butyrate is a wellknown short fatty acid derivative and exhibits good antiinflammatory property through inhibiting HMGB1 expression. It has been proven that sodium butyrate showed protective effect in myocardial ischemia/reperfusion, severe sepsis and ALI by inhibiting HMGB1 (21, 22, 40, 41). Sodium butyrate can also improve the performance of diabetic complications (42, 43). For example, sodium butyrate showed protective effect against diabetic nephropathy (DN) (42). In a high fat diet (HFD) induced type II diabetic model, cardiac function and metabolic dysfunction were improved by sodium butyrate (43). In addition, sodium butyrate can prevent the insulin resistance in HFDinduced obese mice (44).

CD4<sup>+</sup> T helper (Th) cells are major T cell subsets that play a vital role in mediating immune responses. According to cytokine production and specialized functions, the Th cells

healthy group, ##*p* < 0.01 vs. vehicle group.

can be classified into at least four distinct Th phenotypes (Th1, Th2, Th17, and T-regulatory cells). Th1 cells were responded for cellular immunity through secreting interferon (IFN)-γ (45), Th1 cells were thought to be involved in the insulin-producing β-cell destruction in the pancreatic islets (46). Whereas, Th2 cells were mainly involved in mediating humoral immunity and would actually protect against autoimmune disease (47). Induction of Th2 cells could lead to dominant protective effect against T1D development (48). Previous study had shown that Th1/Th2 imbalance could contribute to pathogenesis of some autoimmune diseases (26). Th17 cells, another subset of CD4+T cells, can produce proinflammatory cytokine IL-17 and induce inflammation to mediate autoimmune pathology. Some evidence indicated that Th17 cell and its related cytokines had significant effects on the onset and progression of T1D in both human and animals (25, 27).

In this study, we evaluated the potential anti-diabetic effect of sodium butyrate in a STZ-induced T1D mice model. Multiple low dose injection of STZ led to pronounced pancreatic insulititis, followed by β-cell destruction and plasma glucose elevation (14), then the HMGB1 was passively released from damaged β-cell (11), and it acted as proinflammatory cytokine to enhance the inflammatory response. We found that sodium butyrate exhibited the protective effect on streptozotocin-induced type 1 diabetes in mice. Sodium butyrate treatment decreased the

level of plasma glucose and delayed the onset of diabetes. In order to investigate the mechanism of beneficial effects of sodium butyrate for T1D, we first analyzed the phenotypes of CD4<sup>+</sup> T cells. The results showed that ratio of Th1/Th2 and Th17 were increased in diabetic mice, and sodium butyrate treatment significantly decreased the proportion of Th1/Th2 in spleen and Th17 from PLNs. As aforementioned, Th1 and Th17 cell are closely associated with onset and development of T1D. So it is not difficult to speculate that sodium butyrate could recover the balance of Th1/Th2 and inhibit theTh17 cell to the normal level. Th17 initiated the inflammation through stimulating the production of proinflammatory cytokines, IL-1β, IL-6, and TNFα and in turn accelerated β-cell destruction (49). And our results demonstrated that sodium butyrate treatment did decrease the level of IL-1β in serum. To further address the underlying mechanism of sodium butyrate, we determined the NF-κB p65 expression. Because NF-κB is an important transcription factor that regulates the inflammatory gene expression. An in vitro study showed high glucose mimicking diabetes can lead to the activation of NF-κB and subsequent increased expression of inflammatory chemokines and cytokines (35, 50). In our study, NF-κB p65 expression was increased in diabetic mice, indicating NF-κB signaling pathway is involved in pathogenesis of the streptozotocin-induced type 1 diabetes model. Whereas treatment with sodium butyrate can inhibit the NF-κB p65 levels. Therefore, the main results can be summarized in **Figure 6**. Briefly, STZ directly targets and destroys pancreatic β-cells, then HMGB1 was passively released from damaged β-cells (11), and interacts with DCs or macrophage via the corresponding surface receptor(s), induces a signaling cascade and activates NF-κB pathway (12), therefore leading to the production of proinflammatory cytokines (such as IL-1β), the cytokines together with unbalanced Th1/Th2/Th17 accelerate the islets inflammation and β-cells destruction and finally develop into diabetes, high glucose would promote HMGB1 expression and further aggravate the diabetic condition through positive feedback effect of HMGB1 and high glucose. Whereas, sodium butyrate, as a direct HMGB1 antagonist, could down-regulate the expression of HMGB1 and mediate the balance of Th1/Th2/Th17 paradigm, thus attenuating type 1 diabetes.

Additionally, emerging evidences have implicated that gut bacterial composition may be associated with disease development and progression of T1D in both animal and human (51, 52). And a recently study has also suggested that short chain fatty acids (including sodium butyrate) treatment to rat breeders can ameliorates T1D in the offspring through reshaping the intestinal microbiota (53). So we speculate that sodium butyrate, to some extent, could restore the balance of intestinal flora to maintain metabolic homeostasis in the STZ-induced T1D mice. And we would perform experiments to observe the effect of sodium butyrate on intestinal microbiota composition in the future, such as Lactobacillus and Bifidobacterium that are associated with the progression of T1D (54).

Altogether, our data suggest that sodium butyrate ameliorates STZ-induced type 1 diabetes. The beneficial effects could be attributed to the effects of sodium butyrate on restoring the unbalanced Th1/Th2/Th17 paradigm and inhibiting NF-κBmediated inflammatory pathway. Therefore, sodium butyrate would become a beneficial dietary supplementation for T1D patients.

## AUTHOR CONTRIBUTIONS

YG and ZX searched related literatures and made the mice model of STZ-induced T1D. WY was responsible for blood glucose monitoring. SL and BY performed the ELISA and western blot. YW and JZ performed the flow cytometry analysis and statistical analysis. YZ was responsible for histological and morphological analysis. QG and BR designed the experiment and supervised

## REFERENCES


the project. BZ wrote the manuscript with contribution from all authors.

## FUNDING

This work was supported by Natural Science Foundation of Hubei Province (2015CFA080).

## ACKNOWLEDGMENTS

We gratefully thank Dr. Feng Qian for the critical reading of the manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo. 2018.00630/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 Guo, Xiao, Wang, Yao, Liao, Yu, Zhang, Zhang, Zheng, Ren and Gong. 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.

# Phloretin Prevents Diabetic Cardiomyopathy by Dissociating Keap1/Nrf2 Complex and Inhibiting Oxidative Stress

Yin Ying1†, Jiye Jin2†, Li Ye<sup>3</sup> , Pingping Sun<sup>1</sup> , Hui Wang<sup>1</sup> and Xiaodong Wang<sup>4</sup> \*

*<sup>1</sup> Department of Pharmacy, Tongde Hospital of Zhejiang Province, Hangzhou, China, <sup>2</sup> Department of Rehabilitation, Tongde Hospital of Zhejiang Province, Hangzhou, China, <sup>3</sup> Department of Nursing, Tongde Hospital of Zhejiang Province, Hangzhou, China, <sup>4</sup> Department of Vascular Surgery, Tongde Hospital of Zhejiang Province, Hangzhou, China*

Hyperglycemia induces chronic inflammation and oxidative stress in cardiomyocyte, which are the main pathological changes of diabetic cardiomyopathy (DCM). Treatment aimed at these processes may be beneficial in DCM. Phloretin (PHL), a promising natural product, has many pharmacological activities, such as anti-inflammatory, anticancer, and anti-oxidative function. The aim of this study was to investigate whether PHL could ameliorate the high glucose-mediated oxidation, hypertrophy, and fibrosis in H9c2 cells and attenuate the inflammation- and oxidation-mediated cardiac injury. In this study, PHL induced significantly inhibitory effect on the expression of pro-inflammatory, hypertrophy, pro-oxidant, and fibrosis cytokines in high glucose-stimulated cardiac H9c2 cells. Furthermore, PHL decreased the levels of serum lactate dehydrogenase, aspartate aminotransferase, and creatine kinase-MB, and attenuated the progress in the fibrosis, oxidative stress, and pathological parameters via Kelch-like ECH-associated protein 1 (Keap1)/nuclear factor E2-related factor 2 (Nrf2) pathway in diabetic mice. In additional, molecular modeling and immunoblotting results confirmed that PHL might obstruct the interaction between Nrf2 and Keap1 through direct binding Keap1, and promoting Nrf2 expression. These results provided evidence that PHL could suppress high glucose-induced cardiomyocyte oxidation and fibrosis injury, and that targeting Keap1/Nrf2 may provide a novel therapeutic strategy for human DCM in the future.

#### Keywords: phloretin, Nrf2, Keap1, diabetic cardiomyopathy, oxidative stress

## INTRODUCTION

Diabetes mellitus (DM) is an emerging global health problem. Diabetic cardiomyopathy (DCM) is the one of the major complications of diabetes that causes mortality and morbidity in diabetic patients (1). DCM starts with diastolic dysfunction in patients with type-1 (T1DM) or type-2 (T2DM). Previous studies showed that DCM presented with structural and functional abnormalities of the myocardium, leading to increased risks for myocardial fibrosis, ventricular hypertrophy, and heart failure (2, 3). There are several physiological mechanisms related to the pathogenesis for DCM, including insulin resistance signaling, cardiac inflammation, oxidative stress, endoplasmic reticulum stress, etc. (4). Increasingly, evidence demonstrated that oxidative stress may be the most common feature linking diabetes-induced alterations to the development of cardiac dysfunction.

#### Edited by:

*Jie Chen, Xiamen University, China*

#### Reviewed by:

*Xiaoqiang Tang, Sichuan University, China Hsien-Hui Chung, National Cheng Kung University, Taiwan*

\*Correspondence: *Xiaodong Wang wangxiaodong155@gmail.com*

*†These authors have contributed equally to this work*

#### Specialty section:

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

Received: *04 October 2018* Accepted: *10 December 2018* Published: *20 December 2018*

#### Citation:

*Ying Y, Jin J, Ye L, Sun P, Wang H and Wang X (2018) Phloretin Prevents Diabetic Cardiomyopathy by Dissociating Keap1/Nrf2 Complex and Inhibiting Oxidative Stress. Front. Endocrinol. 9:774. doi: 10.3389/fendo.2018.00774*

The Keap1/Nrf2 pathway is the most crucial anti-oxidative mechanism that protects cells from oxidative stress (5). Nrf2 is a master regulator of redox status and cellular detoxification responses by inducing the expression of multiple downstream anti-oxidant genes, including heme oxygenase (HO-1), nicotinamide adenine dinucleotide phosphate-H (NADPH), quinineoxidoreductase-1 (NQO-1), and glutamate-cysteine ligase catalytic (GCLC) (6). Recently, it was reported that Nrf2 prevented T2DM-induced cardiac injury (7). Furthermore, another study indicated that fibroblast growth factor-21 (FGF21) prevented diabetic cardiomyopathy via AMPK-AKT2-Nrf2 mediated anti-oxidation and lipid-lowering effects in the heart (8). Therefore, controlling cytoprotective oxidative stress response enzymes in DCM via Keap1/Nrf2 pathway targeting is a potential important strategy.

Phloretin (PHL, **Figure 1A**) is a dihydrochalcone flavonoid found in peel and root skin of frequently consumed vegetables and fruits. PHL showed numerous biological and pharmacological activities, such as anti-inflammatory, antioxidant, and anti-cancer in various disease models (9–12). Previously, our group found that PHL significantly inhibited tert-Butyl hydroperoxide (TBHP) induced oxidation. However, the potential of PHL for the treatment of DCM and the molecular mechanisms underlying the actions of PHL remain unclear. In this study, we evaluated the protective effect of PHL in hyperglycemia-induced oxidative stress and cardiac injury in vitro and in vivo. Subsequently, molecular modeling and immunoblotting were used to explore the underlying mechanisms and possible targets of PHL. Our results revealed that PHL could prevent cardiac injury in T1DM by attenuating hyperglycemia-induced oxidative stress and hypertrophy. We identified Keap1, a negative regulator of Nrf2, as the possible target for PHL. Overall, our study provided crucial evidence that PHL may be a novel therapeutic agent for the treatment of DCM.

## MATERIALS AND METHODS

## Reagents, Cell Culture and Treatment

Phloretin (purity 98%, verified by high-performance liquid chromatography; molecular weight = 274.27) was purchased from Aladdin (Shanghai, China). Phloretin was dissolved in DMSO for in vitro experiments and in CMC-Na (0.5%) for in vivo experiments, both stored at 4◦C for further use. H9c2 embryonic rat heart-derived cell line was purchased from the Shanghai Institute of Biochemistry and Cell Biology (Shanghai, China) and cultured in DMEM medium (Gibco, Eggenstein, Germany) including 5.5 mmol/l of D-glucose supplemented with 10% FBS, 100 U/ml of penicillin, and 100 mg/ml of streptomycin. For the high glucose-treated group (HG), cells were incubated with a DMEM medium, which was contained 33 mmol/L of glucose. Glucose and streptozotocin (STZ) were obtained from Sigma-Aldrich (St. Louis, MO). Haematoxylin-eosin (H&E) was purchased from Beyotime (Nantong, China). Masson's trichrome kits was obtained in Solarbio (Beijing, China).

Antibodies for Nrf2 (#12721), TGF-β (#3711), Keap1 (#8047), GAPDH (#5174), and secondary antibodies (mouse #7076, rat #7077) were obtained from Cell Signaling Technology (Danvers, USA). RIPA lysis buffer was purchased from Boster Biological technology (Wuhan, China).

#### Animals and Treatment

Male C57BL/6 mice weighing 20–22 g were obtained from Zhejiang Animal Center (Hangzhou, China). The mice were housed at a constant room temperature with a 12:12 h lightdark cycle and fed with a standard rodent diet and water. All animal experimental procedures complied with the "The Detailed Rules and Regulations of Medical Animal Experiments Administration and Implementation" (Document No. 1998–55, Ministry of Public Health, PR China), and were approved by the Tongde Hospital of Zhejiang Province Animal Policy and Welfare Committee (Approval Document No. SCXK2014-0001).

Eighteen mice were randomly divided into three groups. Twelve mice were received intraperitoneal (i.p.) injection of STZ at the dose of 100 mg/kg formulated in 100 mM citrate buffer (pH 4.5) for 1 time, blood glucose levels were detected using a glucometer, control animals received buffered saline only. Six mice treated with phloretin at 10 mg/kg through i.g. after injection STZ 8 days. At Day 56 after STZ induction, the mice were killed under anesthesia, and then blood samples were collected. At the time of death, the heart tissues were removed.

## Determination of Serum Aspartate Aminotransferase (AST), Lactate Dehydrogenase (LDH) and Creatine Kinase (CK-MB), Malondialdehyde (MDA), and Superoxide Dismutase (SOD)

Serum levels of AST, LDH, and CK-MB and supernatant levels of MDA and SOD were analyzed by commercial ELISA kits refer to the manufacturers' instruction (Nanjing Jiancheng Bioengineering Institute, Nanjing, China).

#### Heart Histopathology

Heart tissue was fixed in 10% formalin for 24 h, embedded in paraffin, and sectioned at 5µm. Then, the heart sections were deparaffinized, rehydrated, and stained with hematoxylin and eosin (H&E). Cardiac fibrosis was tested by Masson's trichrome staining for collagen deposition as described previously. To estimate the extent of damage, the specimen was observed under a light microscope (Nikon, Japan).

## Cell Cytotoxicity

Before PHL treatment, seeding cells into 96-well plates with 5,000 cells/well. Adding PHL into wells with various doses and incubated for 24 h. After treatment, MTT was added to each well (1 mg/ml), incubated at 37◦C for 4 h. The formazan crystal was dissolved with DMSO, 150 µL/well. The absorbance was detected at 490 nm on a microplate reader. Cell cytotoxicity was expressed as the percentage of MTT reduction compared to control.

#### Rhodamine Phalloidine Staining

For hypertrophy, cells were fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X-100, and stained with rhodamine phalloidin at a concentration of 50µg/mL for 30 min. Nuclei were stained with the DAPI at room temperature for

5 min. Immunofluorescence was viewed and captured using Nikon fluorescence microscope (Nikon, Japan).

## RNA Isolation and Real-Time PCR (q-PCR)

Total RNA was extracted from the heart tissues and cells by using Trizol reagent (Invitrogen, Carlsbad, CA) according to each manufacturer's protocol. Both reverse transcription and quantitative PCR were carried out using a two-step M-MLV Platinum SYBR Green qPCR SuperMix-UDG kit (Invitrogen, Carlsbad, CA). Eppendorf Mastercycler ep realplex detection system (Eppendorf, Hamburg, Germany) was applied to q-PCR analysis. The primers of genes including NQO-1, HO-1, Nrf2, GCLC, CTGF, collagen-1, TGF-β, ANP, BNP, β-MyHC, and βactin were obtained from Invitrogen (Shanghai, China). The primer sequences were listed in **Table S1**. Comparative cycle time (Ct) was used to determine fold differences between samples and normalized to β-actin.

## Western Blot Analysis

Heart tissues and harvested cell pellets were homogenized in RIPA lysis buffer (Santa Cruz Biotechnology, Santa Cruz, CA) to obtain total protein or nuclear protein extracted using a nuclei isolation kit (NUC201, Sigma-Aldrich). Western blot assay was performed for target protein quantification, as described previously. The proteins were separated by 10% sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) and transferred to a nitrocellulose membrane. Membranes were blocked with 5% non-fat milk for 1 h and incubated overnight at 4◦C with the specific antibodies over-night in 4 ◦C. Immunoreactive bands were detected by incubating with secondary antibody conjugated with horseradish peroxidase and visualizing using enhanced chemiluminescence reagents (Bio-Rad, Hercules, CA). The amounts of the proteins were analyzed using Image J analysis software and normalized to GAPDH.

#### Co-immunoprecipitation (Co-IP) Analysis

Nrf2 was co-precipitated with Keap1 from cardiac tissues to detect the association of Nrf2 with Keap1. Cardiac tissue extracts (500 µg) were incubated with anti-Nrf2 antibody (0.5 µg) at 4 ◦C overnight and then precipitated with protein A agarose (Beyotime, Shanghai, China) for 3 h. The Keap1/Nrf2 complex level in the beads was further detected by western blot.

## Intracellular ROS Measurement

The ROS production was measured by using the ROSsensitive dye, 2,7-dichlorodihydrofluorescein diacetate (DCFH-DA, Beyotime Biotechnology, China) as an indicator, as described previously. To measure ROS production in tissues and cell, dihydroethidium (DHE) staining was performed as described previously (13).

## Construction of the Initial Structure of the Keap1/PHL Complex

The atomic co-ordinates of mouse Keap1 was download from the Protein Data Bank (PDB code: 5CGJ) (14). Next, the structure of Keap1 was preprocessed by VMD software (15). The AutoDock 4.2.6 package were used to predict the possible binding pose between Keap1 and PHL (16). Before molecular docking, the Keap1 and PHL were prepared by AutoDockTools 1.5.6 package, including adding missing hydrogen atoms and Gasteiger partial charges. After that, a grid box size of 22.5 Å × 22.5 Å × 22.5 Å was assigned, which covered almost the entire binding site of Keap1. During molecular docking, trials of 100 dockings, Lamarckian Genetic Algorithm (LGA) was used to globe conformational sampling, and other parameters were set as default. The lowest predicted binding energy conformation was used to further molecular dynamics (MD) simulation analysis.

## Conventional Molecular Dynamics (MD) Simulation and Gaussian Accelerated Molecular Dynamics (GaMD) Simulation

The structural optimization of PHL was conducted using B3LYP combined with 6 – 31+G <sup>∗</sup> basis set and RESP fitting method was applied for charge derivation based on the optimal conformation. The ff14SB force field was employed for the Keap1 protein and the general amber force field (gaff2) for the PHL (17, 18). The Keap1/PHL complex was solvated into TIP3P water box with boundary extended 12 Å away from any solute atom. The counter ions of Na<sup>+</sup> were added to maintain the electroneutrality.

Before productive simulation, three consecutive minimization stages were applied to relax the system, including 5,000 steps of steepest descent and 5,000 steps of conjugate gradient steps. Firstly, energy minimization of only hydrogen atoms, followed by water molecules and counter ions, were performed with harmonic constraint potential of a 5.0 kcal mol−<sup>1</sup> Å −2 . Thereafter, the whole system was minimized without any constraint. The systems were gradually heated to 310 K in 100 ps, followed by 600 ps density equilibration. Finally, 200 ns conventional MD simulations and 400 ns GaMD simulation were performed in the NPT ensemble and NVT ensemble, respectively. In particular, the dual potential boost method was employed in the GaMD simulation (19, 20). The boost parameters were calculated from an initial ∼4 ns NVT conventional MD simulation. During these simulations, the temperature were maintained by the Langevin temperature equilibration scheme (21). The long-range electrostatic interaction (cutoff = 10.0 Å) was used to evaluate direct space interaction with the particle-mesh Ewald (PME) method (22). All bonds involving hydrogen atoms were constrained by using the SHAKE algorithm (23). The simulation trajectories were processed by CPPTRAJ module in Amber 16 package (24). The Binding free energy decomposition was calculated by molecular mechanics/generalized Born surface area (MM/GBSA) method based on 200 snapshots extracted from the last 40 ns conventional MD simulation trajectory.

#### Statistical Analysis

Statistical analysis was performed using Student's t-test, and ANOVA as appropriate, with Tukey or Bonferroni post hoc-tests. All data were analyzed with GraphPad Prism 5.0 (Graphpad Software, Inc.); p < 0.05 were considered significant.

## RESULTS

## The Cytotoxicity of PHL

We investigated the cytotoxicity of PHL on H9c2 cells by MTT assay. As shown in **Figure 1B**, treatment of H9c2 cells with increasing concentrations of PHL (2.5, 5, 10, 20, 40, and 80µM) for 24 h showed that even at 80µM, PHL was relatively non-toxic to H9c2 cells. These results indicated that PHL had no obvious toxic effect on H9c2 cells. Finally, 10µM PHL was selected for subsequent in vitro experiments.

## PHL Reduced Hyperglycemia-Induced ROS Levels Through Regulation of Nrf2 Antioxidant Responses in H9c2 Cells

A large amount of evidence have shown that reactive oxygen species (ROS) play an important role in the pathogenesis of DCM (25). Therefore, we investigated the effect of PHL on hyperglycemia-induced oxidative stress. Firstly, we measured the effects of PHL on hyperglycemia-induced ROS generation and redox status markers. The hyperglycemia-induced group demonstrated markedly increased ROS generation, which significantly reduced by pre-treatment with PHL in hyperglycemia-induced H9c2 cells (**Figure 2A**). PHL also decreased the levels of MDA, a natural byproduct of lipid peroxidation (**Figure 2B**) and enhanced the enzyme SOD activity (**Figure 2C**). Dihydroethidium (DHE) reacts with superoxide anions to form a red fluorescent product, allowing signal visualization, and measurement of intracellular ROS in H9c2 cells. As shown in **Figure 2H**, PHL attenuated hyperglycemiainduced ROS production. These findings indicated that PHL is a potential inhibitor for hyperglycemia-induced ROS production.

As reported as Foresti et al. (26), hyperglycemia downregulated Nrf2 in vitro and in vivo. Nrf2 regulates the transcription of genes coding for anti-oxidant and detoxifying proteins, such as HO-1, NADPH, NQO-1, glutathione peroxidase-2, GCLC, and glutathione S-transferase. As shown in **Figure 2D**, the hyperglycemia-mediated downregulation of Nrf2 was prevented by the pre-treatment with PHL. Subsequently, we evaluated the expression levels of Nrf2, NQO-1, and HO-1 by pre-treating H9c2 cells with PHL for 30 min, followed by stimulation with high glucose (33 mM) for 24 h. The gene expression showed that Nrf2 (**Figure 2E**), NQO-1 (**Figure 2F**), and HO-1 (**Figure 2G**) were significantly increased in the PHL pre-treated group despite hyperglycemia induction. These results indicated that PHL inhibited the hyperglycemia-induced oxidation through Nrf2 gene in H9c2 cells.

## PHL Attenuated Hyperglycemia-Induced Cell Fibrosis and Hypertrophy in H9c2 Cells

Myocardial fibrosis and hypertrophy are the major mechanisms contributing to cardiomyocyte remodeling in DCM (2, 27), so we assessed the effect of PHL on hyperglycemia-induced cardiac fibrosis and hypertrophy in H9c2 cells. As shown in **Figure 3A**, increased protein levels of hypertrophy marker atrial natriuretic peptide (ANP) and fibrosis marker TGF-β were observed in hyperglycemia-induced cells. These increases were significantly ablated by PHL pre-treatment. Meanwhile, RT-qPCR analysis revealed that PHL inhibited hyperglycemia-induced increases in hypertrophy factors, including ANP (**Figure 3C**), brain natriuretic peptide (BNP, **Figure 3D**) and β-myosin heavy chain (β-MyHC, **Figure 3E**) as well as fibrosis factors, such as transforming growth factor-β (TGF-β, **Figure 3F**), collagen-1 (**Figure 3G**), and connective tissue growth factor (CTGF, **Figure 3H**) gene expression. Rhodamine phalloidin staining also demonstrated an ablated hyperglycemia-induced cell size increase by PHL pre-treatment (**Figure 3B**).

## PHL Prevented Cardiomyocyte Injury in Diabetic Mice

The in vitro experiments clearly showed that PHL could reduce oxidative stress, hypertrophy and fibrosis in H9c2 cells. Next, we evaluated the protected effects of PHL on heart in vivo. As shown in **Figures 4A–C**, PHL significantly inhibited the hyperglycemiainduced upregulation of biochemical markers LDH (**Figure 4A**), CK-MB (**Figure 4B**) and AST (**Figure 4C**) of myocardial injury. The H&E stain revealed that the hyperglycemia-stimulated hearts had structural abnormalities, namely disorganized myofibers. Heart tissues from PHL treatment group showed no significant structural changes compared to the control (CON) group (**Figure 4D**). Investigation of hypertrophy related genes in heart tissues revealed that PHL significantly attenuated the high expression levels of ANP (**Figure 4E**), BNP (**Figure 4F**), and β-MyHC (**Figure 4G**) found in diabetic mice. However, PHL treatment did not significantly affect fasting weight and blood

sugar (**Figures S1A,B**), suggesting that the cardiac protective effects of PHL were not related to metabolic changes.

## PHL Inhibited Cardiac Oxidative Stress and Fibrosis in the Diabetic Myocardium

To identify potential pharmacological activity for PHL protection in vivo, a series of biomarkers related to oxidation and fibrosis were investigated. Firstly, DHE staining showed that PHL attenuated the increase in ROS level in DCM (**Figure 5A**). Next, connective tissue collagen and histopathology of the hearts from PHL treated and untreated diabetic mice were assessed using Masson's trichrome stain. PHL effectively inhibited the fibrotic process of DCM (**Figure 5B**). As shown in **Figures 5C–E**, the expression levels of HO-1, NQO-1, GCLC in DCM tissues were upregulated by PHL. Additionally, in agreement with the histopathology, the high TGF-β (**Figure 5F**), collagen-1 (**Figure 5G**), and CTGF (**Figure 5H**) were significantly downregulated in DCM. These results suggested that the protective effect of PHL on cardiomyocyte in hyperglycemiainduced DCM might be related to its anti-oxidant and antifibrotic function.

## PHL Targeted Keap1 Leading to Dissociation of the Keap1/Nrf2 Complex

Keap1/Nrf2 signaling pathway play a major role in regulating oxidative stress (28). Increasing evidence showed that Keap1/Nrf2 signaling pathway regulates inflammation, fibrosis and endoplasmic reticulum stress (ER stress) in DCM (29, 30). The in vitro study revealed that PHL promoted Nrf2 expression in hyperglycemia stimulation but the target and mechanism remained unclear. Previous studies indicated that Nrf2 maintains an inactive state in the cytoplasm under unstressed conditions bound to its inhibitor Keap1 (29), a vital regulator of the anti-oxidant response. Dissociation of the protein-protein interaction between Nrf2 and Keap1 leads to expression of detoxifying anti-oxidant enzymes. Thus, molecular modeling

and immunoblotting analysis were applied, focusing particularly on the hydrophobic binding site of Keap1.

Molecular docking was used to generate the initial structure and molecular dynamics (MD) simulations were applied to investigate the dynamic behavior. In order to monitor the stability of Keap1/phloretin complex during MD simulation, the root-mean-square deviations (RMSDs) of the backbone atoms (Cα) of Keap1 and the heavy atoms of PHL were analyzed (**Figure S2**). As shown in **Figures S2A,B**, the RMSD values of the backbone atoms of Keap1 and the heavy atoms of phloretin have a small fluctuation during the whole conventional MD simulation. GaMD simulation indicated that the Keap1/PHL complex was relatively stable. To gain an insight into the roles of individual residues in determining the interaction between Keap1 and PHL, the binding free energy decomposition was carried out. As shown in **Figure 6A**, the 10 most contributed residues were Gly-511, Ile-559, Gly-558, Ala-366, Val-512, Val-465, Val-606, Gly-464, Gly-605, and Gly-417. The predominant interactions were hydrogen bonds (residues of Ile-559, Gly-511, and Val-512) and hydrophobic interactions (**Figures 6A,B**).

Due to possible energy barriers between various intermediate states, conventional MD simulations cannot sample the conformational ensembles. Hence, an enhanced sampling technique to speed up the conformational sampling and take conformational samples at various intermediate states is needed. The traditional enhanced sampling methods often require predefined reaction coordinates (RCs), such as RMSD, atom distances, eigenvectors generated from the principal component analysis, which usually requires rich experience of the simulated systems. Nevertheless, the enhanced sampling method of GaMD simulation avoids such a requirement. Compared with conventional MD simulation, GaMD simulation can take samples at various intermediate states by adding a harmonic boost potential to smoothen the system potential energy surface, which are not accessible to conventional MD simulations. Therefore, the enhanced sampling method, GaMD simulation, was carried out to enhance conformational sampling (19, 20). Alignment of the representative structures from conventional MD and GaMD simulations exhibited high similarity with minor adjustments, indicating that PHL in the binding site of Keap1 was sufficiently stable (**Figure 6C**). The principal component analysis (PCA) further supported this observation (**Figure 6D**). Theoretically, when principal components are plotted against each other, similar structures are clustered and each cluster shows a different protein conformational state. As shown in **Figure 6D**, only one cluster was observed, the target for PHL

is likely to be Keap1. To verify this theory, the expressions of Keap1/Nrf2 complex in the heart tissues were assessed that the expressions level of Nrf2 in diabetic heart tissues. The results showed that the expression of Nrf2 was significantly lower and the levels of Keap1/Nrf2 complex was a little higher in diabetic heart tissues, while PHL inhibited this process (**Figure 6E** and **Figure S4**). In summary, the computational and experiment results demonstrated that Keap1 may be the direct target for PHL to exert anti-oxidative effects and promoted Nrf2 expression for the protective effects in DCM.

#### DISCUSSION

It has been well-established that persistent hyperglycemia in diabetic patients induce cardiomyocyte hypertrophy, myocardial inflammation, fibrosis, and apoptosis (1, 2, 5, 8). As observed in our study, high glucose induced oxidative stress, hypertrophy, and fibrosis in myocardial cells and diabetic mice (**Figures 2**– **5**). Current therapies for DCM focus on intensive blood glucose control (31). However, this strategy does not prevent the development of cardiac complications associated with hyperglycemia. Treatment with anti-diabetics, such as miglitol, significantly reduced the blood glucose via different mechanisms in vivo, while the diabetic complications remained (32). Moreover, a few studies showed that anti-inflammatory agents prevented the development of cardiac and renal injury in streptozotocin (STZ)-induced diabetic mice (2). Meanwhile, a large number of targets, such as TLR4 and FGFR as well as fatty acid-induced cardiac remodeling have been investigated (33, 34). Therefore, we speculate that there are unknown mechanisms contributing to cardiac fibrosis, cardiac remodeling, and cell death.

Recently, natural products have been increasingly evaluated for their effects on cardiovascular diseases. Most notably, flavonoids exhibited properties of inflammation, oxidant stress, cell death, and fibrosis by direct or indirect mechanisms (35, 36). It was recently been reported that several flavonoids, such as quercetin, kaempferol, liquiritin, and baicalein exhibited multiple effects in diabetic complications by inhibiting inflammation, oxidation and fibrosis (37–40). PHL, a natural flavonoid compound derived from apples and pears, was shown to reduce myocardial hypertrophy in acute models of heart disease (41). In addition, PHL prevents high-fat diet-induced obesity and improves metabolic homeostasis through anti-oxidative effect (42). However, to-date, no data is available regarding the effects of PHL on DCM. In this study, we investigated the activities and mechanisms of PHL in an in vivo model for DCM. Following STZ injection for 10 weeks, cardiac remodeling including hypertrophy, cardiomyocyte disorganization, and fibrosis were observed in the hearts of diabetic mice. Interestingly, oral PHL

(10 mg/kg) attenuated cardiac remodeling without any effects on the glucose level or body weight of diabetic mice (**Figure S1**, **Figures 2**–**5**). Similar biochemical results were reflected in hyperglycemia-treated H9c2 cells (**Figures 2**, **3**). It is interesting that PHL attenuated MDA level and upregulated Nrf2, HO-1, and NQO-1 expressions in vitro (**Figure 2**). These findings suggested that PHL has a potential therapeutic effect in DCM, likely through decreased cardiac oxidative stress.

Increasing evidence has shown that oxidative stress plays an important role in the pathophysiology of hyperglycemic induction of cardiovascular disease (43, 44). Cardiac oxidative stress is associated with increased cardiac fibrosis and cell death, leading to the development of severe heart failure (43, 44). Excess oxidation results in ROS aggregation in tissues. In metabolic disease, sustained production of ROS is critical to the development of cardiac injuries. ROS induced mitochondrial DNA has been proposed to be particularly susceptible to oxidative damage. In this study, PHL significantly reduced hyperglycemia-induced ROS accumulation in H9c2 cells and heart tissues from diabetic mice (**Figure 2A** and **Figure 5A**). There is an increasing recognition that Keap1/Nrf2 pathway activation can be beneficial for DCM, as it has been recognized as a key regulator of anti-oxidant defense system by mediating the expression of anti-oxidant genes, such as HO-1, GCLC, NAPD-H, and NQO-1. While hyperglycemia reduced the expression of Nrf2, HO-1, and NQO-1, PHL markedly upregulated these genes in vitro and in vivo (**Figure 2**, **Figure 5**). Interestingly, Keap1 was increased by hyperglycemia, but PHL downregulated Keap1

protein levels. Using molecular modeling and immunoblotting, PHL was determined to bind to Keap1 via hydrogen bonds and hydrophobic interactions (**Figures 6A–D**), which subsequently resulted in increased available Nrf2.

In conclusion, this study demonstrated that PHL could prevent the hyperglycemia-induced cardiac injury. The possible mechanism involved, in prevention of oxidative stress and related cytoprotective effect, could be through degradation of Keap1 and upregulation of Nrf2 expressions, leading to downstream regulation of key detoxifying enzymes (**Figures 2**–**6** and **Figure S3**). Thus, PHL could be a candidate for treating diabetic complications, especially for DCM. In addition, conventional MD and GaMD simulations (**Figures 6A–D**) supported the hypothesized mechanism whereby PHL protected cardiomyocyte from hyperglycemia through disruption of the interaction between Keap1 and Nrf2. Co-IP and western blot analyses also supported this hypothesis (**Figure 6E** and **Figure S4**). However, it is still not clear whether PHL induced Nrf2 transcription and led to elimination of ROS or whether PHL inhibited Keap1/Nrf2 complex first and the subsequent antioxidant response. And the mechanism of PHL how to induced Nrf2 transcription is unknown. The approach using shRNA knockdown in vitro with hyperglycemia would be more robust to solve these shortages. This is a limitation of this study.

Overall, this study strongly supports PHL as a promising natural agent via increased Nrf2 expression dissociation of Keap1/Nrf2 complex, leading to decreased cardiac oxidative stress in DCM. At the same time, this study also highlighted Keap1/Nrf2 pathway as a potential therapeutic target for DCM management.

## AUTHOR CONTRIBUTIONS

YY and JJ who originally designed the project, performed the research, analyzed data, and wrote the manuscript draft. PS, HW, and LY performed partial experiments and data collection. XW were responsible for revising the manuscript. All authors approved the final version of the manuscript.

## FUNDING

Financial support was provided by Natural Science Foundation of Zhejiang Province of China (LQ14H280003) and Zhejiang Medical and Health Science Technology Project (2013KYB062), and Zhejiang Chinese Medical and Health Technology Project (2019ZA030).

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo. 2018.00774/full#supplementary-material

## REFERENCES


by suppressing hepatic glucose production. Molecules (2018) 23. doi: 10.3390/molecules23092338


**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 Ying, Jin, Ye, Sun, Wang and Wang. 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.

# IL-33 at the Crossroads of Metabolic Disorders and Immunity

#### Lei Tu<sup>1</sup> and Lijing Yang<sup>2</sup> \*

*<sup>1</sup> Division of Gastroenterology, Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China, <sup>2</sup> Department of Radiation and Medical Oncology, Zhongnan Hospital, Wuhan University, Wuhan, China*

As a cytokine in interleukin-1(IL-1) family, interleukin-33(IL-33) usually exists in the cytoplasm and cell nucleus. When the cells are activated or damaged, IL-33 can be secreted into extracellular and regulate the functions of various immune cells through binding to its specific receptor suppression of tumorigenicity 2 (ST2). Except regulating the function of immune cells including T cells, B cells, dendritic cells (DCs), macrophages, mast cells, and innate lymphoid cells, IL-33 also plays an important role in metabolic diseases and has received an increasing attention. This review summarizes the regulation of IL-33 on different immune cells in lipid metabolism, which will help to understand the pathology of abnormal lipid metabolic diseases, such as atherosclerosis and type 2 diabetes.

#### Edited by:

*Jixin Zhong, Case Western Reserve University, United States*

#### Reviewed by:

*Yanbo Yu, Qilu Hospital of Shandong University, China Takashi Yazawa, Asahikawa Medical University, Japan Fenna Sille, Johns Hopkins University, United States*

> \*Correspondence: *Lijing Yang yanglijing23@126.com*

#### Specialty section:

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

Received: *23 October 2018* Accepted: *15 January 2019* Published: *30 January 2019*

#### Citation:

*Tu L and Yang L (2019) IL-33 at the Crossroads of Metabolic Disorders and Immunity. Front. Endocrinol. 10:26. doi: 10.3389/fendo.2019.00026* Keywords: IL-33, metabolism, diabetes, innate & adaptive immune response, ST2

#### INTRODUCTION

IL-33, a new member of the IL-1 family, was discovered in 2005 (1) while its receptor ST2 containing intracellular domain Toll/IL-1R (TIR) was found in BALB/c-3t3 mouse fibroblasts in 1989 (1, 2). The receptor complex of IL-33 is composed of ST2 and interleukin-1 receptor accessory protein (IL-1RAcP). IL-33 mediates its biological effect through binding to its specific receptor ST2 (2, 3), whereas the expression of ST2 is restricted and determines the cellular responsiveness to IL-33 treatment (3).Two forms of ST2 have been demonstrated, a membrane-bound form (ST2L) and a soluble form (sST2), the latter which prevents its signaling as the decoy receptor for IL-33. IL-33 is mainly expressed in fibroblasts, epithelial cells and endothelial cells, and especially in high endothelial venules (HEV) (4). Indeed, as designated as an "alarmin," IL-33 is usually released after cell injury to alert the immune system and initiate repair processes. In a recent study, islet mesenchymal-cell-derived IL-33 has been identified as an islet immunoregulatory feature (5). As the receptor of IL-33, ST2 is expressed in many immune cells. IL-33 is a dual-function cytokine. In the absence of inflammatory stimulation, IL-33 is located in the nucleus as a nuclear factor. Once the cell is damaged and/or necrotic, IL-33 can be released from the nucleus and then act as an endogenous "alarmin" (4). The activation signal produced by IL-33/ST2 pathway is transmitted to the cell and a series of signal transmissions activate nuclear factor kappa-light-chainenhancer of activated B cells (NF-κB) and mitogen-activated protein kinase (MAPK) pathway to regulate immune response (1, 6). Under normal physiological condition, inflammation induced by a dysregulated lipid metabolism is benefit for the maintenance of homeostasis and is controlled to avoid excessive damage to the host. However, if not properly controlled, the inflammatory response will promote the excessive production of lipid metabolites, inflammatory cytokines and adhesion molecules, which lead to acute or chronic diseases (7), such as obesity, non-alcoholic steatohepatitis (NASH), atherosclerosis, and acute cardiovascular events. To date, an increasing body of evidence has demonstrated that IL-33 plays a critical role in the lipid metabolism. This review highlights the function of IL-33/ST2 axis on different immune cells in the metabolic disorders.

#### IL-33/ST2L SIGNALING IN INNATE IMMUNE RESPONSES

IL-33 and ST2 have been shown to be expressed in human and murine adipose tissue, and IL-33 expression is strongly correlated with leptin expression in human adipose tissue (8). In addition, administration of IL-33 increases browning of white adipose tissue and energy expenditure in mice (9). These observations show that a critical role of IL-33 played in the adipose tissues homeostasis.

Macrophages have functional plasticity in adipose tissue inflammation, which can exhibit pro-inflammatory or antiinflammatory function. According to the phenotypes and secreted cytokines, macrophages can be divided into two categories named as classical activated macrophages (CAM, M1 type) and alternatively activated macrophages (AAM, M2 type), respectively. CAM are generated in response to helper T1 cells (Th1 cells)-related cytokines, such as interferon-γ (IFN-γ) and tumor necrosis factor-α (TNF-α), while AAM polarization is linked to the helper T2 cells (Th2 cells)-related cytokines (IL-4 and IL-13) (10). Previous studies showed that AAM could attenuate adipose tissue inflammation and obesity-induced insulin resistance (11–14). It has been showed that ST2 can be detected on the cell surface of macrophages. IL-33 can promote the expression of lipopolysaccharide (LPS) receptor components such as myeloid differentiation factor 2 (MD2), tolllike receptor (TLR) 4, soluble cluster of differentiation 14 (CD14) and myeloid differentiation primary response gene 88 (Myd88), which result in an enhanced inflammatory cytokine production (15). However, IL-33 administration improves glucose tolerance, which is associated with the accumulation of M2 macrophages in adipose tissue of ob/ob mice that are the mutant mice to construct the model of Type II diabetes (16). As the result of purine metabolism disorder, gout is a very common metabolic disease in human (17, 18). Hyperlipidaemia is common in gout patients including increased low-density lipoprotein (LDL) cholesterol and decreased high-density lipoprotein (HDL) cholesterol (19). The serum IL-33 expression is predominantly increased in gout patients compared to healthy controls and positively correlated with the expression of HDL, while negatively correlated with LDL expression (20). It has been reported that the elevated IL-33 level is considerably reduced in renal impairment when compared with normal renal function in gout patients (20–22). These data suggest that IL-33 may prevent the kidney injury through regulating the lipid metabolism, which may be resulted from the AAM polarization.

Although ST2 can be detected on the cell surface of macrophages, IL-33/ST2 signaling cannot directly promote AAM polarization. The involvement of IL-33/ST2 signaling in the differentiation and activation of AAM is associated with type II cytokines induction (23–25). A previous finding showed that a population of cells expressing ST2 in adipose was potential to produce large amounts of Th2 cytokines in response to IL-33 (26). Recent studies have named this population as group 2 innate lymphoid cells (ILC2s), characterized by expressing ST2 receptor, and secreting type 2 cytokines such as IL-5 and IL-13 in response to IL-33 (27–30). In addition, soluble ST2 can prevent ILC2s from IL-33 stimulation (31). Recent observation has shown that ILC2s activation favors macrophages toward a protective AAM, which lead to a reduced lipid storage and decrease gene expression of lipid metabolism and adiposeness (32). Furthermore, it has showed that IL-13Rα2 may act as a critical checkpoint in the protective effect of the IL-33/IL-13 axis in obesity (33). In addition, IL-33 promotes β cell function through islet-resident ILC2s that elicite retinoic acid (RA)-producing capacities in macrophages and dendritic cells via the secretion of IL-13 and colony-stimulating factor 2 (5). These data suggest that IL-33 plays a protective role in the adipose tissue inflammation through regulating macrophage function, which is closely associated with the activation of ILC2 to produce type 2 cytokine and IL-4Rα signaling.

## IL-33/ST2L SIGNALING ON T CELL IMMUNE RESPONSES

As a subset of T cells, the regulatory T cells (Tregs) play a critical role in suppressing autoimmune reactivity and have gained an increasing attention in the autoimmune diseases (34). It is shown that an impaired Tregs function is investigated in ST2 gene knockout mice with streptozotocin-induced diabetes, where the glycaemia and β cell loss are severe (35). Indeed, the exogenous IL-33 treatment propagates Tregs expressing the ST2 on the cellular surface, which suggests that the Tregs expansion induced by IL-33 administration is likely to be the result of a direct effect of IL-33 on ST2L<sup>+</sup> Tregs (36, 37). Besides, ST2<sup>+</sup> DCs stimulated by IL-33 to secrete IL-2, which promotes the selective expansion of ST2<sup>+</sup> Tregs vs. non-Tregs, are required for in vitro and in vivo Tregs expansion (37, 38). In the Th1/Th17-mediated allograft rejection, IL-33 treatment can prevent allograft rejection through increasing ST2 positive Tregs in mice (39). In the mouse model of trinitrobenzene sulfonic acid (TNBS)-induced colitis, dextran sulfate sodium (DSS)-induced colitis or T cell adoptive transfer induced colitis, IL-33 can increase the number of Foxp3<sup>+</sup> Tregs (40–42).

The Tregs also play a immunosuppressive function in obesityassociated inflammation (43). Interestingly, studies have also demonstrated that IL-33 maintain homeostasis in adipose tissue. A high level of ST2 expression is observed on human adipose tissue Tregs. Furthermore, IL-33 treatment can induce vigorous population expansion of Tregs in obese mice, and the changes of metabolic parameters are significantly correlated with the increased Tregs (44, 45). IL-33 signaling through the IL-33 receptor ST2 and the myeloid differentiation factor MyD88 pathway is essential for the development and maintenance of Tregs in visceral adipose tissue (44). However, ILC2 intrinsic IL-33 activation is required for Tregs accumulation in vivo and is independent of ILC2 type 2 cytokines but partially dependent on direct co-stimulatory interactions via the inducible costimulator ligand (ICOSL)/ICOS pathway

(46). Concordantly, the ST2<sup>+</sup> Tregs population is with a higher expression of activated marker ICOS and CD44 (38). Thus, IL-33 plays a protective role in adipose tissue inflammation through directly and indirectly regulating Tregs function.

It has also been reported that increasing severity of insulin resistance and microalbuminuria is strongly correlated with the decreased level of IL-33 in patients with diabetic nephropathy, where an enhanced Th1 and suppressed Th2 response is observed (47). ST2 is selectively and stably expressed on the surface of Th2 cells, and IL-33 can effectively induce the immune response of Th2 cells and the expression of Th2 related cytokines IL-5 and IL-13 without increasing IFN-γ expression (48, 49). These studies suggest that the ST2/IL-33 axis is closely associated with the Th1/Th2 response imbalance in the development of diabetes. Atherosclerosis is characterized by the formation of fibrotic plaques in the major arteries and increased Th1 immune response, which leads to myocardinfal iarction and stroke (50, 51). It has been shown that Th1-to-Th2 shift can reduce the development of atherosclerosis (52, 53). Due to the effect of IL-33 on Th2-type immune response, IL-33 exhibits a protective role in the pathogenesis of atherosclerosis (54). Previous findings also showed that the reduced level of IL-33 might increase the risk of atherosclerosis development for certain individuals (55). These data suggest a crucial role of IL-33 in the lipid metabolism through regulating T cells differentiation.

## REFERENCES

1. Schmitz J, Owyang A, Oldham E, Song Y, Murphy E, McClanahan TK, et al. IL-33, an interleukin-1-like cytokine that signals via the IL-1 receptor-related protein ST2 and induces T helper type 2-associated cytokines. Immunity (2005) 23:479–90. doi: 10.1016/j.immuni.2005.09.015

## CONCLUSION

Due to the vital role of IL-33 in the metabolic homeostasis, a sound understanding of the production, regulation, and function of IL-33 will facilitate the treatment of metabolic disorders. The potential mechanisms (**Figure 1**) of IL-33/ST2 axis in the metabolic disorders may include: (1) IL-33 promotes the AAM polarization; (2) IL-33 regulates Tregs and Th2 differentiation and function; and (3) IL-33 regulates the function of ILC2. Notably, the AAM polarization induced by IL-33 depends on Type 2 cytokines, which may be released from ILC2. However, most studies in this area were mainly carried out on animal models and there were limited clinical trials. To what extent IL-33 contributes to metabolic disorders in humans still requires further investigation.

#### AUTHOR CONTRIBUTIONS

LT and LY reviewed the literature and wrote the first draft. LY finalized the manuscript. LT and LY have read and approved the final manuscript.

## FUNDING

This work was supported by the National Natural Science Foundation of China 81700490 to LT.


during allergic airway inflammation. Int Immunol. (2011) 23:307–15. doi: 10.1093/intimm/dxr006


**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 Tu and Yang. 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.

# Role of Hippocampal Lipocalin-2 in Experimental Diabetic Encephalopathy

#### Anup Bhusal 1†, Md Habibur Rahman1†, In-Kyu Lee<sup>2</sup> and Kyoungho Suk 1,3 \*

*<sup>1</sup> BK21 Plus KNU Biomedical Convergence Program, Departments of Biomedical Science and Pharmacology, School of Medicine, Kyungpook National University, Daegu, South Korea, <sup>2</sup> Division of Endocrinology and Metabolism, Department of Internal Medicine, School of Medicine, Kyungpook National University, Daegu, South Korea, <sup>3</sup> Brain Science and Engineering Institute, Kyungpook National University, Daegu, South Korea*

Diabetic encephalopathy is a severe diabetes-related complication in the central nervous system (CNS) that is characterized by degenerative neurochemical and structural changes leading to impaired cognitive function. While the exact pathophysiology of diabetic encephalopathy is not well-understood, it is likely that neuroinflammation is one of the key pathogenic mechanisms that cause this complication. Lipocalin-2 (LCN2) is an acute phase protein known to promote neuroinflammation via the recruitment and activation of immune cells and glia, particularly microglia and astrocytes, thereby inducing proinflammatory mediators in a range of neurological disorders. In this study, we investigated the role of LCN2 in multiple aspects of diabetic encephalopathy in mouse models of diabetes. Here, we show that induction of diabetes increased the expression of both *Lcn2* mRNA and protein in the hippocampus. Genetic deficiency of *Lcn2* significantly reduced gliosis, recruitment of macrophages, and production of inflammatory cytokines in the diabetic mice. Further, diabetes-induced hippocampal toxicity and cognitive decline were both lower in *Lcn2* knockout mice than in the wild-type animals. Taken together, our findings highlight the critical role of LCN2 in the pathogenesis of diabetic encephalopathy.

#### Edited by:

*Jixin Zhong, Case Western Reserve University, United States*

#### Reviewed by:

*Hsien-Hui Chung, National Cheng Kung University, Taiwan Xiaoqiang Tang, Sichuan University, China*

> \*Correspondence: *Kyoungho Suk ksuk@knu.ac.kr*

*†These authors have contributed equally to this work*

#### Specialty section:

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

Received: *25 November 2018* Accepted: *15 January 2019* Published: *30 January 2019*

#### Citation:

*Bhusal A, Rahman MH, Lee I-K and Suk K (2019) Role of Hippocampal Lipocalin-2 in Experimental Diabetic Encephalopathy. Front. Endocrinol. 10:25. doi: 10.3389/fendo.2019.00025* Keywords: Lipocalin-2, diabetic encephalopathy, hippocampus, glia, neuroinflammation, cognitive dysfunction

## INTRODUCTION

Diabetic encephalopathy is one of the most severe microvascular complications of diabetes in the central nervous system (CNS). The measurable manifestations of diabetic encephalopathy include electrophysiological and structural changes, as well as cognitive decline (1, 2). Diabetic encephalopathy involves direct neuronal damage caused by persistent hyperglycemia, a phenomenon referred to as glucose neurotoxicity (1, 3). However, the pathogenesis of this disease is poorly understood and its diagnosis is complicated due to multiple pathogenic pathways involved (4, 5). Cognitive impairment is one of the many consequences of such toxicity that reduces the quality of life of patients (6–9). Studies on both humans and rodents have revealed a close association among hyperglycemia, oxidative stress (10–12), mitochondrial dysfunction (13, 14), and activation of inflammatory pathways (11, 15). These phenomena seem to play a crucial role in the structural and functional damage of neurons in the brain, particularly in the hippocampus, thereby leading to diabetic encephalopathy.

Increasing evidence indicates that inflammation plays an important role in the pathogenesis of diabetic encephalopathy (11, 15). Chronic diabetes initiates an inflammatory response through the activation of glial cells and the upregulation of pro-inflammatory cytokines in the hippocampus in rodent models of diabetes (16, 17). This inflammatory reaction has been suggested to lead to hippocampal neuronal loss (18, 19) and cognitive decline in diabetic mice (17). Therefore, neuroinflammation and its associated endogenous factors in the brain, particularly in the hippocampus, might be useful targets for the prevention and treatment of diabetic encephalopathy. However, molecular targets that can be used for clinical purposes have not been well-explored.

Lipocalin-2 (LCN2), also called neutrophil gelatinaseassociated lipocalin (NGAL), is an acute-phase protein that has been reported to be selectively induced in a variety of tissues in diabetic mice (20, 21). LCN2 is a regulator of immune and inflammatory responses in a range of neurological diseases (22–24). Our previous studies have shown that LCN2 causes gliosis, glial activation, and increased expression of inflammatory cytokines (25–27). This inflammatory microenvironment has been correlated with neurodegenerative phenotypes in several disease models. Further, the circulating levels of LCN2 have been reported to be increased in both rodents (28) and patients with diabetes (29, 30). Based on these observations, we hypothesized that LCN2 might play an important role in the development of diabetic encephalopathy. In this study, we used mouse models of diabetes to examine the pathological role of LCN2 in the progression of diabetic encephalopathy. Our results suggest that LCN2 may contribute to the augmentation of hippocampal inflammation and subsequent pathologies associated with diabetic encephalopathy.

## MATERIALS AND METHODS

#### Mouse Breeding and Maintenance

Male C57BL/6 mice (Samtako, Osan, South Korea) and Lcn2-/ mice on pure C57BL/6 background were housed under a 12-h light/dark cycle (lights on from 07:00 to 19:00 h) at a constant ambient temperature of 23 ± 2 ◦C with food and water provided ad libitum. Lcn2-/- mice were kindly provided by Dr. Shizuo Akira (Osaka University, Japan). Lcn2+/+ and Lcn2-/- mice were back-crossed for eight to ten generations onto the C57BL/6 background to generate homozygous animals free of background effects on the phenotypes, as previously described (31, 32). All experiments were conducted in accordance with the animal care guidelines of the National Institute of Health and efforts were made to minimize the number of animals used as well as animal suffering. Male Lcn2 wild-type (WT, Lcn2+/+) and Lcn2-knockout (KO, Lcn2-/-) mice aged 8–10 weeks were used in further experiments.

## Diabetes Induction

Age-matched Lcn2 knockout (KO, Lcn2-/-) and Lcn2 wild-type (WT, Lcn2+/+) mice of the same background strain (C57BL/6) were used to study diabetic encephalopathy. Diabetes was induced using two experimental protocols described previously (33, 34) with slight modifications. First, multiple low-dose intraperitoneal injections of streptozotocin (STZ) dissolved in citrate buffer at pH 4.5 (MLDS; 40 mg STZ/kg body weight on 5 consecutive days) were administered to induce both pancreatic β-cytotoxic effects and STZ-specific T-cell-dependent immune reactions. Second, a single high-dose intraperitoneal injection of STZ (HDS; 150 mg/kg body weight) was administered to effect direct toxicity in pancreatic β-cells, which results in necrosis within 48–72 h and causes permanent hyperglycemia. Control mice were injected with the same volume of the citrate buffer. The day of the first STZ injection was termed day 0. Blood samples were collected from the tail vein 1 week after the injection, and glycaemia was determined using an SD CodeFreeTM glucometer (SD Biosensor Inc., Suwon, Korea). Animals with fasting blood glucose values >260 mg/dl were considered diabetic.

## High-Fat Diet Model

High-fat diet (HFD) fed mice are a robust and efficient model for type 2 diabetes and are commonly used for both mechanistic studies and as a tool for developing novel therapeutic interventions (35–37). Six-week-old male WT (C57BL/6) mice were fed a HFD in which 20% of the calories were derived from carbohydrates and 60% were derived from fat (D12492 pellets; Research Diets, Inc.). Control animals were fed an isocaloric lowfat diet/control diet (CD) in which 70% of the calories were derived from carbohydrates and 10% were derived from fat (D12450B pellets; Research Diets, Inc.). The mice were housed and maintained on a 12-hr light/dark cycle at 22 ± 2 ◦C. Tissues were rapidly collected by sacrificing the mice according to the experimental time-points, and the samples were immediately stored at −80◦C.

## Enzyme-Linked Immunosorbent Assay (ELISA)

The levels of LCN2 protein in the hippocampus, CSF, and plasma were assessed using the mouse LCN2 ELISA kit (R&D systems). The assays were performed in 96-well plates using the hippocampal tissue, CSF, or plasma (1:10 or 1:100 dilution) as per the manufacturer's protocol. Mouse recombinant LCN2 was used as a standard at concentrations ranging from 75 to 2,500 pg/ml, and the absorbance was measured at 450 and 540 nm using a microplate reader (Molecular Devices). All measurements were obtained from duplicated assays.

#### Immunofluorescence Staining

Deeply anesthetized mice were sacrificed and subjected to intracardiac perfusion-fixation using a solution of 0.9% sodium chloride (VWR International, LLC) and 4% paraformaldehyde (Sigma-Aldrich) in 0.1 M phosphate-buffered saline (PBS, pH 7.4). Isolated brains were immersion-fixed in 4% paraformaldehyde for 24 h, and then incubated in 30% sucrose and embedded in optimal cutting temperature (OCT) compound for cryoprotection (Tissue-Tek; Sakura Finetek USA, Torrance, CA). Staining was carried out as described previously (27) with some modifications. Coronal brain sections (20-µm thick) were rinsed in PBS and incubated overnight with the following primary antibodies: rabbit anti-ionized calcium-binding adapter molecule 1 (Iba-1, 1:200; Wako, Osaka, Japan), rabbit anti-glial fibrillary acidic protein (GFAP, 1:1,000; Dako, Carpinteria, CA), and mouse anti-cluster of differentiation 68 (CD68, 1:200; BMA Biomedicals, Switzerland). Following incubation with primary antibodies, the sections were rinsed and incubated with fluorescein isothiocyanate (FITC)-conjugated and Cy3-conjugated secondary antibodies (1:200; Jackson ImmunoResearch, West Grove, PA). Slides were washed and then coverslipped with VECTASHIELD mounting medium (Vector Laboratories, Burlingame, CA). Images of the immunostained tissues were captured using a fluorescence microscope (Leica Microsystems, DM2500, Wetzlar, Germany).

#### Cresyl Violet, 4′ , 6-Diamidino-2-Phenylindole (DAPI), and Hematoxylin and Eosin (H & E) Staining

Neuronal damage was visualized using DAPI (a fluorescent chromophore that binds to double-stranded DNA in the nuclei of all cells), Cresyl violet (a Nissl stain for evaluating neuronal cell body numbers and features) (Sigma-Aldrich) (38, 39), and H & E staining. Following Cresyl violet staining, the number of pyramidal cells showing a distinct nucleus and nucleolus in each CA1 subfield of the hippocampus were counted (27, 40). For DAPI staining, sections were washed in PBS followed by mounting with VECTASHIELD mounting medium (Vector Laboratories, Burlingame, and CA) (41). Similarly, the tissues were stained with Harris' H & E solution, as described previously (42). Further, the Cresyl violet-, H & E-, and DAPI-stained tissues were visualized using bright field microscopy and fluorescent microscopy, respectively.

#### Quantitative Real-Time and Traditional Reverse Transcription-Polymerase Chain Reaction (PCR)

Mice were deeply anesthetized and then perfused with normal saline through the aorta to remove the blood. Hippocampal tissues were rapidly dissected, frozen in liquid nitrogen, and homogenized in Trizol reagent (Life Technologies, Carlsbad, CA) for total RNA isolation. Total RNA (2 µg) from each sample was reverse-transcribed into cDNA using a first strand cDNA synthesis kit (MBI Fermentas, Hanover, Germany). Real-time PCR was performed using the one-step SYBR <sup>R</sup> PrimeScript TM RT-PCR kit (Perfect Real-Time; Takara Bio Inc., Tokyo) and the ABI Prism <sup>R</sup> 7000 sequence detection system (Applied Biosystems, Foster City, CA), according to the manufacturer's instructions. The 2–11CT method was used to calculate relative changes in gene expression (43), with glyceraldehyde 3-phosphate dehydrogenase (GAPDH) used as an internal control. The nucleotide sequences of the primers used in the real-time-PCR were as follows: Lcn2: forward, 5 ′ -CCC CAT CTC TGC TCA CTG TC-3′ ; reverse, 5′ -TTT TTC TGG ACC GCA TTG-3′ ; Tnf-α: forward, 5′ -ATG GCC TCC TCA TCA GTT C-3′ ; reverse, 5′ -TTG GTT TGC TAC GAC GTG-3′ ; Il-6, 5′ -AGT TGC CTT CTT GGG ACT GA-3 ′ (forward) and 5′ -TCC ACG ATT TCC CAG AGA AC-3 ′ (reverse) Gapdh: forward, 5′ -TGG GCT ACA CTG AGC ACC AG-3′ ; reverse, 5′ -GGG TGT CGC TGT TGA AGT CA-3′ . Likewise, the traditional reverse transcription-PCR was performed using a DNA Engine Tetrad Peltier Thermal Cycler (MJ Research, Waltham, MA). To analyze the PCR products, 10 µl of each PCR reaction was electrophoresed on a 1% agarose gel and detected under ultraviolet (UV) light. The nucleotide sequences of the primers used in the traditional reverse transcription-PCR were as follows: Lcn2: forward, 5′ - ATG TCA ACC TCC ACC TGG TC-3′ ; reverse, 5′ -CAC ACT CAC CCA TTC AG-3′ ; Gapdh: forward, 5′ -ACC ACA GTC CAT GCC ATC AC-3′ ; reverse, 5′ -TCC ACC ACC CTG TTG CTG TA-3′ .

## Novel Object Recognition Test

The novel object recognition (NOR) test was performed as described previously (27, 44, 45). The NOR test was performed over 3 days that included a habituation phase (5 min for 1 day), a training phase (10 min for 1 day), and a test phase (10 min for 1 day) for each mouse. The test was conducted using a metal cylinder (diameter, 7 cm; height, 10 cm) and a rectangular plastic cuboid filled with sand (5 × 5 × 10 cm). During the habituation phase, mice were allowed to acclimatize to the testing arena for 5 min in the absence of objects. During the training session, mice were exposed to two identical objects placed in opposite corners of the arena 3 cm away from the walls, and the mice were allowed to explore the objects for 10 min. During the test session, mice were placed in the arena with the familiar object in its previous location and a novel object in the place of the removed object. The total time spent exploring each object was recorded for 10 min. Exploration was defined as being within 3 cm of an object or touching it with the nose. Objects were thoroughly cleaned between trials to eliminate residual odors. The relative exploration time was measured using a discrimination index [(DI) = (time spent at the novel object—time spent at the familiar object)/(time spent at the novel object + time spent at the familiar object)]. N indicated the time spent near the novel object, while F indicated the time spent near the familiar object. Thus, a positive DI value indicates that the mice spent more time exploring the novel object than the familiar object, whereas a DI of 0 indicates that the mice spent equal amounts of time exploring the two objects.

## Y-Maze Test

The Y-maze test was performed as described previously (46) with slight modifications. Spatial cognition was examined using the spontaneous alternation task in the Y-maze apparatus. The Ymaze is a three-arm horizontal maze (length, 40 cm; width, 3 cm; wall height, 12 cm) in which the three arms are at equal angles from one another. The animals were initially placed at the center of the maze and allowed to move freely through it during a 7-min session. The series of arm entries were recorded visually; an arm entry was considered to be completed when the hind paws of the mouse were completely placed within the arm. A spontaneous alternation was defined as an entry into a different arm on each consecutive choice (i.e., ABC, CAB, or BCA, but not BAB). The arms of the maze were cleaned thoroughly with water before each trial in order to remove any residual odors. The percentage of alternations was defined as [(number of alternations)/(total arm entries)] × 100.

#### Passive Avoidance Test

The step-through passive avoidance test was performed as previously described with minor modifications (47, 48). The test was performed over 3 days, including a habituation phase (day 1), a training phase (day 2), and a test phase (day 3) for each mouse. Briefly, the animals were placed in the light compartment of a two-compartment box (one light, one dark; San Diego Instruments, San Diego, CA) with the door to the dark compartment closed. Following 30 s of exploration, the door was opened. When the mouse entered the dark compartment, the door was closed and the mouse received a single electric shock (0.5 mA, 3 s) on the 2nd day. A retention test was conducted after 24 h, and the step-through latency for the animal to enter the dark compartment was recorded without the use of an electric shock. A maximum latency of 300 s was scored if the animal did not enter the dark compartment at all.

## Quantification and Statistical Analysis

For the immunohistochemical analysis, 3–4 tissue sections/animal were used to analyze microscopic images of the hippocampus. For the determination of immunofluorescence intensity, the whole image was selected, and the average intensity was measured using the ImageJ software (National Institutes of Health, Bethesda, MD). The graphs represent an average of all the images. Statistical analyses were performed with GraphPad Software (GraphPad Software, La Jolla, CA, USA). All the results have been presented as the means ± standard errors (SE). Statistical comparisons were performed using the Student's t-test. Differences with p < 0.05 (p < 0.05) were considered statistically significant.

## RESULTS

#### LCN2 Expression Is Increased in the Hippocampus of Diabetic Mice

To study the role of LCN2 in the pathogenesis of diabetic encephalopathy, we used multiple animal models of diabetes in this study, including the multiple low dose STZ injection (MLDS), high dose STZ injection (HDS), and high fat diet (HFD) feeding models. First, intraperitoneal administration of STZ induced diabetes with elevated blood glucose levels in the first week post-injection; the increased blood glucose levels were maintained throughout the study (**Supplementary Figure 1**). Next, we examined the expression of Lcn2 mRNA and protein following STZ injection using conventional RT-PCR and ELISA, respectively. We found a substantial increase in the expression of Lcn2 mRNA in hippocampal tissues at 8 weeks post MLDS and 4 weeks post HDS administration (**Figures 1A,C**). Similarly, LCN2 protein level in the hippocampus was significantly increased at 8 and 4 weeks post STZ injection in both models compared to the vehicle-injected control groups (**Figures 1B,D**). HFDfed mice showed a similar upregulation of Lcn2 mRNA and increased expression of LCN2 protein in the hippocampus (**Figures 1E,F**). The Lcn2 mRNA expression data was further confirmed by real time-PCR; our results were consistent with the conventional RT-PCR data (**Supplementary Figures 2A–C**).

We predominantly used the MLDS model to investigate the role of LCN2 in the pathogenesis of diabetic encephalopathy because it represents a model of mild type 1 diabetes and closely resembles the pathophysiology observed in human patients (34, 49). The MLDS model is less toxic with a slowly progressive pancreatic β-cell death, and causes gradual elevation of blood glucose levels, minimal muscle weakness and movement abnormalities in mice, thereby providing us with an opportunity to accurately measure animal behavior. We used the HDS model to explore molecular changes and to identify the role of LCN2 in the pathogenesis of diabetic encephalopathy. Further, we also measured the LCN2 level in cerebrospinal fluid (CSF) and blood plasma of the MLDS model mice. The enzyme-linked immunosorbent assay (ELISA) revealed an increased level of LCN2 protein in the CSF and blood plasma of diabetic mice at 8 weeks post MLDS injection, when compared with the vehicle-injected control animals (**Supplementary Figures 2D,E**). These data suggest that induction of diabetes enhances the expression of LCN2 in the hippocampus, CSF, and blood plasma of mice.

## Lcn2 Deficiency Reduces Diabetes-Induced Glial Activation, Proliferation, Macrophage Infiltration, and Proinflammatory Cytokine Expression in the Hippocampus

Glial activation and proliferation are well-known pathological features in the CNS across a range of neurological disorders (17, 50). To investigate the role of LCN2 in diabetes-induced changes in glial characteristics, brain sections were immunostained with anti-Iba-1 and anti-GFAP antibodies to label the microglia and astrocytes, respectively (51, 52). Immunofluorescence analysis showed increased immunoreactivity of Iba-1 in the hippocampus of the diabetic mice at 8 weeks (**Figures 2A,B**) and 4 weeks (**Supplementary Figures 3A,B**) post MLDS and HDS injections, respectively. We observed a significant increase in the number of Iba-1-positive microglial cells in the hippocampus of diabetic mice, where microglia exhibited enhanced Iba-1 immunoreactivity with short and thick processes when compared to control mice. These morphological features of the microglia and the increased Iba-1 immunoreactivity in the hippocampus were attenuated in the Lcn2 KO mice. Similarly, the GFAP-positive astrocytes exhibited increased immunoreactivity and hypertrophic morphology with thick processes in the hippocampus of diabetic mice in comparison with the vehicleinjected control animals; the immunoreactivity and hypertrophic morphology were both significantly reduced in Lcn2 KO mice. Further, the immunofluorescence-based analyses of CD68 (a macrophage marker in the brain) (53, 54) and Iba-1(a marker for both microglia and macrophages) (55, 56) in the brain tissues from diabetic mice showed a marked increase in the number of infiltrated macrophages in the hippocampus at 8 weeks post-MLDS injection compared with vehicle-injected control animals. Mice with Lcn2-deficiency showed a significant decrease in the number of infiltrated macrophages in the hippocampus compared with WT-diabetic animals (**Supplementary Figure 4**).

Furthermore, STZ-induced hyperglycemia increased the expression of Tnf-α and Il-6 mRNA in the hippocampus of diabetic mice, whereas the expression levels of these cytokines were significantly attenuated in the Lcn2-deficient mice (**Figure 2C**). These findings indicate that diabetesinduced activation of glial cells and enhanced expression of proinflammatory cytokines in the hippocampus might be direct consequences of pathological inflammation, which is itself modulated by LCN2 activity (52).

## Lcn2 Deficiency Attenuates Diabetes-Induced Loss of Hippocampal Neurons

The increased expression of pro-inflammatory cytokines in the brain under diabetic conditions plays an important role in neuronal damage (2, 57). LCN2-mediated glial activation and subsequent inflammatory responses in the diabetic hippocampus led us to investigate the histological changes in the CA1 neurons of the hippocampus. Histological analysis using Cresyl violet and DAPI staining revealed fewer number of neurons in the CA1 of the hippocampus of diabetic mice than in that of vehicle-injected control animals. Moreover, such changes in the diabetic hippocampus were ameliorated by Lcn2 deficiency (**Figures 3A,B**). Further, the H & E staining data revealed that many of the granular neurons in area CA1 of the hippocampus of WT-diabetic mice were densely stained and showed shrunken appearance with minimal or no cytoplasm compared with control animals. Fewer such degenerative features were found in the granular neurons of Lcn2 KO mice (**Supplementary Figure 5**). These results indicate that the neurotoxicity in the CA1 region of the hippocampus might be related to the diabetes-induced elevation of LCN2 levels and its contribution to augmented inflammatory processes.

## Attenuation of Diabetes-Induced Cognitive Impairment in Lcn2 KO Mice

To investigate whether the cognitive function of the diabetic mice is impaired and if Lcn2 deficiency can ameliorate this impairment, we examined the cognitive function of mice using the NOR test, which is based on differential exploration behavior with respect to familiar and new objects, and used to study short-term declarative memory and attention (58). As shown in **Figure 4A**, the discrimination index of the WTdiabetic mice was significantly lower than that of the vehicleinjected control animals. However, Lcn2-deficient diabetic mice showed an improved discrimination index compared to the WT-diabetic animals. As an alternative method to confirm whether Lcn2 deficiency ameliorates diabetes-induced memory dysfunction, we used the Y-maze test, which measures the natural behavioral tendency of mice to explore new environments. Mice typically prefer to investigate a new arm of the maze rather than returning to one that has already been visited. Many parts of the brain, including the hippocampus, are involved in this task (46). Diabetic mice showed a decline in the percentage of spontaneous alteration compared to control animals, indicating an impairment in spatial memory function. However, Lcn2 deficiency ameliorated this diabetes-induced memory dysfunction (**Figure 4B**). Further, diabetic mice of both genotypes (WT and KO) showed decrease in movement velocity and the number of arm entries at 8 weeks post-MLDS injection, which might be associated with diabetes-induced motor impairments. In line with our findings, several studies have shown a reduction in the number of total arm entries of diabetic mice compared with non-diabetic controls (59–61). Likewise, another study has found a marked reduction in speed and traveling distance of diabetic mice, indicating impairment of exploratory behavior and motor activity (62). In order to address these limitations and to confirm the changes observed in NOR and Y-maze test performances following diabetes, we performed the passive avoidance test, which is fear-aggravated, and requires lesser movement to evaluate learning and memory in rodent models of CNS disorders (48, 63). This behavior requires the association between a normally neutral environment and an aversive stimulus, which is dependent on hippocampal function. In passive avoidance test, we found that the latency during the learning trial did not differ among the experimental and control

groups, indicating that all the mice had similar responses to the testing environment and the electric shock. However, as shown in **Figure 4C**, Lcn2-deficient mice showed longer latency than the WT-diabetic animals 24 h after the training process. However, we found no significant difference between the vehicleinjected control mice groups. These findings suggest that LCN2 plays a crucial role in the development of cognitive impairment associated with diabetic encephalopathy.

## DISCUSSION

of the mean.

Diabetic encephalopathy is one of the most severe complications of diabetes mellitus that involves cognitive dysfunction and an increased incidence of dementia (64–69). Growing evidence points to the role of neuroinflammation in the progression of clinically well-recognized complications of diabetes like encephalopathy (70, 71). Several epidemiological studies have shown a strong link between poor cognitive ability and elevated inflammatory markers in patients with diabetes (72–74). Thus, targeting inflammation might be an important therapeutic strategy for the treatment of diabetic encephalopathy. Inflammation, especially in the hippocampus may lead to impairments in a variety of cognitive domains as it is the center for learning and memory processing in the brain (75, 76). The role of LCN2 in neuroinflammation and cognitive function has already been established. In the current study, we used STZ-induced diabetes and HFD mice models to evaluate the expression level of LCN2 in the

hippocampus. We further employed the MLDS model to study the encephalopathy and associated cognitive impairment in diabetic mice in greater detail. We found that the induction of diabetes through STZ administration and HFD feeding significantly increased the expression of LCN2 in the hippocampus at the level of mRNA and protein. Further, deletion of the Lcn2 gene ameliorated the diabetes-induced reactive gliosis in the hippocampus seen in the WT-diabetic mice. In addition, the hyperglycemia-induced increase in expression of pro-inflammatory cytokines was significantly attenuated in Lcn2 KO mice. Moreover, Lcn2-deficient mice showed decreased neuronal loss in the CA1 region of the hippocampus following diabetes, an effect that was correlated with improved cognitive behavior in these animals. These results suggest that the diabetes-induced increase in the hippocampal LCN2 level causes neuroinflammation, which may play an important role in the development of diabetic encephalopathy and associated impairments in cognitive function (**Figure 5**).

Several in vivo studies have revealed the effect of STZinduced diabetes on LCN2 expression in the peripheral tissues of rodents (77, 78). In these studies, STZ-induced diabetic rodents showed enhanced expression of LCN2 in the kidney and adipose tissues, respectively. However, the expression of LCN2 in the brain has not yet been studied using a mouse model of STZinduced diabetes. To our knowledge, this is the first study to report the upregulation of LCN2 and its pathological role in the mouse hippocampus following STZ injection, which is an insulin-deficient diabetes model. As HFD-fed and STZ-injected mice are characterized by hyperinsulinemia and insulinopenic state, respectively, it is important to understand whether the presence or absence of insulin affects the expression of LCN2.

It has been reported that insulin treatment of STZ-induced diabetic mice leads to decreased LCN2 expression in skin excisional wound tissues, which has been proposed as a potential therapeutic target for improving diabetic wound healing (79). Further, HFD-fed mice show insulin resistance and hyperinsulinemia (80). In this condition, insulin cannot be used to correct the diabetic state. It has been reported that proliferatoractivated receptor-γ agonist rosiglitazone, an insulin sensitizer, markedly decreases LCN2 expression in both obese mice and humans, which is correlated with a decrease in inflammation and improved insulin sensitivity (20, 28). However, upregulation of LCN2 in 3T3-L1 adipocytes by insulin treatment under hyperglycemic condition has also been reported, suggesting that glucose metabolism is required for the effect of insulin on LCN2 expression (81). In line with this, a human study revealed that circulating LCN2 level is increased after insulin intake, and also

deficit was assessed using the Y-maze test. The number of alternations and total arm entries (B) were compared. Cognitive impairment was further evaluated using the passive avoidance test (C). \**p* < 0.05 vs. the vehicle-treated control animals; #*p* < 0.05 between the indicated groups; Student's *t*-test; *n* = 8–10 for each group; data are represented as mean ± SEM. WT, wild type; KO; knockout; MLDS, multiple low dose of STZ; w, weeks; NS, not significant; SEM, standard error of mean.

highlighted the involvement of both phosphoinositide 3-kinase and mitogen-activated protein kinase signaling pathways in insulin-induced LCN2 upregulation (82). These findings suggest a regulatory role of insulin in mediating the effects of LCN2 in metabolic diseases. Future studies may explain the mechanistic correlation between metabolic hormones and LCN2 induction with respect to neuroinflammation.

Several epidemiological studies have found increased levels of markers and mediators of inflammation in diabetic condition (83, 84). In the brain, microglia and astrocytes are the immune cells, and play a nuanced role in neuroprotection and maintenance of homeostasis (85, 86). However, under pathological conditions, such as diabetes, reactive gliosis is a wellknown phenomenon (16, 87, 88). Consistently, in the present study, STZ-induced diabetes caused significant microglial and astrocytic activation 8 weeks after the injection. However, glial activation was attenuated in the Lcn2 KO diabetic mice compared with WT mice. This result can be compared to our previous

finding, in which LCN2 produced phenotypic changes in glia via Rho-associated protein kinase, and affected their migration by secreting chemokines (52). Further, LCN2 is known to act via autocrine and paracrine mechanisms to activate and recruit macrophages at the site of inflammation (89). These macrophages may further induce inflammatory mediators and exacerbate the immune response. In our study, we observed increased number of macrophages in the hippocampus of diabetic mice. A previous study using fluorescence-activated cell sorting revealed an increase in blood-brain barrier permeability, leading to macrophage infiltration in the brain of db/db mice (90). These findings suggest that the gliosis and macrophage infiltration that occur in the hippocampus of diabetic mice is likely mediated by LCN2, and that deletion of Lcn2 prevents the inflammatory reactive gliosis and macrophage infiltration associated with encephalopathy.

Activation of immune cells in the brain of diabetic subject has been associated with several manifestations of diabetic encephalopathy (84, 91). This has been attributed to the effects of poorly controlled hyperglycemia that triggers activation of inflammatory transcription factors, such as nuclear factor kappa B, and the release of proinflammatory cytokines (91– 93). LCN2 is known to stimulate various immune cells like neutrophils (94), microglia, and astrocytes (27, 95) to produce vital proinflammatory mediators, such as IL-6, IL-8, and TNF-α. In addition, the inflammatory factors released by microglia can activate astrocytes, and factors released from astrocytes may, in turn, activate microglia, further aggravating the situation (96, 97). These neurotoxic inflammatory cytokines, once released under diabetic conditions, might act directly on neurons to induce apoptosis and cell death (98, 99). In our study, WT-diabetic mice showed elevated inflammatory cytokines such as Tnf-α and Il-6 in the hippocampus, which may be due to LCN2-mediated activation of glial cells and recruitment of immune cells therein. Cognitive impairment following HFD consumption is associated with neuroinflammation impairing insulin signaling in the hippocampus of experimental animals (100). Similarly, in STZinduced diabetic mice, hyperglycemic stress and inflammation progressively worsen the condition in the hippocampus, leading to neurodegeneration and, ultimately, diabetic encephalopathy (4). Based on these findings, it is suggested that LCN2-mediated neuroinflammation may be a potential mechanism underlying diabetic encephalopathy and cognitive deficit in both type 1 and type 2 diabetes mellitus.

Chronic neuroinflammation plays an important role in the onset and progression of various neurodegenerative diseases (101). Neurodegeneration is a condition in which neuronal structure and function are altered, leading to reduced neuronal survival and increased neuronal death in the CNS (102). Recent evidence points to an important role of LCN2 in the pathophysiology of sterile inflammatory conditions like obesity and diabetes (28, 30, 103, 104). However, LCN2 has been shown to have contradictory roles in the development of obesity or diabetes in rodents. An in vitro study by Zhang et al. recently demonstrated that exposure of primary adipocytes to recombinant LCN2 protein inhibits the expression of Tnf-α mRNA, which is suggested to be mediated by the induction of peroxisome proliferator-activated receptor gamma, a key anti-inflammatory transcription factor (105). On the contrary, Law et al. have shown that LCN2 deficiency attenuates obesity-induced expression and activity of 12-lipoxygenase and production of TNF-α in the mouse fat tissue (103). However, several studies have reported that LCN2 contributes to immune and inflammatory responses in the brain, eventually leading to the development of neurodegenerative diseases (22, 23, 106, 107). Once released, LCN2 binds to its cell-surface receptor LCN2R to regulate neuroinflammation and cell death (31, 108). In our study, Cresyl violet and DAPI staining revealed significant neuronal loss in the CA1 region of the hippocampus of diabetic WT mice compared to Lcn2 KO mice, which is consistent with other studies (27, 41). Further, this result was supported by H & E staining. These results suggest that elevated levels of LCN2 in the hippocampus may potentiate neuroinflammation and cause neuronal loss following diabetes.

Various studies indicate that neuroinflammation and neuronal cell death are implicated in diabetes-associated learning and memory deficits (57, 62). Previously, our research group has reported increased LCN2 levels in patients with cognitive impairment (109); these findings are consistent with the study by Dekens et al. (110). In this study, WT-diabetic mice showed poor cognitive function when compared with Lcn2 KO mice as measured by the NOR, Y-maze, and passive avoidance tests. In chronic conditions such as diabetes, oversecretion of LCN2 in the hippocampus may subsequently aggravate the neural imbalance in the hippocampus, and lead to impaired cognitive function. Our previous findings of LCN2 treatment inducing inflammatory activation of glial cells and having a toxic effect on co-cultured hippocampal neurons supports this notion (27). Other groups have reported a specific role of LCN2 in the decline of cognitive function (111) and suggested that its upregulation following stress reduces dendritic spine density in hippocampal neurons (112). Taken together, these findings suggest that LCN2 can mediate neuroinflammation via activation of glial cells and increased expression of inflammatory cytokines which leads to hippocampal neuronal death and impairs cognitive function across domains in patients with diabetes (**Figure 5**). Thus, LCN2 upregulation in the hippocampus might be a potential pathogenic mechanism leading to further disruption of the hippocampus through neuroinflammation in the diabetic state. Increased level of circulating LCN2 has been considered an inflammatory marker closely associated with insulin resistance and hyperglycemia in patients with diabetes (20). Therefore, the control of hyperglycemia-induced expression of LCN2 or its activity in the hippocampus may be important for neuroprotection in these patients.

In summary, our findings suggest that increased LCN2 expression due to diabetes is critical for the development of several manifestations of diabetic encephalopathy, in which LCN2-mediated inflammatory reaction as well as direct toxicity by interacting with its receptors in the hippocampal neurons of diabetic animals has been proposed as a potential mechanism. The extent to which it can worsen cognitive ability in diabetic mice makes LCN2 a promising target for therapeutic interventions against diabetic encephalopathy.

## ETHICS STATEMENT

This study was carried out in accordance with the recommendations of Kyungpook National University Animal Care Committee. The protocol was approved by Kyungpook National University Animal Care Committee.

## AUTHOR CONTRIBUTIONS

All authors have made a substantial intellectual contribution to this work, and approved submission of the manuscript. AB and MR designed and performed the research, analyzed the data, and prepared the manuscript. I-KL provided essential reagents. KS directed the study and was involved in all aspects of the experimental design, data analysis, and manuscript preparation.

## ACKNOWLEDGMENTS

This work was supported by a grant from the Korea Healthcare Technology R&D Project, Ministry of Health & Welfare, Republic of Korea (HI16C1501) and the Basic Science Research Program through the National Research Foundation (NRF), which is funded by the Korean government (MSIP) (2018R1A2A1A05077118, 2016M3C7A1904148, NRF-2017R1A5A2015391).

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo. 2019.00025/full#supplementary-material

Supplementary Figure 1 | Blood glucose levels in STZ-induced diabetic mice. After intraperitoneal injection of low dose of STZ (MLDS), blood glucose levels significantly increased within the first week and remained significantly high throughout the course of the study as compared to control non-diabetic animals. <sup>∗</sup>*p*< 0.05 vs. the vehicle-treated control animals; Student's *t*-test; *n* = 4–6 for each group; data are represented as mean ± SEM. WT, wild type; KO; knockout; MLDS, multiple low dose of STZ; w, weeks; SEM, standard error of mean.

Supplementary Figure 2 | LCN2 expression following diabetes induction. The expression of *Lcn2* mRNA in the hippocampus at 8 w of MLDS (A) and 4 w of HDS (B) injection, and 24 w post HFD feeding (C) was assessed using real-time PCR. Further, the expression levels of LCN2 protein in the CSF (D) and blood plasma (E) from diabetic mice were estimated using ELISA. <sup>∗</sup>*p*< 0.05 vs. the vehicle-treated control animals; Student's *t*-test; *n* = 3 for each group; data are represented as mean ± SEM. MLDS, multiple low dose of STZ; HDS, high dose of STZ, HFD, high fat diet; LCN2, Lipocalin-2; PCR, polymerase chain reaction; CSF, cerebrospinal fluid; ELISA, enzyme-linked immunosorbent assay; w, weeks; SEM, standard error of mean.

Supplementary Figure 3 | Effect of *Lcn2*-deficiency on diabetes (HDS)-induced gliosis in the hippocampus. Immunoreactivity (IR) of Iba-1 and GFAP was increased in the hippocampus of WT mice after 4 w of HDS injection, whereas *Lcn-2* deficiency attenuated such an increase in IR (A,B). The quantification of relative intensity of IR is presented adjacent to the microscopic images. <sup>∗</sup>*p* < 0.05 vs. the vehicle-treated control animals; #*p* < 0.05 between the indicated groups; Student's *t*-test; *n* = 3 for each group; data are represented as mean ± SEM. Scale bar, 200µm (Iba-1 stained images) and 100µm (GFAP-stained images). WT, wild-type; KO, knockout; NS, not significant; HDS, high dose of STZ; w, weeks; SEM, standard error of mean; Iba-1, ionized calcium-binding adapter molecule 1; GFAP, glial fibrillary acidic protein.

Supplementary Figure 4 | Macrophage infiltration in the hippocampus of diabetic mice. To determine the role of *Lcn2*-deficiency in diabetes-induced infiltration of macrophages in the hippocampus, co-immunostaining of CD68 (red) and Iba-1 (green) was performed using brain tissue sections collected from WT and *Lcn2* KO mice at 8 w post-MLDS/vehicle injection. Quantification of CD68+/Iba-1<sup>−</sup> cells (macrophages) is presented adjacent to the representative images. Arrows indicate the CD68<sup>+</sup> macrophages. <sup>∗</sup>*p* < 0.05 vs. vehicle-treated control animals, #*p* < 0.05 between the indicated groups, Student's *t*-test; *n* = 3 for each group; data are presented as mean ± SEM. Scale bar, 200µm. CD68, cluster of differentiation 68; Iba-1, ionized calcium-binding adapter molecule 1; WT, wild type; KO, knockout; MLDS, multiple low doses of streptozotocin; w, weeks; SEM, standard error of the mean.

#### REFERENCES


Supplementary Figure 5 | *Lcn2*-deficiency confers protection against diabetes-induced degeneration of hippocampal granular neurons. To determine the role of LCN2 in diabetes-induced loss of hippocampal tissue integrity and neuronal degeneration, H and E staining was performed using brain tissue sections collected from WT and *Lcn2* KO mice at 8 w post-MLDS/vehicle injection. Quantification of percentage of degenerated neurons in the hippocampus is presented adjacent to the representative images. <sup>∗</sup>*p* < 0.05 vs. vehicle-treated control animals, #*p* < 0.05 between the indicated groups, Student's *t*-test; *n* = 3 for each group; data are presented as mean ± SEM. Scale bar, 200 and 100µm in the original and magnified images, respectively. LCN2, lipocalin-2; H and E, hematoxylin and eosin; WT, wild type; KO, knockout; MLDS, multiple low doses of streptozotocin; w, weeks; SEM, standard error of mean.


GPR55 and GPR119 in multiple low-dose streptozotocin-induced diabetic and incretin-receptor-knockout mice. Diabetologia (2016) 59:2674–85. doi: 10.1007/s00125-016-4108-z


type 2 diabetes, but not type 1 diabetes. J Diabetes Res. (2018) 2018:1470476. doi: 10.1155/2018/1470476


spine formation and maturation. Proc Natl Acad Sci USA. (2011) 108:18436– 41. doi: 10.1073/pnas.1107936108

**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 Bhusal, Rahman, Lee and Suk. 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.

# Corrigendum: Role of Hippocampal Lipocalin-2 in Experimental Diabetic Encephalopathy

#### Approved by:

*Frontiers in Endocrinology Editorial Office, Frontiers Media SA, Switzerland*

> \*Correspondence: *Kyoungho Suk ksuk@knu.ac.kr*

*†These authors have contributed equally to this work*

#### Specialty section:

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

Received: *25 March 2019* Accepted: *26 March 2019* Published: *09 April 2019*

#### Citation:

*Bhusal A, Rahman MH, Lee I-K and Suk K (2019) Corrigendum: Role of Hippocampal Lipocalin-2 in Experimental Diabetic Encephalopathy. Front. Endocrinol. 10:239. doi: 10.3389/fendo.2019.00239* Anup Bhusal 1†, Md Habibur Rahman1†, In-Kyu Lee<sup>2</sup> and Kyoungho Suk 1,3 \*

*<sup>1</sup> BK21 Plus KNU Biomedical Convergence Program, Departments of Biomedical Science and Pharmacology, School of Medicine, Kyungpook National University, Daegu, South Korea, <sup>2</sup> Division of Endocrinology and Metabolism, Department of Internal Medicine, School of Medicine, Kyungpook National University, Daegu, South Korea, <sup>3</sup> Brain Science and Engineering Institute, Kyungpook National University, Daegu, South Korea*

Keywords: Lipocalin-2, diabetic encephalopathy, hippocampus, glia, neuroinflammation, cognitive dysfunction

#### **A Corrigendum on**

#### **Role of Hippocampal Lipocalin-2 in Experimental Diabetic Encephalopathy**

by Bhusal, A., Rahman, M. H., Lee, I.-K., and Suk, K. (2019). Front. Endocrinol. 10:25. doi: 10.3389/fendo.2019.00025

In the original article, there was a mistake in **Figure 1** as published. The Gapdh band image used in **Figure 1C** was from the pilot experiment. The corrected **Figure 1** appears below.

The authors apologize for this error and state that this does not change the scientific conclusions of the article in any way. The original article has been updated.

Copyright © 2019 Bhusal, Rahman, Lee and Suk. 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.

injection was assessed by conventional PCR (A,C). Further, the expression level of LCN2 protein in the hippocampus of STZ-induced diabetic mice was estimated by ELISA assay (B,D). Similar upregulation of *Lcn2* mRNA and LCN2 protein was detected in the hippocampus at 24 w post HFD feeding (E,F). \**p* < 0.05 vs. the vehicle-treated control animals; Student's *t*-test; *n* = 3 for each group; data are represented as mean ± SEM. STZ, streptozotocin; MLDS, multiple low dose of STZ; HDS, high dose of STZ; HFD, high fat diet; LCN2, Lipocalin-2; w, weeks; SEM, standard error of the mean.

# IL-33 Ameliorates the Development of MSU-Induced Inflammation Through Expanding MDSCs-Like Cells

Ke Shang1†, Yingying Wei 2†, Qun Su2†, Bing Yu<sup>2</sup> , Ying Tao<sup>2</sup> , Yan He<sup>2</sup> , Youlian Wang<sup>1</sup> , Guixiu Shi <sup>2</sup> \* and Lihua Duan1,2 \*

*<sup>1</sup> Department of Rheumatology and Clinical Immunology, Jiangxi Provincial People's Hospital, Nanchang, China, <sup>2</sup> Department of Rheumatology and Clinical Immunology, The First Affiliated Hospital of Xiamen University, Xiamen, China*

#### Edited by:

*Wen Kong, Wuhan Union Hospital, China*

#### Reviewed by:

*Hui Yin, Guangdong Pharmaceutical University, China Carlos Alfaro, NavarraBiomed, Spain*

#### \*Correspondence:

*Lihua Duan lh-duan@163.com Guixiu Shi gshi@xmu.edu.cn*

*†These authors have contributed equally to this work*

#### Specialty section:

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

Received: *20 November 2018* Accepted: *16 January 2019* Published: *26 February 2019*

#### Citation:

*Shang K, Wei Y, Su Q, Yu B, Tao Y, He Y, Wang Y, Shi G and Duan L (2019) IL-33 Ameliorates the Development of MSU-Induced Inflammation Through Expanding MDSCs-Like Cells. Front. Endocrinol. 10:36. doi: 10.3389/fendo.2019.00036* Interleukin-33 (IL-33), a member of the IL-1 superfamily, has been shown to play a critical role in many diseases through regulating the immune cell responses, including myeloid-derived suppressor cells (MDSCs). Our previous study demonstrated that IL-33 might play a protective role in kidney injury in gout patients by regulating the lipid metabolism. However, the role of IL-33in the development of MSU-induced inflammation remains elusive. In this study, an increased IL-33 expression was observed in gout patients, which was positively correlated with inflammatory marker CRP. To explore the effects and mechanisms of the increased IL-33 expression in the gout patients, the anti-ST2 antibody and exogenous recombinant IL-33 were used in MSU-induced peritonitis animal model that mimics human gout. Compared with control group, mice with exogenous recombinant IL-33 significantly ameliorated the inflammatory cells infiltration, while blockage of IL-33 signaling by anti-ST2 had no effect on the development of MSU-induced peritonitis. Furthermore, the crucial inflammatory cytokine IL-1β was markedly decreased in IL-33-treated mice. Besides that, a large number of anti-inflammatory MDSCs with CD11b+Gr1intF4/80<sup>+</sup> phenotype was observed in the IL-33-treated mice, and adoptive transfer of IL-33-induced MDSCs (CD11b+Gr1intF4/80+) markedly inhibited the IL-1β production in MSU-induced peritonitis. In conclusion, our data provide clear evidences that the increased expression of IL-33 in the gout patients might be due to a cause of self-negative regulation, which inhibits the development of MSU-induced inflammation through expanding MDSCs. Thus, IL-33 might serve as a promising therapeutic target for gout.

Keywords: gout, MDSCs, IL-33, MSU, IL-1β

## INTRODUCTION

Gout is the most common inflammatory arthritis caused by inflammatory responses to the deposition of monosodium urate (MSU) crystals. The uric acid levels exceed the physiological saturation concentration, which will lead to monosodium urate crystal (MSU) formation (1). MSU is precipitated in a single crystalline form and deposited in synovial, cartilage, and other tissues around the joint. MSU crystals trigger an intense inflammatory response by activating the resident tissue macrophages and promoting the recruitment of neutrophils to the joint (2). IL-1β is the most important proinflammatory regulatory cytokine from activated macrophage by MSU stimulation in the onset of acute gout (3). Likewise, it has also demonstrated that a high level of IL-1β can be detected in the sera of gout patients. During the development of MSU-induced inflammation, mature IL-1β production by macrophages induces activation of IL-l signaling pathways and medullary differentiation factor (MyD88)-dependent NFκB pathway, resulting in a large number of chemokines and pro-inflammatory factors being produced, and inflammatory immune cells infiltration (4). An interesting character of acute gout is the self-limiting nature of the inflammatory flare. In the absence of clinical intervention, a gouty episode can spontaneously resolve within 7–10 days. However, the mechanism of inflammation self-resolution is still elusive (5, 6).

Interleukin-33 (IL-33), a member of the IL-1 family, is widely expressed in a variety of tissue cells, especially epithelial cells, endothelial cells, and fibroblast cells (7, 8). Numerous studies have shown the complex biological effects of IL-33 in human diseases (9). As an inflammatory factor, IL-33 can promote the pathological development of diseases like asthma, allergic rhinitis and rheumatoid arthritis by inducing type2-immune response and activating mast cells (10–13). In contrast, IL-33 also can act as an anti-inflammatory factor to suppress inflammation by promoting alternatively activated macrophage polarization and Tregs differentiation. Recent studies have shown that IL-33 prevents the development of parasitic infection, allogeneic allograft rejection and atherosclerosis (14, 15). However, the role of IL-33 in the development of MSU-induced inflammation remains unclear.

Myeloid-derived suppressor cells (MDSCs) are a group of heterogeneous cells discovered in recent years, including immature myeloid cells, immature granulocytes, monocytesmacrophages, dendritic cells, and myeloid precursor cells. MDSCs are commonly divided granular cell-like MDSC (CD11b+Gr-1high) and mononuclear cell-like MDSC (CD11b+Gr-1int) (16, 17). Numerous studies have shown that MDSCs play a role in the regulation of effector T cells by nitric oxide production, arginase I synthesis and reactive oxygen species (ROS) production, and promoting Treg cells expansion through indoleamine-2,3-dioxygenase (IDO) production (18). In addition, MDSCs also regulate macrophages and dendritic cells function by the production of IL-10 and TGF-β (16). It is known that varieties of factors could affect the induction and activation of MDSCs, such as IL-6, IL-4 and IL-13 (17, 19). In allogeneic transplantation and tumor environments, it has been shown that IL-33 can induce MDSCs expansion, which inhibits transplant rejection and promotes the progression of mouse breast cancer through inhibition of T cell responses (20–22). Recent literature reported that LPS can induce mononuclear cell-like MDSC, which can block the action of neutrophils causing tissue damage by phagocytose them, thereby alleviating bacterial septic shock (23).

In our study, we found that the expression of IL-33 in gout patients was significantly higher than that of the healthy control group, which was positively correlated with inflammatory indicator C-reactive protein (CRP). However, mice treated with exogenous recombinant IL-33 could significantly alleviate MSUinduced inflammation through expanding CD11b+Gr1intF4/80<sup>+</sup> MDSCs. However, blockage of endogenous IL-33 signaling by anti-ST2 had no effect on the development of MSU-induced peritonitis. Taken together, our data demonstrated that the increased levels of IL-33 may be a cause of self-negative regulation which inhibits the development of MSU-induced inflammation, and IL-33 might offer an alternative therapy to our current approaches of managing gout.

## MATERIALS AND METHODS

## Human Subjects and Animals

Fifty-two cases of acute gout patients were collected from the First Affiliated Hospital of Xiamen University in accordance with the diagnosis criteria of American Rheumatology Association &&(1977). Fifty-eighty healthy controls who matched their age and sex were also collected. The research program has been approved by the Ethics Committee of Xiamen University, and all subjects have signed an informed consent form in accordance with the Declaration of Helsinki. Male 6–8 week C57BL/6 were obtained from the Animal Laboratory Center of Xiamen University. All animal studies were conducted in accordance with the guidelines of the Animal Care and use committee of Xiamen University.

## Culture of Human Synovial Fibroblasts

In sterile conditions, we isolated the synovial fibroblasts from the joint fluid of the acute gout patients. Cells were cultured in Dulbecco's modified Eagle's medium plus 10% Fetal Bovine Serum (Hyclone, United States) and seeded at the flask, fed after 48 and 72 h. When the cells grown to sub-confluence (85%) of the culture dishes, used trypsin (Hyclone, United States) to passage cell, rinsed with PBS and plated into the dish. We then collected non-adherent cells after 2 h culture in Dulbecco's modified Eagle's medium plus 10% Fetal Bovine Serum. Repeated passages to the third generation, fibroblast cells can be used for subsequent experiments.

## Histological and Immunohistochemical Analysis

The synovial fibroblasts were put on cover slides in the 6-well plate and were treated with MSU, TNFα, or IL-1β for 24 h. Next, the cell slides were taken out for immunohistochemistry analysis. Cell slides were fixed for 5 min in 4% paraformaldehyde solution**,** rinsed twice with PBS, followed by 0.5% Triton-100 solution treatment. The slides were treated with 3% BSA (Sangon Biotech, Shanghai) for 30 min at 37◦C for blocking non-specific staining. After that, the slides were then incubated with goat anti-human IL-33 Ab (Minneapolis, MN, United States) or control gout IgG at 4◦ overnight, and were then used hypersensitive two-step immunohistochemical detection reagent (ZSGB-BIO, China) to detect the IL-33 expression levels by microscope.

#### Establishment of Acute Gout Animal Model

Male C57BL/6 mice (6–8 weeks old) were grouped into PBS, MSU, IL-33, and IL-33 plus MSU. The mice of PBS group and MSU group were daily administered an intraperitoneal (IP) injection of PBS for 4 days; the mice of IL-33 group and IL-33 + MSU group were daily administered an intraperitoneal (IP) injection of 2 µg rIL-33 for 4 days. On the fourth day, the mice of MSU group and IL-33 plus MSU group were administered an intraperitoneal (IP) injection of 3 mg MSU after inoculation; the mice of PBS group and IL-33 group were administered an intraperitoneal (IP) injection of PBS after inoculation. Expression and purification of mouse recombinant IL-33 were carried out as previously described (24). For blockage of endogenous IL-33 activity, a neutralizing anti-ST2 antibody (DIH4; Biolegend, San Diego, CA, United States) or control IgG (200 µg/mouse) was given into MSU-treated mice 1 h before MSU administration. At 16 h after MSU stimulation, peritoneal exudate cells were harvested by lavage with 3 ml PBS. Cells were analyzed by flow cytometry and lavage fluids were retained for cytokine assay. In the MDSCs adoptive transfer experiments, the CD11b+Gr-1 highF4/80+, CD11b+Gr-1intF4/80+, and CD11b+Gr-1intF4/80<sup>−</sup> were isolated from IL-33-treated mice, and then the 3 mg MSU per mouse were administered, the lavage fluids were harvested for cytokine assay after 16 h.

#### Flow Cytometry

1 × 10<sup>6</sup> peritoneal exudate cells were obtained, and the cells were incubated with the fluorescent-conjugated monoclonal antibodies in the staining buffer. Antibodies used for flow cytometry are as follows: FITC anti-mouse F4/80, Percy5.5 antimouse CD11b and PE/Cy7 anti-mouse Gr-1. All antibodies are purchased from Biolegend (San Diego, CA, United States).

## ELISA (Enzyme-Linked Immunosorbent Assay)

The gout patients and healthy volunteers were collected 3∼5 ml through elbow vein in the morning, and put into nonanticoagulant test tube, the serum samples were separated in 2 h. The concentration of IL-33 was determined by ELISA Kit according to the Manufacturer's instruction. IL-1β, IL-6, IL-10, IL-5, and IL-13 in the peritoneal lavage fluid of model mice were also determined by ELISA Kit. All kits were purchased from R&D (Minneapolis, MN, United States).

#### Quantitative Real-Time Reverse Transcriptase-Polymerase Chain Reaction

The total RNA was extracted from the collected human PBMC and mouse PECs by TRIzol lysis (Invitrogen) and used Reverse transcription kit (Roche) to acquire cDNA according to the manufacturer's protocol. Then the outcomes were used to analyze the expression level of the target gene arginine1. The sequences were used for the amplification of cDNA derived from mRNA transcripts of the arginine1 gene as shown below: (Forward: 5′ - TTG GCA ATT GGA AGC ATC TCT GGC-3′ ; Reverse: 5′ - TCC ACT TGT GGT TGT CAG TGG AGT-3′ ), and the primer sequences for β-actin gene (Forward: 5′ -AGA AAA TCT GGC ACC ACA CC-3′ ; Reverse: 5′ -AGA GGC GTA CAG GGA TAG CA-3′ ). The sequences were used for the amplification of cDNA derived from mouse PECs mRNA transcripts as shown below: mouse β-actin gene (Forward: 5′ -ACC TTC TAC AAT GAG CTG CG-3′ ; Reverse: 5′ -CTG GAT GGC TAC GTA CAT GG-3 ′ ), mouse IL-1β gene (Forward: 5′ -ACG GAC CCC AAA AGA TGA AG-3′ ; Reverse: 5′ -TTC TCC ACA GCC ACA ATG AG-3′ ), mouse IL-6 gene (Forward: 5′ -AAA CCG CTA TGA AGT TCC TCT C-3′ ; Reverse: 5′ -GTG GTA TCC TCT GTG AAG TCT C-3 ′ ), mouse nlrp3 gene (Forward: 5′ -ACC TTT GCC CAT ACC TTC AG-3′ ; Reverse: 5′ -TGC CAC AAA CCT TCC ATC TAG-3 ′ ), mouse caspase-1 gene (Forward: 5′ -TCT GTA TTC ACG CCC TGT TG-3′ ; Reverse: 5′ -GAT AAA TTG CTT CCT CTT TGC CC-3′ ). The obtained data were analyzed on the ABI7500. Relative expression levels for cytokines were normalized by βactin and calculated by using the 2−11Ct method.

## Statistical Analysis

Data are presented as mean ± SEM. Group comparisons were performed using Student's t-test by GraphPad Prism software; p-values (two-tailed) below 0.05 were considered as significant. The Mann-Whitney U-test and Spearman's correlation analysis were used to calculate the clinical sample significance. Statistical significance was accepted for p < 0.05.

## RESULTS

#### Positive Correlation of Increased Serum IL-33 With Disease Activity Index CRP in Gout Patients

Our previous study has shown that the serum IL-33 level was predominantly increased in gout patients when compared to healthy controls, and the increased IL-33 expression might play a protective role in kidney injury by regulating lipid metabolism in gout (25). In this study, we recruited more participants to compare levels of IL-33 in gout patients and healthy volunteers. Consistent with our previous study, an increased expression of IL-33 was observed in the sera of gout patients compared with healthy control (data not shown). It has been reported that IL-33 was expressed in synovial fibroblasts from patients with rheumatoid arthritis (RA), and expression was markedly elevated in vitro by TNFα and IL-lβ stimulation (13, 26, 27). Deposition of MSU in the articular cavity can stimulate resident tissue macrophages to produce inflammatory factors TNFα and IL-lβ. Therefore, synovial fibroblasts from gout patients with gouty arthritis were separated and treated with MSU or TNFα/ILlβ. Consistently, TNFα and IL-lβ also induced the up-regulation of IL-33 expression in the synovial fibroblasts from gout patients. In addition, we also found that MSU can directly induce the expression of IL-33 in synovial fibroblasts (**Figure 1A**). CRP was an acute time-phase reaction protein and the most common inflammatory marker for disease activity index in acute gout. Although a protective role of IL-33 in the kidney injury of gout was observed, we here found a positive correlation between the increased IL-33 expression and inflammatory indicator CRP (r = 0.38, p = 0.005; **Figure 1B**). Our data suggested that IL-33 might modulate MSU-induced inflammation.

## IL-33 Reduces the Development of Experimental Gout in Mice

Next, we sought to determine the role of increased expression of IL-33 in gout by using MSU-induced peritonitis experimental model. The exogenous IL-33 was given intraperitoneally daily before MSU treatment for 4 continuous days. The infiltrated leukocytes in the peritoneal cavity were harvested to analyze after MSU administration. Because neutrophils are the important effector cells in MSU-induced inflammation, the peritoneal exudate cells were subjected to analyze the neutrophils by flow cytometry. The CD11b+Gr-1highF4/80<sup>−</sup> cells were considered as neutrophils (**Figure 2A**). The percentage of neutrophils in mice treated with PBS was very low, and exogenous IL-33 treatment slightly elevated the percentage of neutrophils. As expected, the percentage of neutrophils was significantly increased after MSU treatment. However, the percentage of neutrophils induced by MSU administration was significantly decreased in the mice with IL-33 treatment (**Figure 2B**). In addition, we also analyzed the absolute number of neutrophils in these mice. In keeping with the percentage, the absolute number of neutrophils in the MSU-treated mice was also significantly decreased in the mice with IL-33 administration (**Figure 2C**). Collectively, these results indicated that IL-33 can prevent the recruitment of neutrophils in MSU-induced acute inflammation.

## IL-33 Shapes the Cytokines Profiles in MSU-Induced Inflammation

The above data showed that IL-33 played a protective role in the development of gout, we here further explored whether exogenous IL-33-induced amelioration of MSU-induced inflammation correlated with reduction of IL-1β, which is the critical cytokine in the development of MSU-induced inflammation. The levels of IL-1β in the peritoneal cavity lavage were analyzed in our study. As expected, the expression of IL-1β was significantly decreased in the gout model mice with IL-33 treatment (**Figure 3A**). Besides, the inflammatory cytokine IL-6 was also inhibited by IL-33 treatment in the gout animal model (**Figure 3B**). IL-10 is an anti-inflammatory cytokine and plays a role in regulating inflammatory response. Here, we found that IL-33 administration evidently upregulated the anti-inflammatory cytokine IL-10 production when compared with the PBS group (**Figure 3C**). In consistent with previous studies, we also observed an increased level of IL-5 and IL-13 in the peritoneal cavity after exogenous IL-33 treatment (**Figures 3D,E**). Taken together, IL-33 could upregulate the expression of anti-inflammatory cytokines and inhibit the pro-inflammatory cytokines production in MSUinduced inflammation.

## Blockage of Endogenous IL-33 Signaling Has No Effect on MSU-Induced Inflammation

Exogenous recombinant IL-33 treatment significantly inhibited the development of MSU-induced inflammation. To investigate the functional effects of endogenous increased IL-33 production in MSU-induced inflammation, the endogenous IL-33 signaling was blocked by the administration of anti-ST2 antibodies. However, mice treated with anti-ST2 antibody were unaffected by the neutrophils infiltration and MDSCs expansion in the MSU-induced inflammation. Furthermore, the expressions of IL-1β, IL-6, NLRP3, Caspase-1 were detected by RT-PCR, while no significant difference was observed between control group and anti-ST2 group (**Figure 4**). Therefore, we speculate that an increased expression of IL-33 in the gout patients might be due to a cause of self-negative regulation, while the increased amount of IL-33 expression was inadequate to induce a potent protective effect to reduce the development of gout.

## IL-33 Inhibits MSU-Induced Inflammation Through Expanding MDSCs

MDSCs are a group of heterogeneous cells, which consist of granular cell-like MDSCs and mononuclear cell-like MDSCs. By flow cytometry analysis of the collective peritoneal cells, we observed a large number of MDSCs in the PECs of the mice with IL-33 treatment. Specially, the increased number of MDSCs induced by IL-33 in the animal gout model was characterized by surface markers CD11b+Gr-1intF4/80<sup>+</sup> (**Figures 5A–C**). Arginine1 was also recognized as an important

marker for MDSCs, we detected the expression of arginine1 in the gout and healthy control PBMCs at the mRNA level by RT-PCR. We observed an increased expression of arginine1 in the gout patients when compared with healthy controls (**Figure 5D**). These results demonstrated that IL-33 can induce MDSCs recruitment, which might be involved in the selflimiting of the MSU-induced inflammation in the gout patients. Furthermore, three groups of cells induced by IL-33 were isolated and adoptively transferred to the gout model mice (**Figure 6A**). The expression levels of cytokines IL-1β in the peritoneal cavity were analyzed. Compared to other groups, the expression of IL-1β cytokine in the MSU-inflammation was decreased by CD11b+Gr-1intF4/80<sup>+</sup> cells transfusion (**Figure 6B**). Thus, the CD11b+Gr-1intF4/80<sup>+</sup> expansion induced by IL-33 exerts an important role in the resolution of MSU-induced inflammation.

#### DISCUSSION

In this study, we found an increased level of IL-33 in gout patients, which was positively correlated with inflammatory marker CRP. However, exogenous recombinant IL-33 significantly ameliorated the development of animal gout model. Interestingly, a large number of CD11b+Gr1intF4/80<sup>+</sup>

MDSCs was detected in the IL-33-treated mice, and adoptive transfer of IL-33-induced MDSCs markedly inhibited the IL-1β production in MSU-induced peritonitis. This is the first to document the role of IL-33 in a mouse model of human gout. Current results provide evidences for a novel mechanism by which IL-33 ameliorates the MSU-induced inflammation via expanding CD11b+Gr1intF4/80<sup>+</sup> MDSCs.

IL-33 is a cytokine that is widely expressed in a variety of tissue cells, especially fibroblasts (8). Its production can be upregulated by inflammatory cytokines, such as such as IL-1β and TNF-α. Unlike conventional cytokines, IL-33 might be also secreted via unconventional pathways, and can be released upon cell injury as an alarmin. Here, we also observed an increased production of IL-33 in fibroblasts from gout patients, which were stimulated by inflammatory cytokines or MSU. Recent studies have shown that IL-33 played a complex effect in many diseases by interacting with its specific receptor ST2 (7). It has been reported that IL-33 plays a deleterious role in Th2-type immune mediated asthma and Th17-mediated autoimmune arthritis (11, 13). Conversely, a protective role of IL-33 was observed in atherosclerosis, sepsis, allograft transplant and parasite infection (9). Indeed, the IL-33/ST2 signaling also played a dichotomous role in inflammatory bowel disease pathogenesis (28). Our previous studies showed that IL-33 was markedly increased in

the mice with TNBS-induced colitis, while rIL-33 treatment had a significant beneficial effect on Th1/Th17-mediated experimental colitis through promoting alternatively activated macrophage polarization and Tregs differentiation (24, 29). Interestingly, in our present study, although the level of IL-33 expression was increased in the gout patients and positively correlated with inflammatory marker CRP, the exogenous IL-33 treatment significantly inhibited the development of inflammation induced by MSU administration. In addition, blockage of IL-33 signaling by anti-ST2 antibody had no effect on the development of MSUinduced peritonitis. An interesting character of acute gout is the self-limiting nature of the inflammatory flare. In the absence of clinical intervention, a gouty episode can spontaneously resolve within 7–10 days. However, the mechanism of inflammation selfresolution is still elusive. Here, we speculate that an increased expression of IL-33 in the gout patients might due to a cause of self-negative regulation, which inhibits the development of MSU-induced inflammation, while the increased amount of IL-33 expression was inadequate to induce a potent protective effect to reduce the development of gout.

Acute gouty attacks are by nature self-limiting, and multiple mechanisms have been proposed for the spontaneous resolution

mice were sacrificed after 16 h of MSU treatment, and the cells from the peritoneal cavity were harvested (A). The peritoneal cells were stained with anti-CD11b, anti-Gr-1, and anti-F4/80 antibodies. CD11b+Gr-1+F4/80<sup>−</sup> cells were considered to be neutrophils, while CD11b+Gr-1+F4/80<sup>+</sup> cells represent MDSCs (B). Data are presented as the percentage of CD11b+Gr-1+F4/80<sup>+</sup> cells in CD11b+Gr-1<sup>+</sup> cells (C). The absolute number of CD11b+Gr-1+F4/80<sup>+</sup> cell was also quantified. The results shown are from one of three independent experiments. \*\*\**p* < 0.0001 (D). The PBMCs collected from gout patients and healthy control (HC) were subjected to analyze the Arginase-1 expression by RT-PCR method. The Mann-Whitney *U* test was used to calculate the significance. \*\**p* < 0.001.

of acute gout. Previous studies have shown that activated neutrophil impedes the interaction between immunoglobulin G (IgG) and MSU, leading to decrease the inflammation response. Besides that, ApoB and ApoE also exerts a key role in the resolution of acute gout through coating MSU crystal (30). Recently, the formation of neutrophil extracellular traps Shang et al. IL-33 Inhibits MSU-Induced Inflammation

(NETs) has gained increased attention in self-limiting gout inflammation. NETs can neutralize bacteria as well as other danger signals such as MSU crystals by the rapid extrusion of DNA, and can cleave inflammatory cytokines within minutes (5). Numerous studies have showed that MDSCs play an important role in regulating immune responses through producing arginase I, IL-10, and TGF-β (16). In allogeneic transplantation and tumor environments, IL-33 could regulate the induction and migration of MDSCs to inhibit transplant rejection and promotes the progression of mouse breast cancer, the mechanism of which is related to its inhibition of T cell responses (31). Here, we observed a large number of MDSCs (CD11b+Gr1intF4/80+) in MSU-treated mice, which was different from neutrophil (CD11b+Gr1highF4/80−) phenotype. Furthermore, adoptive transfer of IL-33-induced MDSCs (CD11b+Gr1intF4/80+) markedly inhibited IL-1β production in MSU-induced peritonitis. Our study here presented a new clue for the self-limiting acute gout, and the increased IL-33 expression in gout patients might be a negative feedback mechanism on MSU-induced inflammation. Consistently, a recent study reported that LPS induces mononuclear cell-like MDSCs which can prevent tissue damage in acute lung injury through clearing apoptotic neutrophils (23). The clearance of apoptotic neutrophils by MDSCs was dependent on IL-10. In MSU crystal-induced inflammation, the IL-10 expression was increased after IL-33 treatment. We speculate that IL-33-induced MDSCs might alleviate the development of gout inflammation through phagocytosing neutrophils.

Our data demonstrated that increased IL-33 expression in gout patients might be a negative feedback mechanism on the MSU-induced inflammation response, because the inflammatory

#### REFERENCES


cytokine IL-1β production and the number of neutrophils were both markedly decreased in IL-33-treated mice. Furthermore, we reveal a requirement for MDSCs (CD11b+Gr1intF4/80+) in resolution of MSU-induced inflammation. In conclusion, our data provide clear evidences that the increased expression of IL-33 in gout patients was involved in the self-resolution of MSU-induced inflammation, and IL-33 might be a promising therapeutic target for gout.

#### AUTHOR CONTRIBUTIONS

KS, YYW, and LD analyzed data and wrote the manuscript. YYW, KS, QS, and BY performed the experiments. KS and YT contributed to the clinical patients' sample collection. YH provided technical support. KS, YLW, and GS provided critical revision of the manuscript. LD and GS designed the research and revised the article. All authors have read and approved the final manuscript.

#### FUNDING

This research was supported by grants from the National Natural Science Foundation of China (NSFC 81671544 and 81871286 to LD, 81701556 to YH, 81471534 to GS), Fujian Province health planning of young outstanding talents training project no.2016- ZQN-82 to Lihua Duan, Natural Science Foundation of Fujian Provincial Department of Science and Technology 2017J01356 to LD.

### ACKNOWLEDGMENTS

We thank all the volunteers who participated in this study.


**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 Shang, Wei, Su, Yu, Tao, He, Wang, Shi and Duan. 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.

# IgG Anti-ghrelin Immune Complexes Are Increased in Rheumatoid Arthritis Patients Under Biologic Therapy and Are Related to Clinical and Metabolic Markers

Mildren Porchas-Quijada<sup>1</sup> , Zyanya Reyes-Castillo<sup>1</sup> \*, José Francisco Muñoz-Valle<sup>2</sup> , Sergio Durán-Barragán3,4, Virginia Aguilera-Cervantes <sup>1</sup> , Antonio López-Espinoza<sup>1</sup> , Mónica Vázquez-Del Mercado<sup>4</sup> , Mónica Navarro-Meza<sup>1</sup> and Patricia López-Uriarte<sup>1</sup>

#### Edited by:

Wen Kong, Wuhan Union Hospital, China

#### Reviewed by:

Giulia Ricci, Università degli Studi della Campania Luigi Vanvitelli Caserta, Italy Xue Xu, The Affiliated Drum Tower Hospital of Nanjing University Medical School, China

\*Correspondence:

Zyanya Reyes-Castillo zyanya.reyes@cusur.udg.mx

#### Specialty section:

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

Received: 20 August 2018 Accepted: 01 April 2019 Published: 18 April 2019

#### Citation:

Porchas-Quijada M, Reyes-Castillo Z, Muñoz-Valle JF, Durán-Barragán S, Aguilera-Cervantes V, López-Espinoza A, Vázquez-Del Mercado M, Navarro-Meza M and López-Uriarte P (2019) IgG Anti-ghrelin Immune Complexes Are Increased in Rheumatoid Arthritis Patients Under Biologic Therapy and Are Related to Clinical and Metabolic Markers. Front. Endocrinol. 10:252. doi: 10.3389/fendo.2019.00252 1 Instituto de Investigaciones en Comportamiento Alimentario y Nutrición, Centro Universitario del Sur, Universidad de Guadalajara, Ciudad Guzmán, Mexico, <sup>2</sup> Instituto de Investigaciones en Ciencias Biomédicas, Centro Universitario de Ciencias de la Salud, Universidad de Guadalajara, Guadalajara, Mexico, <sup>3</sup> Departamento de Reumatología, Clínica de Investigación en Reumatología y Obesidad, Guadalajara, Mexico, <sup>4</sup> Instituto de Investigación en Reumatología y del Sistema Músculo Esquelético, Centro Universitario de Ciencias de la Salud, Universidad de Guadalajara, Guadalajara, Mexico

Rheumatoid arthritis (RA) is a systemic autoimmune disease associated with increased risk of cardiovascular disease and metabolic alterations. The mechanisms underlying these alterations remain unclear. Ghrelin is a gastrointestinal hormone with potent effects on food intake, body weight, metabolism, and immune response. Recent studies reported the presence of anti-ghrelin autoantibodies in healthy subjects and the levels and affinity of these autoantibodies were altered in anorectic and obese individuals. In this cross-sectional study we analyzed anti-ghrelin autoantibodies in RA patients and evaluated its relationship with clinical, body-composition and metabolic parameters. Clinical measurements of RA patients included the disease activity score-28 (DAS-28), inflammatory biomarkers, autoantibodies (RF and anti-CCP), body composition, glucose and lipid profile. Serum ghrelin levels were measured by enzyme-linked immunosorbent assay (ELISA). Free and total anti-ghrelin autoantibodies quantification (IgG and IgA isotypes) was performed by in-house ELISA. RA patients had lower IgG anti-ghrelin autoantibodies levels and higher immune complexes percentage (IgG+ghrelin) compared to the control group, while the IgA anti-ghrelin autoantibodies showed no significant differences. In the bivariate analysis, the percentage of IgG anti-ghrelin immune complexes positively correlated with BMI and ghrelin whereas in the multivariate regression model, the variables associated were DAS-28, body weight, visceral fat, LDL-C and TG (R <sup>2</sup> = 0.72). The percentage of IgA anti-ghrelin immune complexes positively correlated with RF and anti-CCP and the multivariate regression model showed an association with RF and body fat percentage (R <sup>2</sup> = 0.22). Our study shows an increased percentage of IgG anti-ghrelin immune complexes in RA patients despite ghrelin levels were similar in both groups, suggesting an increase in the affinity of these autoantibodies toward ghrelin. The associations found in the multiple regression analysis

**56**

for anti-ghrelin immune complexes support the previously reported functions of these natural autoantibodies as carriers and modulators of the stability and physiological effect of the hormone. However, in RA both the disease activity and the RF appear to influence the formation of these anti-ghrelin immune complexes.

Keywords: ghrelin, rheumatoid arthritis, metabolic alterations, clinical activity, body composition

#### INTRODUCTION

Rheumatoid arthritis (RA) is a systemic inflammatory autoimmune disease, associated with high incidence of cardiovascular disease (CVD) and metabolic alterations, reporting an increase of 50% in deaths related to CVD in RA patients compared to the general population (1). This elevated risk is not only related to the classical CV risks like hypertension and dyslipidemias, but also to the chronic systemic inflammation occurring in the patients (2, 3). In addition, the inflammatory status in RA is associated with modifications of body composition, specifically an increase of body fat mass and depletion of lean mass, known as rheumatoid cachexia that can be present simultaneously with obesity (4, 5). Both, cachexia and obesity have been associated with higher disability scores and elevated risk of progressive disability (6–9). It has been recognized that the treatment of RA patients with disease-modifying anti-rheumatic drugs (DMARDs) such as methotrexate (MTX) and biological drugs including TNF-α and IL-6 blockers, as well as co-stimulation inhibitors such abatacept exert beneficial effects on the metabolic profile (10). Nevertheless, the precise mechanisms underlying the body-composition and metabolic alterations in RA as well as its relationship with the pharmacological therapy remain poorly understood.

Research on appetite and metabolism regulating hormones has increased substantially in rheumatic autoimmune diseases such as RA during the last years. Accumulating data strongly indicate that these hormones also exhibit potent actions on the regulation of immune and inflammatory responses and may play a role both in the pathogenesis and development of comorbidities in RA (11–13). Ghrelin is a 28 amino acid gastrointestinal peptide with potent orexigenic effects as well as anti-inflammatory properties including the inhibition of cytokines such as TNF-α, IL-1β, and IL-6 produced by T lymphocytes and monocytes (14–17). It has been shown that ghrelin increases the expression of adhesion molecules and exerts anti-proliferative effects on microvascular endothelial cells (18, 19). Also, ghrelin has demonstrated regulatory effects on bone metabolism, as it promotes osteoblast differentiation and proliferation and inhibits apoptosis (20, 21). Nevertheless, the concentrations and clinical relevance of ghrelin in RA are still controversial considering that some investigations reported lower (15), higher (22) or even similar (23, 24) levels of this hormone in patients under DMARDs and/or biological therapy in comparison to healthy controls.

Recent studies described the presence of autoantibodies directed against ghrelin in healthy subjects, and altered levels and affinity of these natural autoantibodies were reported in anorectic and obese individuals sera; suggesting that anti-ghrelin autoantibodies, specifically immunoglobulins of the G isotype (IgG) may affect the hormone transport and function according to its affinity, as well as regulate its stability by protecting the hormone from degradation (25–28). Co-administration of ghrelin and ghrelin-reactive IgG extracted from plasma of ob/ob mice to male lean C57B16 mice, increased their daily food intake and induced a tendency to increase their body weight (28). This may indicate that the presence of these natural autoantibodies directed against ghrelin and other appetite-regulating peptide hormones are associated with metabolism and body weight alterations (29). However, the presence and the possible role of these autoantibodies in autoimmune diseases have not been addressed, especially in RA where the occurring metabolic and body-composition alterations are prominent and the underlying mechanisms remain barely explored. Therefore, in this study we analyzed serum samples of RA patients and controls to characterize the circulating anti-ghrelin autoantibodies of IgG and IgA isotypes and evaluate its relationship with metabolic profile, body-composition and clinical parameters in RA patients undergoing biological therapy.

#### MATERIALS AND METHODS

#### Subjects

A cross-sectional study of RA patients and control subjects was performed. RA patients were previously classified according to the American College Rheumatism (ACR)/European League Against Rheumatism (EULAR) 2010 criteria (30). Those who had known history of other autoimmune, diabetes mellitus, renal, hepatic or CV disease, as well as those under 18 years of age or pregnant were excluded. Control subjects reported not having any diagnosed autoimmune, cardiovascular, hepatic, renal, infectious, or thyroid disease as well as not being under lipid-lowering medication. Patients were recruited from the Obesity and Rheumatology Research Center, located at Guadalajara, Jalisco, Mexico.

The study was approved by the Ethics Research Committee of University of Guadalajara (CEICUC, Review Board registry number CONBIOETICA14CEI03420150130) and was conducted according to the principles of the declaration of Helsinki. All participants were adults and voluntarily signed an informed consent before their inclusion in the study.

## Methods

#### Clinical Data

All patients and controls answered a demographic and clinical interview at the moment of the blood sample collection. Patients were evaluated by a rheumatologist that performed a general physical examination and assessed the number of painful and swollen joints for determining the RA clinical activity index by the disease activity score 28 (DAS-28) (31). The functional disability was assessed through the Health Assessment Questionnaire-Disability Index (HAQ-DI, Spanish version) (32). The erythrocyte sedimentation rate (ESR) was assessed through the Wintrobe method. C-reactive protein (CRP) and rheumatoid factor (RF) were measured by turbidimetric assays (Cat. No. COD31029 and COD31030A25, respectively, A25 Biosystems, Barcelona, Spain) using automatized equipment (BS-10; Mindray, Shenzhen, China). The cutoff value for RF positivity was 30 IU/mL, with a 95% specificity. Anti-cyclic citrullinated peptide antibodies (anti-CCP) were assessed using an ELISA kit (Cat. No. COD-FCCP600; Axis-Shield Diagnostics, Dundee, UK) following the instructions provided by the manufacturer; 5 U/mL were set as the cut-off point for positivity, with a 100 % specificity.

#### Body Composition and Anthropometry

All body-composition and anthropometric measurements were made by certified nutritionists. Body weight, musculoskeletal mass, body fat percentage (BF%) and visceral fat were measured by a bioelectrical impedance equipment (HBF-514C Omron Healthcare, Inc. Lakeside Drive Bannockburn, Illinois, USA) following the manufacturer instructions.

The participant's height was measured using a 2.05 m ± 1 mm scale portable stadiometer (Holtain Limited, Crynich, Difed, Britain Ltd. UK) following the technique described by Jellife and Jellife (33). The body mass index (BMI) was calculated by dividing weight in kg by the square of height in m. The waist circumference (WC) was measured midway between the lower rib and the iliac crest, at the end of a normal expiration to the nearest 0.1 cm.

#### Glucose and Lipid Profile

The blood samples were taken after an 8–12 h overnight fasting and centrifuged at 3,500 rpm during 15 min for serum separation. Samples were aliquoted and stored at −20◦C until the day of the assay. Fasting glucose (Cat. No .1001190), total cholesterol (TC, Cat. No. 41022), low-density lipoprotein cholesterol (LDL-C, Cat. No. BSIS51-E), high-density lipoprotein cholesterol (HDL-C, Cat. No. BSIS37-E) and triglycerides (TG, Cat. No.1001313) were determined by colorimetric enzymatic methods using commercial kits (all reagents by Spinreact, Girona, Spain). Dyslipidemias and impaired fasting glucose were defined according to the Adult Treatment Panel III (ATPIII) guidelines (34) as TC ≥240 mg/dL, LDL-C ≥160 mg/dL, TG ≥200 mg/dL, HDL-C <40 mg/dL, and glucose ≥110 mg/dL.

#### Quantification of Ghrelin and Anti-ghrelin Autoantibodies

The fasting total ghrelin concentrations were assessed using an ELISA kit (Cat. No. EZGRT-89K, Upstate Chemicon Linco, Millipore), according to the manufacturer's instructions. A subsample of the RA patients (n = 25) was used to assess the serum ghrelin levels since not all were fasting.

To measure the autoantibodies against ghrelin of both IgA and IgG isotypes, an in-house ELISA test was performed based on a published protocol (35). This test allows the quantification of free (autoantibodies unbound to ghrelin) and total (autoantibodies forming immune complexes with ghrelin plus the free form) using two types of sample dilution buffers that create normal (pH 7.4) or dissociative conditions (pH 8.9), respectively, giving information about relative autoantibodies levels and affinities. Slight modifications were made to the method after thorough standardization; including the RA patients and controls serum dilution (1:1,000), the time of serum incubation (2 h) on ELISA plate, and the detection antibody dilution (1:8,000 for IgG and 1:15,000 for IgA). Both detection antibodies were conjugated to horseradish peroxidase (Cat. No. MBS674609 and Cat. No. MBS176676 for anti-human IgG and IgA, both from MyBioSource, California, USA) and incubated for 2 h. Subsequently, the plates were washed 4 times and 100 µl of the substrate tetramethylbenzidine (TMB, Sigma Aldrich) were added to the plate and incubated for 15 min for color development. Finally, 50 µl of stop solution was added and the optical density (OD) was read at 450 nm on a plate spectrophotometer (Bio-Rad, California, USA). For each ELISA plate, 2 wells were set as blank and the mean OD values were subtracted from the mean OD values of samples. Blank OD values in ELISA tests were all below 0.1, indicating that there was no unspecific binding of the detection antibody. Patient and control samples were run in duplicate, obtaining values with a mean variation <5% between duplicates for both IgA and IgG. IgG and IgA anti-ghrelin autoantibodies serum levels were expressed as OD, whereas IgG- and IgAghrelin immune complexes percentage were calculated using a ratio between free and total immunoglobulins by the following formula:

$$\text{Immune complexes percentage} = 100 - \left(\frac{\text{Free OD}}{\text{Total OD}}\right) \times 100$$

#### Statistical Analyses

The data distribution was verified by D'Agostino-Pearson normality test and were reported as mean ± standard deviation (s.d.) for parametric data, and median (25–75th centiles) for variables with non-parametric distribution. Categorical variables were expressed as percentage and absolute frequency. Differences between two groups were assessed using Student's t-test or Mann-Whitney U-test for independent samples, according to data normality. To evaluate the relationship between clinical and metabolic variables with the anti-ghrelin autoantibodies a correlation analysis was carried out using Pearson's or Spearman's correlation tests in accordance with the data normality. Multivariate linear regression analysis was performed to analyze the association of clinical, biochemical and bodycomposition variables with the IgG and IgA anti-ghrelin immune complexes. Analyses were carried out using GraphPad Prism 6.0 (GraphPad Software, USA) and NCSS 2007 software (Number Cruncher Statistical System for Window, USA). The significance level was set at p ≤ 0.05.

## RESULTS

#### Patients Characteristics

Forty-nine RA patients and 32 control subjects were included in the study. The demographic, body composition, biochemical and clinical characteristics of both groups are shown in **Table 1**. The RA patients had a mean age of 50 ± 15 years, of which 87.7% were females. Controls had a mean age of 43 ± 8 years, of which 93.7% were females. The 30.6% of the RA patients reported past, and 8.2% current smoking habits while 9.4% of the control group reported past and 9.4% current smoking habits. Mean disease duration among patients was 8.5 ± 8.46 years, the ESR levels were higher in patients than in controls. The CRP levels in patients were found within normal values, whereas positivity for RF was 75 and 95% for anti-CCP. According to the DAS-28 score and the HAQ-DI questionnaire, on average, patients had moderate activity (3.43 ± 1.23) and low disability (0.54 ± 0.51). Most patients (91.8 %) were treated with politherapy, incorporating biological DMARDs (DMARDb).

The body-composition parameters were similar between RA patients and controls, except the visceral fat level which was higher (p < 0.01) among controls. On average, both groups were overweight (>25 kg/m<sup>2</sup> ) (36) according to the BMI (26.69 ± 4.05 kg/m<sup>2</sup> for RA patients and 27.92 ± 5.26 kg/m<sup>2</sup> for controls). The WC mean value was 89 cm in both groups, indicating abdominal obesity according to the ATP III WC cutoff point for women (>88 cm) (37). Both groups showed an excess of body fat (>30%) (38). Alike, the lipid profile and glucose levels were similar between RA patients and controls being within the normality ranges, excluding the LDL-C median levels which were higher among controls (141.50 mg/dL) than patients (113.50 mg/dL). The median total ghrelin levels were 636.20 pg/mL for RA patients and 642 pg/mL for controls, showing no significant differences.

#### Anti-ghrelin Autoantibodies Analysis

Anti-ghrelin autoantibodies levels of both IgG and IgA isotypes are shown in **Figure 1**. Both free and total IgG anti-ghrelin autoantibodies were significantly higher in controls than in RA patients (p < 0.05) and on the contrary, the IgG immune complexes percentage was lower in controls than in RA patients (p < 0.05). While the IgA free and total antighrelin autoantibodies levels, as well as the immune complexes percentage, showed no significant differences between controls and RA patients.

#### Correlations Between Anti-ghrelin Autoantibodies With Clinical Parameters and Metabolic Profile in RA

We assessed the correlation between the clinical parameters in RA with the free and total fractions of IgG and IgA anti-ghrelin autoantibodies. Anti-CCP antibodies were positively correlated with total IgA anti-ghrelin autoantibodies (r = 0.326, p = 0.022) and the DAS-28 activity score showed a positive correlation with free IgA anti-ghrelin autoantibodies (r = 0.296, p = 0.050).When addressing the relationship between metabolic profile in RA with free and total fractions of anti-ghrelin, we detected a



Data are shown as mean ± s.d. or median (25–75th centile). TC, total cholesterol; HDL-C, high density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; TG, triglycerides; BMI, body mass index; WC, waist circumference; BF, body fat; ESR, erythrocyte sedimentation rate; RF, rheumatoid factor; Anti-CCP, anti-cyclic citrullinated peptide antibodies; CRP, C-reactive protein; DAS-28, disease activity score 28; HAQ-DI, health assessment questionnaire-disability index (Spanish version); MTX, methotrexate; DMARDb, biological disease-modifying anti-rheumatic drugs. Difference between groups was assessed by Student's t-test or Mann-Whitney U-test, p ≤ 0.05 values are highlighted in bold. <sup>a</sup> Including TNF, CD20, and IL-6 receptor inhibitors.

negative correlation between visceral fat level and total IgG anti-ghrelin autoantibodies (r = −0.519, p = 0.000). LDL-C levels were found positively correlated with free IgA anti-ghrelin autoantibodies (r = 0.404, p = 0.040). Ghrelin levels showed a negative correlation with free IgG anti-ghrelin autoantibodies (r = −0.534, p = 0.006).

The percentage of ghrelin-immune complexes also displayed significant correlations with both the clinical and metabolic parameters in RA (**Figure 2**); the RF and anti-CCP antibodies

showed positive correlations with the IgA anti-ghrelin immune complexes percentage (r = 0.300, p = 0.042 and r = 0.372, p = 0.008, respectively) while the BMI was negatively correlated with IgG anti-ghrelin immune complexes percentage (r = −0.307, p = 0.035). Ghrelin levels were found positively correlated with the IgG anti-ghrelin immune complexes percentage (r = 0.432, p = 0.030).

## Multiple Regression Analyses of the Clinical, Biochemical, and Body-Composition Variables Associated With Anti-ghrelin Immune Complexes in RA

The IgG and IgA anti-ghrelin immune complexes percentage were further analyzed in a multiple regression model. We focused on the analysis of these complexes as they were significantly increased in RA patients and are likely implicated in the transport and modulation of biological effects of ghrelin (28). Thus, once we analyzed the simple correlations between the clinical, biochemical and body-composition parameters with both IgG and IgA anti-ghrelin autoantibodies levels, we selected the variables showing a significance of p < 0.20 as well as those previously reported to affect ghrelin concentrations.

The IgG anti-ghrelin immune complexes percentage was found associated with the DAS-28 score, weight, visceral fat level, LDL-C and TG concentrations explaining in a 72% its variance. While the IgA anti-ghrelin immune complexes were associated with RF and the BF percentage which explained a 22% of its variance (**Table 2**).

TABLE 2 | Clinical, body composition, and biochemical parameters associated with IgG and IgA anti-ghrelin immune complexes in rheumatoid arthritis patients.


DAS-28, disease activity score 28; RF, rheumatoid factor; LDL-C, low density lipoprotein cholesterol; TG, triglycerides. Multivariate analysis results from multiple linear regression analysis. The total explained variance of the model was <sup>a</sup>R <sup>2</sup> = 0.72, <sup>b</sup>R <sup>2</sup> = 0.22.

#### DISCUSSION

RA is associated with metabolic alterations, changes in body composition as well as increased risk of developing CV diseases (1, 2). Ghrelin, the main orexigenic hormone, is a gastrointestinal peptide with regulatory effects on metabolism and antiinflammatory properties (15, 16). Recent studies reported the presence of natural autoantibodies directed against ghrelin in healthy individuals and altered levels and affinity of these autoantibodies in appetite-related pathologies including obesity and anorexia nervosa, suggesting a physiological role of these autoantibodies in ghrelin regulation (25, 27). In the present study, we evaluated for the first time the presence of anti-ghrelin autoantibodies of both IgG and IgA isotypes in RA patients under biological therapy and analyzed its relationship with body composition, metabolic profile and clinical activity parameters.

Analysis of ghrelin serum levels showed no significant differences in our cohort of RA patients in comparison to controls. Similarly to our findings, no significant differences in ghrelin levels were found among patients with established RA receiving traditional DMARDs treatment (17, 18) neither in patients under anti-TNF-α therapy (39) when compared to healthy controls. Conversely, a decrease in acyl-ghrelin levels in RA patients under DMARDs and/or biologics was reported in comparison to controls (15, 22), whereas higher levels of total ghrelin were observed after treatment with a TNF-α blocker in comparison to controls (22, 40). These apparently conflicting results may be explained by differences in the form of ghrelin measured in the studies (acyl, des-acyl, or total hormone levels), the patient's clinical characteristics as well as the treatment scheme.

Despite the lack of differences in ghrelin serum levels, we found that free and total IgG anti-ghrelin autoantibodies were decreased in our cohort of RA patients compared to controls. This decrease is probably an effect of the patient's immunosuppressive therapy as all of them were under MTX and/or biological therapy. MTX and abatacept have potent anti-inflammatory properties, and several clinical studies have demonstrated its effects on reducing serum immunoglobulins (41–44). Furthermore, MTX has demonstrated to lower total and transitional B cells as well as total T cell numbers in peripheral blood (41, 42, 45). Similarly, a decrement in both free and total IgG anti-ghrelin autoantibodies levels, as well as a decrease on their binding affinity with ghrelin, were observed in a rat model of methotrexate chemotherapy-induced anorexia (46). To support and extend the data obtained in this cross-sectional study it will be important to conduct longitudinal studies to further determine the effects of different RA treatment schemes on these natural anti-ghrelin autoantibodies.

In contrast to the decrease in free and total IgG antighrelin autoantibodies in RA, we detected a higher percentage of IgG-ghrelin immune complexes in the patients as compared to controls. This is indicative of an increase on the affinity these autoantibodies toward ghrelin, probably as an adaptive mechanism in response to the reduction of the total fraction of anti-ghrelin autoantibodies among treated patients. We speculate that such affinity-mediated augment in the formation of anti-ghrelin immune complexes in RA, favor the stability and transport of ghrelin thereby enhancing its biological effects. Despite in the present study we did not measured the kinetic affinities of these autoantibodies in RA, previous studies performed by Fetissov and co-workers in other pathologies including obesity and anorexia nervosa can support our theory (28, 47). They demonstrated that ghrelin-reactive antibodies of the IgG isotype present a very variable Fab region showing distinctive affinities in the context of different pathologies (28). In obese individuals, increased kinetics affinities of IgG antibodies for the orexigenic hormone ghrelin were found and the transfer of these antibodies to mice promoted food intake and body weight gain. Similarly, IgG antibodies directed to the anorexigenic hormone leptin showed decreased affinity in obese humans (28, 48). It is worth mentioning that although there can be changes in the affinity of these ghrelin-reactive autoantibodies its affinity remains at the micromolar range and these can be classified as low-affinity, not being capable of neutralizing the hormone, but instead modulate its transport and/or protect it from degradation by serum enzymes (28). Even though the precise mechanisms fine-tuning the affinity of these natural autoantibodies remain unknown, these findings provide evidence that IgG anti-ghrelin autoantibodies undergo affinity modification in the context of metabolic pathologies as well as in immune-mediated diseases such as RA.

The IgG anti-ghrelin immune complexes percentage correlated with BMI and ghrelin levels in the bivariate analysis. This positive correlation with total ghrelin levels is reasonable, since it suggests that an increase in serum ghrelin levels cause more availability for the ghrelin-reactive antibodies to bound and form immune complexes. While in the multivariate regression model we found an interaction between DAS-28 score, body weight, visceral fat level, LDL-C, and TG concentrations, which explained a 72% of the variation in the percentage of these complexes. It should be noted that body weight, visceral fat level, LDL-C and TG have been previously correlated with ghrelin levels in several studies. The BMI and visceral fat mass were reported negatively correlated with plasma ghrelin (49, 50), which may indicate that the secretion of this hormone is reduced as a physiological adjustment in response to the positive energetic balance. In addition, it is recognized that ghrelin participates in lipid metabolism as it enhances adiposity by promoting the expression of several fat storage-related proteins in adipocytes (51). In concordance with our multivariate model, LDL-C and TG concentrations were reported negatively correlated with ghrelin in a study in obese children (52). In addition, Beaumont et al. (53) described that HDL-C particles may directly interact with ghrelin serving as a circulatory carrier, however, in our model HDL-C did not show an association with the percentage of ghrelin-immune complexes, suggesting that these antibodies are an independent carrier of the hormone.

The association of anti-ghrelin immune complexes with the DAS-28 in the multivariate analysis was striking, considering the fact that serum ghrelin levels alone were not associated with the disease activity nor with other clinical biomarkers in RA (data not shown); this suggests that anti-ghrelin autoantibodies but not ghrelin, are affected by the disease activity of the patients. Taken together, our results show that IgG-ghrelin immune complexes percentage exhibit the same associations previously reported for ghrelin and strongly support the idea of these autoantibodies as low-affinity carriers modulating the stability and biological effects of the hormone. However, in RA the IgG-ghrelin immune complexes formation appears to be negatively influenced by the disease activity. According to the multiple regression model it can be predicted that an increase on the disease activity as well as a rise in LDL-C and TG concentrations could affect the formation of IgG-ghrelin immune complexes. Based on the observation that patients in our study had an overall controlled RA (showing low to moderate activity according to DAS-28 score and low inflammatory biomarkers) this model may also explain the significant increase observed in the IgG anti-ghrelin immune complexes percentage in the patients, besides the previously discussed affinity-mediated compensatory mechanism that may promote ghrelin immune complexes formation.

While for the IgA anti-ghrelin autoantibodies levels, no differences were found between controls and patients. We hypothesize that this is related to the sample type used to quantify these isotype because IgA is not predominant in serum (in contrast to the IgG class), but is rather predominantly secreted in mucous membranes (54). Furthermore, the presence of IgA anti-ghrelin autoantibodies in serum, may indicate that these autoantibodies are triggered by antigens present in the intestinal lumen either under physiological or pathological conditions (27, 55). Since the gastrointestinal microbiota is a major physiological source of antigens, there has been suggested that bacterial epitopes can trigger cross-reactivity to regulatory peptides such as ghrelin, which displays sequence homology with commensal bacteria and viruses (27).

An interesting finding was the positive correlation observed between IgA anti-ghrelin immune complexes with RF and anti-CCP autoantibodies. The role of rheumatoid factors in the formation of immune complexes has been widely acknowledged in RA; this mechanism can enhance the inflammatory status through the release of inflammatory cytokines such as TNF-α and IL-6 via activation of Fc receptors expressed by innate immune cells such as macrophages. Nevertheless, this observation does not imply that anti-ghrelin complexes are considered pathogenic but rather regulatory, as these are detected in healthy individuals. In addition, autoantibody systems in RA are frequently correlated with each other; as described previously for anti-CCP and RF, and for RF and anti-PAD4 antibodies (56). Similarly, the multivariate regression model showed an association of the IgA anti-ghrelin immune complexes with RF and body fat percentage but only explained a 22 % of its variance. The negative association with body fat percentage again supports the hypothesis that the secretion of this hormone is reduced as a physiological adjustment in response to the positive energetic balance.

In summary, we evaluated for the first time the presence of IgG and IgA anti-ghrelin autoantibodies in RA patients and confirmed the previously reported presence of these natural autoantibodies in healthy individuals (26). Increased percentage of IgG anti-ghrelin immune complexes was found in RA patients under biological therapy with a low to moderate disease activity. This is indicative of an increase in the affinity of these autoantibodies toward ghrelin, probably as an adaptive mechanism in response to the reduction of the total fraction of anti-ghrelin autoantibodies among treated patients. In the multivariate regression analyses, the IgG-ghrelin immune complexes were associated with DAS-28, body weight, visceral fat, LDL-C, and TG while the IgA-ghrelin immune complexes were only associated to RF and body fat percentage, supporting the idea of these anti-ghrelin natural autoantibodies acting as carriers and modulators of the stability and physiological effect of the hormone. However, in RA both the disease activity and the hallmark autoantibody (RF) appear to influence the formation of these anti-ghrelin immune complexes. Future longitudinal studies should address the effects of different treatment schemes on these natural antibodies as well as monitor its changes in relation to clinical activity and metabolic changes in RA. In addition, analysis of autoantibodies directed to acyl-ghrelin or des-acyl ghrelin specific forms would be also informative.

#### REFERENCES


#### ETHICS STATEMENT

The study was approved by the Ethics Research Committee of University of Guadalajara (CEICUC, Review Board registry number CONBIOETICA14CEI03420150130) and was conducted according to the principles of the declaration of Helsinki. All participants were adults and voluntarily signed an informed consent before their inclusion in the study.

#### AUTHOR CONTRIBUTIONS

ZR-C was involved in the conception, design and performance of experiments, interpretation of data, and revision of the manuscript. MP-Q performed experiments, analyzed and interpreted the data, and wrote the article. JFM-V and SD-B aided in the design of the study, interpretation of data, and critical revision of the manuscript. MN-M provided help on data acquisition and interpretation. VA-C, AL-E, MV-D, and PL-U helped with the interpretation of data and critical revision of the paper. All authors read and approved the submitted version of the manuscript.

#### FUNDING

This study was supported by Instituto de Nutrición y Salud Kellogg's (INSK) by the funding program Apoyo a Proyectos de Investigación en Nutrición (APIN-2016) granted to ZR-C. The funding source had no involvement on the design nor any steps of the study.

#### ACKNOWLEDGMENTS

The authors would like to thank Instituto de Investigación en Ciencias Biomédicas and Centro de Investigación en Biología Molecular de las Enfermedades Crónicas, for kindly providing the facilities for performing some of the experiments of the study.


study of patients with rheumatoid arthritis. Arthritis Rheum. (1995) 38:44–8. doi: 10.1002/art.1780380107


with plasma leptin and markers of obesity and diabetes. Nutr Diabet. (2018) 8:32. doi: 10.1038/s41387-018-0044-y


**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 Porchas-Quijada, Reyes-Castillo, Muñoz-Valle, Durán-Barragán, Aguilera-Cervantes, López-Espinoza, Vázquez-Del Mercado, Navarro-Meza and López-Uriarte. 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.

# Thyroid Hormone Action on Innate Immunity

#### María del Mar Montesinos and Claudia Gabriela Pellizas\*

Facultad de Ciencias Químicas, Centro de Investigaciones en Bioquímica Clínica e Inmunología (CIBICI-CONICET) and Departamento de Bioquímica Clínica, Universidad Nacional de Córdoba, Córdoba, Argentina

The interplay between thyroid hormone action and the immune system has been established in physiological and pathological settings. However, their connection is complex and still not completely understood. The thyroid hormones (THs), 3,3′ ,5,5′ tetraiodo-L-thyroxine (T4) and 3,3′ ,5-triiodo-L-thyronine (T3) play essential roles in both the innate and adaptive immune responses. Despite much research having been carried out on this topic, the available data are sometimes difficult to interpret or even contradictory. Innate immune cells act as the first line of defense, mainly involving granulocytes and natural killer cells. In turn, antigen presenting cells, macrophages and dendritic cells capture, process and present antigens (self and foreign) to naïve T lymphocytes in secondary lymphoid tissues for the development of adaptive immunity. Here, we review the cellular and molecular mechanisms involved in T4 and T3 effects on innate immune cells. An overview of the state-of-the-art of TH transport across the target cell membrane, TH metabolism inside these cells, and the genomic and non-genomic mechanisms involved in the action of THs in the different innate immune cell subsets is included. The present knowledge of TH effects as well as the thyroid status on innate immunity helps to understand the complex adaptive responses achieved with profound implications in immunopathology, which include inflammation, cancer and autoimmunity, at the crossroads of the immune and endocrine systems.

#### Edited by: Wen Kong,

Wuhan Union Hospital, China

#### Reviewed by:

Lan Wu, Vanderbilt University Medical Center, United States Taisen Iguchi, National Institute for Basic Biology, Japan

\*Correspondence:

Claudia Gabriela Pellizas claudia@fcq.unc.edu.ar

#### Specialty section:

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

Received: 08 January 2019 Accepted: 15 May 2019 Published: 04 June 2019

#### Citation:

Montesinos MM and Pellizas CG (2019) Thyroid Hormone Action on Innate Immunity. Front. Endocrinol. 10:350. doi: 10.3389/fendo.2019.00350 Keywords: thyroid hormones, innate immunity, neutrophils, natural killer cells, macrophages, dendritic cells

## INTRODUCTION

Growing evidence compiled over recent decades has revealed a bidirectional crosstalk between thyroid hormones (THs) and the immune system. This interplay has been demonstrated for several pathophysiological conditions of the thyroid functioning and the innate and adaptive immunity. Many situations primarily affecting the action of THs have an impact on the characteristics and/or functions of immune cells, and are translated to host defense status and related disorders. In turn, immune-related disorders conduct to the most frequent thyroid dysfunctions, which have an autoimmune origin. The connection between these systems is complex and not well-understood. This article reviews the current evidence supporting the contribution of THs to the modulation of innate immunity at the cellular level.

#### Thyroid Hormone Action

THs exert a pivotal role for normal development and function. The thyroid produces 3,3′ ,5,5′ tetraiodo-L-thyroxine (T4) and 3,3′ ,5-triiodo-L-thyronine (T3), mainly under thyrotropin (TSH) regulation. While this gland secretes 100% of circulating T4, it provides only a low percentage of serum levels of the most physiologically active TH: T3, which for the major part derives from peripheral 5′ deiodination of T4 (1). At the target cell level, the action of THs is genomic (nuclear) and non-genomic. The former requires T3 and the specific nuclear receptors (TRs): TRα1, TRβ1, TRβ2, and TRβ3 (2) and is controlled by a multiprotein complex comprising both corepressors and coactivators (3).

Translocation of TRs from their synthesis in the cytosol to the nucleus is a functionally active process (4). In this regard, non-genomic effects exerted intracellularly by TRs and truncated variants occur rapidly, can be observed in the cytoplasm, mitochondria and other organelles, and are independent of nuclear receptor activity and protein synthesis. Many effects conducted by cytoplasmic TRs involve PI3K-dependent Akt activation (5). Furthermore, non-genomic actions of THs are also initiated at the plasma membrane through different proteins. The best studied is the integrin ανβ3, which binds mainly T4 and tetraiodothyroacetic acid (tetrac), a derivative of T4, inducing activation of AMPK, PI3K/Akt, and MAPK (6, 7). Overall, THs interact with a wide variety of signaling pathways that are not yet fully deciphered.

Circulating levels of THs are not representative of what each cell type detects. Instead, the action of THs requires an appropriate interplay among membrane TH transporters, TH deiodinases and TR expression, and thus there is a fine-tuned cellular TH responsiveness. The main TH transporters include monocarboxylate transporters (MCT) 8 and 10, organic anion transporter polypeptides (OATPs) and large neutral amino acid transporters (LATs), with MCT8, MCT10, and LATs having a higher affinity for T3 than T4 uptake. Additionally, the cellular concentrations of THs are regulated by the activity of the 1, 2, and 3 iodothyronine deiodinases: D1, 2, and 3. D2 is an "activating" enzyme, responsible for the peripheral production of 50–80% of the body pool of T3 from T4. In contrast, D3 restrains T3 action, converting T4 and T3 into inactive metabolites. TH transporters and deiodinases exhibit a particular expression profile that is cellular and metabolic state specific (8, 9). Newly discovered actions of T4 and T3 metabolites, such as 3,5-diiodothyronine (3,5-T2), and 3-iodothyronamine (T1AM) are emerging (10).

#### Innate Immunity

The immune system includes cells that protect the organism from foreign antigens, such as microbes, cancer cells, toxins, and damage signals. It is simplistically referred to as innate and adaptive immunity. The former offers immediate protection against intruders, with specific cells being able to fight a wide range of pathogens, with the latter being specific and antigendependent (11). Moreover, adaptive immunity is orchestrated and directed by its innate counterpart.

The main innate cells are polymorphonuclear leukocytes (PMNL, mainly neutrophils), innate lymphoid cells (ILCs) including natural killer (NK) cells and cytokine-producing helper-like ILCs, innate T-like cells comprising NKT and γδ T cells, monocytes, macrophages and dendritic cells (DCs). Their complete classification and plethora of functions have been extensively reviewed (12–15).

The belief that innate immunity is non-specific was challenged after the description of pattern-recognition receptors and molecules that recognize pathogen and damage-associated molecular patterns from intruders (16, 17). Furthermore, the concept of exclusive memory for adaptive responses was weakened after the description of "trained innate memory," involving a heightened response upon re-exposure to a certain stimulus (16, 18) under the control of the cellular metabolism (19). Moreover, innate immune tolerance has also been demonstrated (20).

This review article focuses on the state-of-the-art of the TH mechanism of action and its effects on innate immunity at cellular level, with the pathophysiological role of the reported findings also discussed. The main effects of T3 and/or T4 in Neutrophils, NK cells, Macrophages and DCs are depicted in **Figure 1** and considered below.

### Neutrophils

Neutrophils are the first line of defense against bacteria and fungi, and also help to combat parasites and viruses (21). They travel from the blood to the inflammatory site where they engage and kill microorganisms and clear infections through chemotaxis, phagocytosis, and cytokine synthesis, and the release of reactive oxygen species (ROS) and granular proteins such as myeloperoxidase (MPO) (22). Classical concepts of neutrophil biology are being increasingly challenged by recent findings (23, 24).

Administration of T3 to rats increased the respiratory burst activity of isolated PMNLs with enhanced NADPH oxidase and MPO activities (25, 26). Accordingly, increased mitochondrial oxygen consumption and ROS production were reported in PMNLs from both Graves' disease and toxic adenoma patients (27). Moreover, T3 administration to euthyroid subjects induced ROS generation by PMNLs (28). However, a decrease in oxidative metabolism was registered in human PMNLs during hypothyroidism, which was reversed upon L-T4 substitution therapy (29). The authors suggest that this effect was unlikely to result from direct actions of THs on PMNLs, considering that T3 showed no appreciable effect on superoxide anion (O<sup>−</sup> 2 ) generation in in vitro experiments with PMNLs from healthy donors. In addition, hypothyroidism causes changes in the lipid composition of PMNLs' membranes that may be involved in their impaired function (30). To note, human neutrophils express TR (31).

T4 and the TH metabolite 3,5-T2 as well as T3 induced respiratory-burst activity and stimulated MPO activity in human PMNLs. These effects were mediated by a non-genomic mechanism initiated at the plasma membrane, dependent on PKC and Ca<sup>+</sup> levels. Moreover, O<sup>−</sup> 2 production in resting PMNLs of hyperthyroid patients was elevated compared with either controls or hypothyroid subjects (32). Furthermore, PMNLs

express receptors for T1AM, a T4 derivative, involved in the chemosensory migration toward T1AM (33).

TH metabolism plays an important role in neutrophil function during infection. It has been demonstrated that D3 is strongly expressed in murine neutrophils during chronic chemical inflammation and in acute bacterial infection. Accordingly, human neutrophils express D3, D1, MCT10, and TRα1, which could therefore be involved in TH action in this cell type. Furthermore, evidence has supported the notion that D3 plays a role in the bacterial killing capacity of neutrophils, either through generation of iodide for the MPO system or through modulation of intracellular TH bioavailability (34). Recent results have demonstrated that intracellular TH levels are regulated by D3, playing a key role in neutrophil function in zebrafish, mice and humans (35).

#### Natural Killer Cells

NK cells mediate cytolytic activities against tumor and virusinfected targets. Of note, NK cells also possess traits of adaptive immunity and can acquire functional qualities associated with immunological memory (36). The studies of the effects of THs on these cells have produced conflicting results. A positive correlation between serum T3 concentration and NK cell activity in healthy elderly subjects was recorded but exogenous T3 administration increased NK cell activity only in old individuals who had T3 concentrations at the lower end of the reference range (37). Although NK cell functionality was impaired in Graves' patients and restored in the euthyroid state (38, 39), in vitro treatment with T4 to peripheral blood lymphocytes from these patients did not show any increase in NK cell activity (40). In agreement, hyperthyroxinemia induced in mice reduced NK cell capacity to lyse target cells (41) whereas exogenous T4 or T3 administered to mice increased NK cell lytic activity (42), as well as during protein starvation (43), or aging (44).

Endogenous IFN-γ plays a relevant role in the host defense against infectious and neoplastic diseases by mechanisms that involve modulation of the NK cell function (45, 46). Both T3 and T4 boosted IFNγ-response in murine NK cells (44, 47), while T4 amplified the effect induced by both IFN-γ and IL-2 (48). These findings suggest a role for THs in the modulation of NK cell sensitivity to IFN-γ.

A recent study linked uterine NK cells (the most prominent leukocytes at the maternal-fetal interface) with THs. These cells express MCT8 and MCT10, as well as TRα1 and β1 in the first trimester of human pregnancy. An increase of IL-6 secretion after T3 exposure in vitro was also reported (49).

#### Monocytes—Macrophages

Macrophages are strategically positioned in all tissues of the body and can recognize and remove pathogens, toxins, cellular debris, and apoptotic cells. Tissue-resident macrophages in adulthood rely on replenishment by bone marrow (BM) derived blood monocytes, with circulating monocytes being recruited to tissues by specific chemotactic factors. Among other names, tissue-resident macrophages are referred to as "microglia" in the central nervous system and "Kupffer cells" in the liver (50–52). Depending on the signal and the dose, a second stimulation can result in tolerance or trained immunity (53, 54). In response to stimuli, differentiated macrophages polarize to classically activated M1 or alternatively activated M2 macrophages, although a spectrum of phenotypes across the M1/M2 continuum is recognized. M1 macrophages phagocytize and destroy microbes, eliminate tumor cells, and present antigens to T cells through ROS production, expression of inducible nitric oxide synthase (iNOS) and release of proinflammatory cytokines, thereby promoting T helper (Th) 1 responses (55). In contrast, M2 macrophages show an immunosuppressive phenotype characterized by a decreased antigen presentation to T cells and production of cytokines that stimulate Th2 responses. These regulatory cells are involved in tissue repair, promote tumor growth and exert antiparasitic effects (56).

In spite of controversial results concerning the expression of TR isoforms, macrophages express TRα and β (57–61). In addition, murine and human macrophage cell lines express D2, MCT10, and MCT8 (59). Over the past decade, it has become clear that shifts in cellular metabolism are determinants of macrophage function and phenotype (62). The activities of key enzymes of glycolysis are regulated by THs in these cells, affecting macrophage metabolism and function (63). Stimulation of the immune system in hyperthyroid rats revealed that monocyte migration and ROS production by macrophages were suppressed. In contrast, hypothyroidism enhanced ROS release, whereas monocyte migration was not affected (64).

THs enhanced the phagocytic activity of intraperitoneal macrophages from hypothyroid rats (64). Moreover, T4 administration to old mice also increased their phagocytic capacity (65). In agreement, a stimulatory effect of T4 (but not T3) on the phagocytosis process of cultured peritoneal mouse macrophages was reported (66). However, both THs enhanced bacteria-cell interaction and intracellular killing in mice RAW 264.7 and human THP-1 monocyte-derived macrophage cell lines (67). This mechanism involved the integrin αvβ3, THinduced iNOS expression, generation of NO and triggering of the PI3K and ERK1/2 signaling pathways.

The inflammatory response exerted by macrophages was stimulated during hypothyroid condition and inhibited in the course of hyperthyroidism (68). T4 inhibited the migration inhibitory factor (MIF) in macrophages (67, 69), and in agreement, low plasma T4 concentrations augmented plasma MIF levels in both patients and rats with severe sepsis (69). Although T4 attenuated proinflammatory responses in vivo, no significant changes in IL-6 and TNFα levels could be detected in T4-treated peritoneal macrophages from mice, or in mouse and human cell lines (67).

The "euthyroid sick syndrome" (or "nonthyroidal illness") is distinctive of critically ill patients with severe infections or sepsis, being characterized by low serum T3 and in serious cases by also low serum T4 without the expected increase in TSH (70). Interestingly, supplementation of T4 to rats and mice in bacterial infectious models enhanced animal survival and attenuated septicemia and inflammatory responses (67, 71). In agreement, hypothyroid mice exhibited increased mortality during inflammation induced by LPS, whereas circulating T3, through TRβ1 signaling, protected animals from endotoxemia (57). However, it was reported that hyperthyroidism increased mice mortality in response to LPS. Noteworthy, Signal Transducer and Activator of Transcription 3 (STAT3) activation induced by LPS or IL-6 was inhibited by T3 through TR signaling in RAW 264.7 cells and in primary cultures of BM-derived macrophages. These authors suggested that inhibition of IL-6 signaling induced by T3 has potent regulatory functions during infection and inflammation (72).

Switching from M1 to the M2 phenotype protects the organism from excessive inflammation, whereas switching from M2 to M1 prevents allergic and asthmatic Th2 reactions, decreases the bactericidal properties of macrophages and favors the resolution of inflammation (63). In this regard, T3 reduced monocyte differentiation into macrophages and induced a M1 signature. In agreement, T3 decreased the expression of genes regulated by M2-activated macrophages through a TRβ1-mediated mechanism (58). Although comparable results were registered in RAW264.7 macrophages, a TRα-dependence was revealed (73). In contrast, in a model of kidney obstruction, ligand-bound TRα inhibited the NF-κB pathway and proinflammatory cytokines in macrophages isolated at the inflammatory site (61).

The role of intracellular TH metabolism in macrophages has been extensively reported and reviewed by Boelen group (34), and is therefore not covered in this review. More recently, a reduction of intracellular T3 concentration due to a lack of D2 activity with impaired macrophage function was reported. Also, primary BM-derived macrophages treated with LPS decreased phagocytosis and proinflammatory cytokines in D2 KO mice (73), consistent with earlier results in RAW264.7 cells (59).

Modifications in the homeostatic conditions of the nervous tissue promote microglia activation, release of inflammatory mediators and phagocytosis of degenerating cells (74). Lima et al. (75) reported that rat microglial cells in culture express TRα and TRβ, whereas other authors did not observe the latter (76). αVβ3 integrin has also been described in these cells (77), and in mice microglia, the TH transporters OATP4A1, LAT2, and MCT10 were also found (78, 79). It is known that T3 modulates microglial development (75) and functions such as migration and phagocytosis by genomic and non-genomic pathways (80). The molecular mechanism involves T3 uptake by TH transporters and binding to TRs, thus triggering multiple signaling pathways (80, 81). Moreover, T3 increased the release of soluble factors by the microglia through STAT3 activation, promoting glioma growth (82).

Liver is one of the most relevant TH target tissues. T3 induced acceleration of cellular O<sup>−</sup> 2 consumption, resulting in elevated ROS and NO (83). In agreement, T3-stimulated free radical activity reduced the cellular antioxidant defenses leading to oxidative stress in rats, a phenomenon also observed in human hyperthyroidism (84, 85). Kupffer cells are main scavengers constantly clearing gut-derived pathogens from the blood, preventing liver diseases (86). T3 promoted hyperplasia and hypertrophy of these cells, with a resulting enhancement in the respiratory burst activity. Furthermore, T3-induced calorigenesis resulted in transient elevations in serum TNF-α, determined by actions exerted in Kupffer cells and involving activation of NF-kB (87). The hepatic response induced by T3 involved cell proliferation associated with TNF-α generation by Kupffer cells (88).

## Dendritic Cells

DCs are the main antigen presenting cells in the interface between innate and adaptive immunity. They integrate signals derived from infection or damage, and present processed antigen to naive T cells to tailor the appropriate T cell program. Recent advances in DC immunobiology have led to a clearer understanding of how T cell responses are shaped (89). The main DCs include conventional (classical or myeloid) DCs (cDCs, referred as DCs from now on) and plasmacytoid DCs (pDCs). The genetic signature of DCs from different tissues is similar, but differs from that of pDCs, monocytes and macrophages. To note, DCs are functionally different to macrophages (89, 90). Immature DCs (iDCs) have substantial endocytic activity but lower surface expression of major histocompatibility complex (MHC) class I and II proteins. After encountering any stimulus, DCs mature to undergo considerable cytoplasmic reorganization, transporting peptide-MHC complexes to the cell surface and upregulating costimulatory molecules (90). Recent studies highlighted the relevance of DC migration in the maintenance of immune surveillance. Immature DCs are rather immotile, and after processing foreign and self-antigens or damage signals undergo an activation process, leading to an increase in motility corresponding to upregulation of CC-chemokine receptor 7 (CCR7). The interaction of CCR7 with its ligand guides DCs toward secondary lymphoid organs (91).

The role of THs in the initiation of adaptive immunity remained uncertain for many years, with Mooij et al. providing the earliest clues that THs and other iodinated derivatives, mainly T3, favored the maturation of human peripheral blood monocytes into functional DCs (92). Many years later, our laboratory initiated a study on the effects of THs at the DC level (**Figure 2**). We observed the expression of TRs in BM-derived mouse DCs, principally the TRβ1 isoform, and mainly in the cytoplasm of both iDCs and LPS-matured DCs. The ability of physiological concentrations of T3 to induce phenotypic and functional activation of DCs and to drive a Th1 profile was also demonstrated (93). Mechanistically, this effect involved activation of the Akt and NF-kB pathways (94) and was counteracted by glucocorticoids (95). The requirement for an intact TRβ-T3 signaling in T3-induced DC activation was confirmed by in vitro and in vivo studies (94, 96).

Interestingly, we showed that T4, the main circulating TH, did not reproduce T3-dependent effects in DCs. The characterization of the mechanisms of TH transport and metabolism in DCs supports the notion of a homeostatic balance to prevent unspecific systemic activation of DCs. In this regard, DCs express MCT10 and LAT2 TH transporters, and mainly transport T3 with a favored involvement of MCT10, as its inhibition almost prevented T3 saturable uptake mechanism and reduced T3 induced IL-12 production. In addition, DCs express D2 and D3, and exhibit both enzymatic activities with a prevalence toward TH inactivation (97).

Immunotherapy has become the fourth pillar of cancer care, complementing surgery, cytotoxic therapy, and radiotherapy (98). In this context, DCs have been the subject of numerous studies seeking new immunotherapeutic strategies against cancer. However, despite initial enthusiasm, disappointing results including a short half-life of DCs in circulation and induction of tolerogenic responses by death cells, have raised doubts regarding these approaches. Nevertheless, the increased understanding of DC immunobiology and the search for optimization strategies are allowing a more rational development of DC-based immunotherapies (99, 100). A new role for THs in this field has arisen, with T3 binding to TRβ increasing mice DC viability and augmenting CCR7 expression, thereby driving migration of DCs to lymph nodes. Moreover, T3 stimulated the antigen cross-presentation ability of DCs, boosting antigen-specific cytotoxic T-cell responses. Also, vaccination with T3-stimulated DCs in mice bearing B16 melanoma inhibited tumor growth and prolonged host survival (96, 101). Overall, these results established the adjuvant effect of T3-TRβ signaling in DCs, identifying a DC vaccination approach in cancer immunotherapy.

Further recent in vitro and in vivo evidence has shed light on the molecular and cellular mechanisms driven by T3-conditioned murine DCs (102). Findings revealed an induction of a proinflammatory cytokine profile and a down-modulation of PDL expression in DCs. In co-cultures, these cells increased the frequency of IL-17-producing splenocytes, mainly by the γδ-T population. Thus, downregulation of tolerogenic T regulatory (Treg) cells and PD1 expression were induced, limiting the inhibitory signals and emphasizing the relevance of T3 as an additional immune-endocrine checkpoint.

The understanding of the effect of THs in human DCs is still limited. Dedecjus et al. (103) reported that the thyrometabolic state influenced the major human peripheral blood DCs, pDCs, and cDCs, with T4 substitution to thyroid cancer patients after surgery increasing the frequency of these cells and the expression of CD86 and HLA-DR (activation markers). In hypothyroid patients with Hashimoto's Thyroiditis, T4 supplementation exerted changes of peripheral blood DC subpopulations, with increased expression of costimulatory molecules (104). Although TRs in human DC populations have not yet been found, increased expression of CD86 by T3 addition to cell cultures of human peripheral blood pDCs was reported (103). Also, T3 increased the ability of human DCs to upregulate the proliferative response and secretion of IL-12 by peripheral blood mononuclear cells, similar to our findings in mice splenocytes co-cultured with T3-stimulated DCs (93).

The proinflammatory role of IL-12 and its involvement in Th1-mediated organ-specific autoimmune diseases (105) confer potential clinical relevance of the aforementioned studies. An increased synthesis of IL-12 by DCs obtained from hyperthyroid mice has been reported (106). Furthermore, patients with

Graves' disease exhibited elevated IL-12 circulating levels (107). Considering that DCs are involved in the pathogenesis of autoimmune thyroid diseases (108) and also their potential application for the treatment of these pathologies (109), further research should shed light in this field.

#### CONCLUDING REMARKS

The relationship between THs and innate immune cells is complex, with an improved knowledge still necessary. Cellular and molecular signaling pathways involved in the crosstalk between THs and innate immune functions, and their role directing adaptive immunity have profound implications in immunopathology, including cancer and autoimmune manifestations of the thyroid gland, at the crossroads of the immune and endocrine systems. The etiopathogenic mechanism involved in both immune-related thyroid pathologies and immune disorders due to thyroid dysfunctions are now better understood. With a focus on particular cell subsets, further research will provide valuable tools for manipulating the immunogenic potential of innate immune cells to positively regulate the development of protective immunity, or negatively control the generation of autoimmune thyroid inflammation.

## AUTHOR CONTRIBUTIONS

MM and CP: conception and design, analysis and interpretation of available data, writing, review, and revision of the manuscript. MM: design of figures. CP: general supervision.

#### REFERENCES


#### FUNDING

This study was supported by grants from Agencia Nacional de Promoción Científica y Tecnológica (ANPCyT, PICT-2016- 0254 y PICT-2016-0037), Secretaría de Ciencia y Tecnología de la Universidad Nacional de Córdoba (SeCyT, 2018-2020), and Fundación Sales.

#### ACKNOWLEDGMENTS

The authors are extremely grateful to Isabel Maria Montesinos for her excellent assistance in the design and drawing of the figures.


human polymorphonuclear leukocytes. J Endocrinol. (2005) 185:121– 9. doi: 10.1677/joe.1.05968


receptor beta1 expression controls dendritic cell function via Akt signaling. J Biol Chem. (2010) 285:9569–82. doi: 10.1074/jbc.M109.071241


**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 Montesinos and Pellizas. 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.

# Corrigendum: Thyroid Hormone Action on Innate Immunity

#### Approved by:

Frontiers Editorial Office, Frontiers Media SA, Switzerland

> \*Correspondence: Claudia Gabriela Pellizas claudia@fcq.unc.edu.ar

#### Specialty section:

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

> Received: 02 July 2019 Accepted: 04 July 2019 Published: 19 July 2019

#### Citation:

Montesinos MM and Pellizas CG (2019) Corrigendum: Thyroid Hormone Action on Innate Immunity. Front. Endocrinol. 10:486. doi: 10.3389/fendo.2019.00486 Facultad de Ciencias Químicas, Centro de Investigaciones en Bioquímica Clínica e Inmunología (CIBICI-CONICET) and Departamento de Bioquímica Clínica, Universidad Nacional de Córdoba, Córdoba, Argentina

Keywords: thyroid hormones, innate immunity, neutrophils, natural killer cells, macrophages, dendritic cells

#### **A Corrigendum on**

#### **Thyroid Hormone Action on Innate Immunity**

María del Mar Montesinos and Claudia Gabriela Pellizas\*

by Montesinos, M. M., and Pellizas, C. G. (2019). Front. Endocrinol. 10:350. doi: 10.3389/fendo.2019.00350

An author name was incorrectly spelled as "Claudia Pellizas." The correct spelling is "Claudia Gabriela Pellizas."

The authors apologize for this error and state that this does not change the scientific conclusions of the article in any way. The original article has been updated.

Copyright © 2019 Montesinos and Pellizas. 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.

# "Beige" Cross Talk Between the Immune System and Metabolism

Krisztina Banfai 1,2, David Ernszt 2,3, Attila Pap<sup>4</sup> , Peter Bai 5,6,7,8, Kitti Garai 1,2 , Djeda Belharazem<sup>9</sup> , Judit E. Pongracz 1,2 and Krisztian Kvell 1,2 \*

<sup>1</sup> Department of Pharmaceutical Biotechnology, Faculty of Pharmacy, University of Pécs, Pécs, Hungary, <sup>2</sup> Szentagothai Research Center, University of Pécs, Pécs, Hungary, <sup>3</sup> Department of Physiology, Medical School, University of Pécs, Pécs, Hungary, <sup>4</sup> Department of Biochemistry and Molecular Biology, Faculty of Medicine, University of Debrecen, Debrecen, Hungary, <sup>5</sup> Medical Chemistry, Faculty of Medicine, University of Debrecen, Debrecen, Hungary, <sup>6</sup> MTA-DE Cell Biology and Signaling Research Group, Debrecen, Hungary, <sup>7</sup> MTA-DE Lendulet Laboratory of Cellular Metabolism, Debrecen, Hungary, <sup>8</sup> Research Center for Molecular Medicine, University of Debrecen, Debrecen, Hungary, <sup>9</sup> Department of Pathology, University Hospital of Mannheim, Mannheim, Germany

With thymic senescence the epithelial network shrinks to be replaced by adipose tissue. Transcription factor TBX-1 controls thymus organogenesis, however, the same TBX-1 has also been reported to orchestrate beige adipose tissue development. Given these different roles of TBX-1, we have assessed if thymic TBX-1 expression persists and demonstrates this dualism during adulthood. We have also checked whether thymic adipose involution could yield beige adipose tissue. We have used adult mouse and human thymus tissue from various ages to evaluate the kinetics of TBX-1 expression, as well as mouse (TEP1) and human (1889c) thymic epithelial cells (TECs) for our studies. Electron micrographs show multi-locular lipid deposits typical of beige adipose cells. Histology staining shows the accumulation of neutral lipid deposits. qPCR measurements show persistent and/or elevating levels of beige-specific and beige-indicative markers (TBX-1, EAR-2, UCP-1, PPAR-gamma). We have performed miRNome profiling using qPCR-based QuantStudio platform and amplification-free NanoString platform. We have observed characteristic alterations, including increased miR21 level (promoting adipose tissue development) and decreased miR34a level (bias toward beige adipose tissue differentiation). Finally, using the Seahorse metabolic platform we have recorded a metabolic profile (OCR/ECAR ratio) indicative of beige adipose tissue. In summary, our results support that thymic adipose tissue emerging with senescence is bona fide beige adipose tissue. Our data show how the borders blur between a key immune tissue (the thymus) and a key metabolic tissue (beige adipose tissue) with senescence. Our work contributes to the understanding of cross talk between the immune system and metabolism.

Keywords: thymus senescence, beige adipose tissue, TBX-1, UCP-1, PPARgamma

#### INTRODUCTION

In human the degenerative process of thymic adipose involution is already detectable in childhood and accelerates with puberty due to hormonal (sex-steroid) induction (1–3). The process shows identical kinetics in mouse. Also, we have developed a model whereby TECs are treated by a steroid (using Dx or dexamethasone) thus both in vivo and in vitro model systems are readily available (4) As for all adipose tissues subtypes, thymic adipose involution is orchestrated by transcription

Edited by:

Jie Chen, Xiamen University, China

#### Reviewed by:

Lluis Tort, Autonomous University of Barcelona, Spain Mark Klitgaard Nøhr, University of Copenhagen, Denmark

> \*Correspondence: Krisztian Kvell kvell.krisztian@pte.hu

#### Specialty section:

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

Received: 11 March 2019 Accepted: 24 May 2019 Published: 18 June 2019

#### Citation:

Banfai K, Ernszt D, Pap A, Bai P, Garai K, Belharazem D, Pongracz JE and Kvell K (2019) "Beige" Cross Talk Between the Immune System and Metabolism. Front. Endocrinol. 10:369. doi: 10.3389/fendo.2019.00369 factor PPARgamma (5–7). It is estimated that by the age of 50 years in human (approx. 12 months in mouse), the thymus loses approx. Ninety percent of its function: naïve Tcell production (8, 9). The consequences of impaired thymus function are profound: elevated incidence of infections, cancer and autoimmune disorders observed at senior ages (10, 11). This poses a significant burden on health-care and health-insurance systems, while simultaneously lowering the quality of life in the elderly.

Transcription factor TBX-1 is a key molecular player in the formation of the third pharyngeal pouch involved in thymus organogenesis during embryonic development (12). Human patients with 22q11.2DS impairing TBX-1 often have thymus hypoplasia or aplasia. In accordance, Tbx-1null mice develop severe pathologies in tissues derived from the third pharyngeal pouch, including hypoplasia of the thymus (13, 14). In these cases, impaired thymus organogenesis leads to deficient thymocyte development, naive T-cell production, and immune functions (15). However, recently it has also been reported that the role of TBX-1 in thymus organogenesis is more complex. Ectopic expression of TBX-1 may suppress transcription factor FoxN1, the mastermind of thymic epithelial identity (16). The issue was investigated in the embryonic setting, but the potential role of persistent TBX-1 expression during adulthood has not been addressed.

TBX-1 has another pivotal role in the development and function of a recently described subtype of adipose tissue: beige adipose tissue (17–20). White adipose tissue stores energy, brown adipose tissue generates heat (via NST or non-shivering thermogenesis), while beige adipocytes act as intermediates. Beige adipocytes respond to adrenergic stimuli by thermogenesis (21). TBX-1 is considered as a beige-specific marker, but other beige-indicative markers have also been described. Mitochondrial uncoupling proteins (mostly UCP-1) have been reported to be expressed by brown / beige adipose tissue. EAR2 (also known as Nr2f6) was reported to efficiently promote adipose tissue development with beige bias, while CD137 (also known as Tnfrsf9) is an acknowledged beige adipocyte surface marker (22).

The adult thymus expresses TBX-1 and UCP-1 in the stromal compartment, both known to promote beige adipose tissue development. Yet to date thymic adipose tissue that develops with age has not been accurately positioned on this white-beigebrown continuum of adipose tissue subtypes, despite recent cellular analysis from an adipocyte perspective (23–26). For this reason, we have characterized senescence-related thymic adipose tissue using molecular, cellular and histological markers, at structural and ultra-structural levels, using both mouse and human samples. Additionally, we have also performed metabolic profiling and complete miRNome analysis using both PCR-based and amplification-free platforms.

#### METHODS

#### Cell Cultures

For in vitro experiments primary-derived (BALB/c) thymic epithelial cells were used (TEP1) as reported previously (cell source: Prof. G. Anderson, University of Birmingham, UK) (27). Briefly, the cells were cultured in DMEM (Dulbecco's Modified Eagle's medium Lonza) supplemented with 10% FCS, penicillin, streptomycin and β-mercapto-ethanol. Human thymus-derived 1889c thymic carcinoma cells were cultured in RPMI 1640 (Roswell Park Memorial Institute medium, Lonza) containing 10% FCS, penicillin, streptomycin, L-Glutamine and Hepes (28, 29). Adipose differentiation of TEP1 and 1889c cells was induced by steroid treatment. Briefly, experiments differentiation was induced by dexamethasone alone (Dx) as added to complete DMEM and RPMI medium. Cells were treated with Dx at a final concentration of 1µM for 1 week.

## Animal Samples

Thymus lobes were used from C57BL/6J mice at 1, 6, 8, 12, 14, 18, and 21 months of age. Mice were housed under minimal disease (MD) conditions. Animal rooms were ventilated 15 times/h with filtered air, mice received autoclaved pellet diet (Altromin VRF1) and tap water ad libitum. The cages contained sterilized bedding. Room lighting was automated with 12 h light and 12 h dark periods. Room temperature was 21 ± 2 ◦C, relative humidity was between 30 and 60%. Mice were kept in the Laboratory Animal Core Facility of the University. Experimental procedures were carried out according to the "1988/XXVIII act of the Hungarian Parliament on Animal Protection (243/1988)" which complies with recommendations of the Helsinki Declaration. All animal experiments were performed with the consent of the Ethics Committee on Animal Research of the University (ref. no.: #BA02/2000-46/2016).

#### Enrichment of Primary Cells

Mouse thymic epithelial cells were isolated by MACS cell separation. Briefly, mouse thymic lobes (1 month-old or 12 month-old) were digested with type F collagenase from C. hystolyticum (3mg/ml, Sigma-Aldrich) for 2 h, with stirring in every 20 min, then washed with DMEM. Cell suspensions were then labeled with anti-EpCAM1 antibody (1:100, rat monoclonal antibody clone: G8.8) and washed with MACSbuffer (2% FCS, 1mM EDTA in PBS) followed by incubation with Dynabeads sheep anti-rat IgG-coated beads (Invitrogen) The EpCAM+-cells were separated with EasySep column-free cell isolation platform (Stemcell Technologies) according to the manufacturer's instructions. Isolated cells were used for total RNA isolation and subsequent qPCR analysis.

#### Human Thymus Samples

Formalin-fixed, paraffin-embedded (FFPE) human thymus samples from 18, 23, 42, 44, and 58 years of age were provided by the Department of Pathology, Faculty of Medicine, University of Pecs, Hungary. Experiments involving human samples were performed with the consent of the Regional and Local Ethics Committee of Clinical Center of the University (ref. no.: 6069/2016) according to their guidelines. All subjects gave written informed consent in accordance with the Declaration of Helsinki.

#### Transmission Electron Microscopy

Cells were harvested and pelleted then fixed with PBS containing 2.5% glutaraldehyde overnight at 4◦C. Following fixation, pellets were mixed in 3% porcine gelatin (Sigma-Aldrich). Hardened small blocks of approximately 1 mm<sup>3</sup> were cut. Blocks were post-fixed in 1% osmium-tetroxide in PBS for 1 h at 4◦C and dehydrated with increasing concentration of ethanol. Uranyl-acetate (1%) was added in 70% ethanol to increase contrast. After complete dehydration in ascending ethanol series, blocks were transferred to propylene-oxide twice for 4 min. Then blocks were immersed in the mixture of propylene-oxide and Durcupan resin (Sigma-Aldrich) for 30 min. Later blocks were placed into Durcupan-containing tinfoil boats overnight, and embedded into gelatin capsule filled with Durcupan resin (Sigma-Aldrich). Following polymerization and hardening of the resin at 56 ◦C for 72 h, semi thin sections were cut with Leica Ultracut ultramicrotom, mounted on glass slides, stained with toluidine-blue and examined with Olympus BX50 light microscope. Then serial ultrathin sections were cut by ultramicrotom, and mounted on mesh grids. Ultra-thin sections were contrasted by uranylacetate and lead-citrate, and examined using JEOL 1200EX-II electron microscope.

#### Immune-Histochemistry

Human thymus lobes were fixed in paraformaldehyde (4% PFA in PBS) then paraffin embedded. 5µm thick sections were stained with immunohistochemistry method as described earlier (30). First, slides were rinsed in heated xylene then washed with a descending series of alcohol. After deparaffinization slides were rehydrated and antigen retrieval was performed in Target Retrieval Solution (pH 6 DAKO) at 97◦C for 20–30 min. Following wash in dH2O and endogenous peroxidase activity was blocked with 3% H2O<sup>2</sup> in TBS (pH 7.4) for 15 min. Then slides were washed with TBS containing Tween (0.05%, pH 7.4). Preblocking was carried out with 3% BSA in TBS for 20 min followed by overnight incubation with a-TBX-1 (1:100, rabbit polyclonal antibody, Sigma-Aldrich) primary antibody at 4◦C. After the incubation slides were washed with TBS then incubated with peroxidase conjugated secondary antibody (1:100, Polyclonal Goat Anti-Rabbit IgG, DAKO) for 90 min. Labeling was visualized with liquid DAB Substrate Chromogen System (DAKO). Hematoxylin served for nuclear counterstaining. Slides were mounted with Faramount Aqueous Mounting Medium (DAKO). Histological evaluation was performed with Panoramic MIDI digital slide scanner (3DHistech) and images were captured with CaseViewer. Image analysis was made with ImageJ / IHC toolbox.

#### Immune-Fluorescent Staining

Immune-fluorescent staining was performed on 8µm cryostat thymus sections. Cytospin technique was used to spin TEP1 and 1889c cells onto glass slides (4). Slides were fixed in cold acetone, then dried and blocked using 5% BSA in PBS for 20 min before staining with fluorochrome conjugated or primary antibodies: a-EpCAM-FITC (1:100, clone: G8.8,), a-UCP-1 (1:100, rabbit polyclonal antibody, Abcam) a-TBX-1 (1:100, rabbit polyclonal antibody, Sigma-Aldrich), a-PPARgamma (1:100, rabbit monoclonal antibody, Cell Signaling Technology). For secondary antibody Alexa-555 conjugated arabbit goat IgG (1:200, Life Technologies) was used. Fluorescent lipid staining was performed on paraformaldehyde (4%) fixed TEP1 and 1889c cytospin slides with LipidTOX Red dye (1:200, Invitrogen). For nuclear counterstain DAPI (Life Technologies) was used. Sections were imaged using a Nikon Eclipse Ti-U microscope equipped with a CCD camera (Andor Zyla 5.5) and NIS-Elements software.

## Metabolic Profiling

The use of TEP1 cells for Seahorse metabolic profiling was started by pilot experiments for optimal starting cell number, duration of differentiation, differentiation medium etc. Accordingly, 15,000 cells/well were cultured for 9 days using standard MDI differentiation protocol (31). This was followed by the evaluation of their metabolic profile using the Seahorse XF 96 platform (Seahorse Bioscience). Cells were plated into Seahorse cell plates at confluence and were left to attach overnight. The next day, cells were subject to oxymetry measurement. After recording baseline oxygen consumption cells were treated with butyril-cAMP (500µM), oligomycin (2µM), and antimycin (10µM). Antimycin-resistant oxygen consumption was considered as background and was subtracted from all values. Baseline oxygen consumption, membrane leak (OCR, after oligomycin treatment) was calculated. Glycolysis was assessed through the extracellular acidification value (ECAR, before oligomycin treatment) and the ECAR/OCR values were calculated. Negative values were omitted in calculations.

#### RNA Isolation, cDNA Preparation, qRT-PCR, TaqMan Array

Total RNA of enriched thymic epithelial cells, TEP1 and 1889c cells was isolated with the NucleoSpin RNAII kit (Macherey-Nagel). cDNA was prepared using the High Capacity cDNA Reverse Transcription kit (Applied Biosystems). For qPCR analysis the StepOnePlus (Applied Biosystems) platform was used with SensiFAST SYBR Hi-ROX Mix (Bioline) as well as PikoRealTM Real-Time PCR System (Thermo Fisher Scientific) using Luminaris Color HiGreen qPCR Master Mix (Thermo Fisher Scientific) (for primer list see **Table 1**). Gene expression was normalized to β-actin, GUSB and HPRT1 housekeeping genes. Reverse transcription of 1889c RNA samples for miRNA analysis was completed with MegaplexTM RT Primers specific to human Pool A (Cat. No.: 4399966) and Pool B (Cat. No.: 4444281). MiRNA profiling was performed on Applied Biosystems QuantstudioTM 12K Flex Real-Time PCR System platform using TaqManTM Array Human MicroRNA A (Applied Biosystems, Cat. No.: 4398965) and B Card (Applied Biosystems, Cat. No.: 4444910) containing 6 housekeeping genes (RNU44, RNU48, ath-miR159a and 4 U6 snRNAs) and 377 human miRNAs. Additionally 600 ng of total RNA was mixed with TaqManTM Fast Universal PCR Master Mix (2X), no AmpEraseTM UNG (Applied Biosystems, Cat. No.: 4364103) for each array

TABLE 1 | List of mouse and human primer sequences.


card. Gene expressions were analyzed using Expression Suite Software version 1.1.

#### NanoString System Assay

One hundred nanogram of total RNA was used to detect up to 800 miRNA targets with nCounter SPRINT Profiler (NanoString Technologies) using nCounter <sup>R</sup> Human v3 miRNA Expression Assay. Sample preparation was performed with nCounter <sup>R</sup> CodeSet (NanoString Technologies) following annealing, ligation and purification. Hybridization protocol was completed at 65◦C and 12 h long according to the manufacturer's instructions. Quantified data was analyzed using nSolverTM Analysis Software version 4.0. Threshold count was determined using negative controls as background noise. Gene expression changes were visualized on heat map using GraphGad Prism version 7.04.

#### Statistical Analysis

All experiments were performed at least on three occasions, representative experiments are shown. Measures were obtained in triplicates, data are presented as mean and ±SD as error bars. For statistical analysis GraphPad Prism software and SPSS Statistics version 22.0 was used. To evaluate the kinetics of TBX-1 expression with age in both mouse and human samples normality distribution was tested using Shapiro-Wilk test (n < 50). In case of human samples our data met the assumption of homogeneity of variances, so parametric one-way ANOVA with Tukey's honestly significant difference (HSD) post hoc test was used. To determine the significant differences of mouse samples non-parametric Kruskal-Wallis test was used. For further cases two-tailed student's t-test was applied. Significant differences are shown by asterisks (ns for p > 0.05, <sup>∗</sup>p ≤0.05, ∗∗p ≤ 0.01, ∗∗∗p ≤ 0.001).

#### RESULTS

#### Aging and Steroid-Induced TECS Show Beige Adipocyte Markers

Thymic senescence is accompanied by the appearance of adipose tissue. Mediastinal location and local FGF21 production are characteristic to the thymus and both were reported to promote beige adipose tissue development (23–25). For this reason we searched for the up-regulation of beige adipocyte markers in the adult thymus tissue and its model system: steroid-induced TECs.

#### Aging Up-Regulates key Beige Adipocyte Marker in Human Thymus Tissue

Using a pilot set of human thymic FFPE samples of various adult ages (18, 23, 42, 44, and 58 years) we performed immunehistochemistry staining for beige adipose tissue-specific marker TBX-1 (**Figures 1A–E**). TBX-1 expression (enzyme reaction in brown) appears to persist throughout adulthood. Normalization to hematoxylin nuclear counterstain (in blue) shows that TBX-1 staining intensity transiently decreases at young adult age (23 years) to show rebound at later ages (**Figure 1F**). In other words: TBX-1 expression may present a bimodal nature with elevations at both young and adulthood ages and an in-between transient decrease during young adulthood. The histological appearance of adipocytes is observed from 44 years of age onwards in this series.

#### Further Beige Adipocyte Markers Are Also Up-Regulated in Steroid-Induced Human TECs

As reported previously molecular level events are similar in the aging thymus and steroid-induced TECs in the mouse setting (4, 6). Accordingly, Dx-treatment significantly up-regulated (p < 0.01) pan-adipocyte marker PPAR-gamma expression in the human 1889c TEC line (**Figures 2A,B,I**). We have evaluated 1889c human TECs for the expression of beigespecific and beige-indicative protein markers as well following Dx-treatment. Similar to human thymus sections above, 1889c cells showed persistent and increasing (p < 0.05) TBX-1 expression following Dx-treatment (**Figures 2C,D,I**). UCP-1 expression showed only indicative (not significant) increase upon Dx-treatment (**Figures 2E,F,I**). Lipid accumulation was also tested, using a fluorescent dye (LipidTox Red) specific for neutral lipid deposits. The staining showed that Dxtreatment triggers significant (p < 0.05) accumulation of neutral lipid deposits (**Figures 2G–I**) in harmony with our previous reports (4, 6).

#### Aging Up-Regulates key Beige Adipocyte Marker in Mouse Thymus Tissue

Using a pilot set of mouse thymic cryosections of various ages (1, 6, 8, 12, 14, 18, and 21 months) we performed immune-fluorescent staining for beige adipose tissue-specific marker TBX-1 (**Figures 3A–G**). TBX-1 expression (in red) appears to persist throughout adulthood in the mouse similar to human above. EpCAM-1 staining (in green) shows medullary areas to demonstrate histological organization. Normalization

to DAPI nuclear counterstain (in blue, not shown here) reveals that TBX-1 staining intensity transiently decreases at adult mid-term (12–14 months) to show rebound at senior ages (**Figure 3H**). In other terms: TBX-1 expression potentially appears to be bimodal in the mouse as well showing elevation at both young and senior ages with a transient in-between decrease during adulthood. Murine kinetics of TBX-1 expression resembles the previously shown human kinetics but with higher resolution in time. Please note medullary involution observed from 14 months of age onwards in line with our previous reports (4, 6).

#### Further Beige Adipocyte Markers Are Also Up-Regulated in Steroid-Induced Mouse TECs

As reported previously focusing on PPARgamma expression molecular level events are similar in the aging thymus and steroid-induced TECs in the mouse setting (4, 6). We have evaluated TEP1 mouse TECs for the expression of beigespecific and beige-indicative protein markers after Dx-treatment. TEP1 cells showed persistent, unchanged TBX-1 expression following Dx-treatment (**Figures 4A,B,I**). UCP-1 expression showed significant (p < 0.01) increase following Dx-treatment (**Figures 4C,D,I**). Lipid accumulation was also tested (LipidTox Red as above). The staining showed that Dx-treatment results in significant (p < 0.05) accumulation of neutral lipid deposits (**Figures 4E,F,I**) in accordance with our previous reports (4, 6). Ultra-structural imaging by TEM shows the appearance of multi-locular intracellular lipid deposits (indicated by asterisks) upon Dx-treatment, reminiscent of beige adipose tissue (**Figures 4G,H**).

## Steroid-Induced TECs Show Beige Adipocyte Metabolic Profile

There is a significant difference between white, brown and beige adipose tissue metabolic traits. In search of further evidence we have characterized the metabolic fingerprint of Dx-induced mouse TECs (TEP1).

The metabolic fingerprint of TEP1 cells treated with Dx (as part of MDI differentiation medium) was assayed using the Seahorse platform (**Figure 5**). MDI cells showed significantly higher basal OCR values compared to control cells (p < 0.001) (**Figure 5A**). Of note cAMP-induced OCR was rapid (30 min post-treatment) and lasted shorter than in previous reports (32, 33). In line with elevated UCP-1 expression oligomycinresistant respiration was significantly higher in MDI cells than in control cells (p < 0.001) (**Figure 5B**). Although we have recorded increased glycolysis marked by significantly increased ECAR values (p < 0.001) (**Figure 5C**), the significantly increased ratio of basal OCR and ECAR (p < 0.001) (**Figure 5D**) in MDI cells suggests their dependence on mitochondrial oxidation.

as fluorescent nuclear counter stain in all cases. PPAR-gamma, TBX-1, UCP-1 and LipidTOX staining relative to DAPI staining is also shown by histograms (I). PPAR-gamma and TBX-1 show significant increase, UCP-1 remains unchanged, while neutral lipid deposits show significant increase following Dx-induction. Significant differences are shown by asterisks (\*p ≤ 0.05, \*\*p ≤ 0.01). Data were calculated from six slides, representative slide is shown. For exact numerical values and standard error of mean please refer to Supplementary Data Sheet.

were evaluated by immune-fluorescent staining (A–G), respectively. Epithelial network is shown in green (EpCAM1-FITC) while TBX-1 expression is shown in red (TBX1-Alexa555) fluorescence. Please note signs of degeneration (auto-fluorescence) at elevated ages. Please also note that TBX-1 staining pattern localizes to both nuclear and cytoplasmic bodies in accordance with The Human Protein Atlas: http://www.proteinatlas.org/ENSG00000184058-TBX1/cell. TBX-1 staining was normalized to DAPI nuclear counter-stain (not shown) and is presented as relative value (H). Please note that relative TBX-1 expression shows a transient decrease at adult age (12 months of age). Significant differences are shown by asterisks (\*p ≤ 0.05, \*\*\*p ≤ 0.001). Data were calculated from three slides, representative slide is shown. For exact numerical values and standard error of mean please refer to Supplementary Data Sheet.

Taking the observed increase in basal OCR value, OCR/ECAR ratio, cAMP-response, and oligomycin-resistant respiration into consideration, these suggest that MDI-differentiated TECs possess a beige metabolic fingerprint in accordance with the up-regulation of beige-specific and beige-indicative markers shown above.

## Aged and Steroid-Induced TECs Show Beige Adipocyte Gene Expression Profile

Adult human and mouse thymus sections showed similar histological changes with age. Likewise, mouse and human steroid-induced TECs were also similar by immune-fluorescent staining. Next, TECs enriched from adult mice or Dxtreated (murine or human) TECs were subjected to gene expression analysis.

Changes in gene expression were further tested at the mRNA level in EpCAM1-enriched primary murine thymic epithelial cells from senior adult age (12 m) and steroid-induced TEP1 or 1889c cells for beige-specific (TBX1) and beige-indicative genes (UCP1, CD137, EAR2) (21–26). Enriched cells showed the up-regulation of both beige-specific and beige-indicative genes with age (1 vs. 12 months) as TBX1, UCP1, and EAR2 all showed significant elevation (p < 0.05 for all, **Figure 6A**), while CD137 activity remained unchanged. Gene expression analysis of mouse TEC line following Dx-treatment showed a similar tendency. as significant increase of TBX1 and UCP1 expression was detected (p < 0.01 and p < 0.05, respectively, **Figure 6B**), while CD137 and EAR2 were not altered. Likewise, the steroidinduced human TEC line showed significantly increased PPARgamma expression (p < 0.01) as reported previously for mouse TECs (4) and also significant increase of UCP-1 expression (p < 0.05) (**Figure 6C**), while CD137 and EAR2 remained identical. Please note the harmony of in vivo and in vitro data in both mouse and human species supporting our observations.

#### Steroid-Induced TECs Show Beige Adipocyte miRNA Profile

There is a significant difference between white, brown and beige adipose tissue miRNA profile. Seeking further evidence we have characterized the miRNA profile of Dx-induced human TECs (1889c).

We have elaborated two distinct platforms (**Figure 7**) for complete human miRNome analysis. For both platforms increased copy numbers are shown in red, while decreased copy numbers are shown in green (heat map). QuantStudio miRNA (QS) panels (A and B, **Figures 7A,B**) evaluate 768 miRNA entities, while the NanoString (NS) cartridge measures copy numbers of 880 miRNA entries (**Figure 7C**). Of note QS is amplification- (PCR) based while NS is amplification free. QS provides enhanced sensitivity, NS ensures unmatched signal-to-noise ratio. Accordingly, QS identified more miRNA species with occasional out-of-scale activities (shown in white) while NS recognized less species with a compressed scale of activities relative to QS. An overlap of the recognized miRNA species identified by at least one platform or evaluated by both

platforms similarly is summarized by **Table 2**. The table connects the identified miRNA species with context-relevant function based on literature search. Of note, several of the recognized species have relevance to thymus senescence with special focus on adipose tissue development, epithelial-to-mesenchymal transition, cell proliferation and senescence.

## DISCUSSION

## Thymic Tissue Samples and Steroid-Induced TECs Show Beige Adipocyte Markers

TBX-1 has been extensively studied for its role in the thymic context during embryonic organogenesis, but not in the adult thymus undergoing adipose involution (11–15). Using human and mouse thymus sections we show that TBX-1 expression persists throughout adulthood with a transient decrease in expression (23 years of age in human and 12 months of age in mouse) based on our pilot studies. This persistence of thymic TBX-1 expression raises the possibility of an alternative role in adulthood. This hypothesis is supported by reports showing that (1) once the thymus has been formed TBX-1 suppresses FoxN1 (key transcription factor of thymic epithelial identity) and (2) TBX-1 has a key and specific role in beige adipose tissue development (17–21). This plausible connection is supported by our results as further beige-indicative markers (UCP-1, EAR2) show increasing mRNA levels with age. This is in harmony with the fact that the thymus resides in the mediastinum and secretes FGF21, both reported to promote the emergence of beige adipose tissue (25). The in vitro model system of aging (Dx-treated mouse TEP1 or human 1889c cells) show similar molecular and cellular changes. Immune-fluorescent histology shows the presence of TBX-1 both in control conditions and following Dx-treatment. UCP-1 protein expression, on the other hand, significantly increases following Dx-treatment. Protein level data are in accordance with mRNA results as both TBX-1 and UCP-1 showed an increase following Dx-treatment (both in mouse and human). The above molecular changes are accompanied by evident phenotypical changes: the appearance of typical intracellular multi-locular neutral lipid deposits (characteristic to brown/beige adipose tissue) as shown by LipidTox staining and transmission electron microscopy. Taken together, these data suggest that thymic adipose tissue emerging with senescence and modeled by steroid-induced TECs show beige adipocyte features.

## Steroid-Induced TECs Show Beige Adipocyte Metabolic Profile

There is profound difference between white and brown/beige adipose tissues with respect to metabolic traits (23). Basal respiration (OCR) is significantly lower in white fat cells than in brown/beige fat cells. Also, UCP-mediated uncoupled respiration rate (resistant to inhibition by oligomycin) is characteristic to brown/beige fat cells and not observed in white fat cells (32). Furthermore, in brown/beige fat cells cAMP-induced mitochondrial oxidation is elevated compared to white fat cells (33). Having analyzed these metabolic parameters, our data suggest that adipocyte differentiation in our model system shows beige bias as indicated by elevated basal OCR, increased OCR/ECAR ratio, enhanced cAMP-response and oligomycinresistant respiration. Our metabolic readouts are in accordance with the recorded beige adipose tissue markers, morphological characteristics and gene expression profiles.

## Steroid-Induced TECs Show Beige Adipocyte miRNA Profile

Unbiased dual platform complete miRNome analysis identified a number of context-relevant miRNA copy number alterations. Of note, miR-27a and miR-106b are beige adipose tissue regulators and miR-155 is an inhibitor of brown/beige adipose tissue formation (36, 46, 53). From a broad pan-adipocyte perspective, miR-128a-3p, miR-1825, miR-301a-5p, miR-30d, miR-425-5p, miR-550a, and miR92b-3p also influence adipose tissue formation and show changes in the current experimental setting (40, 43, 48, 49, 54, 55, 62, 66, 67). Furthermore, a cornerstone of thymus adipose involution: epithelial-tomesenchymal transition (EMT), operates via miR-105-5p, miR-200a-3p, miR-597-5p, miR-888, and miR-99b, all demonstrating

changes in copy number in steroid-induced TECs (35, 36, 50, 63, 65, 69, 70). Taking a final expansion of interest, from a senescent perspective miR-125a-3p, miR-125a-5p, miR-15b-5p, miR-181a-5p, miR-323-3p, and miR-331-3p affect cellular / tissue level senescence with focus on the thymus and also show significant changes (36, 39, 40, 45, 47, 56–58). In summary, steroid-treatment in TECs affects the same miRNA species that were reported in connection with senescence-related thymus adipose involution that apparently yields beige adipose tissue.

## Expanding Overlap Between Metabolism and Immunity

Overlap of metabolism and immunity has already been raised decades ago, and this inter-disciplinary field has recently become a prominent research area. For example, both previous and recent papers discussed overlap between the neuroendocrine and immune systems with regard to melatonin (71–73). Melatonin mainly produced by the pineal gland, but also expressed by the thymus in small amounts—has been reported to have an immune-modulatory effect, enhancing immune functions with Th1 bias. Accordingly, anti-viral and anti-cancer defense is boosted by melatonin and age-related loss of melatonin production partly explains elevated incidence of infection and cancer observed with senescence. Protection from cancer metastasis development in the central nervous system (CNS) implies proper blood-brain barrier (BBB) function (74–76). BBB function, CNS function and immune status are all controlled by metabolic interplays involving small molecules such as lactate. Local tissue lactate concentration has been reported to have important role in immune regulation, its accumulation promoting autoimmune reactions (77, 78). Intercellular immune modulatory signals triggered by metabolically active small molecules are transmitted in cells through signaling pathways. An important pathway connecting metabolism and immunity utilizes mechanistic target of rapamycin (mTOR). It was shown that mTOR senses certain small nutrients (amino acids) and thus affects immune tolerance through regulatory T-cells (79, 80). Mammalian immunity heavily relies on both the innate and the adaptive branch. Within innate immunity macrophages have an important role in connecting metabolism and immunity. It has also been reported that carbohydrate metabolism significantly affects inflammation via macrophages (81).

Carbohydrates are also basic metabolic fuels. The mammalian immune system is a costly defense system with regard to T-cell

#### TABLE 2 | Overlap of QS- and NS-based miRNA results with functional and literature annotation.


development and selection taking place in the thymus, where approx. Ninety-five percent of developing thymocytes are deleted being useless or potentially autoimmune. However, the adaptive branch heavily relies on the constant supply of fresh naïve and scrupulously selected T-cells to prevent infection, cancer and autoimmunity from developing. Severe negative imbalance in energy expenditure (due to fasting or malnutrition) has long been known to hamper thymus function and immunity (82). In contrast, currently, global human population is more threatened by obesity than fasting/malnutrition along with its reported negative effects on thymus function (1, 83). Fashionable countermeasures of obesity include e.g., applying diet to induce ketosis. Ketosis has been reported to enhance FGF21 secretion, known to promote white adipose tissue browning especially in the mediastinal context, where the thymus also resides (25, 84). Further options of white adipose tissue browning include interventions e.g., irisin (exercise hormone) treatment (32). However, since irisin promotes beige adipose tissue development it may also impair thymus function via promoting adipose involution identical to thymus senescence.

Our study highlights another potential intersection of immunity and metabolism via the dual role of TBX-1 during thymus development and senescence. TBX-1 shows bimodal expression (high expression in early and late ages, with a transient decrease in-between) in both mouse and human. It is conceivable that TBX-1 plays a role in thymus organogenesis early on (early "immune" peak) and thymic adipose involution later on (late "metabolic" peak). This dualism may be unique to the thymus due to the observed "beige" adipose involution process.

With senescence the thymus suffers adipose involution. Impaired thymic niche leads to decreased naïve T-cell output. This in turn weakens T cell-mediated anti-viral and anti-cancer defense, and elevates the chances of autoimmune disorders due to dysfunctional T-cell selection. Therefore, thymic adipose tissue emerging with age impairs immune homeostasis and the maintenance of tolerance. Our results indicate that thymic adipose tissue shows "beige" characteristics by molecular, cellular and metabolic profiling. Our research contributes to the breadth of overlap between metabolism and immune homeostasis.

#### DATA AVAILABILITY

All datasets generated for this study are included in the manuscript and/or the **Supplementary Files**.

#### ETHICS STATEMENT

Mice were kept in the Laboratory Animal Core Facility of the University. Experimental procedures were carried out according to the 1988/XXVIII act of the Hungarian Parliament on Animal Protection (243/1988) which complies with recommendations of the Helsinki Declaration. All animal experiments were performed with the consent of the Ethics Committee on Animal Research of the University (ref. no.: #BA02/2000-46/2016). Formalin-fixed, paraffin-embedded (FFPE) human thymus samples from 18, 23, 42, 44, and 58 years of age were provided by the Department of Pathology, Faculty of Medicine, University of Pecs, Hungary. Experiments involving human samples were performed with the consent of the Regional and Local Ethics Committee of Clinical Centre of the University (ref. no.: 6069/2016) according to their guidelines. All subjects gave written informed consent in accordance with the Declaration of Helsinki.

#### AUTHOR CONTRIBUTIONS

KB performed most histological, molecular biology, and statistics work in the project and was involved in manuscript preparation. DE performed all human IHC work. AP and PB were responsible

#### REFERENCES

1. Yang H, Youm Y-H, Vandanmagsar B, Rood J, Kumar KG, Butler AA, et al. Obesity accelerates thymic aging. Blood. (2009) 114:3803–12. doi: 10.1182/blood-2009-03-213595

for preparative Seahorse measurements. KG performed statistical analysis. DB was involved in experiments performed on human 1889c cells. JP was involved in planning experiments and manuscript preparation as well as local supervision of respective department. KK was involved in histological, molecular biology, and statistics work, also in planning experiments and manuscript preparation, and supervised the project.

#### FUNDING

Scientific research support was provided by the Hungarian National Science Foundation (No. 78310) and PTE AOK KA-2016-16 to KK. The project was also supported by the University of Pecs in the frame of Pharmaceutical Talent Center program and the Viral Pathogenesis Talent Center program via KK. The Janos Bolyai Scholarship of the Hungarian Academy of Sciences and Bolyai+ 2018/2019 (UNKP-18-4 2018/2019 new national excellence program of the ministry of human capacities) also supported KK. JP was also supported by the European Union and the State of Hungary, co-financed by the European Social Fund in the framework of, TAMOP-4.2.2. A-11/1/KON-2012- 0024 and TAMOP-4.2.4.A/2-11/1-2012-0001 National Excellence Program and PTE AOK-KA-2013/22. Further grant support was provided to PB by NKFIH K108308, C120732, TAMOP-4.2.2.A-11/1/KONV-2012-0025. The project was also supported by the UNKP-18-3 2018/2019 new national excellence program of the ministry of human capacities to KG. Research funding was also provided by University of Pecs Biomedical Engineering Project to JP and KK.

#### ACKNOWLEDGMENTS

The authors wish to thank Hajnalka Abraham MD, PhD (Central Electron Microscope Laboratory, University of Pecs, Hungary) for the technical aid in taking transmission electron microscope images. The thymic carcinoma cell line 1889c was kindly provided by Prof Ralf J. Riecker (Institute of Pathology, University Hospital, Heidelberg, Germany). The authors wish to thank Ricky Odedra (Humeltis Ltd) and prof. Mary Keen (University of Birmingham, UK) for improving the manuscript using their native speaker skills. The present scientific contribution is dedicated to the 650th anniversary of the foundation of the University of Pecs, Hungary.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo. 2019.00369/full#supplementary-material


depending on its anatomical peri-atrial, peri-ventricular, or peri-coronary location. Cardiovasc Res. (2015) 108:62–73. doi: 10.1093/cvr/cvv208


**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 Banfai, Ernszt, Pap, Bai, Garai, Belharazem, Pongracz and Kvell. 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.

# Regulation of Adaptive Immune Cells by Sirtuins

#### Jonathan L. Warren<sup>1</sup> and Nancie J. MacIver 1,2,3 \*

*<sup>1</sup> Department of Pediatrics, Duke University School of Medicine, Durham, NC, United States, <sup>2</sup> Department of Immunology, Duke University School of Medicine, Durham, NC, United States, <sup>3</sup> Department of Pharmacology and Cancer Biology, Duke University School of Medicine, Durham, NC, United States*

It is now well-established that the pathways that control lymphocyte metabolism and function are intimately linked, and changes in lymphocyte metabolism can influence and direct cellular function. Interestingly, a number of recent advances indicate that lymphocyte identity and metabolism is partially controlled via epigenetic regulation. Epigenetic mechanisms, such as changes in DNA methylation or histone acetylation, have been found to alter immune function and play a role in numerous chronic disease states. There are several enzymes that can mediate epigenetic changes; of particular interest are sirtuins, protein deacetylases that mediate adaptive responses to a variety of stresses (including calorie restriction and metabolic stress) and are now understood to play a significant role in immunity. This review will focus on recent advances in the understanding of how sirtuins affect the adaptive immune system. These pathways are of significant interest as therapeutic targets for the treatment of autoimmunity, cancer, and transplant tolerance.

## Edited by:

*Jie Chen, Xiamen University, China*

#### Reviewed by:

*Terrence Deak, Binghamton University, United States Xiaoqiang Tang, Sichuan University, China*

> \*Correspondence: *Nancie J. MacIver nancie.maciver@duke.edu*

#### Specialty section:

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

Received: *02 February 2019* Accepted: *26 June 2019* Published: *11 July 2019*

#### Citation:

*Warren JL and MacIver NJ (2019) Regulation of Adaptive Immune Cells by Sirtuins. Front. Endocrinol. 10:466. doi: 10.3389/fendo.2019.00466* Keywords: adaptive immunity, T cells, epigenetics, sirtuins, metabolism

## INTRODUCTION

The adaptive immune system is critical for responding to and eliminating foreign pathogens. T cells are important members of the adaptive immune system, and are generally responsible for recruiting additional inflammatory machinery to the site of infection or tumor. T cells develop within the thymus and, upon maturation, are classified broadly by their expression of either CD4 or CD8 receptor. Both CD4<sup>+</sup> and CD8<sup>+</sup> T cells exist as a number of subsets that perform unique functions within the immune milieu and exhibit unique surface receptors, produce lineage-specific cytokines, and express lineage-defining transcription factors. CD4<sup>+</sup> T effector cells (Teff) are a broad class of T cells that are further divided into unique subsets with distinct effector functions. T helper 1 (Th1) cells are necessary for combating intracellular bacteria and viruses and for producing cytokines (most notably IFNγ and IL-2, but also TNFα) to promote cellular immunity, macrophage activation, and phagocytosis (1). Th2 cells are important for responses to helminthic and other gastrointestinal parasitic infections, producing IL-4, IL-5, and IL-10, and stimulating B cell differentiation (2). Th9 cells are a relatively newly defined subset, and are known to primarily produce IL-9 and facilitate the immune response against intestinal worms (3). Th17 cells are broadly involved in inflammation as well as host response to infection, producing IL-17, IL-6, and TNFα to recruit additional immune cell types to the site (4). In contrast with the pro-inflammatory nature of Teff cells, regulatory T (Treg) cells are responsible for immunosuppression, preventing overactive inflammatory responses and autoimmunity (5). This subset secretes key anti-inflammatory cytokines, notably TGFβ and IL-10, and Treg differentiation is driven by the

Warren and MacIver Adaptive Immunity and Sirtuins

transcription factor Foxp3. CD8<sup>+</sup> cytotoxic T cells are primarily responsible for killing infected or malignant cells through the release of cytotoxic cytokines (TNFα, IFNγ) and granules (perforin, granzymes), and by initiating apoptotic processes mediated by the caspase cascade (6). Lastly, B cells are lymphocytes derived from bone marrow and are also major players within the adaptive immune system, supporting humoral immunity upon activation by producing large quantities of antibodies (7).

The metabolic profile of each of these specialized T cell subsets is optimized to support their unique functions (8). For example, activated CD4<sup>+</sup> Teff cells, including Th1, Th2, Th17, and CD8<sup>+</sup> cytotoxic T cells upregulate glucose uptake and glycolysis to promote rapid growth, proliferation, and effector function. Teff cells also rely on increased glutamine uptake and metabolism to support cell growth and proliferation, although the requirement for glutamine metabolism varies among T cell subsets (9). In contrast, Treg cells rely primarily on lipid oxidation to support their suppressive activity. While considerably less is known about B cell metabolism relative to T cell metabolism, there are some similarities such that naïve B cells are relatively quiescent, but following stimulation, have increased metabolic demand, likely to support proliferation and antibody production (10). Additionally, B cell subsets tend to display unique metabolic phenotypes as a product of their environment and function (11). Thus, the pathways that control adaptive immune cell function and metabolism are intimately linked (12–14). A number of recent advances indicate that immune cell identity, function, and metabolism are controlled, at least in part, via epigenetic mechanisms.

#### EPIGENETICS

While DNA sequence is the same from cell to cell within an organism, the transcription (or lack thereof) of certain genes contributes greatly to the differentiation of the myriad of cell types that are present within an organism. This variation is controlled in large part by epigenetic mechanisms that result in dynamic but heritable changes in gene expression that do not involve changes in DNA sequence (15). There are several mechanisms by which this kind of transcriptome regulation may occur. One example of these mechanisms is DNA methylation, the addition of a methyl group to cysteine residues within DNA by various DNA methyltransferases typically in regions rich in cysteine-guanine dinucleotides. Generally, DNA methylation can act to either inhibit gene transcription (if methylation occurs within a promoter region) or promote transcription (if methylation occurs within the gene body) (16). Another mechanism of epigenetic regulation occurs via non-coding RNA. These single-stranded RNA fragments seek out complementary sites within the mRNA of target genes and degrade RNA, ultimately preventing translation (17). Lastly, the modification of histones is a highly prevalent mechanism of epigenetic control and contributes to genetic regulation by altering the physical structure of chromatin to improve or impair the accessibility of DNA to various transcription factors and transcription machinery. Histones form the backbone of the nucleosome, providing structure and stability. In contrast with the stability of DNA methylation, histone modifications can be more fluid over acute periods of time (18). Deacetylated histones form a densely packed chromatin structure, known as heterochromatin, physically preventing transcription. Histone acetylation maintains a more loose and fluid chromatin structure. Other common histone modifications can occur by methylation, phosphorylation, and deamination. Moreover, all of these mechanisms likely work synergistically to regulate the epigenome (19).

## SIRTUINS

Among the four defined classes of histone deacetylases, sirtuins (class III) are a unique family of highly conserved, NAD+ dependent protein deacetylases with important implications on the epigenome. In addition to their deacetylase activity, sirtuins display some additional enzymatic function on other substrates, including ADP-ribosyltransferase and desuccinylase activity (20, 21). Mammalian sirtuins are orthologs of the silent information regulatory 2 (Sir2) protein, which was first identified in yeast as a significant contributor to the life-span extending effects of calorie restriction (22). These effects were further observed in C. elegans and Drosophila (23), suggesting that this pathway is conserved across species. Given the NAD+-dependent activity of sirtuins, they are activated in periods of catabolism and low nutrient availability, and were thus thought to be a novel target for mimicking the life-span extending effects of calorie restriction in humans.

Mammals ubiquitously express seven sirtuins (SIRT1-SIRT7) with different subcellular locations and functions. Sirtuins are currently gaining widespread attention in the context of a number of disease states associated with inflammation, including autoimmunity (24), cardiometabolic diseases (25), and cancers (26). Further, the mammalian sirtuins have been found to mediate cellular metabolism and adaptive responses to a variety of stresses, including calorie restriction and other metabolic stress (27). Given the relationship between nutrient availability and the function of the adaptive immune system (28), sirtuins are currently of great interest as mediators of tumor proliferation, autoimmunity, and the ability of an organism to respond to foreign pathogens.

The idea that sirtuins might be involved in immunity was posited over a decade ago (23), following the early finding that SIRT1 can regulate NF-κB (29), a transcription factor well-known to regulate inflammation and immune cell proliferation (30). Indeed, further study has begun to elucidate the link between this family of proteins and immune cell function. While each of the sirtuins has since been broadly studied, the most intense attention has been given to SIRT1 (primarily localized to the nucleus) and SIRT3 (primarily in mitochondria) with respect to adaptive immune cells, and will thus be the focus of this review (**Figure 1**). Further, considerations for the use of sirtuin-modifying drugs to manipulate immune activity are explored.

## EFFECTOR CD4<sup>+</sup> T CELLS

Activation of T cells occurs seconds after stimulation of the T cell antigen receptor (TCR) by a ligand, along with co-stimulatory signals, which initiate a number of signaling pathways that promote differentiation and growth. This activation is supported by a transition from a relatively quiescent oxidative metabolism to an intense glycolytic metabolic signature to support proliferation and cytokine production (12). This metabolic switch is likely driven, at least in part, by sirtuin activity, however there presently appears to be a number of incongruous effects reported across the various CD4<sup>+</sup> T cell subsets.

SIRT1 appears to play a significant role in the regulation of Teff cell activation (31, 32). Early studies into the function of SIRT1 on Teff cells indicated that SIRT1 inhibits the immune response by acting as an antagonist against transcription factors that support IL-2 production (32) and thereby decreasing Th1 cell activation. This relationship may contribute in part to the link between fasting/calorie restriction and poor immune performance (28), given that SIRT1 is generally activated in response to fasting (33). The direct role of sirtuins within adaptive immune cells has also begun to be studied using knockout (KO) animal models. T cells from a SIRT1 KO mouse model are more proliferative, produce more IL-2 both in vitro and in vivo, and these mice are more susceptible to experimental autoimmune encephalomyelitis (EAE), indicating a more inflammatory immune phenotype (32). Studies in SIRT1 KO animals also found that T cells without SIRT1 can be activated solely via the T cell receptor (TCR), without co-stimulation by CD28, suggesting a hypersensitivity to activation signals when SIRT1 is not present (32). Follow-up studies indicated that IL-2 is involved in a feedback response to reduce further SIRT1 gene transcription and allow for proliferation in response to the activation cascade (34). Hyper-responsive Teff cells can contribute to an environment prone to autoimmune disease. In fact, earlier studies in SIRT1-null mice detailed the development of a mild autoimmune condition that resembled systemic lupus erythematosus, characterized by deposition of immune complexes within liver and kidneys, with some mice going on to spontaneously develop a diabetes insipidus-like autoimmune disorder after 2 years of age, altogether suggesting a preventative role of SIRT1 in autoimmunity (35).

SIRT1 also inhibits Bcl-2 Associated Transcription Factor 1 (Bclaf1) (36). Bclaf1 was originally identified as a promoter of apoptosis (37); however, subsequent studies revealed further reaching effects of Bclaf1 on T cell development, activation, and proliferation (38), perhaps by promoting hypoxia-inducible factor 1-α (HIF-1α) transcription (39). SIRT1-mediated inhibition of Bclaf1 is thought to occur by the binding of SIRT1 to the promoter region of Bclaf1 after stimulation of the TCR, suppressing acetylation of the histone 3 lysine 56 residue (H3K56) (36). When SIRT1 was knocked out of T cells, there was greater expression of the gene coding for Bclaf1, and specific knockdown of Bclaf was able to suppress the increase in IL-2 production and proliferation seen in SIRT1 KO mice (36).

SIRT1 inhibition can also depress the adaptive response and differentiation of Th2 cells. Pharmacological SIRT1 inhibition contributed to decreased allergic inflammation in BALB/c mice exposed to ovalbumin via aerosol (40). In addition, mice that exhibit KO of a transcriptional activator essential for Th2 differentiation (B-cell lymphoma/leukemia 11B; Bcl11b) have been found to be protected against EAE (41). Bcl11b is a transcriptional repressor and likely functions by recruiting SIRT1 for histone deacetylase activity (42). Although T helper 9 (Th9) cells exhibit a number of similarities to Th2 cells, SIRT1 inhibition has been found to promote Th9 cell differentiation and IL-9 production by these cells (43).

The implications of SIRT1 on Th17 cells have been more equivocal. Pharmacological induction of SIRT1 using resveratrol, low dose metformin, or the inhibitor SRT1720 has been shown to impair Th17 cell differentiation with decreased expression of IL-17 and RORγt, in a STAT3-dependent manner (44). The same study further describes anti-tumor effects of metformin by its action in reducing Th17 differentiation and STAT3 acetylation. In a separate study, in vivo activation of SIRT1 through treatment with NAD<sup>+</sup> contributed to a delayed onset of EAE. This protection was hypothesized to be conferred by enhanced SIRT1 expression within the spinal cord of mice exposed to the EAE stimulus, which may suppress inflammatory responses by Th1 and Th17 cells (45). However, others have shown that SIRT1 is necessary for the production of pro-inflammatory Th17 cells through the deacetylation of transcription factor RORγt, which suggests that SIRT1 inhibitors could confer protection against autoimmunity (46). Clearly, more studies are needed to dissect out the role of SIRT1 in Th17 cell differentiation, proliferation, and cytokine response.

SIRT3 is a mitochondrial sirtuin that supports the structure, function, and biogenesis of the mitochondria (47). SIRT3 is elevated in fasting and calorie restriction in liver, muscle, and brown adipose tissue, and is known to modify cellular metabolism in those tissues (48, 49). In a model of experimental allogenic bone marrow transplantation, total T cells from donor animals that exhibit a whole-body SIRT3 KO were less likely to promote graft-vs.-host disease relative to T cells from control mice, but did not affect the graft-vs.-tumor effect, suggesting that targeted inhibition of SIRT3 in allogenic T cells can improve outcomes after transplant (50). Further, while SIRT3 KO did not affect the composition of peripheral naïve T cell subsets, it was determined that SIRT3 KO Teff cells were less proliferative and produced less reactive oxygen species (ROS) in response to non-specific TCR stimulation (50). However, a SIRT3 KO mouse model did not affect the development of immune cells or immune responses to various endotoxins (51), suggesting SIRT3 may play a limited role in Teff cell function.

Little is known about the role of the other sirtuins on Teff cell development and function. SIRT6 may be a negative regulator of glycolytic activity, notably through the inhibition of glucose transporter 1 (GLUT1) and the transcriptional regulator HIF-1α (52), suggesting a potential role for SIRT6 in downregulating Teff cell activation. HIF-1α is a transcriptional regulator of glycolysis and is known to regulate the production of a number of cytokines (53) and enhance Th17 cell differentiation (54). Relatively little work has been done on SIRT2 in adaptive immunity. However, SIRT2 has been identified as a potential suppressor of colitis through its deacetylase activity on NF-κB within bone marrow-derived macrophages in a mouse model (55). Further, this study observed a greater proportion of activated (CD4+CD69+) T cell populations at the mesenteric lymph nodes of SIRT2 KO mice in response to DSS-induced colitis, indicative of enhanced inflammatory action (55), and ascribing a role for SIRT2, similar to SIRT1, in limiting CD4<sup>+</sup> Teff cell inflammation.

## REGULATORY CD4<sup>+</sup> T CELLS

SIRT1 has been found to reduce the activity of Foxp3, contributing to an overall more inflammatory immune phenotype (46). Further, inhibiting SIRT1 can promote greater Treg suppressive activity (56, 57). While the regulation of Treg metabolism by sirtuins has not been widely studied, Treg function has been shown to be regulated in part by sirtuin activity. In addition, Foxp3 itself has been found to be a regulator of epigenetic activity to support the Treg phenotype, and is regulated in part by the deacetylase activity of SIRT1 (56). A previous review has outlined the role of demethylation and histone modifications that occur in order to promote and stabilize the expression of Foxp3 during Treg cell development (58). Briefly, three conserved non-coding sequences are primary targets for epigenetic mechanisms that regulate Foxp3 expression in response to external environmental stimulus.

Given the localization of SIRT3 to the mitochondria and its role in oxidative metabolism and mitochondrial function, it is not surprising that the loss of SIRT3 in Treg cells has been shown to impair their suppressive activity (59). Indeed, deletion of histone deacetylase 9 was found to be sufficient to increase Treg suppressive activity, by increasing the expression of SIRT3. Further, Treg cells from SIRT3 KO mice had impaired suppressive function both in an in vitro suppression assay and an in vivo cardiac allograft model, likely due to the role of SIRT3 in promoting oxidative metabolism (59).

## CD8<sup>+</sup> T CELLS

Activation and differentiation of CD8<sup>+</sup> T cells leads to markedly variable chromatin accessibility (60), which is likely critical to facilitate the transition between naïve, effector, and memory CD8<sup>+</sup> T cells. SIRT1 appears to play a crucial role in CD8<sup>+</sup> T cell differentiation. Basic leucine zipper ATF-like transcription factor (BATF) has been shown to inhibit the expression of SIRT1, contributing to increased histone acetylation, particularly at the T-bet locus (61). This has been shown to affect CD8<sup>+</sup> T cell differentiation and activity, as CD8<sup>+</sup> T cells from BATF KO animals exhibited lower ATP production and lower mRNA expression of perforin and IFNγ (61). SIRT3 is also involved in CD8<sup>+</sup> T cell function. Toubai et al. found that SIRT3-null activated CD8<sup>+</sup> T cells produced less ROS upon activation (50). This impairment in ROS production may lead to impairments in sulfenylation, a process known to play a role in the regulation of histone deacetylases (62). Further SIRT3-null donor cells were able to attenuate graft-vs.-host disease within the gastrointestinal tract and the authors hypothesize this may be due, in part, to decreased CD8<sup>+</sup> T cell trafficking to site (50).

## MEMORY T CELLS

Following the primary immune response, a portion of T cells (CD4<sup>+</sup> and CD8+) can become memory cells that remain ready to respond rapidly in the event that they re-encounter their antigen. These cells are relatively long-lived and therefore exhibit a relatively quiescent oxidative metabolism similar to that of a naïve immune cell until they are re-activated. Little is known about the role of sirtuins in mediating the generation or longevity of memory T cells; however, given the role of sirtuins in promoting oxidative metabolism, this is a potentially interesting area for further study. Though not specific to memory T cells, SIRT1 has distinct effects on PGC-1α and PGC-1β, both proteins with roles in facilitating mitochondrial biogenesis and oxidative metabolism. Thus, the return to a more oxidative metabolism in memory cells may be mediated in part by SIRT1, given the role of SIRT1-mediated deacetylation on transcription of PGC-1α and PGC-1β (63, 64). In support of this hypothesis, SIRT1 expression has been shown to be decreased in terminally differentiated CD8+CD28<sup>−</sup> memory T cells, driving the downregulation of forkhead box protein O1 (FoxO1), a transcription factor that mediates T cell homing and differentiation (65). Further, these authors demonstrate that these SIRT1-low CD8+CD28<sup>−</sup> memory cells have an enhanced glycolytic capacity in the resting state, which can support effector function upon reactivation.

## B CELLS

Naïve B cells exhibit a relatively inert epigenetic status with high levels of DNA methylation and histone deacetylation. However, upon activation and maturation toward a germinal center B cell phenotype, B cells exhibit dramatic shifts in methylation status and become hypomethylated with increased histone acetylation and expression of various miRNAs (66, 67). Thus far, the limited literature on sirtuin activity in B cells indicates sirtuins support B cell viability, proliferation, and function. SIRT1 overexpression by viral transfection in BaF3 B cells (a murine B cell line) has been shown to support enhanced viability (mediated in part by a decrease in p53) and increased cytokine production (68). A short, non-coding microRNA, miR-132, is increased in B cells of patients with multiple sclerosis (MS) concurrent with a reduced expression of SIRT1 (69). SIRT3 also has been found to be a tumor suppressor in the context of B cell malignancies, as a number of malignant B cell lines display decreased SIRT3 protein expression and higher ROS levels, and overexpression of SIRT3 in these lines decreased proliferative activity (70). SIRT4 has also been shown to act as a tumor suppressor by inhibiting glutamine metabolism, which is necessary to conserve resources for repairing DNA damage (71). Further, SIRT4 overexpression can inhibit proliferation of Burkitt lymphoma cells, a model of B cell lymphoma (72). Inhibition of SIRT1 and SIRT2 increased apoptotic activity and ROS production in cells from patients with B cell chronic lymphotic leukemia (73). While not specific to sirtuins, histone deacetylase inhibitors have been shown to be effective in preventing B cell malignancies (74).

## MODULATING SIRTUIN ACTIVITY TO ALTER IMMUNE OUTCOMES IN VIVO

A number of novel drugs with sirtuin modulatory activity have been studied in the context of immune function. However, it will be critical to determine whether to induce or inhibit sirtuin activity and, further, how to target specific sirtuins (perhaps even within a particular lymphocyte subset), in order to reach desired immune outcomes. For instance, promoting SIRT3 activity in Treg cells to improve suppressive capabilities and temper inflammation could be a novel means to treat autoimmunity. On the other hand, SIRT1 generally inhibits Teff inflammation, suggesting that activators of SIRT1 could be useful for the treatment of autoimmune disease; however, the effects of SIRT1 on T cells vary by subset and are context-dependent. As sirtuins are proteins with functions in a wide array of cell types, targeting specific sirtuins in specific tissues (and immune cell subsets) will remain an immense challenge.

Despite the challenges with tissue-specificity, drugs to modify sirtuin activity have been studied in cell culture and animal models. Inhibition of SIRT1 in vivo (using a SIRT1-specific inhibitor, EX-527) increased complications of sepsis at 12 h despite conferring dramatic protection at 24 h (75). EX-527 and sirtinol (another commonly studied sirtuin inhibitor with specificity against SIRT1 and SIRT2) also have been found to reduce platelet count (76). Metformin, classically prescribed as a medication for the management of type 2 diabetes, has been well-documented to have anticancer effects (77); however, the precise mechanism of action is largely unknown, though there is speculation that these effects may be due to the stimulatory effect of metformin on sirtuins. SIRT1 activation by metformin has been shown to decrease Th17 cell populations promoting a less inflammatory environment (44). SIRT1 knockdown has also been shown to promote apoptotic processes in leukemia cells (78), suggesting the exact mechanisms for sirtuin modulation in cancers is still being determined.

Resveratrol, a polyphenol compound with anti-inflammatory properties, is a compound with sirtuin modifying effects that is currently of intense interest within both the scientific and lay communities. Resveratrol increases SIRT1 activity and impedes acetylation of c-Jun, thereby limiting T cell activation (79). Further, resveratrol has been shown to improve outcomes in two well-characterized murine models of autoimmunity: EAE (80) and colitis (81). Additionally, resveratrol confers protection against a murine model of rheumatoid arthritis, by inhibiting Th17 expansion and IL-17 production, as well as autoantibody production from B cells (82). Resveratrol has also been shown to increase the ratio of CD4<sup>+</sup> to CD8<sup>+</sup> T cells and increase total Treg cells in the context of diet-induced obesity in mice while also conferring benefits on glucose homeostasis by activating phosphoinositide 3-kinase (PI3K) signaling pathways (83). Dosages necessary to produce these effects in humans are likely impossible to obtain exclusively through diet, but could realistically be obtained through supplementation.

The role of sirtuins in promoting organ transplant tolerance is also an area of intense investigation. While advances have been made in long-term survival following transplantation, current immunosuppressive therapies are known to promote infections and cancer. There is speculation that SIRT1 inhibitors may enhance the function of Treg cells to support immune suppression and allograft tolerance (84). Additionally, SIRT1 inhibitors may also confer prolonged allograft survival through the suppression of Th17 activity, as evidenced by decreased IL-17A (85); however, these results are in direct opposition to the SIRT1-activating and anti-tumor properties of metformin described above (44), and further studies are needed.

## CONCLUDING REMARKS

The wide-ranging effects of sirtuins and the availability of a number of sirtuin-modifying compounds provide a significant opportunity for future study to improve immune cell phenotypes. However, there are significant challenges ahead in developing drugs with targeted tissue-specific effects given the ubiquity of these mechanisms within the body. The development of tissue-specific and sirtuin-specific therapies remains an intriguing possibility to treat the myriad of autoimmune diseases, cancers, and other chronic diseases associated with inflammation that are now understood to be regulated by some degree of protein acetylation.

## AUTHOR CONTRIBUTIONS

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

### REFERENCES


#### FUNDING

This work was supported by the National Institutes of Health (R01-DK106090) and the Derfner Foundation.


of high-fat diet-induced obesity in C57BL/6 mice. Food Funct. (2014) 5:1452–63. doi: 10.1039/C3FO60714C


**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 Warren and MacIver. 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 Association of Hypoglycemia Assessed by Continuous Glucose Monitoring With Cardiovascular Outcomes and Mortality in Patients With Type 2 Diabetes

Wei Wei <sup>1</sup> , Shi Zhao<sup>1</sup> \*, Sha-li Fu<sup>1</sup> , Lan Yi <sup>1</sup> , Hong Mao<sup>1</sup> , Qin Tan<sup>2</sup> , Pan Xu<sup>3</sup> and Guo-liang Yang<sup>4</sup>

*<sup>1</sup> Department of Endocrinology, Tongji Medical College, The Central Hospital of Wuhan, Huazhong University of Science and Technology, Wuhan, China, <sup>2</sup> Department of Cardiology, Tongji Medical College, The Central Hospital of Wuhan, Huazhong University of Science and Technology, Wuhan, China, <sup>3</sup> Department of Neurology, Tongji Medical College, The Central Hospital of Wuhan, Huazhong University of Science and Technology, Wuhan, China, <sup>4</sup> Department of Information, Tongji Medical College, The Central Hospital of Wuhan, Huazhong University of Science and Technology, Wuhan, China*

*Xiamen University, China*

Edited by: *Jie Chen,*

#### Reviewed by:

*Youlian Wang, Jiangxi Provincial People's Hospital, China Edward Narayan, Western Sydney University, Australia*

> \*Correspondence: *Shi Zhao zhaoshiwuhan@126.com*

#### Specialty section:

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

Received: *23 April 2019* Accepted: *18 July 2019* Published: *06 August 2019*

#### Citation:

*Wei W, Zhao S, Fu S, Yi L, Mao H, Tan Q, Xu P and Yang G (2019) The Association of Hypoglycemia Assessed by Continuous Glucose Monitoring With Cardiovascular Outcomes and Mortality in Patients With Type 2 Diabetes. Front. Endocrinol. 10:536. doi: 10.3389/fendo.2019.00536* Objective: Hypoglycemia has been shown to promote inflammation, a common pathogenic process, in many chronic health conditions including diabetes and cardiovascular disease. The aim of this study was to investigate the association of hypoglycemia, assessed by continuous glucose monitoring (CGM) with major adverse cardiovascular event (MACE) outcomes and all-cause mortality.

Methods: A retrospective cohort study was conducted with 1,520 patients with type 2 diabetes mellitus (T2DM). The severity of hypoglycemia event was assessed by CGM system.

Results: Three hundred and forty-seven participants experienced hypoglycemia events (323 with mild hypoglycemia and 24 with severe hypoglycemia). A fraction of 72.62% hypoglycemia was asymptomatic. During a median follow-up of 31 months, 380 participants reached the primary outcome of MACE (61 cardiovascular death, 50 non-fatal myocardial infarction [MI], 116 non-fatal stroke, 153 unstable angina requiring hospitalization), 80 participants died before the end of the study. In multivariate Cox regression models, hypoglycemia was associated with cardiovascular death (HR 2.642[95CI% 1.398–4.994]), non-fatal stroke (HR 1.813 [95CI% 1.110–2.960]) and all-cause mortality (HR 1.960 [95 CI% 1.124- 3.418]) after the full adjustment. Hypoglycemia was not associated with non-fatal MI and unstable angina. The HR of severe hypoglycemia was higher than mild hypoglycemia for cardiovascular death. Patients with symptomatic and asymptomatic hypoglycemia had similar MACE outcomes and all-cause mortality.

Conclusions: CGM is effective to detect asymptomatic and nocturnal hypoglycemia. Hypoglycemia is associated with an increased risk of non-fatal stroke, cardiovascular related death, and total mortality. The cardiovascular mortality is dose-dependent on the severity of hypoglycemia.

Keywords: T2DM, hypoglycemia, CGM system, MACE, all-cause mortality

## INTRODUCTION

Chronic low grade inflammation is a common feature underlying many chronic diseases and conditions such as insulin resistance and cardiovascular disease. Studies have suggested that hypoglycemia is able to promote inflammatory processes in patients with or without diabetes (1, 2). Hypoglycemia is a frequent adverse effect of anti-diabetic therapy in diabetic patients, and severe hypoglycemia has been identified as a potential risk factor for cardiovascular events in patients with type 2 diabetes mellitus (T2DM). Previous studies have reported an increased hazard ratio (HR) for adverse cardiovascular outcomes and total mortality in diabetic patients with hypoglycemia (3–6). However, these conclusions were drawn from the secondary analyses of randomized clinical trials (7, 8) or retrospective analyses of medical claims databases (9, 10), which a clear definition of hypoglycemia was lacking. Continuous glucose monitoring (CGM) system is an effective tool for the assessment of hypoglycemic status, particularly for the asymptomatic nocturnal hypoglycemia (11). Using CGM system, researchers found out that hypoglycemia is more common in T2DM patients than previously thought (12), and the frequency of hypoglycemia in those previous studies may be underestimated.

In the present study, using glucose data collected from CGMs, we investigated the relationships of hypoglycemia with total mortality and major adverse cardiovascular events (MACE), including non-fatal stroke, non-fatal myocardial infarction (MI), unstable angina leading to hospitalization, and cardiovascular death.

## MATERIALS AND METHODS

#### Subjects

In this retrospective cohort study, we recruited 1,520 patients admitted to our hospital between January 2013 and December 2017.

The inclusion criteria were: (1) T2DM according to 2013 American Diabetes Association standards (13). Patients with other types of diabetes, such as type 1 diabetes, gestational diabetes, and secondary diabetes were excluded from the study. (2) Patients without the acute phase of illness, such as acute coronary syndrome, uncontrolled infection, etc. (3) CGM (iPro <sup>R</sup> 2 CGM System Gold, Medtronic MiniMed, Inc., Northridge, CA) was used to monitor blood glucose within 3 days of admission without changing medications before the patients were discharged from hospital.

The CGM devices used in this study continuously measured the glucose level in the interstitial fluids within the range of 2.2– 22.2 mmol/L (40–400 mg/dL). The glucose level was determined every 5 min, 288 times maximum per day. The CGM system was calibrated with blood glucometer measurements (ACCU-CHEK <sup>R</sup> Aviva, Roche, Mannheim, Germany) four times daily according to the manufacturer's instruction. Participants were instructed to keep a diary about the occurrence of hypoglycemic symptoms. The CGM system used in our study measured interstitial glucose in a blinded manner, and data analysis was performed after disconnection of the device.

This clinical study was approved by the ethics committee boad of The Central Hospital of Wuhan. All participants had signed the informed consent form during the enrollment and before the start the study.

## Clinical Data Collection

A clinical data warehouse was created as a collaborative program between Wuhan Central Hospital (Wuhan, China) and Shanghai Lejiu Healthcare Technology Co., Ltd. Data were extracted every 24 h by Lex Clinical Data Application 3.2 (Shanghai Lejiu Healthcare Technology Co., Ltd) from the Hospital Health Information System (HIS) to a designated clinical data warehouse including admission/transfer/discharge, laboratory orders/results, medication orders, discharge summary, administration events, flow sheet entries, procedures, medical reports, etc. All original unstructured data (i.e., pathology report, radiology report, admit/discharge summary etc.) were exclusively converted to a uniformed structured format. Core elements of the data warehouse were completely de-identified so that all queries and analytics could be carried out without exposing the confidential health data, allowing the investigators with sufficient privilege to re-identify data. Lex Clinical Data Application 3.2 was a self-service data access tool designed to query the clinical data warehouse and return tabular data for analysis and visualization.

#### Definition of Hypoglycemia

A hypoglycemia event was defined as an interstitial glucose level below 3.9 mmol/L (70 mg/dl) for at least 15 min and recovery when the interstitial glucose concentration had been continuously above the threshold for 15 min or more (11). Percent time at interstitial glucose level below 3.9 mmol/L was evaluated. A severe hypoglycemia event was defined as cognitive impairment requiring external assistance for recovery. A mild hypoglycemia event was defined as a glucose level below 3.9 mmol/L without cognitive impairment and external assistance for recovery. Nocturnal hypoglycemia was defined as an episode occurring between 00:00 a.m. and 06:00 a.m.

#### Follow-Ups and Outcomes

After discharging from the hospital, participants were invited to join the out-patient blood glucose management system, which was run by the professional medical staff. Routine self-blood glucose (at least four times per week) and HbA1c (every 3 month) monitoring were demanded in order to know the glucose control condition and clinical medication. The median duration of follow-ups was 31 months (inter-quartile range, 22–56). Eighteen patients (1.18%) with new episodes of severe hypoglycemic events during follow-up were excluded from the study, and 41 (2.70%) patients were lost before follow-ups due to various reasons.

The primary outcome was the first occurrence of an adjudicated MACE, including non-fatal MI, non-fatal stroke, cardiovascular death, and unstable angina leading to hospitalization. The secondary outcome was death of any cause. The diagnoses of MACE outcomes were ascertained according to the hospitalization records, discharge summary and certification of death, which were adjudicated by an independent committee. The members of the committee were from the cardiovascular and neurology departments of our hospital, who were unaware of the CGM results. Follow-up time was calculated from the date of hypoglycemia event to the onset date of the MACE event, death, or end of study (31 August 2018). Cause of death was classified as cardiovascular death and all other causes of death.

#### Statistical Analysis

The differences between groups were compared using t-test for continuous variables and Chi-square test for categorical data. Cox proportional models were used to evaluate the association between hypoglycemia and either MACE or all-cause mortality. We progressively adjusted the models for potential confounders. Model 1 was a crude model. Model 2 included age, sex, eGFR, HbA1c, BMI, and duration of diabetes. Model 3 included all variables in model 2 plus smoking status, alcohol history, past medical history (hepatic disease, renal disease, malignancy, coronary heart disease, and stroke), all diabetic medications (insulin, sulfonylureas, metformin, alpha-glucosidase inhibitors, pioglitazone, glinides, and DPP-4 inhibitors), hypertension medication, lipid-lowering medication, and antiplatelet agents.

TABLE 1 | Baseline characteristics of the study population (*n* = 1,520).


*Dates are presented as numbers (percentages). MACE, major adverse cardiovascular event.*

TABLE 2 | Clinical characteristics of participants by the occurrence of hypoglycemia.


*Continuous variables are shown as mean* ± *SD. Categorical data are presented as numbers (percentages).*

*CGM, continuous glucose monitoring; SD, standard deviation; FPG, fasting plasma glucose; BMI, body mass index; GFR, glomerular filtration rate; TG, triglyceride; TC, total cholesterol; LDL, low density lipoprotein; HDL, high density lipoprotein; DPP-4, dipeptidylpeptidse 4.*

For further analysis, we evaluated the risk of MACE outcomes and total mortality according to the severity of hypoglycemia and the appearance of hypoglycemic symptoms. The severity of hypoglycemia was classified into three categories: no hypoglycemia, mild hypoglycemia, and severe hypoglycemia. The group of no hypoglycemia served as the reference group, and adjusted for models described before. The survival curves of the three groups were estimate by Kaplan-Meier method, and the homogeneity between survival curves was tested by log-rank test.

All analyses were performed using State software (version 13.1, Stata Corp, College Station, TX). P < 0.05 was considered as statistical significance.

#### RESULT

#### Characteristics of Study Population

Baseline characteristics of the study population were presented in **Table 1**. The total number of patients included in the study was 1,520, with 347 (22.83%) patients experiencing hypoglycemic events. A total of 1,028 hypoglycemic events were recorded, corresponding to 250 h in hypoglycemia status. Of all the hypoglycemia, 24 patients were with severe hypoglycemia and 323 with mild hypoglycemia. The overall fraction of asymptomatic hypoglycemia was 72.62%, and the fraction of nocturnal hypoglycemia was 44.67%.

Compared to patients without hypoglycemia, those who experienced hypoglycemia were significantly older, had a longer duration of diabetes, and a lower eGFR. They were more likely to have experienced hypoglycemia events previously. They also took less anti-diabetic medications of metformin, pioglitazone, and DPP-4 inhibitors; meanwhile they were more often treated with insulin and anti-platelet agents (**Table 2**). Patients with severe hypoglycemia were even older, and more likely to be a smoker. They experienced hypoglycemia earlier than those with mild hypoglycemia (**Supplementary Table 1**).

Patients with hypoglycemia were further divided into two groups according to the appearance of hypoglycemic symptoms. Compared to patients with symptomatic hypoglycemia, patients experiencing asymptomatic hypoglycemia had lower mean glucose of CGM and smaller glycemic variability, were more likely to experience nocturnal hypoglycemia, and had longer periods of hypoglycemic events (**Supplementary Table 2**). The other baseline clinical characteristics of patients with symptomatic hypoglycemia were similar to those with asymptomatic hypoglycemia, except that patients experiencing asymptomatic hypoglycemia had a higher proportion of diabetic peripheral neuropathy (**Supplementary Table 3**).

#### Association Between Hypoglycemia and MACE Outcomes

During the follow-up, 380 diabetic patients had developed MACE (61 cardiovascular deaths, 153 unstable angina requiring hospitalization, 50 non-fatal MI, 116 non-fatal strokes). Eighty patients died before the end of our study. Of the 347 patients with hypoglycemia, the median time between hypoglycemia events and MACE outcomes was 21 (inter-quartile range, 11–38) months.

The crude incidence of MACE outcomes in people with hypoglycemia was higher than the incidence in people without hypoglycemia. These estimated results still remained significant (model 2: HR 1.592, 95%CI, 1.233-2.056; model 3: HR 1.615, 95% CI, 1.239-2.106) after further adjustments for potential confounding factors.

TABLE 3 | Association between hypoglycemia and MACE outcomes and all-cause mortality.


*Hypoglycemia was modeled as a time-dependent exposure.*

*Model 1 was a crude model.*

*Model 2 included age, sex, eGFR, HbA1c, BMI, and duration of diabetes.*

*Model 3 included all variables in model 2 plus smoking status, alcohol history, past medical history (hepatic disease, renal disease, malignancy, coronary heart disease, and stroke), all diabetic medications (insulin, sulfonylureas, metformin, alpha-glucosidase inhibitors, pioglitazone, glinides, and DPP-4 inhibitors), hypertension medication, lipid-lowering medication, and antiplatelet agents.*

*MACE, major adverse cardiovascular event; MI, myocardial infarction.*

Furthermore, we examined the findings by subtypes of MACE outcomes. Compared to patients without hypoglycemia, those with hypoglycemia had a higher rate for non-fatal stroke, cardiovascular death and total mortality (**Table 3**). The associations were still persistent after additional adjustment in model 2 and model 3. In the minimally adjusted models, hypoglycemia was associated with an increased risk of nonfatal MI, which was no longer observed after further adjustment (model 3: HR 1.549, 95%CI 0.768–3.124). No association with hypoglycemia was found for unstable angina requiring hospitalization in any model.

Patients with severe hypoglycemia had a higher risk of cardiovascular death than those with mild hypoglycemia (**Figure 1A**). For subtypes of MACE outcomes, the values of HRs had a trend of rising in the severe hypoglycemia group compared with those in the hypoglycemia group, but the difference did not reach statistical significance (**Figure 2**). Patients with symptomatic and asymptomatic hypoglycemia had similar MACE outcomes and all-cause mortality (**Figures 1B**, **3**).

#### DISCUSSION

The results of our study showed that hypoglycemia events detected by CGM were strongly associated with subsequent MACE outcomes and all-cause mortality. This association persisted after adjustment for a wide range of confounders. Furthermore, the risk of cardiovascular death and all-cause mortality was the highest after severe hypoglycemia events during CGM monitoring, especially in the first year, suggesting that health care providers should pay particular attention to the potential for morbidity and mortality after a severe hypoglycemic event. CGMs also detected a high proportion of asymptomatic hypoglycemic events, which appeared to have a similarly effect on MACE outcomes and all-cause mortality like the symptomatic hypoglycemia.

The Action to Control Cardiovascular Risk in Diabetes (ACCORD) study reported that intensive glycemic control was associated with increased risk of cardiovascular-related death (3, 4). Since the premature closure of the ACCORD study, the hypoglycemia-related cardiovascular adverse outcomes have led to considerable debate. The impacts of hypoglycemia on cardiovascular events in diabetic patients have been evaluated in several large prospective clinical trials, which have different conclusions (5, 6, 14). Also, there were many observational studies, with inconsistent results of the association between hypoglycemia and MACE outcomes (7, 15). A subsequent metaanalysis including 10 studies suggested that severe hypoglycemia was associated with an almost two fold increased risk of cardiovascular events (17). Consistent with parts of those previous studies, the results of our analyses showed that hypoglycemia was associated with cardiovascular-related death and all-cause mortality. However, due to the insufficient cases of non-cardiovascular mortality (n = 19), we were unable to analyze the association between the cause-specific mortality and hypoglycemia like the Trail Comparing Cardiovascular Safety of Insulin Degludec vs. Insulin Glargine in Patients with Type 2

FIGURE 1 | (A) The HR and 95% CIs of MACE outcomes and all-cause mortality according to the severity of hypoglycemia. (B) The HR and 95% CIs of MACE outcomes and all-cause mortality according to the appearance of hypoglycemic symptoms. The severity of hypoglycemia was classified into three categories: no hypoglycemia, mild hypoglycemia, and severe hypoglycemia. The appearance of hypoglycemic symptoms was classified into three categories: no hypoglycemia, asymptomatic hypoglycemia, and symptomatic hypoglycemia. The group of no hypoglycemia served as the reference group, and adjusted for models including age, sex, duration of diabetes, HbA1c, BMI, eGFR, past medical history, diabetes medications, hypertension medication, lipid-lowering medication, and antiplatelet agents. HR, Hazard Ratios; MACE, major adverse cardiovascular event; MI, myocardial infarction.

Diabetes at High Risk of Cardiovascular Events (DEVOTE)study (16). In analysis of other subtypes of MACE outcomes, we can only see the association between hypoglycemia and an increased risk of developing non-fatal stroke. There was only a trend between hypoglycemia and non-fatal MI and unstable angina requiring hospitalization, which contrasted with the previous analyses (10, 14).

cardiovascular death. (F) For all cause-mortality.

This can be explained by the different definitions of hypoglycemia, resulting in variable frequencies of hypoglycemia across studies. For most large epidemiological studies, hypoglycemia cases were collected by self-report (5, 6) or ICD-codes from medical electronic data (9, 10, 14), which may underestimate the prevalence of hypoglycemia. In our study, a fraction of 76.37% hypoglycemic episodes were asymptomatic, which was also supported by earlier studies using CGM system (18, 19). Such a high proportion of asymptomatic hypoglycemia in diabetic patients with reduced awareness is worth our serious attention in clinical management. CGM is an effective way to detect hypoglycemia events, especially nocturnal and asymptomatic hypoglycemia, which could play an important role in reducing hypoglycemia events and is worth promoting in clinical applications.

We found some evidence of a dose-dependent relationship between the severity of hypoglycemia and cardiovascular death and all-cause mortality. Our assumption is that the cardiovascular outcomes of severe hypoglycemia may be worse than that of mild hypoglycemia. A sub-analysis of the ACCORD

study found that the protective effect of recurrent mild hypoglycemia was more pronounced than severe hypoglycemia (20). It is suggested that exposure to mild hypoglycemia may offer better preparation against the adverse cardiovascular outcomes caused by severe hypoglycemia through prior blunting of sympathetic responses (21). In contrast to the findings of our study, several observational studies (22, 23) found out that mild hypoglycemia events have no association with mortality. Differences in methods of defining mild hypoglycemia may contribute to discrepancies. However, given the small number of participants with severe hypoglycemia (n = 24) in our study, these results may be limited by low statistical power.

The clinical management of T2DM emphasizes the importance of glycemic control to reduce the risk of chronic complications associated with diabetes (24). However, a toointensive glucose management therapy also puts patients at increased risk of hypoglycemia, which could be life-threatening. Given the concern that hypoglycemia might be a risk factor for cardiovascular disease, avoiding hypoglycemia remains a significant goal in optimizing glucose control. Individualizing glycemic targets should be considered for people with T2DM who are at high risk for hypoglycemia.

Currently, the standard of care in clinical practice is self-monitoring of capillary blood glucose (SMBG), which only provides a single point of time measurement and often fails to detect nocturnal and asymptomatic hypoglycemia. With the ability to measure glucose levels continuously and reflect glycemic variability, CGM technology is gaining increasing interest in clinical management. Numerous studies using CGM have demonstrated significant improvements in reducing hypoglycemia (25, 26). Future incorporation of CGMs in large clinical trials may provide precise information on the severity of hypoglycemia, as well as the glucose level at the occurrence of a hypoglycemic event.

Our study has two important strengths. First, to our knowledge, this is the first study of applying CGM to reveal the relationship between hypoglycemia and the increased risk of CVD. Previous epidemiologic investigations without an accurate definition of hypoglycemia may underestimate the prevalence of hypoglycemia events. Both the severity and time of hypoglycemia can be collected precisely through the CGM system to make a precise diagnosis of hypoglycemia. Second, we were able to adjust for numerous standardized and high-quality covariates, including the duration of diabetes, personal habits (smoking and drinking), BMI, and kidney function. In most large clinical trials, the ICD-code extracted from electronic medical records may be inaccurate, possibly leading to misclassification of exposure and confounding factors.

There were several limitations in our research. Firstly, the number of events for some outcomes may limit the precision of our estimations. Secondly, our study had a relatively short duration of follow-up with a median time of 31 months, which may limit the power to detect a significant association. Thirdly, the study only recruited participants of T2DM during hospitalization, which may not be generalizable beyond this population. Finally, the retrospective nature of our study precludes the possibility to explain the direct causal effect between hypoglycemia and MACE outcomes. Thus, welldesigned prospective cohort studies with the primary intention are needed to evaluate the association between hypoglycemia and cardiovascular outcomes.

In conclusion, through the analysis of glucose data collected by using CGMs, our results add to the accumulating evidence that hypoglycemia is associated with an increased risk of non-fatal stroke, cardiovascular death, and all-cause mortality. We also revealed a dose-dependent relationship between the severity of hypoglycemia and cardiovascular outcomes. Therefore, effective measurements should be taken to prevent severe hypoglycemia

#### REFERENCES


in patients with T2DM, especially those at high risk of cardiovascular problems.

## DATA AVAILABILITY

The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.

#### ETHICS STATEMENT

This clinical study was approved by the ethics committee broad of The Central Hospital of Wuhan. All participants had signed the informed consent form during the enrollment and before the start the study.

#### AUTHOR CONTRIBUTIONS

WW and SZ conceived and designed the study. WW conducted the statistical analyses and wrote the manuscript. SZ provided guidance for the statistical analysis and made critical revisions to the manuscript for important intellectual content. HM provided guidance for the statistical analysis. GY collaborated with Shanghai Lejiu Healthcare Technology Co., Ltd. and took responsibility for the integrity of the data. QT and PX adjudicated the MACE outcomes and the cause of death. LY and SF took charge of the patients' follow-up.

#### FUNDING

This work was supported by the National Natural Science Foundation of China [No. 81370942] and the project of Regenerative Medicine Clinical Research Center of Hubei Province.

#### ACKNOWLEDGMENTS

The authors thank the participants of this study for their contributions.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo. 2019.00536/full#supplementary-material


diabetes assessed by continuous glucose monitoring. J Clin Endocrinol Metab. (2018) 103: 2329–35. doi: 10.1210/jc.2018-00142


**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 Wei, Zhao, Fu, Yi, Mao, Tan, Xu and Yang. 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.

# Characteristics of the Urinary Microbiome From Patients With Gout: A Prospective Study

Yaogui Ning1,2, Guomei Yang<sup>1</sup> , Yangchun Chen<sup>3</sup> , Xue Zhao<sup>1</sup> , Hongyan Qian<sup>3</sup> , Yuan Liu<sup>3</sup> , Shiju Chen<sup>3</sup> \* and Guixiu Shi <sup>3</sup> \*

*<sup>1</sup> School of Medicine, Xiamen University, Xiamen, China, <sup>2</sup> Department of Intensive Care Unit, The First Affiliated Hospital of Xiamen University, Xiamen, China, <sup>3</sup> Department of Rheumatology and Clinical Immunology, The First Affiliated Hospital of Xiamen University, Xiamen, China*

#### Edited by:

*Wen Kong, Huazhong University of Science and Technology, China*

#### Reviewed by:

*Zhiguang Su, Sichuan University, China Xue Xu, The Affiliated Drum Tower Hospital of Nanjing University Medical School, China*

#### \*Correspondence:

*Shiju Chen shiju@xmu.edu.cn Guixiu Shi gshi@xmu.edu.cn*

#### Specialty section:

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

Received: *16 July 2019* Accepted: *14 April 2020* Published: *20 May 2020*

#### Citation:

*Ning Y, Yang G, Chen Y, Zhao X, Qian H, Liu Y, Chen S and Shi G (2020) Characteristics of the Urinary Microbiome From Patients With Gout: A Prospective Study. Front. Endocrinol. 11:272. doi: 10.3389/fendo.2020.00272* The role of host microbes in the pathogenesis of several diseases has been established, and altered microbiomes have been related to diseases. However, the variability of the urinary microbiome in individuals with gout has not been evaluated to date. Therefore, we conducted the present prospective study to characterize the urinary microbiome and its potential relation to gout. Urine samples from 30 patients with gout and 30 healthy controls were analyzed by Illumina MiSeq sequencing of the 16S rRNA hypervariable regions, and the microbiomes were compared according to alpha-diversity indices, complexity (beta diversity) with principal component analysis, and composition with linear discriminant analysis effect size. The most significantly different taxa at the phylum and genus levels were identified, and their potential as biomarkers for discriminating gout patients was assessed based on receiver operating characteristic (ROC) curve analysis. Compared with the healthy controls, there was a dramatic decrease in microbial richness and diversity in the urine of gout patients. The phylum Firmicutes and its derivatives (Lactobacillus\_iners, Family\_XI, and Finegoldia), the phylum Actinobacteria and its derivatives (unidentified\_Actinobacteria, Corynebacteriales, Corynebacteriale, Corynebacterium\_1, and Corynebacterium\_tuberculostearicum), and the genera *Prevotella* and Corynebacterium\_1 were significantly enriched in the urine of gout patients. ROC analysis indicated that the top five altered microbial genera could be reliable markers for distinguishing gout patients from healthy individuals. These findings demonstrate that there are specific alterations in the microbial diversity of gout patients. Thus, further studies on the causal relationship between gout and the urinary microbiome will offer new prospects for diagnosing, preventing, and treating gout.

Keywords: gout, urine, microbiome, 16S rRNA, high-throughput sequencing, biomarker

#### INTRODUCTION

The human microbiome is increasingly considered an important contributor to both health and diseases (1, 2). Advanced technologies for the sequencing of bacterial 16S rRNA alleles have revealed the existence of culture-independent microbes in areas of the human body previously considered to be microbe-free, including urine (3–6). Indeed, several studies have demonstrated a diverse microbiome in the human mouth, skin, respiratory system, gastrointestinal tract, and urine,

including previously uncharacterized and uncultivated species (7, 8). Moreover, accumulating evidence since the completion of the National Institutes of Health Human Microbiome Project points to an association between dysbiosis of the microbiome and the pathogenesis of multiple diseases, including gastrointestinal diseases, pulmonary diseases, cancer, metabolic diseases, and inflammatory diseases (1, 9, 10). Thus, further comprehensive study of the microbiome characteristics related to diseases can provide new insights into the pathogenic mechanism or inform new strategies for diagnosis and treatment.

Gout is one of the most common auto-inflammatory diseases, which is characterized by elevated levels of serum uric acid (UA) with consequent deposition of urate in and around the joints. The prevalence of gout is around 3.9% in the USA and 2.5% in Europe (11–13). Although the specific pathogenic mechanism of gout remains unclear, disruption of purine metabolism and inflammation regulation has been implicated. Treatment options are also limited, and patients with gout require the use of long-term drugs to decrease the UA level. UA is excreted through the kidneys and intestine, and several studies have indicated the abnormal excretion of UA in gout, which could potentially alter the microbiome. Moreover, the microbiome itself might contribute to the abnormal UA metabolism or inflammation regulation in gout. Thus, understanding the characteristics of the microbiome in gout patients might provide potential new strategies for diagnosis and help elucidate the pathogenic mechanisms.

A previous study demonstrated the presence of intestinal bacterial dysbiosis in gout patients compared with a healthy population (14). However, the characteristics of the microbiome in the urine of gout patients remain unclear. In this study, Illumina MiSeq sequencing was employed to investigate the microbial community in urinary extracts from gout patients and healthy volunteers to explore the urinary microbiome alterations associated with gout. These results can serve as a resource for further research to gain a better understanding of the role of bacterial dysbiosis in the pathogenic mechanism and identify candidate diagnostic biomarkers for gout.

## MATERIALS AND METHODS

#### Study Design and Subject Recruitment

The study cohort consisted of two groups (**Table 1**): the gout group (n = 30) and the healthy group (n = 30). Gout was diagnosed by rheumatologists according to the American College of Rheumatology (ACR)/EULAR 2015 criteria (15). The patients and healthy volunteers were (all males) were recruited from the First Affiliated Hospital of Xiamen University, China, from March to October 2017. The exclusion criteria were follows: (1) any comorbid disorders and (2) taking antibiotics within 1 month prior to enrolment in the study. The study was approved by the Ethics Committee of the First Affiliated Hospital of Xiamen University, and informed consent was obtained from all


*Values are presented as mean* ± *SD. BMI, body mass index; UA, ureic acid; BUN, urea nitrogen; Cr, creatinine.*

participants. Demographic characteristics, including age, body mass index, smoking, alcohol intake, or dietary habits, and laboratory data were recorded for all subjects.

#### Sample Preparation and DNA Extraction

Mid-stream urine samples freshly collected from each individual were immediately frozen at −20◦C and transported to the laboratory with an ice pack. Total bacterial DNA from samples was extracted at Novogene Bioinformatics Technology Co., Ltd. using a TIANGEN kit according to the manufacturer's protocols. The quality of the extracted DNA was determined by 1% agarose gel electrophoresis, and the optical density value at 260/280 nm was measured on a spectrophotometer. According to the concentration, the DNA was diluted to 1 ng/µL using sterile water and stored at −20◦C for Illumina MiSeq sequencing analysis.

#### Polymerase Chain Reaction (PCR) Amplification of the Bacterial 16S rRNA V3–V4 Region and Illumina Pyrosequencing

The V3–V4 hypervariable region of the 16S rRNA gene was amplified from the diluted DNA extracts with the forward primer 341F and reverse primer 806R. All PCRs were carried out in 30 µL reactions with 15 µL of Phusion <sup>R</sup> High-Fidelity PCR Master Mix (New England Biolabs), 0.2µM of forward and reverse primers, and about 10 ng of template DNA. The thermal cycling program consisted of initial denaturation at 98◦C for 1 min, followed by 30 cycles of denaturation at 98◦C for 10 s, annealing at 50◦C for 30 s, and elongation at 72◦C for 30 s, followed by a final extension step at 72◦C for 5 min. The same volume of 1X loading buffer (containing SYBR Green) was mixed with the PCR products and subjected to electrophoresis on a 2% agarose gel for detection. Samples with a bright main strip between 400 and 450 bp were chosen for further analysis.

The PCR products were mixed at equidensity ratios and purified with GeneJET Gel Extraction Kit (Thermo Scientific). Sequencing libraries were generated using TruSeq <sup>R</sup> DNA PCR-Free Sample Preparation Kit according to the manufacturer's recommendations, and index codes were added. The library quality was assessed on a Qubit <sup>R</sup> 2.0 fluorometer (Thermo Scientific) and an Agilent Bioanalyzer 2100 system. Finally, the

**Abbreviations:** UA, ureic acid; ROC, receiver operating characteristic; LDA, linear discriminant analysis; LEfSe, linear discriminant analysis effect size; PCoA, principal coordinates analysis; AUC, area under the curve.

library was sequenced on an Illumina HiSeq 2500 system, and 250-bp paired-end reads were generated.

#### Statistics

#### Clinical Data Analysis

Quantitative demographic data with a normal distribution are expressed as mean ± standard deviation, and the t-test was used for comparisons between the two groups. All statistical tests were two-sided, and P < 0.05 was regarded as statistically significant. Statistical analyses were performed using SPSS 19 (SPSS, Chicago, IL, USA).

#### Bioinformatics Analysis

Alpha diversity was used to analyze the complexity of species diversity for a sample through six indices: observed-species, Chao1, Shannon, Simpson, ACE, and Good's coverage. All indices in our samples were calculated with QIIME (version 1.7.0), and the results were displayed with the R software (version 2.15.3). Chao1 and observed-species represent bacterial richness, whereas the Shannon and Simpson indices are quantitative measures of bacterial diversity reflecting both species richness and evenness. The difference in alpha diversity was evaluated by the Kruskal–Wallis test.

Beta diversity was used to evaluate differences in species complexity among samples. Beta diversity on both weighted and unweighted UniFrac distances was calculated by the QIIME software (version 1.7.0). Principal coordinates analysis (PCoA) was performed to obtain principal coordinates and reduce the complex, multidimensional data for visualization of patterns and to assess whether urinary microbial species could be differentiated between gout patients and healthy controls. A distance matrix of weighted or unweighted UniFrac distances among samples was transformed to a new set of orthogonal axes, by which the maximum variation factor is reflected by the first principal coordinate, the second maximum factor is reflected by the second principal coordinate, and so on. PCoA results were displayed using the WGCNA package, stat packages, and ggplot2 package in the R software (version 2.15.3).

Each sample was mapped based on the overall microbial composition and assessed for similarities. To identify significantly different bacteria as biomarkers between groups, the online software linear discriminant analysis effect size (LEfSe) (16) was utilized to select and demonstrate differentially abundant taxonomic groups based on the Kruskal–Wallis test and linear discriminant analysis (LDA) score.

Moreover, receiver operating characteristic (ROC) curves were constructed using the statistical programming language R (V.3.1.2) to visualize the potential of significantly altered genera as gout biomarkers. ROC curves are used to evaluate the performance or the quality of diagnostic tests and are widely used to evaluate the performance of many microbial biomarkers in gut microbiome analyses (17, 18). Area under the curves (AUCs) of ROC was generated to evaluate the performance of the fitted logistic regression models. It was based on the predicted probability of gout for each individual by the multivariate logistic regression coefficient estimates and the individual's transformed relative abundances for each bacterial taxon included in the analysis to predict the probability of gout for each individual.

## RESULTS

#### Subject Characteristics

The basic characteristics of the patients (n = 30) and healthy controls (n = 30) are summarized in **Table 1**. All of the subjects were male, with no significant differences between the two groups in terms of age, smoking history, alcohol intake, or dietary patterns. No significant difference between the groups was found in the laboratory data except for the serum levels of UA and blood urea nitrogen, which were both significantly elevated in the gout group.

## Urinary Dysbiosis in Gout Patients

The urine samples of all 60 subjects were analyzed to assess overall differences in the microbial community structure in urine between gout patients and healthy individuals. After optimization, a total of 4,755,380 sequence reads were included in the final analysis. The results of Illumina MiSeq sequencing showed at least 64,967 valid reads of each sample for operational taxonomic unit (OTU) analysis.

The urinary microbiota of gout patients was significantly different from that of healthy controls. Most of the alpha-diversity indices (ACE, Chao1, Shannon, Simpson, and observed-species) of the urinary microbiota from the gout group were significantly lower than those of the healthy group (**Figure 1**). Good's coverage was higher in the gout patients than in the healthy controls, but there was no statistical difference (P = 0.058). These results indicate that gout patients have a lower diversity and richness in the urinary microbiome, but the evenness of the urinary microbiome is similar to that of healthy controls.

The difference in beta diversity according to the weighted UniFrac distance was assessed by PCoA, which indicated that most of the samples from the two groups clustered together (**Figure 2A**). However, the ordination plot demonstrates difference between the gout patients and healthy controls. Further examination showed that a total of 1606 OTUs were shared between the gout patients and healthy controls, which accounted for 84.34 and 78.84% of the total OTUs in healthy controls and gout patients, respectively (**Figure 2B**).

#### Major Differences Between the Microbiomes of Healthy Individuals and Gout Patients

The range of sequence reads obtained per subject varied from 64,967 to 90,070. A total of 30,488 OTUs were observed across all subjects, and the range of OTU numbers varied from 255 to 769. A total of 10 bacterial phyla accounted for >90% of all sequence reads in the two groups (**Figure 3A**). At the phylum level, the relative abundances of Firmicutes and Actinobacteria in the gout group were significantly higher than those in the controls (P < 0.001) (**Figures 3B,C**). However, the other dominant phyla

FIGURE 1 | Variation in alpha-diversity indices between gout patients (red) and healthy controls (green). Six indices of alpha-diversity were analyzed. The results demonstrate that ACE, Chao1, Shannon, Simpson, and observed-species of the urinary microbiota from the gout group were significantly lower than those of the healthy group (A–C,E,F), while no statistical difference in Good's coverage was found (D).

showed lower abundances in the gout group, especially the relative abundance of Proteobacteria (**Figure 3D**).

At the genus level, Corynebacterium\_1 was clearly enriched in the gout patients (P = 0.004), which possibly inflated the total Actinobacteria abundance observed at the phylum level. Another minor genus that was also significantly more abundant in the gout patients was Prevotella, which belongs to the phylum Bacteroidetes (P = 0.04). The relative abundances of other genera, including Novosphingobium, Derxia, Curvibacter, Methylobacterium, Caulobacter, Skermanella,

FIGURE 3 | Microbiota composition at the phylum and genus levels. (A) Relative abundances of the 10 dominant bacterial phyla found across the two groups shown as histograms. A *t*-test was used to detect the difference between the two groups. (B–D) Phyla with significantly different relative abundances between two groups, \**P* < 0.05. (E) Genus with significantly different relative abundances between two groups. Different genera were assigned only to those presenting minimum variation at a significance level of *P* < 0.05.

unidentified\_Chloroplast, and Rikenellaceae\_RC9\_gut\_group, were significantly decreased in gout patients (**Figure 3E**).

## Potential Biomarkers to Differentiate Gout Patients From Healthy Patients

To explore the potential gout-associated biomarkers, the urinary microbiome sequence data were subjected to LEfSe analysis. A cladogram representative of the structure of the urinary microbiota and the predominant bacteria is shown in **Figure 4**, which also displays the taxa with the greatest differences between the two groups and the discrepant microbial species with a reduced significance threshold (LDA score > 2). The LEfSe method revealed that the phylum Firmicutes and its derivatives (Lactobacillus\_iners, Family\_XI, and Finegoldia), the phylum

Actinobacteria and its derivatives (unidentified\_Actinobacteria, Corynebacteriales, Corynebacteriale, Corynebacterium\_1, and Corynebacterium\_tuberculostearicum), and the genus Prevotella all showed higher relative abundances in the urinary microbiota from the gout patients, suggesting these taxa as candidate biomarkers for potential distinguishing between gout patients and healthy controls.

Five of the significantly altered genera (Pandoraea, Curvibacter, Skermanella, Novosphingobium, and Sulfuritalea) were screened for their abilities in distinguishing gout patients and healthy controls in the ROC analysis. Complete results of the training and test validation subjects using leave-one-out cross-validation are shown in **Figure 5**. This combination of significantly altered genera could effectively differentiate between gout patients and healthy controls with AUC values of 0.971 [95% confidence interval (CI) 89.22–100%] in the test validation subjects (**Figure 5A**) and 0.934 (95% CI 86.28–100%) in the training validation subjects (**Figure 5B**).

#### DISCUSSION

In this prospective pilot study, we first characterized the urinary microbiome of male patients with gout and healthy individuals using high-throughput sequencing of the V3–V4 region of the 16S rRNA gene. The results showed clear differences in the microbiomes between the groups, with lower diversity and richness in the gout patients, which might contribute to the abnormal UA metabolism and inflammation regulation in gout. The combination of significantly altered genera showed good ability to differentiate gout patients from healthy controls, which might provide a new non-invasive biomarker for improving gout diagnosis.

Our results further highlight the polymicrobial composition of human urine, with evident individual variation. The OTU analysis suggested that the microbiota of gout samples showed over 79% similarity with that of healthy controls. The alpha-diversity (observed-species, Chao1, ACE, Shannon, and Simpson) indices of the urinary microbiota from the gout group were all lower than those of the healthy controls, suggesting a decrease in the overall richness and ecological diversity in gout patients. These slight differences spanned from the phylum level to the species level. All of the urine samples were predominantly composed of bacteria from four phyla, Firmicutes, Proteobacteria, Actinobacteria, and Bacteroidetes, which is consistent with the findings of Karstens et al. (19). The most abundant phylum in healthy controls was Proteobacteria, whereas Firmicutes was the most abundant phylum in the gout group. Firmicutes is considered as the most abundant bacteria in normal urine samples (19–23), although Proteobacteria has also been reported to be the most common in healthy urine (24). This overall similarity at the phylum level could suggest a relatively stable bacterial community in gout. However, at the genus level, more specific and clear shifts were observed, especially for genera in Actinobacteria and Bacteroidetes. The predominant ecological structure of the urine microbiomes showed slight variations in relative abundance in the two groups. In particular, Prevotella increased in the gout patients, which belongs to Bacteroidetes, and was reported to be the second and/or third most abundant or

even lower in rank in healthy individuals (21, 22, 25). A previous study also demonstrated that the family Bacteroidaceae and its genus Bacteroides were enriched in the gut of male gout patients (26). Prevotella has been regarded as pathogenic in the vaginal microbiome (27). This genus was also found to be was enriched in patients with type 2 diabetes (24) and in general non-healthy individuals (21, 25). Prevotella was also found to be increased in the gut microbiome of patients with kidney stones compared with non-stone controls (28). Bacteria selectively aggregate to crystals in the urinary tract, which suggests a proper mechanistic role for bacteria in stone formation. Gout is a high-risk factor for the formation of kidney stones; however, the role of Prevotella in gout and/or uric acid stone formation remains unknown.

Corynebacterium\_1 was also a predominant genus detected in the gout group, whereas Novosphingobium was more predominant in the urine of healthy controls. The abundance of Corynebacterium\_1 was previously demonstrated to be correlated with serum concentrations of interleukin-6 and Creactive protein in cancer patients (29). Gout not only is a metabolic condition but is also an auto-inflammatory disease associated with an increased inflammatory reaction. Therefore, these findings could suggest an association between urinary microbiota such as Corynebacterium\_1 and an inflammatory reaction. The relative abundance of Corynebacterium\_1 was previously reported to be increased in patients with prostate cancer (30). Both gout and prostate cancer are clinical conditions that mainly occur in male patients; therefore, it would be interesting to investigate a potential sex-specific pattern of Corynebacterium\_1 abundance. The urine samples of the present study were all from male subjects since gout is predominant in men. However, previous studies have demonstrated a difference in the healthy urine microbiome of males and females (31). Therefore, these conclusions should be interpreted with caution when applied to elderly female patients suffering from gout.

There is previous evidence of an association of the gut microbiota with gout (14, 26, 32), suggesting that the altered metabolites of gout patients may play a role not only in inflammation disorders but also in purine metabolism and UA excretion. The intestinal microbiota composition also markedly varies according to dietary intake (33), which could induce differences in both the microbiota composition and microbial metabolites (34). Consequently, the urinary microbiota can be indirectly influenced by gut microbiota alterations caused by dietary patterns (35).

Here, we found some specific alterations of the urinary microbiome with potential to distinguish gout patients from healthy controls. In particular, the combination of five genera that were significantly altered in the urine of gout patients, Pandoraea, Curvibacter, Skermanella, Novosphingobium, and Sulfuritalea, could effectively distinguish gout patients from healthy controls with a high predictive value. To the best of our knowledge, this is the first attempt to characterize the urinary signatures of gout by integrating the microbiome. The significantly altered urine microbiome could serve as a biomarker to discriminate between gout patients and healthy controls, thereby improving diagnosis and allowing for early intervention. Further exploration of the underlying mechanisms to explain these associations could also provide insight into the pathogenesis of gout and suggest new treatment strategies.

#### DATA AVAILABILITY STATEMENT

Raw Illumina data reads are available in the NCBI Sequence Read Archive database of GenBank under accession ID SRP153570.

#### ETHICS STATEMENT

The studies involving human participants were reviewed and approved by the Ethics Committee of the First Affiliated Hospital of Xiamen University. The patients/participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.

#### AUTHOR CONTRIBUTIONS

GS participated in the research design and data analysis, reviewed and revised the manuscript, and provided general supervision and financial support. YL revised the manuscript. YN performed data acquisition, sample collection and analysis, and

#### REFERENCES


interpretation and contributed to article drafting and revision. SC performed sample collection, data analysis, and article revision and provided financial support. GY, XZ, HQ, and YC performed data acquisition and sample collection. All authors contributed to the study and approved the final submitted manuscript.

#### ACKNOWLEDGMENTS

We appreciate all of the patients and volunteers for their participation. We also thank the postgraduate researchers Xining Liao, Mi Zhou, Xiaoqin Yuan, and Yingling Xiong for their contribution in sample collection. This work was partly supported by the Natural Science of China (Grant Numbers 81501369, 81971536, and 81601384), and Young Grant from Health and Family Planning Commission of Fujian Province (Grant Number 2015-2-40).


and correspondence analysis. Infect Genet Evol. (2016) 44:479–86. doi: 10.1016/j.meegid.2016.08.001


**Conflict of Interest:** 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 © 2020 Ning, Yang, Chen, Zhao, Qian, Liu, Chen and Shi. 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.