# CANCER METABOLISM: CURRENT KNOWLEDGE AND PERSPECTIVES

EDITED BY : Leonardo Freire-de-Lima, Lucia Mendonça-Previato and Luciana Boffoni Gentile PUBLISHED IN : Frontiers in Oncology

#### Frontiers Copyright Statement

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

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

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

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

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

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

The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use. ISSN 1664-8714 ISBN 978-2-88945-989-6 DOI 10.3389/978-2-88945-989-6

#### 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

# CANCER METABOLISM: CURRENT KNOWLEDGE AND PERSPECTIVES

Topic Editors:

Leonardo Freire-de-Lima, Universidade Federal do Rio de Janeiro, Brazil Lucia Mendonça-Previato, Universidade Federal do Rio de Janeiro, Brazil Luciana Boffoni Gentile, Universidade Federal Fluminense, Brazil

Citation: Freire-de-Lima, L., Mendonça-Previato, L., Gentile, L. B., eds. (2019). Cancer Metabolism: Current Knowledge and Perspectives. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-989-6

# Table of Contents


Gopinath Prakasam, Mohammad Askandar Iqbal, Rameshwar N. K. Bamezai and Sybille Mazurek


Ana Carolina B. Sant'Anna-Silva, Gilson C. Santos, Samir P. Costa Campos, André Marco Oliveira Gomes, Juan Alberto Pérez-Valencia and Franklin David Rumjanek

*80 Targeting Ongoing DNA Damage in Multiple Myeloma: Effects of DNA Damage Response Inhibitors on Plasma Cell Survival*

Ana Belén Herrero and Norma Carmen Gutiérrez

# Editorial: Cancer Metabolism: Current Knowledge and Perspectives

Leonardo Freire-de-Lima<sup>1</sup> \*, Lucia Mendonça-Previato<sup>1</sup> \* and Luciana Boffoni Gentile<sup>2</sup> \*

<sup>1</sup> Laboratorio de Glicobiologia, Instituto de Biofisica Carlos Chagas Filho, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil, <sup>2</sup> Laboratório de Patologia Clínica Veterinária, Hospital Universitário de Medicina Veterinária Professor Firmino Mársico Filho (Huvet), Universidade Federal Fluminense, Rio de Janeiro, Brazil

Keywords: cancer, metabolism, glycolysis (glycolytic pathway), multidrug resistance phenotype, invasiviness

**Editorial on the Research Topic**

#### **Cancer Metabolism: Current Knowledge and Perspectives**

Over the first half of the twentieth century, it has been demonstrated that an important hallmark of cancer cells is the reprogramming of cellular metabolism. The capacity to obtain nutrients from an extremely poor microenvironment and utilize it to both generate biomass and maintain the cellular viability is a major feature of transformed cells. Several papers have proposed new mechanisms by which cancer cells undergo substantial metabolic reprogramming to endure within the hostile microenvironment. However, the metabolic profile adopted by cancer cells with invasive and multidrug-resistance phenotypes is still poorly understood. The research topic "Cancer Metabolism: Current Knowledge and Perspectives" provides exciting articles on the valuable field of cancer metabolism, where one hypothesis and theory, two original researches and five outstanding reviews are presented.

#### Edited and reviewed by:

Paolo Pinton, University of Ferrara, Italy

#### \*Correspondence:

Leonardo Freire-de-Lima leolima@biof.ufrj.br Lucia Mendonça-Previato luciamp@biof.ufrj.br Luciana Boffoni Gentile luboffoni@id.uff.br

#### Specialty section:

This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology

> Received: 12 March 2019 Accepted: 29 March 2019 Published: 16 April 2019

#### Citation:

Freire-de-Lima L, Mendonça-Previato L and Gentile LB (2019) Editorial: Cancer Metabolism: Current Knowledge and Perspectives. Front. Oncol. 9:287. doi: 10.3389/fonc.2019.00287

In their original research, Herrero and Gutierrez demonstrated that multiple myeloma cells with ongoing endogenous DNA damage depend on a homologous recombination (HR) pathway, which may be used as therapeutic proposals. They suggested that obstructing HR after the initial step of end resection might be more suitable to promote cell death, since it avoids a compensatory nonhomologous end joining repair mechanism. The authors claim that these preclinical observations afford the basis for its clinical assessment.

Sant'Anna-Silva et al. analyzed the metabolic changes to the metastatic phenotype of human tongue squamous cell carcinoma lineages. By using both metabolomic and fluorescence lifetime imaging microscopy analysis, the authors demonstrated that many pathways linked to lipid metabolism seem to be associated to metastatic phenotype. On the other hand, amino acid metabolism and cell cycle regulation are most correlated to cells with low invasive phenotype.

In their review article, Turgeon et al. addressed evidence that link DNA damage/repair mechanisms with cancer cell metabolism. The authors claim that such connection is progressively apparent, granting opportunities to better understand the metabolic susceptibilities of a considerable fraction of tumors.

Prakasam et al., summarized recent progress in the understanding of many posttranslation modifications (PTMs) in cancer, in particular the PTMs in the M2 isoform of pyruvate kinase (PKM2). The authors believe that such knowledge will be crucial to evaluate their therapeutic potential for the treatment of different types of cancer.

Morrot et al. provided a snapshot of metabolic reprogramming in cancer cells, describing how, even in aerobic conditions, transformed cells opt for glycolysis instead of oxidative phosphorylation (OXPHOS). They discussed how this metabolic reprogramming is able to induce a high-lactate output, then promoting mmunosuppressive events. The authors believe that further studies are needed to better understand the effect of lactate and other "waste" metabolites on cancer progression.

**4**

Coelho et al. presented an overview of the metabolic reprogramming in thyroid cancer, emphasizing factors that promote enhanced glycolysis in transformed cells. The authors also discussed about promising metabolic targets that might be useful to treat patients with thyroid cancer.

Snyder et al. discussed about the metabolic phenotype of cancer stem cells (CSCs), especially taking into account both glycolytic and OXPHOS pathways. Since CSCs present metabolic peculiarities when compared to most of cancer cells in a tumor, the authors believe that such singularities might offer a great potential for developing improved treatments for cancer patients.

Finally, in their hypothesis and theory article, Vidal et al. addressed the connection between multidrug-resistance phenotype and metabolic reprogramming in cancer cells, taking into consideration the functions mediated by ATPbinding cassette transporters, as well as the numerous non-metabolic roles mediated by enzymes that are part of the glycolytic pathway, with special attention to glyceraldehyde-3-phosphate dehydrogenase.

Taken together, the published papers in this research topic strengthen the concept that cancer cell metabolism is an important field in cancer biology, and further studies in this lively research area might provide important information, which may be useful for treatment, diagnostic and therapeutic purposes.

## AUTHOR CONTRIBUTIONS

LF, LM-P and LG wrote the paper. All the authors read and approved the final version of the manuscript.

### FUNDING

The authors are supported by grants from the Brazilian agencies: CAPES, FAPERJ, CNPq, and Cancer Foundation.

### ACKNOWLEDGMENTS

The authors would like to express their deep gratitude to all authors who kindly contributed in this Research Topic.

**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 Freire-de-Lima, Mendonça-Previato and Gentile. This is an openaccess 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.

# Cancer Stem Cell Metabolism and Potential Therapeutic Targets

*Vusala Snyder1 , Tamika C. Reed-Newman2 , Levi Arnold1 , Sufi Mary Thomas1,3,4\* and Shrikant Anant2,4\**

*1Department of Otolaryngology, University of Kansas Medical Center, Kansas City, KS, United States, 2Department of General Surgery, University of Kansas Medical Center, Kansas City, KS, United States, 3Anatomy and Cell Biology, University of Kansas Medical Center, Kansas City, KS, United States, 4Cancer Biology, University of Kansas Medical Center, Kansas City, KS, United States*

#### *Edited by:*

*Leonardo Freire-de-Lima, Universidade Federal do Rio de Janeiro, Brazil*

#### *Reviewed by:*

*Jane Frimodig, University of Louisville, United States Dhiraj Kumar, University of Texas MD Anderson Cancer Center, United States Federica Barbieri, Università di Genova, Italy*

*\*Correspondence:*

*Sufi Mary Thomas sthomas7@kumc.edu; Shrikant Anant sanant@kumc.edu*

#### *Specialty section:*

*This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology*

*Received: 04 March 2018 Accepted: 21 May 2018 Published: 05 June 2018*

#### *Citation:*

*Snyder V, Reed-Newman TC, Arnold L, Thomas SM and Anant S (2018) Cancer Stem Cell Metabolism and Potential Therapeutic Targets. Front. Oncol. 8:203. doi: 10.3389/fonc.2018.00203*

Malignant tumors contain heterogeneous populations of cells in various states of proliferation and differentiation. The presence of cancer stem or initiating cells is a well-established concept wherein quiescent and poorly differentiated cells within a tumor mass contribute to drug resistance, and under permissive conditions, are responsible for tumor recurrence and metastasis. A number of studies have identified molecular markers that are characteristic of tissue-specific cancer stem cells (CSCs). Isolation of CSCs has enabled studies on the metabolic status of CSCs. As metabolic plasticity is a hallmark of cancer cell adaptation, the intricacies of CSC metabolism and their phenotypic behavior are critical areas of research. Unlike normal stem cells, which rely heavily on oxidative phosphorylation (OXPHOS) as their primary source of energy, or cancer cells, which are primarily glycolytic, CSCs demonstrate a unique metabolic flexibility. CSCs can switch between OXPHOS and glycolysis in the presence of oxygen to maintain homeostasis and, thereby, promote tumor growth. Here, we review key factors that impact CSC metabolic phenotype including heterogeneity of CSCs across different histologic tumor types, tissue-specific variations, tumor microenvironment, and CSC niche. Furthermore, we discuss how targeting key players of glycolytic and mitochondrial pathways has shown promising results in cancer eradication and attenuation of disease recurrence in preclinical models. In addition, we highlight studies on other potential therapeutic targets including complex interactions within the microenvironment and cellular communications in the CSC niche to interfere with CSC growth, resistance, and metastasis.

#### Keywords: stem cells, metabolism, microenvironment, targets, cancer stem cell markers

### INTRODUCTION

Despite the advances in modern medicine, some of the major challenges currently confronted in treating cancer patients include the development of therapeutic drug resistance and disease recurrence. Traditional treatments target cancer cells as a means to eradicate tumors and treat the patients. These methods are largely based on the stochastic model—a theory that suggests that cancer cells can arise from a cell that undergoes gene mutations resulting in the acquisition of a highly proliferative state (1, 2). Each progenitor cell bears the mutation and phenotypic profile of the parent cell and is capable of reconstituting a tumor. Several studies have since challenged this theory by demonstrating the existence of a subpopulation of cells called cancer stem cells (CSCs) or tumor-initiating cells, which are typically quiescent but under certain conditions, capable of proliferating to self-renew

**6**

the CSC population and generate progenitor tumor cells (3, 4). CSCs are resistant to therapies that target rapidly proliferating tumor cells and are primarily responsible for tumor relapse. In the 1990s, the theory of a hierarchical organization within tumors was introduced in acute myeloid leukemia (AML), identifying leukemia-initiating cells *via* their expression of a CD34++CD38<sup>−</sup> phenotype. This hierarchical model postulates that individual tumor cells have distinct mutational profiles and epigenetic modifications contributing to cellular heterogeneity. In the years to follow, researchers have used molecular markers to identify and isolate CSCs of various solid tumors (5–7).

Currently, there are more than 40 established CSC markers (**Table 1**); however, much controversy surrounds the scientific techniques employed to identify surface markers. Moreover, majority of the markers established for the identification of CSCs were previously described in human embryonic stem cells and/or adult stem cells of normal tissue cells (5, 8). This shared feature may suggest two possibilities: CSCs could originate from genetic alterations in normal stem cells or could be the result of dedifferentiation of mutated cancer cells into stem-like cells. Despite the shared properties, CSCs differ from normal stem cells in that unlike CSCs, cell proliferation is rigidly controlled in normal stem cells (9). Glycosylation of glycoprotein markers has also been suggested to impact the biological behavior of CSCs (8). It is important to focus future investigation on the mutations, metabolic phenotype, and other aspects of the microenvironment that distinguish CSCs from normal stem cells.

Normal stem cells are unique in their ability to self-renew, proliferate, and differentiate into various tissue types, as well as reproduce progeny essential to maintain and repair the organ system in which they are found (35, 36). Embryonic stem cells, hematopoietic stem cells, and mesenchymal stem cells have a low mitochondrial DNA copy number, as well as poorly developed mitochondrial morphology and reduced oxidative capacity. On the other hand, glycolytic pathways are highly active in these stem cells. Hypoxia-induced glycolysis in pluripotent stem cells and inhibition of mitochondrial respiration promote stemness, whereas inhibition of glycolysis disrupts proliferation and promotes cell death (37).

Although CSCs share many of the characteristics of normal stem cells, the differ in that, they contribute to tumor progression, drug resistance, and recurrence (38). In addition, several reports suggest that CSCs preferentially use glycolysis. However, other reports suggest a propensity for mitochondrial oxidative phosphorylation (OXPHOS) suggesting a possible metabolic plasticity. The aim of this review is to summarize and emphasize some of


*AML, acute myeloid leukemia; ALL, acute lymphocytic leukemia; CRC, colorectal carcinoma; HNSCC, head and neck squamous cell carcinoma; HCC, hepatocellular carcinoma; OXPHOS, oxidative phosphorylation; ROS, reactive oxygen species; CSCs, cancer stem cells: PKM2, pyruvate kinase M2.*

the key aspects currently known about CSC metabolism and the potential therapeutic targets contributing to cancer progression.

#### CELLULAR AND CANCER METABOLISM

All cells require energy for growth, division, and survival, which they acquire through the absorption of nutrients including glucose that are broken down in a series of metabolic reactions involving glycolysis and cellular respiration through OXPHOS. Under normal physiological conditions, cells rely on both glycolysis and OXPHOS for efficient energy production (39). The process of glycolysis involves the breakdown of glucose through a series of reactions to produce pyruvate, two molecules of adenosine 5′-triphosphate (ATP), and nicotinamide adenine dinucleotide (NADH). Under normoxic conditions (oxygen is readily available), pyruvate is transported to the mitochondria where it is converted into acetyl coenzyme A. Acetyl CoA enters the tricarboxylic acid cycle to produce high amounts of energy in the form of NADH and flavin adenine dinucleotide (FADH2) molecules. The hydrogen ions from NADH trigger the electron transport chain and generation of up to 32 molecules of ATP through OXPHOS (40, 41).

Since OXPHOS generates more ATP molecules than glycolysis, normal cells rely primarily on OXPHOS as an efficient source of energy. This process, however, is impaired in hypoxic conditions due a dearth in oxygen.

Rapidly proliferating cancer cells outpace angiogenesis resulting in areas of low oxygen. However, increasing evidence suggests that cancer cells engage in glycolysis even in the presence of oxygen (42). As a result, tumor cells demonstrate enhanced glycolytic production of ATP (43). Otto Warburg first described this phenomenon now known as the Warburg effect of aerobic glycolysis (41, 44, 45). The increased glycolysis was attributed to mitochondrial damage in cancer cells. Subsequent studies found that most cancer cells do not demonstrate mitochondrial damage, but rather suggest that aerobic glycolysis can occur simultaneously to enhance energy production for the maintenance of cancer cell homeostasis (43). In fact, several studies demonstrate that acceleration of glycolysis provides a source of metabolites and other essential factors required for rapidly dividing cells (41, 43, 46, 47).

### METABOLIC PHENOTYPE OF CSCs

Due to the highly proliferative, tumorigenic, and drug-resistant properties of CSCs, in-depth investigation of CSC metabolic phenotype has comprised the cornerstone of numerous recent studies. Although metabolic adaptation or plasticity is one of the hallmarks of cancer, the majority of reports suggest that CSCs are primarily glycolytic (48–55). However, examination of CSCs isolated from patient tumors suggests that OXPHOS is the main source of energy (56, 57). We describe other multifactorial causes contributing to the apparent differences in CSC metabolism across tumor types in the following sections. Emerging evidence suggests the existence of specific metabolic phenotypes of CSCs based on their location, such as those in actively growing regions of the tumor that have adequate levels of oxygen, hypoxic areas of the tumor, or those in a distant metastatic site summarized in **Figure 1**.

### GLYCOLYTIC PATHWAY

A number of studies performed in various tumor types, such as glioblastoma, lung cancer, osteosarcoma, breast cancer, ovarian cancer, and colon cancer, suggest that CSCs more strongly favor the glycolytic pathway than other differentiated cancer cells *in vitro* and *in vivo* (58–62). Rationale for investigating the role of glycolytic metabolism in CSCs is due to its proposed phenotypic similarity to normal stem cells with self-renewal characteristics. Earlier studies paved the way by illustrating the low activity of mitochondrial respiration in brain tumor CSCs, as well as higher rates of glycolysis in CSCs than other tumor cells (63, 64). Further investigations revealed that upregulation of glycolytic enzymes (GLUT1, HK-1, and PDK-1) and stimulation of glycolysis are necessary for cell immortalization and is sufficient to increase cellular lifespan (65). Comparing glucose utilization by CSCs and non-CSCs has revealed differentially elevated glucose consumption, lactate synthesis, and ATP content in CSCs, thus suggesting distinct metabolic profiles of CSCs in comparison to non-CSCs (66–68). Glycolysis has also been identified as the preferred metabolic pathway of CSCs in nasopharyngeal carcinoma and of tumor-initiating stem-like cells in hepatocellular carcinoma (69, 70). In addition, cellular metabolism is thought to control stemness characteristics; in particular, the glycolytic switch has a causal relation in induced pluripotent stem cell reprogramming and acquisition of pluripotent markers (71). Reprogramming the metabolic switch from OXPHOS to glycolysis was shown to enhance stemness and CSC properties in CD44<sup>+</sup>CD24lowEPCAM<sup>+</sup> cells of basal-like breast cancer by reducing reactive oxygen species (ROS) levels (48). Glycolysis-driven induction of pluripotency is consistent with the finding that hypoxia maintains the stem cell state and a hypoxic environment promotes the reprogramming process (72).

### OXPHOS PATHWAY

Growing evidence suggests mitochondrial oxidative metabolism as the preferred form of energy production in CSCs. Several studies in numerous tumor types, such as CD133<sup>+</sup> cells of glioblastoma and pancreatic ductal adenocarcinoma, ROSlow quiescent leukemia stem cells, lung cancer side population cells, and breast cancer, strongly support an OXPHOS phenotype and less glycolytic profile (49, 50, 54, 73). In contrast to the non-CSC cancer cells, which mainly utilize glycolysis for energy production, CSCs have an enhanced mitochondrial ROS, higher rates of oxygen consumption, and overall increased mitochondrial function, as evidenced by increased mitochondrial mass and membrane potential (50, 52, 53, 73–76). Moreover, this increased mitochondrial bulk in a subpopulation of breast cancer cells induces stem-like characteristics and confers metastatic potential and resistance to DNA damage (77). In addition, CSCs may depend on mitochondrial fatty acid oxidation (FAO) for the generation of ATP and NADH. A population of isolated ovarian CSCs revealed upregulated expression of genes associated with

FAO and OXPHOS (52). FAO is instrumental in self-renewal processes of hematopoietic stem cells and leukemia-initiating cells, as in the survival of ablation-resistant pancreatic CSCs and survival of epithelial cancer cells subsequent to matrix detachment (78–80). An oxidative phenotype confers resistance to treatment modalities and evasion of apoptosis as evidenced by the vastly tumorigenic and chemoresistant metabolism found in hepatocellular CSCs, upon NANOG-induced expression of FAO genes (70). The powerful antioxidant defense mechanism of CSCs contributes to therapy resistance, by maintaining a significantly lower ROS levels and preserving stemness and tumorigenic properties of CSCs (52, 81, 82).

#### FACTORS AFFECTING THE METABOLIC STATUS OF CSCs

The reported differences in the metabolic profile of CSCs from various tumor types are due to multifactorial causes. One such explanation is the suggested plasticity of these cells and the potential harvest of them at various stages of differentiation/ dedifferentiation during experiments (2). Another cause may be the lack of uniformity and precision in definition of CSCs and varying techniques utilized to isolate CSCs, such as specific markers, Hoechst staining-based sorting, chemoresistance-based isolation, and reoxygenation sorting post-hypoxic exposure (83–87). This is due to a vast heterogeneity of CSCs across various histologic tumor types. Another potential contributing factor that may explain the contradictory results is the contribution of the microenvironment. We broadly distinguish the metabolic status of CSCs in three locations, namely, regions with normoxic tumor, hypoxic tumor, and at metastatic sites (**Figure 1**).

As mentioned in the preceding sections, under normoxic conditions, CSCs can engage in glycolysis and/or OXPHOS. Furthermore, the metabolic status of CSCs can be affected by crosstalk between CSCs and cancer-associated stroma in the microenvironment. For example, cancer-associated fibroblast-secreted metabolites including lactate and ketone bodies drive OXPHOS in cancer cells (88). The role of cancer-associated stroma in regulating CSC metabolism is unknown.

Similar to embryonic stem cell maintenance, tumor hypoxia promotes the persistence of an undifferentiated, stem cell phenotype (89). Ductal breast carcinoma cells revert to a stem cell-like phenotype through dedifferentiation under hypoxic conditions (90). Other studies showed that hypoxic exposure altered gene expression in human neuroblastoma cells toward a neural crest-like, immature profile and caused upregulation of the stem cell surface marker CD-133 in medulloblastoma (90, 91). Under hypoxic conditions, overexpression of hypoxia-inducible factor-1α (HIF-1α) promotes upregulation and potentiated activity of several glycolytic proteins, such as transporters (GLUT1 and GLUT3) and various isoforms of glycolytic enzymes (92). In addition, HIF-1α regulates pyruvate dehydrogenase kinase 1 levels which facilitate glycolysis in breast CSCs under hypoxic conditions (93).

Metastatic cancer cells undergo epithelial-to-mesenchymal transition (EMT) upregulating a number of factors associated with a stem-like phenotype. EMT-associated factors including HIF-1α, Wnt, and Snail regulate cellular metabolism (94). Furthermore, EMT-associated metabolites—glutamine, glutamate, and alanine—as well as high lactate concentrations are associated with poor survival and higher metastatic potential in breast cancers (95, 96). CSCs have augmented utilization of extracellular catabolites, such as pyruvate, lactate, glutamine, glutamate, alanine, or ketone bodies to support OXPHOS (97–99). In nutrient poor states, quiescent disseminated tumor cells rely on alternative energy sources such as autophagy, yet metabolic plasticity demonstrated by their ability to produce energy through various pathways is instrumental for metastatic growth and proliferation (100–102). Finally, a recent study of metabolic dependencies of non-small cell lung cancers highlighted the significant contribution of the microenvironment as a determinant of the metabolic phenotype of cancer cells, as evidenced by varying profiles *in vitro* and *in vivo* settings. *KRAS*-driven lung cancer cells in mice models showed preferential glutamine utilization *in vitro*, but did not depend on glutamine metabolism *in vivo* (103).

#### TARGETING CELLULAR METABOLISM

A strong association between tumors with high CSC fractions and recurrence, poorer overall survival, and higher incidence of metastasis, underscores the significant prognostic and therapeutic implications of CSCs (104, 105). Defining characteristics of CSCs such as surface markers, metabolic phenotypes, resistance to chemoradiotherapy, and regulatory factors in microenvironment compile the bulk of therapeutic targets. For instance, CD44, a receptor for hyaluronic acid-mediated motility, is shown to induce CSC attachment to extracellular matrix and cell migration, promoting metastasis and invasion (106). Treatment of breast, colon, esophageal, gastric, lung, and ovarian cancers overexpressing CD44, with ONCOFIDTM-S which is a conjugate of hyaluron and chemotherapeutic agent SN38 (7-ethyl-10-hydroxycamptothecin, active metabolite of CPT-11) revealed a strong *in vitro* anti-proliferative activity (107, 108). In addition, use of anti-CD44 antibodies H90 and A3D8 inhibited proliferation and induced apoptosis, by promoting the differentiation of AML blasts (108–111). Finally, CD44 interacts with pyruvate kinase M2 (PKM2), enhancing the glycolytic profile of cancer cells deficient in p53 or exposed to hypoxia. Subsequent ablation of CD44 led to inhibition of glycolysis, increase in ROS and enhancement of chemotherapeutic drug effect in these cancer cells (110). Therefore, preferentially targeting of identified CSC markers, such as CD44, can be utilized for an effective cytotoxic drug delivery. In addition, inhibition of glycolysis can be achieved by targeting various glycolytic enzymes, transporters, and other complex regulators, such as GLUT 1–4, hexokinase, PKM2, and lactate dehydrogenase A (111–113).

Previously discussed evidence for OXPHOS dependence of CSCs in numerous cancer lines proposes mitochondrial metabolism to be a potential target for an effective elimination of CSCs. Inhibition of the OXPHOS pathway reduces sphere formation and tumor formation potential demonstrating vulnerability of CSC to mitochondria-targeted therapies (54, 114, 115). Pharmacological agents targeting CSCs through inhibition of mitochondrial biogenesis and OXPHOS are currently under investigation for cancer treatment. Several FDA-approved compounds known to inhibit mitochondrial function have been reported to achieve a more effective eradication of CSCs. Salinomycin, erythromycins, tetracyclines, and glycylcyclines are some of the approved agents to have already demonstrated efficacy in eradicating CSCs *via* reduction of stemness properties (115–118). Metformin, an inhibitor of OXPHOS complex I, has demonstrated anti-tumoral activity by reducing mammosphere formation, delaying *in vivo* tumor growth, and inducing apoptosis in pancreatic CSCs unable to switch to glycolysis (54, 119, 120). However, emergence of a small subset of resistant CSCs with an intermediate glycolytic/ OXPHOS phenotype could be prevented/reversed by utilizing a mitochondrial ROS inducer such as menadione (54). Dual mechanism of menadione inhibition of Complex I and induction of mitochondrial ROS points out the superior efficacy of multi-modal targeted therapy. Studies have shown that inhibition of mitochondrial respiration not only induces apoptosis in pancreatic CSCs with OXPHOS phenotype but also effectively eliminates primarily glycolytic breast and nasopharyngeal CSCs (53, 54, 121). These data highlight the extended role of mitochondria beyond energy production in CSCs, such as acquiring metabolites from glutamine *via* reductive carboxylation to support growth in tumor cells with defective mitochondria (122). A novel compound 3,5-bis(2,4-difluorobenzylidene)-4-piperidone (DiFiD) has been shown to inhibit pancreatic cancer growth by targeting a CSC marker, doublecortin and CaM kinase-like-1 (DCLK-1) (59, 66, 123). However, the role of DCLK-1 and the impact of DiFiD on CSC metabolism have not been studied.

Evident from the data reviewed, the CSC phenotype varies between cancer subtypes and among populations of the same subtypes. Preferred energy-producing metabolic pathways depend on various factors, including metastatic site highlighting vast metabolic variability and patterns (124). In addition to studies supporting metabolic plasticity, simultaneous enhancement of glycolysis and OXPHOS pathways was observed in highly metastatic breast cancer cell lines relative to non-metastatic cell lines (49, 124). Consequently, dual inhibition of glycolytic and mitochondrial energy pathways has proven to be effective against tumor growth in a number of preclinical cancer models (125). One such study elegantly demonstrated sarcoma cells to be twofold to fivefold more sensitive than normal cells to dual inhibition of glycolysis with 2-deoxyglucose and OXPHOS with oligomycin or metformin (126). Therefore, dual inhibition of metabolic pathways may be a superior approach to eradicating heterogeneous CSCs rather than singularly targeting glycolysis or OXPHOS pathways. Finally, other factors directly affecting metabolic status of CSCs may represent potential targets for pharmacological treatments. These developments may include promoting CSC differentiation, targeting complex interactions within the microenvironment, and disrupting cellular communications in the CSC niche to interfere with CSC growth, resistance, and metastasis (97, 127–130).

### CONCLUSION

Substantial evidence suggests that the CSCs are pluripotent, selfrenewing, "original cells" of a tumor capable of differentiation into more specialized cancer cell types. CSCs are responsible for tumor formation, differentiation, maintenance, spread, and recurrence, making them an attractive therapeutic target for a potential permanent cure or long-term disease-free survival (127, 131). Regardless of the controversy about the metabolic phenotype of CSCs, metabolism is not only a key player but also a regulatory instigator of stemness.

Metabolic singularities that distinguish CSCs need to be further investigated, as they offer a great potential for developing improved treatments to eradicate them. In particular, streamlining

#### REFERENCES


and standardization of CSC identification methods is important. Development of CSC marker combinations would contribute to better delineation of CSCs from non-CSC cancer cells and normal stem cells. Interactions between CSCs and their microenvironment also provide a fertile ground for advance investigations. Chronicity and causality of these complex interactions needs to be established. Moving forward, CSC metabolic pathways and principal players of metabolism comprise potential therapeutic targets with a great promise for improved cancer treatments.

### AUTHOR CONTRIBUTIONS

VS carried out the literature review and wrote and edited the manuscript. TR-N carried out the literature review and wrote a section of the manuscript. LA carried out literature review and prepared the figure and table included in the manuscript. SA and ST conceptualized the manuscript, supervised the writing, and edited the manuscript.

#### FUNDING

This work was supported by the NIH grants CA135559 and CA109269, the Kansas Bioscience Authority, Tom O'Sullivan Foundation Supported Research and Braden's Hope Foundation Grant to SA. This study was supported in part by University of Kansas Cancer Center under CCSG 1-P30-CA168524-0 (to SA and ST), an NIH Clinical and Translational Science Award grant (UL1 TR000001, formerly UL1RR033179) awarded to the University of Kansas Medical Center, and an internal Lied Basic Science Grant Program of the KUMC Research Institute (to ST).


endothelial growth factor via hypoxic-induced factor-1α. *Int J Mol Sci* (2015) 16(1):439–51. doi:10.3390/ijms16010439


loss of matrix attachment. *Nature* (2009) 461(7260):109–13. doi:10.1038/ nature08268


with clinical outcome but may favor distant metastasis. *Clin Cancer Res* (2005) 11(3):1154–9.


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

# Metabolic reprogramming during Multidrug resistance in Leukemias

Raphael Silveira Vidal <sup>1</sup> , Julia Quarti 1,2, Mariana Figueiredo Rodrigues <sup>1</sup> , Franklin D. Rumjanek <sup>1</sup> and Vivian M. Rumjanek <sup>1</sup> \*

*1 Instituto de Bioquímica Médica Leopoldo de Meis, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil, 2 Instituto de Nutrição Josué de Castro, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil*

Cancer outcome has improved since introduction of target therapy. However, treatment success is still impaired by the same drug resistance mechanism of classical chemotherapy, known as multidrug resistance (MDR) phenotype. This phenotype promotes resistance to drugs with different structures and mechanism of action. Recent reports have shown that resistance acquisition is coupled to metabolic reprogramming. Highgene expression, increase of active transport, and conservation of redox status are one of the few examples that increase energy and substrate demands. It is not clear if the role of this metabolic shift in the MDR phenotype is related to its maintenance or to its induction. Apart from the nature of this relation, the metabolism may represent a new target to avoid or to block the mechanism that has been impairing treatment success. In this mini-review, we discuss the relation between metabolism and MDR resistance focusing on the multiple non-metabolic functions that enzymes of the glycolytic pathway are known to display, with emphasis with the diverse activities of glyceraldehyde-3 phosphate dehydrogenase.

Keywords: multidrug resistance, glycolysis, glyceraldehyde-3-phosphate dehydrogenase, leukemia, reactive oxygen species

### BACKGROUND

Multidrug resistance (MDR) in cancer is the major factor impairing the success of conventional chemotherapy (1). Originally, many hopes were placed on the possibility that by inhibiting the activity of ABC transporters a reversal of the resistance would be attained. This was based on the knowledge that the transporters were shown to be capable of mediating the efflux of many chemotherapeutic drugs. Among the efflux transporters of the ABC superfamily, ABCB1 (P-gp), ABCC1 (MRP1), and ABCG2 (BCRP or MXR) have been described as major players in the development of MDR (2–4), with particular emphasis being given to ABCB1. However, it soon became apparent that inhibition of these transporters was not an effective approach since normal cells may also express these transporters and, therefore, inhibitors could sometimes generate an unacceptable toxicity. Furthermore, in experimental situations, where the down regulation of ABCB1 was achieved, a number of other effects were still observed. Now, it is quite clear that MDR is a multifactorial phenomenon that is involved in the regulation of survival and apoptosis, as well as a number of other cellular pathways. ABCB1 transporter expression, in cells with the MDR phenotype, is but one factor linked to pharmacological evasion of chemotherapeutic drugs (1, 5).

A case in point is the observation that hypoxia participates in the regulation of drug resistance. For instance, *ABCB1* gene expression and synthesis of functional proteins are induced by hypoxic environments (6). Furthermore, ABC transporters are expressed not only in MDR cancer cells, but

#### *Edited by:*

*Michael Breitenbach, University of Salzburg, Austria*

#### *Reviewed by:*

*Johannes A. Mayr, Paracelsus Medical University Salzburg, Austria Hamid Morjani, Université de Reims Champagne Ardenne, France Carlo Vito Bruschi, CVBiotechnologies, E.U., Austria*

#### *\*Correspondence:*

*Vivian M. Rumjanek vivian@bioqmed.ufrj.br*

#### *Specialty section:*

*This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology*

> *Received: 10 January 2018 Accepted: 15 March 2018 Published: 04 April 2018*

#### *Citation:*

*(2018) Metabolic Reprogramming During Multidrug Resistance in Leukemias. Front. Oncol. 8:90. doi: 10.3389/fonc.2018.00090* Vidal RS, Quarti J, Rodrigues MF, Rumjanek FD and Rumjanek VM

**15**

also in a number of stem and progenitor cells. Additionally, it has been reported that hypoxia promotes an undifferentiated cell state in various stem and precursor cell populations, as well as in cancer stem cells (7–9). In this respect, it has also been suggested that NOTCH signaling is involved. However, it must be recalled that when cells are under hypoxic conditions, there is a metabolic shift from oxidative phosphorylation to glycolysis (10). This situation contrasts with cells under normoxia, in which glucose is first anaerobically catabolized to pyruvate which is then further catabolized along the Krebs cycle where NADH and FADH2 are reoxidized by the respiratory chain associated to the electron transport system. Incidentally, glycolysis is a hallmark in many types of tumor cells (11). This phenotype is in fact the basis of the so-called Warburg effect, also known as aerobic glycolysis. The Warburg effect describes a situation in which the glycolytic pathway is fully activated even in the presence of adequate oxygen supply (12). Although Warburg originally proposed that cancer was due to an impairment of mitochondrial function, it is accepted today that these organelles retain full oxidative capacities.

It must be mentioned, however, that apart from red blood cells, aerobic glycolysis is prevalent in highly proliferative cells, whether tumoral or not. Stem cells are a case in point (13). The common belief that cells undergoing glycolysis selected an inefficient form of energy production is misguided. Barring the comparative stoichiometry of ATP formation between glycolysis and OXPHOS, aerobic glycolysis is in fact an efficient form of ATP production due to the kinetic properties of the enzymes participating in the pathway which afford very fast fluxes compatible with the ATP demand of the rapidly growing cells. Beyond its role in bioenergetics, glycolysis constitutes a branch of the pentose phosphate pathway (PPP), since glucose-6-phosphate is also the substrate for glucose-6-phosphate dehydrogenase, the first enzyme of that pathway. Thus, glycolysis also contributes to the production of precursors for the biosynthesis of nucleotides (*via* generation of ribulose-5-phosphate). In addition, the PPP pathway promotes the formation of NADPH, an essential coenzyme for reductive biosynthetic processes such as that of fatty acids. NADPH also has an important role in maintaining the redox equilibrium. Similarly, glycolysis can be considered as an anaplerotic pathway by way of its participation in amino acids synthesis (*via* 3-phosphoglycerate or *via* pyruvate). Thus, from an energetic stand point glycolysis more than compensates the relatively small amounts of ATP produced when compared with oxidative phosphorylation.

However, it must be emphasized that tumors are in fact constituted by a mosaic of different cellular subpopulations. As such, from the biochemical point of view tumors can also be envisaged as being functionally heterogeneous. Accordingly, within the context of types of metabolism, tumors can be perceived as composed of subsets of resistant quiescent/slow-cycling cells that occasionally rely more on mitochondrial respiration and less on glycolysis. Likewise the same tumor could also harbor cells that are exclusively glycolytic (14, 15). Interestingly, the possibility of a switch that regulates mitochondrial function in the case of metastasis has been proposed. The results of Porporato et al. showed that overburdening the electron transport system may be an essential step in enhancing migration of cells *in vitro* and *in vivo* (16). The authors concluded that in order to achieve metastasis, mitochondria must be active, although not necessarily functional. By extension such findings suggest that in tumor cells there may be switches that constantly activate/inactivate mitochondrial function depending on changes dictated by the microenvironment that affect, for example, the availability of metabolites. The intermittent switching between anaerobic and oxidative metabolism seems to be a feature of metastasis. According to this scheme, accumulating data show that there is a tradeoff involving growth versus migration, i.e., cells which are proliferating prevalently exhibit a glycolytic type of metabolism, whereas migrating cells which proliferate less, rely more on mitochondria (16). Within this framework, it is known that the switch between the two main types of energy metabolism may be regulated by ATP demand. For example, cells expressing ABC transporters on their surfaces require a considerable amount of ATP in order to sustain the drug efflux activity. Additionally, it has been suggested that the transporter activity might suffer the impact of alterations in pH gradient due to the glycolytic phenotype (17). Normal cells maintain a gradient between acidic vesicular compartments and an alkaline cytoplasm a situation not observed in tumor cells that have an acidic cytoplasm. Nevertheless, it has been observed that MCF-7adr, that has an MDR phenotype, presents a similar pH gradient to that of normal cells (18, 19). This gradient contributes to the sequestration of drugs in acidic organelles and subsequent extrusion from the cell (19).

The mitochondrial electron transport chain generates reactive oxygen species (ROS) (20). In some instances, such as partial disintegration of complex I (21–23), oxidative stress may result from excessive production of ROS which is buffered by redox homeostasis. In turn, homeostasis is achieved by the participation of a number of enzymes such as catalase, superoxide dismutase, gluthatione peroxidase, and cofactors such as reduced gluthatione and NADPH. Control of redox equilibrium is important since an imbalance between the amount of ROS produced and of antioxidant systems may lead to DNA damage, particularly mitochondrial DNA, and other cell lesions.

Notwithstanding, the effect of ROS as agents of oxidative stress affecting the expression of ABCB1 is controversial. Both, downregulation and upregulation of the transporter have been reported (24, 25). Antioxidants (26), as well as products resulting from glycolysis can act as scavengers of free radicals. In this way, the energy metabolism could also play an indirect role as modulators of ABCB1 expression. For instance, it is known that under conditions leading to glycolysis inhibition, ABCB1 expression was observed to be downregulated. Conversely, when exogenous pyruvate was added to the tumor cells there was increased drug resistance and transporter expression (25).

Therefore, there is consensus that under normal conditions, glycolysis, and mitochondrial oxidative phosphorilation operate in concert in so far as energy production in the form of ATP is concerned.

Originally, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was thought to function exclusively as part of the glycolytic process in the cytoplasm, where its role is well established. However, under conditions of oxidative stress GAPDH may also redirect glyceraldehyde-3-P to glucose 6 phosphate (G6P) as a result from the reversal of part of the glycolytic reactions. The formed G6P then becomes the initial substrate of the PPP leading to the increase of NADPH production (27). GAPDH is particularly sensitive to H2O2-induced oxidation and it has been suggested that cytosolic GAPDH might function as a sensor for redox signals and an information hub to transduce these signals (28). Oxidative stress may also promote GAPDH aggregation leading to mitochondrial dysfunction and necrotic cell death *via* the permeability transition pore (29). Interestingly, under some circumstances GAPDH was shown to bind to band 3 protein, an anion transporter located on the inner side of the red blood cell membrane. In this membrane bound state, GAPDH along with other glycolytic enzymes contributed toward ATP channeling thus allowing its direct consumption by ion pumps without release into the cytoplasm (30). Such a role for GAPDH would be in keeping with an accessory function within the context of drug efflux. GAPDH, however, plays a number of other roles in different cell compartments, where several pools of GAPDH sense cellular stresses and activate cognate pathways to maintain homeostasis or activate cell death (31). It has also been reported that following exposure to stressors GAPDH translocate into the nucleus (32) where it may suffer ADP-ribosylation by NO (33). Apart from GAPDH, other enzymes of the glycolytic pathway are known to display multiple non-metabolic functions. In fact, of the 10 enzymes that constitute glycolysis, at least 7 have been shown to display extra-glycolytic activities that may bear on the MDR phenotype (34). For example, hexokinase II (HKII) binds to the mitochondrial voltage-gated anion channels and thus relieves the negative feedback effect of G6P (35). HKII overexpression is regulated by HIF-1α and also by c-Myc oncogene. HKII is also under the control of many miRNAs. When phosphohexoseisomerase, or phosphoglucoisomerase (PGI) is secreted by cells it acquires the status of a cytokine and is renamed as the autocrine motility factor (AMF). As such, AMF stimulates cell motility and so PGI is thought to be one of the factors driving metastasis (36). Besides, PGI is involved in many other activities such as apoptosis and EMT. Other glycolytic enzymes such as phosphofructokinase 1, aldolase and triose phosphate isomerase, phosphoglycerate mutase, and pyruvate kinase have all been reported to take part in several cellular functions that have a direct relation to tumorigenesis. Hence, it is entirely plausible that individually or collectively, the glycolytic enzymes and in particular GAPDH may constitute integral parts of the MDR phenotype by acting in a non-canonical fashion (37).

Alternatively, the metabolic rewiring of tumor cells may also result from gene rearrangement. Using yeast as a model system, genomic instability and the reprogramming of central metabolism have been approached (38). Regarding gene rearrangement, factors that normally contribute toward the integrity of the replicative process may be compromised. Drug resistance has been observed in yeast in which genome translocants were investigated (39). A parallel between yeast and tumor cells could thus be established even though the detailed mechanisms are still not understood. Indeed in MDR cells the over expression of mini chromosome maintenance 7 (MCM7) was detected (40).

Oxidative stress induces severe damage to proteins, lipids, and DNA. Normally, the generation of oxyradicals is prevented by the mitochondrial antioxidant system; however, the degree of damage will also depend on the repair capacity of the cell. DNA damage leads to the activation of Poly(ADP-ribose) polymerases or PARPs that are nuclear enzymes responsible for catalyzing the attachment of ADP-ribose units from nicotinamide adenine dinucleotide (NAD+) to acceptor proteins involved in the recognition and repair of DNA strands breaks. After repair is completed the Poly(ADP-ribose) (PAR) chains are degraded (41). During caspase mediated apoptosis PARP is cleaved into fragments that in turn inactivate the enzyme inhibiting repair to proceed (42, 43). It has been suggested that PARP inhibitors (PARPi) could be used in combination with chemotherapeutic drugs, but PARPi tends to be extruded from the cell by the ABCB1 transporter in MDR tumors (44).

Cell survival and energy metabolism in MDR tumor cells will be discussed in the next section.

### MDR, CHRONIC MYELOID LEUKEMIA, AND ENERGY METABOLISM

Multidrug resistant cells may be very heterogeneous in their characteristics and survival pathways. In the present review, we will highlight, including our own data, some aspects related to the energy metabolism of two MDR chronic leukemia cell lines derived from the CML cell line, K562. The two cell lines were selected after exposure to different chemotherapeutic drugs. One cell line was exposed to vincristine, originating Lucena-1 (45), and the other exposed to daunorubicin, originating FEPS (46). Despite arising from the same parental cell, they were quite distinct. Comparative microarray analysis identified 130 differentially expressed genes between K562 versus Lucena-1, 1,932 between K562 versus FEPS, and 1,211 between Lucena-1 versus FEPS. *ABCB1* was overexpressed in both MDR cell lines, but highly overexpressed in FEPS, which is the most resistant line (47). Similarly, comparative proteomics of the parental cell line and the MDR counterparts indicated that K562 presented 560 unique proteins, Lucena-1 had 38 and FEPS 63 unique proteins. Lucena-1 and FEPS shared 929 proteins. From the latter, 112 were common only to Lucena-1 and FEPS (48). Results from another survey investigating the proteomic profiles of K562 and Lucena-1, identified 36 differentially expressed proteins between these two cell lines (40). From those, the leucine-rich PPR motif-containing protein and MCM7, as well as the expression of ABCB1 could be used as markers to identify patients that would respond or fail to therapy with the tirosine kinase inhibitor, Imatinib (40). The importance of ABCB1 in the resistance to Imatinib in the clinical setting has been demonstrated (49, 50), but it is not the only transporter involved. Tumor cells expressing ABCG2 are capable of Imatinib extrusion with great affinity (51, 52). In a different CML model, comparing the proteomic analysis of another MDR cell line derived from K562 selected with doxorubicin, Qinghong et al. highlighted 44 differentially expressed proteins. Some of the differentially expressed proteins were common to those observed in FEPS (53).

The possibility that overexpression of ABCB1 in these cell lines might reflect stem cell characteristics was analyzed by looking at the NOTCH pathway. Notch was overexpressed in FEPS, the most resistant cell line displaying the highest expression of *ABCB1*, and slow replication time (47). Another pathway present in both, normal stem cells and CML stem cells is the canonical Wnt pathway. This signaling pathway was shown to be more strongly activated to positively regulate ABCB1 in Lucena-1 cells when compared with the non-MDR K562. However, FEPS was not studied on this occasion (40). Conversely, both MDR cell lines had an increased expression of carbonic anhydrase and hemoglobin (47). This was confirmed by proteomics (48) suggesting that these cells are more differentiated compared with the parental erythroleukemic cell line K562 (54). Furthermore, they have the potential to maintain the intracellular pH (17). It has been reported that Imatinib induces erythrocytic differentiation in K562 cells and this is independent of blockade of apoptosis being also observed in resistant clones (55).

One characteristic of the two MDR cell lines described above (Lucena-1 and FEPS) is their antioxidant capacity. The elevated catalase activity observed in Lucena-1 provides these cells with protection against cytotoxic chemicals as well as UV radiation (56). Catalase activity is also elevated in FEPS and these cells also present increased Glucose 6 phosphate dehydrogenase (G6PD) activity (Vidal RS, Faria G, Maia RC, and Rumjanek VM, unpublished data). Despite the well-known role in redox homeostasis, G6PD is also involved in cell growth and signaling and this might be an equally important role in resistant cells (57). Many redox changes are now perceived to allow, in a localized compartment, a rapid and physiological signaling event that may regulate the activity of certain proteins (26, 58). Important players in regulating these intracellular effects are members of the thioredoxin family, including thioredoxin that has increased levels in Lucena-1 and FEPS (48), glutaredoxins, and peroxiredoxins. The third most upregulated gene in Lucena-1 when compared with K562 is *SESN3* that catalyzes peroxiredoxins leading to ROS detoxification (47). Furthermore, several members of peroxiredoxins, a family of antioxidant enzymes, are increased in FEPS (PRDX1, PRDX2, PRDX3, and PRDX6) or Lucena-1 (PDRX1) compared with K562 (47, 48). Therefore, ROS generated by the MDR tumor cell lines Lucena-1 and FEPS, are rapidly reduced. In this way, many anticancer drugs that act *via* generation of oxidative stress become ineffective.

When the MDR cell lines and their parental counterpart, K562, were tested for oxygen consumption measured by highresolution respirometry, the most resistant cell line, FEPS, reproducibly displayed comparatively lower values in all parameters measured (**Figure 1**). This result suggested that in FEPS, energy was probably being obtained *via* glycolysis. Gene expression (47) and protein expression (48) indicate an increase in pyruvate kinase levels in FEPS, whereas in Lucena-1 only increased protein expression was observed (48). Differences in the glycolytic pathway, with higher expression of hexoquinase 2, GAPDH and LDH, were observed using SKOV3 TR, an ovarian cell line transfected with the *ABCB1* gene (59).

Proteomic analysis of another MDR line obtained by doxorubicin selection using K562 as the parental cell, described upregulation of fructose-biphosphate aldolase A, fructose-biphosphate aldolase C, transaldolase, and alpha-enolase suggesting that the cells need more energy to survive chemical stress (53). FEPS cells,

Respiratory parameters of intact K562, Lucena-1, and FEPS cells. Routine respiration—basal respiration of intact cells; leak respiration—rate of oxygen consumption after the addition of oligomycin, that is, uncoupled respiration; coupled respiration (routine–leak); ETS—maximum respiratory capacity (induced by the addition of FCCP); ROX—rate of oxygen consumption in the presence of rotenone and antimycin A, that is, respiration not associated with electron transport system. All the parameters were corrected by ROX values. The bars marked with asterisks denote the values that are significantly different with reference to K562 cells.

selected with daunorubicin also presented upregulated fructosebiphosphate aldolase A (48).

### GAPDH, MDR, AND CELL DEATH

Among the differences observed in the present study was the variation found in the levels of GAPDH between K562 and its MDR counterparts. Using a proteomic approach (48), it was possible to verify that in relation to K562, the levels of GAPDH were 2.4 times higher in FEPS and 1.2 times in Lucena-1. These results were validated by western blot (**Figure 2**). Overexpression of GAPDH in Imatinib-resistant cells has been observed by other authors (60, 61). However, these results differ from those found by Cerezo et al. where no difference in GAPDH was observed in daunorubicin resistant cell lines overexpressing ABCB1 (62). Presumably the inconsistency reflects the well-known intrinsic differences encountered in tumor cells of different origin.

Interestingly, a recent report describes a strategy to circumvent the ABCB1 transporter activity by transferring constructs that specifically inhibited GAPDH into target tumor cells by using liposomes. Such treatment was effective *in vitro* and *in vivo* (63).

Therefore, the possibility exists that GAPDH plays a role in the resistance observed in some MDR tumors. Various non-glycolytic roles have been assigned to GAPDH (64). Post-translational modifications of GAPDH may dictate subcellular localization and different functions (31, 65). It has been proposed that Sirtuin1 expression retains GAPDH in the cytosol (66). However, no differences in gene or protein expression of Sirtuin 1, were observed in Lucena-1 or FEPS when compared with their parental cell line K562. As mentioned before, GAPDH translocates to the nucleus where it binds to DNA and participates in a number of DNA-dependent processes (67). However, the lack of differential expression of Sirtuin 1 in Lucena-1 and FEPS does not invalidate the hypothesis of GAPDH mediated action in the nucleus. GAPDH trafficking may occur by a number of mechanisms.

were loaded in each lane for the detection of glyceraldehyde-3-phosphate dehydrogenase (GAPDH), pro-caspase-7, cleaved PARP, and Hsp70 (loading control) by western blot. Densitometric analysis of each lane was calculated using ImageJ Software. The data are expressed as arbitrary units and represent the mean of two independent experiments.

For example, it is known that GAPDH can bind to microtubules and may thus get access to several intracellular organelles, including the nucleus (68).

It has been suggested that GAPDH is an inhibitor of caspase independent cell death (CID) (61, 69, 70). Death evasion to a number of chemotherapeutic drugs is the characteristic of cells displaying the MDR phenotype. Defects in the apoptotic process involving caspase activation to induce cell death have been observed in a number of multidrug resistant cell lines (62, 71).

Apoptosis is a result of caspase activation as a consequence of mitochondrial permeabilization and cytochrome c release. However, cells might also be killed following mitochondrial permeabilization even when caspase activation is inhibited, in the process known as CID. It has been suggested that, in such cases, cell death might occur as a result of a collapse of mitochondrial function, or the release of other proteins that could mediate the death process such as apoptosis-inducing factor (AIF), Smac/ Diablo, HtrA2/Omi, Endonuclease G (72).

Mitochondrial permeabilization leading to cytochrome c release, and subsequent caspase activation, involves Bcl-2 family members having pro and anti-apoptotic properties. No difference in Bcl-2 levels could be detected when Lucena-1 and FEPS were compared with their parental non-MDR cell line K562 (46). On the other hand, the inhibitor of apoptosis survivin is increased in these cells, similarly to what has been described in K562/ ADR (73) and other MDR cell lines (74, 75). Survivin suppresses cell death *via* caspase inhibition. Therefore, in a situation where caspases are inhibited, cell death could be a result of CID.

In a study, using CML cells, Lavallard et al. described that the tyrosine kinase inhibitor, Imatinib, was able to induce cell death in Bcr-Abl-positive cells by both caspase-dependent and independent manner (61). To induce CID, Imatinib was added to CML cells treated with caspase-inhibitors. In such situation cells transfected with GAPDH were protected from CID, however, these same transfected cells were not protected from Imatinib-induced apoptosis when no caspase-inhibitors were used. Furthermore, Imatinib-resistant K562 cells spontaneously overexpressed GAPDH compared with parental K562 and were protected from CID (61). This finding is in agreement with the results obtained with both Lucena-1 and FEPS MDR cell lines that also showed resistance to Imatinib and displayed higher levels of GAPDH compared with the parental K562 (**Figure 2**).

In experiments where caspase activity was inhibited but cytochrome c release occurred, it has been suggested that GAPDH acted by increasing the glycolytic metabolism and generating ATP as well as translocating to the nucleus where it is involved in the expression of Atg12 (70). Using GAPDH mutants, that either supported ATP production, but did not translocate to the nucleus or presented nuclear function albeit unable to produce ATP, these authors verified the requirement for the dual role played by GAPDH (70). The induction of Atg12 expression following mitochondrial permeabilization leads to autophagy with the removal of damaged mitochondria and subsequent cell survival (69).

Another death process involves the release of AIF from the mitochondrial intermembrane followed by its translocation to the nucleus. This step is a result of overactivation of the nuclear enzyme poly (ADP-ribose) synthetase 1 (PARP-1) (76). In the mitochondrial intermembrane space, AIF co-localizes with Hsp60. Due to its oxyredutase activity, AIF might act as a scanveger in the mitochondrial electron transport. However, AIF nuclear activity is independent of oxyredutase activity as mutations in the oxyredutase domain do not inhibit death-induction when it translocates to the nucleus (77). Once present in the nucleus AIF induces chromatin condensation and large-scale DNA fragmentation. The use of AIF mutants lacking the DNA-binding property abrogated cell death in spite of the preservation of AIF's nuclear translocation (78).

In response to oxidative/nitrosative stress GAPDH binds to Siah and translocates to the nucleus where it activates PARP-1 (79). In the nucleus PARP-1 functions as a sensor to regulate celular DNA repair. PARP-1 overactivation, leading to AIF translocation to the nucleus, is a result of an attempt to restore damaged DNA. Usually, after repair is completed, the PAR chains are degraded. However, in cells with severely damaged DNA, nuclear PARP-1 is extensively activated and promotes the synthesis of an excess of PAR polymer. This reaches toxic levels and in the cytosol constitutes a death signal inducing AIF nuclear translocation (76, 80). The activation of PARP-1 results in the depletion of the cellular NAD<sup>+</sup> and ATP pools. PARP-1 may be cleaved *in vitro* and *in vivo* by caspase-3 and caspase-7 originating two fragments an 89-kD catalytic fragment and a 24-kD DNA binding domain, capable of arresting the process (81). Similar fragments of 89 and 24 kD were detected in the MDR cell line FEPS, whereas only the 89 was observed in Lucena-1 cells (**Figure 2**). The relationship between PARP-1 activity and drug resistance is unclear but ABCB1 activity occurs at the expense of ATP. It has been described that cells from animals knockout for PARP-1 showed and increased ABCB1 expression and activity (82).

#### INTRACELLULAR CALCIUM AND MDR

The homeostatic control of cytosolic calcium concentration is of fundamental importance and changes in intracellular microenvironmental calcium levels can impact on cell survival, diverse cell functions, and cell death.

Proteins of the S100 family are calcium sensing proteins important in maintaining the homeostasis of the cell (83). Reports show that the calcium binding protein S100A6, also known as calcyclin or CACYBP, interacts with GAPDH (84). S100A6 is present in tumor cells as well as in normal fibroblasts, smooth, and heart muscle cells (83). This protein, as well as its gene expression, is increased twofold in FEPS compared with K562 (48). However, it is not clear how the interaction of S100A6 and GAPDH might play a role inhibiting CID.

Sorcin is another calcium–protein associated with MDR (85–87). Using K562 cells overexpressing sorcin, it was observed that ABCB1 was upregulated. The opposite was also true. In MDR-K562 cells selected by exposure to doxorubicin it was found that sorcin was upregulated, suggesting that ABCB1 and sorcin may regulate the expression of each other (85). When sorcin was analyzed using the MDR cell line Lucena-1 selected by vincristine, an increased gene and protein expression was also observed compared with the parental K562 (47, 48). This result is in agreement with the proteomics of another doxorubicininduced K562 where the protein sorcin was also increased (53).

Despite the fact that the relationship between sorcin and ABCB1 has been known for a number of years (88), the meaning of such relation is still unknown. Differences in intracellular calcium levels have been reported in cells presenting the MDR phenotype related to ABCB1 overexpression (89). Some of these differences could be attributed to increased sorcin levels but in most cases, a causal relationship was not established.

Cells transfected with cDNA for *ABCB1* did not mobilize calcium when the SERCA inhibitor thapsigargin was used (90). The explanation proposed by Gutheil et al. was that thapsigargin was being extruded. However, no calcium mobilization could be induced in Lucena-1 cells using thapsigargin in a condition in which the inhibitor could not be extruded from the cell, suggesting that drug efflux could not fully explain the lack of mobilization (91). Considering that the main intracellular calcium store is the endoplasmic reticulum (ER) and that thapsigargin elicits ER-stress inhibiting the ER calcium pump, this is another deathinducing mechanism that is altered in MDR cells. Similarly, other workers using two types of resistant cells, one selected with vincristine and one with a stable transfection with a gene encoding ABCB1, also described the reduced sensitivity of MDR cells to thapsigargin (92, 93). Moreover, using immunofluorescence, they observed differences in the localization of the ER proteins ryanodine receptor (RyR), inositol 1,4,5-trisphosphate receptor and calnexin, between the MDR cells and their parental (92). Despite the fact that no statistical difference was observed in the amount of GAPDH comparing the two types of resistant cells with the parental one, treatment with thapsigargin decreased the protein content of GAPDH as well as ABCB1, suggesting a relationship between the two (93).

### CONCLUSION

The MDR phenotype is a very complex phenomenon. There is now a growing awareness that drug resistance transcends the ABCB1 transporters and involves other elements associated to metabolic reprogramming. Accordingly, the enzymes of the glycolytic pathway have been shown to exert several regulatory roles that bear on drug resistance. Therefore, pharmacological interference studies on the MDR phenotype may have a better chance to succeed if they are expanded by include as potential targets GAPDH and other enzymes of the glycolytic pathway. In this review, many ancillary roles of such enzymes have been commented and underlined those that affect, for instance, the supply of ATP for drug extrusion purposes. Furthermore, attention was called to the fact that cells displaying the MDR phenotype also displayed enhanced expression/activity of enzymes involved in the maintenance of redox homeostasis of tumor cells. Thus, chemotherapy aiming at the combined abrogation of drug resistance and careful modulation of the redox equilibrium may open up alternative avenues for the control of tumor growth and metastasis.

### AUTHOR CONTRIBUTIONS

VR wrote and reviewed the article. RV and JQ contributed to the writing and data gathering. JQ contributed to the editing and submission of this manuscript. FR contributed to the writing and reviewed the article. MR contributed with the experiments involving the high resolution respirometry (oxygen consumption by the cells) which was included in the paper as **Figure 1.**

### FUNDING

This work received financial support from the following Brazilian agencies: Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ), and Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES-PROCAD). RV and JQ were recipients of PhD fellowships from CNPq.

#### REFERENCES


cell line. *Jpn J Cancer Res* (1989) 80:469–74. doi:10.1111/j.1349-7006.1989. tb02338.x


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

The reviewer CVB and the handling Editor declared their past shared affiliation.

Copyright © 2018 Vidal, Quarti, Rodrigues, Rumjanek and Rumjanek. 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.

# Metabolic Reprogramming in Thyroid Carcinoma

#### *Raquel Guimaraes Coelho1 , Rodrigo S. Fortunato2 and Denise P. Carvalho1 \**

*<sup>1</sup> Laboratório de Fisiologia Endócrina, Instituto de Biofísica Carlos Chagas Filho, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil, 2 Laboratório de Radiobiologia Molecular, Instituto de Biofísica Carlos Chagas Filho, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil*

Among all the adaptations of cancer cells, their ability to change metabolism from the oxidative to the glycolytic phenotype is a hallmark called the Warburg effect. Studies on tumor metabolism show that improved glycolysis and glutaminolysis are necessary to maintain rapid cell proliferation, tumor progression, and resistance to cell death. Thyroid neoplasms are common endocrine tumors that are more prevalent in women and elderly individuals. The incidence of thyroid cancer has increased in the Past decades, and recent findings describing the metabolic profiles of thyroid tumors have emerged. Currently, several drugs are in development or clinical trials that target the altered metabolic pathways of tumors are undergoing. We present a review of the metabolic reprogramming in cancerous thyroid tissues with a focus on the factors that promote enhanced glycolysis and the possible identification of promising metabolic targets in thyroid cancer.

#### *Edited by:*

*Simona Pisanti, Università degli Studi di Salerno, Italy*

#### *Reviewed by:*

*Anna Krzeslak, University of Łódz*′*, Poland Anna Maria Porcelli, Università degli Studi di Bologna, Italy Sybille Mazurek, Justus Liebig Universität Gießen, Germany*

> *\*Correspondence: Denise P. Carvalho dencarv@biof.ufrj.br*

#### *Specialty section:*

*This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology*

*Received: 17 November 2017 Accepted: 09 March 2018 Published: 23 March 2018*

#### *Citation:*

*Coelho RG, Fortunato RS and Carvalho DP (2018) Metabolic Reprogramming in Thyroid Carcinoma. Front. Oncol. 8:82. doi: 10.3389/fonc.2018.00082*

Keywords: glycolysis, glutaminolysis, Warburg effect, thyroid cancer, hypoxia-inducible factor, hexokinase, AMP kinase, mammalian target of rapamycin protein

### INTRODUCTION

Thyroid cancers are the most common endocrine tumors and are more prevalent in women and elderly individuals (1, 2). Although the incidence of thyroid tumors can be high in a population (1), epidemiological studies indicate that only a small fraction of tumors are malignant (1, 3). Some rare thyroid malignancies that derive from the follicular thyroid epithelia are poorly differentiated and frequently metastasize early (4). In contrast, differentiated thyroid carcinomas (DTCs) generally exhibit a good prognosis and excellent outcomes (3, 5, 6).

The therapy for intermediate and high-risk DTCs, includes a combination of surgery, radioiodine ablation, and thyroid stimulating hormone (TSH) suppressive therapy. However, although DTCs are slow-growing tumors, disease recurrence can occur (4–6). In approximately 10% of DTC recurrence cases, tumor progression leads to a more aggressive phenotype, metastatic spread, and further loss of iodide uptake ability (4–6).

In the past several years, targeted therapeutic approaches have been developed as an option to control disease progression. Unfortunately, multikinase inhibitors that target angiogenesis and oncogenic pathways have deleterious side effects and do not result in a cure. Although a significant increase in the progression-free survival rate has been observed with the use of multikinase inhibitors, the diversity of tumor types, and tumor resistance that develops during progression impede this unique therapeutic strategy (5, 6). However, tumor metabolic behavior is known to become quite different as cells to transform into malignant cells. Interestingly, some metabolic feature changes are observed in several tumor types. Although the physiological function of the thyroid gland is very well described, its metabolic control and adaptations remain elusive, especially in thyroid cancer. In this review, we discuss some metabolic adaptations identified in thyroid carcinoma that could be used as future therapeutic targets in this disease.

### MAIN MOLECULAR EVENTS RELATED TO THYROID CARCINOGENESIS

The incidence of thyroid cancer has increased in many countries compared to that of other human cancers (7). Approximately 90% of non-medullary thyroid malignancies that originate from thyroid follicular cells are classified as well-DTCs. DTCs are subdivided into follicular thyroid carcinoma (FTC) and papillary thyroid carcinoma (PTC), the latter of which is more prevalent, accounting for approximately 80% of DTCs (8, 9). The oncocytic or Hurthle cell tumors represent approximately 3–5% of follicular thyroid neoplasms (10), and they may be benign (variant of follicular adenoma) or malignant (variant of follicular carcinoma and variant of papillary carcinoma). The main characteristic of Hurthle cell carcinomas is the presence of at least 75% large oxyphilic cells that are characterized by abundant mitochondria (11), and previous exposure to radiation might be a risk factor for the development of Hurthle cell carcinoma and some subtypes of papillary thyroid cancers (8–16).

Undifferentiated thyroid carcinomas represent less than 5% of thyroid malignancies and are frequently associated with disease recurrence and death (4, 8, 9). Finally, medullary thyroid carcinoma, which derives from parafollicular C cells, produces calcitonin and accounts for approximately 5% of thyroid carcinomas (9).

The different morphologic subtypes of DTC are due to specific genetic alterations. RAS is a proto-oncogene that encodes a family of GTPases that are activated through tyrosine kinase receptor pathways involved in the regulation of cell differentiation and proliferation. RAS mutations can be found in 20–25% of all human tumors and in up to 90% of pancreatic cancers (17). Regarding thyroid cancer, RAS mutations are found in approximately 10% of thyroid cancer cases, mainly the follicular variant (18–20). Mutations of the proto-oncogene RAS induce changes in Ras protein, leading to its constitutive activation inside the cell. Although the prevalence of RAS mutations in the thyroid is low, they are associated with aggressive behavior in several other types of cancer (21–23).

Another human gene involved in thyroid carcinogenesis is BRAF. The B-Raf protein is a serine/threonine kinase that is activated downstream of Ras and is involved in cell growth control (24). Mutations in B-Raf induce its constitutive activation, subsequently activating the downstream mitogen-activated protein kinase (MAPK) signaling pathway (20, 24). Although other mutations have been described, BRAFV600E (the substitution of valine for glutamic acid in residue 600) is the most frequent mutation (24). In PTC, BRAF is the predominant mutation (30–40%) and is considered an initiating event in papillary thyroid carcinogenesis (18, 25–27).

Genetic alterations in the RET gene have also been found in several types of cancers (28). RET encodes a transmembrane protein receptor with an intracellular portion containing a tyrosine kinase that triggers its autophosphorylation, initiating intracellular signaling related to the stimulation of the RAS/ ERK and PI3 kinase/AKT cascades (25, 28, 29). In addition to BRAF, RET mutations are also responsible for thyroid cancers (9, 16, 25). Somatic point mutations in RET are associated with familiar or sporadic medullary thyroid cancer, since RET is normally expressed in C cells, but not in follicular thyroid cells (29). In PTC, RET translocations (RET/PTC) can be identified in approximately 20% of the cases (9).

In some tumors, PAX-8, which encodes a transcription factor associated with thyroid development, has been implicated in carcinogenesis (30). Tacha et al. (30) found that mutated PAX8 is expressed in some follicular thyroid cancers due to somatic rearrangement leading to the fusion of PAX-8 with PPARγ1 (peroxisome proliferator-activated receptor gamma 1) (30). In FTC, the frequency of PAX8/PPARγ1 rearrangement is estimated to be approximately 30%, but this is not observed in PTCs. RAS mutations are also found in FTC (9, 20, 21).

The therapeutic approach for thyroid cancer may depend not only on the tumor initial mutational status, which leads to different cell biology characteristics, but also to hallmarks related to tumor progression. Some of these molecular changes result in specific metabolic alterations that might contribute to metastasis and a worst prognosis.

### CANCER CELL METABOLISM

In the past several years, there has been significant interest in the metabolic reprogramming of cancer cells. In general, non-tumor cells use energy substrates, such as glucose and fatty acids, to generate energy under aerobic conditions. Glucose metabolism is initiated by glycolysis, the pathway that converts one glucose molecule into two molecules of pyruvate, which are transported to the mitochondria for oxidation. The glycolytic pathway is generally coupled with the mitochondrial tricarboxylic acid (TCA) cycle due to the action of the pyruvate dehydrogenase (PDH) protein complex that converts pyruvate into acetyl-CoA. The TCA cycle consists of successive reactions that lead to the transfer of electrons to NAD<sup>+</sup>/FAD<sup>+</sup> for the generation of NADH/ FADH2, forming a wide range of metabolic intermediaries that are involved in various biosynthetic routes. The TCA cycle allows electron transfer to oxygen and generates a proton gradient across the inner mitochondrial membrane that is necessary for ATP synthesis in a process called oxidative phosphorylation (OXPHOS), a metabolic strategy that enables the cellular production of a greater amount of ATP (**Figure 1**).

Under physiological stress conditions, such as hypoxia or rapid intracellular ATP decreases, the cell increases its anaerobic metabolism, producing higher amounts of lactate from pyruvate. Interestingly, most cancer cells show a constitutive stress metabolic phenotype due to their high proliferation rates, which induces an elevated ATP demand compared to non-tumor cells. Therefore, cancer cells show variable energy substrate selection and a metabolic shift occurs to maintain cell proliferation and survival (31–35). **Figure 1** summarizes the major metabolic modulation on energy flux in tumor cells.

Decades ago, Otto Warburg described the first tumor-specific metabolic characteristic, the so-called Warburg effect or aerobic glycolysis, which is considered the key metabolic hallmark of cancer (31). The Warburg effect is an alteration of cellular metabolism in which the glycolysis pathway is upregulated even in the presence of normal or high oxygen tension, resulting in

the phenomenon of aerobic glycolysis. However, such metabolic reprogramming in cancer cells represents an energy compensation strategy, since the efficiency of ATP production by glycolysis is much lower than that by OXPHOS (36). Part of this strategy is the upregulation of plasma membrane glucose transporters (GLUTs), a feature that is relatively common in many tumor types and is easily identified by non-invasive imaging positron emission tomography (PET) using fluorodeoxyglucose (34–37). In addition to increased glucose uptake, changes in key enzymes involved in glucose utilization can also be observed (34, 35). Some tumors show increased expression and activity levels of hexokinase (HK) isoforms, phosphofructokinase (PFK1), 6-phophofructo-2-kinase/fructose-2,6-biphosphatase (PFK2), aldolase (ADO), phosphoglycerate kinase (PGK), enolase (ENO), and pyruvate kinase (PK) (37). All these changes can increase pyruvate production from glucose breakdown.

The higher glycolytic flux observed in cancer cells, is not accompanied by increased rates of pyruvate oxidation, but lactate fermentation seems to be higher (**Figure 1**). Although this phenomenon is not fully understood, in recent years, significant progress has been made regarding the underlying molecular mechanisms related to neoplastic transformation and the Warburg effect (32–39). First, the lactate dehydrogenase enzyme (LDH) consumes part of the pyruvate formed by glycolysis to regenerate NAD<sup>+</sup> from the NADH produced by glucose breakdown, allowing a higher rate of glycolysis. Second, the LDH is a reversible enzyme that can generate NADH and pyruvate, thus contributing to mitochondrial OXPHOS. Third, both lactate and pyruvate can be transported from the cytosol to the mitochondria, or they can be secreted out of the cell. Lactate transport is mainly executed by monocarboxylate transporters (MCTs), a family of more than 14 types of transporters (38). Secreted lactate and pyruvate can be taken up by adjacent cancer cells and provide a feedforward mechanism for tumor growth, a phenomenon that is called as the reverse Warburg effect (39–41).

In adverse conditions, such as fluctuating oxygen tension, which is observed in solid tumors in the setting of poor blood vessel irrigation, glycolysis allows cancer cells to live in hypoxic conditions. However, survival at a lower oxygen tension has consequences, such as excessive lactate production and decreased extracellular pH, which leads to a microenvironment that favors the extrusion of tumor cells from primary tissues (33, 36, 40, 41). Therefore, aerobic glycolysis can generate lactate, an important metabolite that favors tumor invasion and progression, which is advantageous for proliferating cells (33, 35, 36, 39). Consequently, the idea that the Warburg effect is due to mitochondrial dysfunction has changed. In many tumor models, OXPHOS changes are important to maintain growth and progression, indicating that OXPHOS may be an important metabolic target in cancer treatment (42–46).

During the process of tumor metabolic reprogramming, many cancer cells show greater glutamine dependence for their survival and proliferation (47–49). The elevated consumption of glutamine, a non-essential amino acid, has been documented in some tumors by assays that evaluate the uptake of two radionuclides, 18F or 11C (49). High 18F-glutamine uptake was related to increased sodiumdependent neutral amino acid transporter type 2 (SLA1A5) expression and by upregulated glutaminase (GLS) in several tumor models (47–49). GLS initiates glutaminolysis by converting glutamine to glutamate. The destination of glutamate depends on divergent routes. Interestingly, this pathway is involved in the maintenance of the TCA cycle and anabolic processes through the synthesis of non-essential amino acids through transamination, nucleotides (purines and pyrimidines), and fatty acids. Glutamate formed in the cytosol is transported into mitochondria, where it can be converted into α-ketoglutarate by distinct reactions catalyzed by: (a) glutamate–pyruvate transaminase, producing alanine and α-ketoglutarate; (b) glutamate–oxaloacetate transaminase (GOT), which transfers the amino group from glutamate to oxaloacetate producing aspartate and α-ketoglutarate; and finally

(c) glutamate dehydrogenase (GLUTD). Together, these reactions represent the major anaplerotic pathways for the synthesis of TCA cycle intermediaries secondary to glutamine metabolism. Glutamine consumption allows the cyclic resynthesis of citrate, which is directed to the formation of fatty acids or the synthesis of amino acids (**Figure 2**). Moreover, glutamine metabolism participates in the generation of antioxidant agents and can also act in cell signaling (47, 49).

All these alterations of glucose and glutamine metabolism observed in cancer cells are synergic. The high glucose uptake linked to energy generation and lactate production reduces oxygen consumption. Furthermore, mitochondrial function is maintained by glutaminolysis and can support biosynthetic processes. Several studies have provided evidence that oncogenic alterations in cancer cells reprogrammed glucose and glutamine metabolism, leading to energy stress that sustains anabolic processes, which are crucial to cancer cell proliferation and progression (31–36, 40, 41, 44, 47, 49).

Figure 2 | Glutaminolysis and glucose metabolism in cancer cells. The higher glycolytic pathway contributes not only to the production of ATP per glucose consumed, but also feeds other biosynthetic pathways. Deviation of glyceraldehyde-3P to glycerol-3P is important for lipogenesis. Glucose-6P can also shift toward the phosphate pentose pathway that provides ribulose-5-phosphate and NADPH to nucleotide synthesis. On the other hand, glutamine uptake maintains the anaplerotic process in the TCA cycle. Glutamine is taken up *via* the transporter ASCT and is converted into glutamate. Glutamate together with pyruvate can be metabolized by GPT producing α-ketoglutarate and alanine; glutamate is metabolized producing α-ketoglutarate and aspartate by GOT; or glutamate is metabolized by glutamate dehydrogenase (GLUTD) forming α-ketoglutarate. All these reactions contribute to support the TCA cycle. Citrate outside mitochondria contributes to the *de novo* formation of fatty acids and aminoacids. Cancer cell metabolism is also characterized by the upregulation of lactate dehydrogenase to facilitate the conversion of pyruvate to lactate, which is then secreted to the tumor microenvironment *via* the MCT. Abbreviations: ASCT, Asc-type amino acid transporter; ETC, electron transport chain; GLUT1/3, glucose transporter 1 or 3; TCA, tricarboxylic acid; GOT, glutamate-oxaloacetate transaminase; GPT, glutamate–pyruvate transaminase; MCT, monocarboxylate transporter.

#### THYROID CANCER AND METABOLISM

Extensive documentation is available describing TSH as the main regulator of the function, proliferation, and metabolism of normal thyroid follicular cells, and well-differentiated thyroid cancer (50–56). In thyrocytes, the signaling network of TSH involves intermediates, such as protein kinase A, protein kinase C (PKC), phosphatidylinositol 3-kinase (PI3K), and MAPK. TSH activation increases glucose metabolism and oxygen consumption to support iodide transport and thyroid hormone (T3 and T4) synthesis (50–54).

Despite the importance of aerobic glycolysis, it is estimated that the ATP content produced by normal thyroid cells is mainly derived from mitochondrial respiration with low glucose consumption (55, 56). Moreover, Mulvey et al. (56) showed that glycolysis seems to be more important to sustain the pentose phosphate pathway (PPP) than ATP production in thyroid cells. The deviation of glycolysis to the PPP in the thyroid could be important to maintain the balance of NADH/NADPH generated, which is crucial for thyroid hormone synthesis.

Regarding thyroid tumors and cellular metabolism, a major aspect is the effect of oncogenes on cell metabolic shift (32). Mutated RAS induces constitutive PI3K/AKT pathway activation independently of TSH stimulation (21, 57). In many tumors, the constitutive PI3K activation results in increased glycolysis flux (58, 59), and the PI3K/AKT pathway is crucial to translocate GLUT1 from the cytoplasm to the plasma membrane in thyroid cells (53). Recently, significant increases in glycolysis, the PPP, glutamine metabolism, and the phosphoserine biosynthetic pathway were identified in colorectal cancers with the KRAS point mutation compared to wild-type cells (59).

Guo et al. (23) showed the impact of RAS mutations on the oxidative profile, which can lead to autophagy induction *in vitro* and *in vivo* in tumors. The autophagy process is characterized by catabolic cellular self-degradation in response to periods of nutrient limitations through macromolecular intracellular recycling (60). According to Guo et al. (23), in addition to providing energy substrates, the autophagy process also preserves the mitochondrial function required for cell growth, especially in models of aggressive cancers. Several years ago, it was demonstrated that in TRβ PV/PV mice, which spontaneously develop well-differentiated FTC, synergism between the KRASG12D mutation and TRβ PV occurs, leading to MYC oncogene activation and the development of the UTC phenotype (61). Interestingly, a prior study showed that in 40% of all human cancers, deregulated MYC expression could be involved in metabolic reprogramming (62). This gene encodes the Myc transcription factor (c-Myc), a multifunctional protein that plays a role in cell-cycle progression, apoptosis, and cellular transformation (62–64). Recently, Qu et al. (64) showed that BRAFV600E signaling also increases c-Myc expression in the human PTC cell lineage.

In addition to thyroid cancer, c-Myc overexpression has been identified in various cancers (62–64) and it upregulates the expression of genes involved in glucose metabolism (**Figure 3**). The first link found between c-Myc and glycolysis was the positive regulation of lactate dehydrogenase A (LDHA), the enzyme that converts pyruvate from glycolysis to lactate (65). Subsequently, GLUT-1, HK2, PFKM, and ENO1 were also identified as c-MYC targets (66–69).

Furthermore, mutated c-MYC also increases the anaerobic status of tumors, probably due to higher glycolytic flux and downregulation of OXPHOS (62, 69, 70). Pyruvate dehydrogenase kinase (PDK) is the main regulatory enzyme of mitochondrial pyruvate consumption. This protein can phosphorylate and inactivate PDH, blocking the conversion of pyruvate into acetyl-CoA and its subsequent oxidation in the TCA cycle. In hypoxic conditions, a decrease in oxidative metabolism occurs due to the inhibition of PDH by PDK. This process is modulated by hypoxiainducible factor-1 (HIF-1), a transcription factor that promotes many cell changes in response to oxygen deprivation (66). Interestingly, cancers that harbor mutated c-Myc also increase the activity of PDK under normoxic conditions (62, 66, 69). Together, these effects contribute to increased glycolysis dependence and the development of the Warburg effect (64, 66–70). **Figure 3** summarizes HIF and MYC targets on glycolytic metabolism.

The ability of tumor cells to maintain growth under hypoxic conditions is crucial to tumor progression, and the crosstalk between HIF-1 and c-Myc has been well documented (63, 66, 69). As a tumor grows, cells that can shift their metabolism in response to differences in blood supply are selected. Low oxygen tension increases HIF-1 expression and stabilization (70). Furthermore, HIF-1 is also activated by inflammatory processes, energy deprivation, and oxidative stress (71–75). Acting together, HIF-1 and c-Myc regulate several adaptations to hypoxic environments (70). As one compelling concept of the Warburg phenotype, HIF-1 enhances glycolysis not only by increasing the transcription of all glycolytic enzymes, but also by increasing their affinity for substrates (73). Moreover, HIF-1 also increases glucose transporter expression and stimulates the inhibitors of mitochondrial metabolism (71–74). In PTC, MTC, and FTC, HIF-1 expression level has been associated with a poor prognosis and metastasis (74).

The upregulation of GLUT1, HK1, HK2, PFK1, PFK2, ENO, PKM2, LDHA, and MCTs is highlighted as the most important action of HIF-1 to increase glycolysis (72, 74–76). Furthermore, HIF-1 also inhibits the PDH complex through PDK 1 overexpression, compromising the synthesis of electron donors (NADH and FADH2) for the respiratory chain complex and promoting the accumulation of TCA cycle intermediates (77). Therefore, HIF-1 impairs OXPHOS, reinforcing the Warburg effect (74–77). Finally, HIF-1 also cooperates with the c-Myc oncogene by reducing mitochondrial biogenesis (77, 78). For this reason, HIF-1 has been described as a central agent that promotes metabolic reprogramming in many cancer cells (73, 75–78).

In addition to its metabolic effects, HIF-1 also stimulates the formation of new blood vessels, a process called angiogenesis (70–74). The steps of tumor angiogenesis induced by HIF-1 are distinct. One possible explanation for tumor HIF-1 overexpression is the loss of von Hippel–Lindau protein activity, which is a tumor-suppressor ubiquitin ligase complex responsible for HIF-1 proteasome degradation (70, 73). In this case, HIF-1 would be able to stimulate anaerobic metabolism even under conditions with minimal fluctuations in oxygen or under normoxia (70). Gatenby and Gilles (34) suggested that the blood vessels recruited

to the tumor microenvironment are disorganized and may not result in efficient cell delivery of oxygen. HIF-1 stimulates angiogenesis predominantly through the increased expression of vascular endothelial growth factor (VEGF). VEGF recruits new microvessels that allow the delivery of nutrients and expansion of the tumor mass. Some previous studies have indicated that VEGF as a risk factor of developing PTC or tumor progression (79, 80).

### METABOLIC MARKERS IN THYROID CANCER

#### GLUT1

The molecular mechanisms related to the upregulation of glucose metabolism in thyroid cancer are not yet completely understood. It has been demonstrated that 18F-FDG uptake can be stimulated by TSH in thyroid cancer tissue *in vivo* (81–84). However, some works have shown that, depending on the thyroid cancer subtype, 18F-FDG uptake can increase through a TSH-independent pathway (85, 86). In a non-tumor thyroid cell model, the action of TSH includes the activation of adenylate cyclase and PI3K (50–55). In the presence of RAS mutations, PI3K is constitutively activated, which may be partially responsible for the increased glucose uptake (57). Haber et al. (86) analyzed GLUT1 protein expression in 38 benign thyroid lesions, including follicular adenomas, Hurthle cell adenomas, nodular goiters, Hashimoto's thyroiditis, Graves' disease, and 28 cases of papillary, follicular, Hurthle cell, anaplastic, and medullary thyroid cancers. The authors showed that GLUT expression is frequently upregulated in thyroid cancers, but it is weakly expressed in benign nodules and in normal thyroid tissue. In addition, the localization of GLUT1 among thyroid cancers shows distinct patterns: (a) a circumferential plasma membrane focally present within the tumor in papillary carcinomas, (b) asymmetric distribution in the basilar membrane of tumor cells adjacent to the stroma and capillary blood supply, or (c) focally in the center of a tumor in metastatic or anaplastic carcinomas. Therefore, the degree and the localization of GLUT1 expression in thyroid cancers may have prognostic significance. Matsuzu et al. (87) studied the differential expression of GLUT genes in normal and pathologic thyroid tissues and demonstrated that the mRNA expression of GLUT 1, 3, 4, and 10 was evident in all thyroid tissues, but no differences were found between normal tissues and those from benign diseases. Recently, we demonstrated that GLUT1 is the predominant glucose transporter expressed in two non-tumor cell lines, PCCL3 (rat origin) and NTHY-ori (human origin) (88). In addition, the PTC cell line showed higher GLUT1 mRNA levels and protein expression compared to non-tumor cells, which may contribute to the elevated glucose uptake found in these cells (88). Recently, Naham et al. (89) analyzed 566 thyroid cancers, including PTC BRAFV600E and ATC, and showed that PTC not only exhibited higher GLUT1 expression, but also higher GLUT3 expression. Moreover, the highest levels of GLUT1 expression were found in ATC, indicating that GLUT expression levels may be related to tumor aggressiveness.

#### Hexokinase

In addition to the expression of GLUTs, HK expression and activity are altered in many cancers, contributing to the increased glycolytic flux (32–34, 88–92). HK catalyzes the first irreversible reaction of glycolysis and its enzymatic product is glucose-6-phosphate (G6P), which is a substrate for glycolysis or the PPP. Although there are four HK isoforms, the isoforms HK1 and HK2 seem to be overexpressed in cancer cells (92, 93). Some biochemical characteristics of HK2 are advantageous to cancer cells. First, HK2 does not have a negative regulatory site, which allows greater activity (94). Second, HK2 can bind to outer mitochondrial membrane porins and voltage-gated anion channels (VDAC), facilitating its access to newly synthesized ATP and decreasing the negative feedback of G6P for glucose phosphorylation (94–98). Third, the binding of HK2 to the mitochondria increases its activity, enhancing ATP production through the glycolysis pathway (94–97). Fourth, the binding of HK2 to VDAC improves the stabilization of the mitochondrial membrane, leading to decreased reactive oxygen species (ROS) generation (96, 97). Finally, the binding of HK2/VDAC prevents Bax/Bak unbinding from the mitochondria and apoptosis (97, 98). Therefore, HK2 seems to be important for sustained cancer growth and has been suggested to be a marker of progression and tumor aggressiveness (91–103).

The earliest studies documenting the relationship between HK activity and thyroid carcinogenesis date back to the 1980s (99, 103). Rijksen et al. (103) showed no differences in HK biochemical characteristics when comparing MTC and ATC. Only the affinity of HK for its substrate was higher in ATC than in MTC. Nahm et al. (89) studied 342 PTC samples and found higher HK2 levels in 50% of PTC samples harboring the BRAFV600E mutation. Recently, we demonstrated that HK activity is higher in the BCPAP and TPC1 cell lines than in non-tumor cells (88). Interestingly, HK activity in the cytosolic and mitochondrial fractions was significantly different between the two thyroid cancer cell lines. TPC1 cells that have RET/PTC translocation, showed equally distributed HK activity in the two subcellular fractions, while BCPAP (BRAF mutated) cells had higher HK activity in the mitochondrial fraction (88). According to Hooft et al. (100, 101), HK expression is similar between metastatic and primary DTC tumors, and positive 18FDG uptake on PET is associated with higher HK1 expression, however, mitochondriabound HK was not evaluated in this study.

#### Pyruvate Kinase

The PK enzyme catalyzes the last reaction of the glycolytic pathway. It is responsible for the conversion of phosphoenolpyruvate and ADP into pyruvate and ATP, respectively. PK monomer is composed of one active site, three main domains (denominated A, B and C), and a small N-terminal domain (104, 105). The C domain is the dimerization interface of the enzyme, and enzyme dimers can interact in a dimer–dimer configuration forming a tetrameric protein (104).

The PK isoform M1 (PKM1) is a constitutive tetramer exhibiting the highest activity that is expressed in tissues with high metabolic demand, such as brain, heart, and skeletal muscle (106, 107). PK isoform M2 (PKM2) is found in normal proliferating cells, but it is predominantly expressed in tumor cells and seems to be important for cancer cell metabolic adaptation (104, 108–110). The PKM2 isoform, in contrast to PKM1, can occur as dimers or tetramers, depending on the presence or absence of allosteric regulators (104, 105). The main positive allosteric regulator of PKM2 is fructose-1,6-biphosphate (Fructose-1,6-P2) that stabilizes the active tetrameric form of the enzyme (104, 105). PKM2 activity is negatively regulated by acetylation, phosphorylation, and oxidation. The phosphorylation of PKM2 at tyrosine 105 interferes with fructose-1,6-P2 binding and induces transformation from tetrameric to dimeric state. Also, the acetylation of PKM2 at lysine 305, or its oxidation at cysteine 358 decreases PKM2 activity (104, 105, 110). Decreased PKM2 activity leads to the accumulation of upstream glycolysis intermediates and consequently results in the deviation of metabolites to the PPP biosynthetic pathway and improved hexamine formation, nucleotide synthesis, and NADPH/NADP<sup>+</sup> formation, contributing to the maintenance of redox homeostasis (110, 111). Therefore, when dimeric PKM2 is present in a tumor, less pyruvate is produced, limiting the mitochondrial substrate, what contributes to the metabolic shift from OXPHOS to aerobic glycolysis (104, 110).

Our group has shown that the human PTC cell lines, BCPAP and TPC1, express higher PKM2 mRNA levels compared to non-tumor cells, but no differences were found in PKM1 mRNA levels. However, the total activity of PK in PTC cells carrying the BRAF mutation (BCPAP) was higher than that in both nontumor (NTHY-ori) and TPC1 (RET/PTC) cell lines, indicating that PKM2 enzymatic responses depend on the PTC driver mutation (88). It is believed that PKM isoform expression and activity change with tumor progression are linked to an increased tumor growth rate (112, 113). Feng et al. (113) showed that PKM2 expression in human PTC was associated with advanced tumor stages and lymph node metastasis. In addition, more intensive immunostaining of PKM2 was detected in PTCs harboring the BRAF mutation (113). Recently, Bikas et al. (114) showed that some thyroid cancer cells (FTC133 and BCPAP) characterized by glycolysis dependency overexpress PKM2. Although there are few studies in the literature, higher PKM2 expression in thyroid carcinomas appears to be significantly associated with the BRAF mutation, suggesting that this enzyme may be a potential therapeutic target in this type of cancer.

The relationship between dimeric and tetrameric PKM2 states has been described as a key factor for cell proliferation (111, 112, 115, 116). Several studies have shown that dimeric PKM2 can be translocated to the cell nucleus, where it directly interacts with multiple transcriptional factors and acts as a transcriptional coactivator involved in the upregulation of glycolytic genes, cell migration, and adhesion. The STAT3 signaling pathway seems to be involved in these effects of PKM2, which could be responsible for metastatic progression (112, 115, 117–119).

#### Lactate Dehydrogenase (LDH)

Lactate production plays a critical role in tumor biology. Due to their high glucose consumption, cancer cells display increased lactate production regardless of oxygen availability (31–36). Lactate is formed by the conversion of pyruvate and NADH in a reversible reaction catalyzed by lactate dehydrogenase (LDH). Although the isoforms of LDH are expressed in several tissues, LDHA is upregulated in a wide range of tumor tissues (120, 121). The LDHA converts pyruvate into lactate preferentially, while the lactate dehydrogenase B (LDHB) acts in opposite way (120). When PKM2 activity decreases, a change in the cytosolic NAD<sup>+</sup>/ NADH ratio occurs (104), what negatively impacts on the pyruvate to lactate conversion by LDHA. The downregulation of LDHA produces energy imbalance and oxidative stress leading to cell death (121, 122). Mirebeau-Prunier et al. (123) showed a lower LDHA/LDHB ratio in thyroid oncocytomas and follicular thyroid tumors. The downregulation of LDHA expression is related to the upregulation of estrogen-related receptor alpha, leading to changes in the oxidative metabolic profile of the tumor (123). In contrast, Kachel et al. (124) showed that LDHA is overexpressed in FTC and PTC compared to non-tumor tissues and its levels were even higher in UTC, suggesting that LDHA could be used as a biomarker of tumor aggressiveness. Comparing two PTC cell lineages, BCPAP and TPC1, we did not find differences in LDHA mRNA expression when compared to non-tumor cells. However, both tumor cell lineages had higher LDH activity and lactate production rates (88).

#### Monocarboxylate Transporter (MCT)

New evidence has identified MCT as an essential factor in thyroid cancer phenotype (38). MCTs are part of a family of transporters with more than 14 defined members. In the thyroid, MCT10 and MCT8 have been characterized. However, only MCTs 1–4 act in the transport of monocarboxylates, such as lactate, pyruvate, and ketone bodies (38). MCT isoform 1 (MCT1) is a bidirectional lactate transporter present in the plasma membrane that mediates the influx of lactate into the cell. MCT1 is also found in the outer mitochondrial membrane, transporting lactate from the cytosol to the mitochondrial matrix, which increases ATP production *via* OXPHOS (38). MCT isoform 4 (MCT4) is a low-affinity lactate transporter that mediates lactate efflux from cells. In tumor cells, these transporters are important for the maintenance of glycolysis under hypoxic conditions or in normoxia so that tumor cells can utilize lactate and other high-energy substrates produced (38, 121, 124). The lower lactate levels found in the media of some cancer cells in culture suggest a higher lactate uptake *via* MCTs, allowing them to generate large amounts of ATP *via* OXPHOS (121, 124, 125). On the other hand, higher lactate levels outside of cells indicate that MCT4 is responsible for the export of lactate in some cancer cells (38, 125). Therefore, according to Curry et al. (125), MCT1 could be used as an indicator of higher OXPHOS, and MCT4 can be used as a marker of glycolytic metabolism. In head and neck cancers, the expression of MCT4 has been associated with a higher tumor stage and poorer clinical outcomes (125). Curry et al. (125) also described crosstalk between PTC thyrocytes and adjacent fibroblasts with a glycolytic phenotype, resulting in the production of high amounts of lactate, which is transported outside the cell by MCT4. On the other hand, PTC cells showed greater MCT1 staining, which allows lactate intake and consumption by mitochondrial oxidation. From these adjustments, PTC cells may obtain the energy to survive, proliferate, and metastasize (126).

### ROLE OF ROS IN THYROID CANCER METABOLISM

During the process of tumor progression, some metabolic changes are associated with high levels of ROS (126, 127). High levels of ROS can generate oxidative stress due to an imbalance between ROS production and antioxidant defenses. The major source of ROS seems to be the mitochondria where they are produced as a consequence of OXPHOS. Therefore, decreased mitochondrial metabolism may be important for decreasing ROS production and protecting cancer cells from death. However, improved antioxidant defenses have been observed as a compensatory mechanism in response to ROS generation, which is often increased in several tumors, including thyroid carcinomas (126, 127).

The NADPH oxidase/dual oxidase enzymes, also called NOXs and DUOXs, are specialized sources of ROS that are widely expressed in a variety of tissues, including the thyroid gland (128, 129). In the thyroid, DUOX1 and DUOX2 are the main producers of H2O2, although thyrocytes also express NADPH oxidase 4 (NOX4) that is prominently expressed in PTC and corresponds to an important source of ROS (128–130). In fact, there is a significant positive association between BRAF oncogene activation and NOX4 expression (130).

The oncocytic tumor cells are characterized by the presence of a high number of mitochondria probably due to an imbalance between mitochondria biogenesis and destruction; these cells depend on OXPHOS for energy conservation and produce high ROS levels (131). According to Maximo and Sobrinho-Simoes (132), the increased ROS production of oncocytic cells could be secondary to the decreased activities of complexes I and III of the electron transport chain (133). In fact, the oncocytic phenotype is associated with disruptive mutations in complex I subunits genes encoded by mitochondrial DNA (134), which might be involved in tumor cell death due to inefficient metabolic adaptation, since the induction of the Warburg phenotype through the stabilization of HIF-1 alpha depends on normal complex I function that sustains tumor growth (135).

Interestingly, Paik et al. (136) showed that ROS can also increase glucose metabolism through HIF-1 activation. Using endothelial cells, they showed that increased ROS levels are accompanied by higher glucose uptake and lactate production when these cells are subjected to hypoxia. However, the glycolytic phenotype is blocked by HIF-1 stabilization, suggesting that ROS-driven HIF-1α accumulation accelerates glycolysis in endothelial cells (136).

Reactive oxygen species also change cell metabolism through AMP kinase (AMPK) protein activation (137). Intracellular ROS can stimulate AMPK, a metabolic stress-sensing cytosolic enzyme that regulates energy consumption and production processes (137). Although controversy remains in the literature relative to AMPK pathway involvement in tumorigenesis and cancer progression, several studies have demonstrated that activated AMPK causes cell-cycle arrest and has a strong antiproliferative effect in different cancer cell lines (137–139). AMPK activation also increases GLUT-1 protein expression, glucose uptake, and utilization of the glycolytic pathway in both non-tumor models and in tumor models, including PTC models (137–139). Some studies have shown that disruption of AMPK activity induces the Warburg effect in tumor cells (137, 140, 141). In an AMPKdeficient mouse model of Peutz–Jeghers syndrome, mammalian target of rapamycin protein (mTOR) is upregulated and HIF-1 promotes higher HK2 and Glut1 expression and increased glucose utilization by tumors (142). According to Bikas et al. (114), AMPK activation seems to be involved in the glucose metabolism dependence that is observed in some PTC cells. We described greater expression of the active phosphorylated form of AMPK in PTC tissue samples and in PTC tumor cell lineages in culture compared to that in non-tumor tissues (139, 143). However, new studies are necessary to understand the role of AMPK in human thyroid cancer, especially in terms of metabolic control, cell growth, apoptosis, and survival.

#### TARGETING METABOLISM IN THYROID CANCER

Considering the diversity of thyroid tumors and their distinct metabolic requirements, establishing a unique strategy for cancer therapy is not an easy task. However, some drugs for cancer therapy that target the tyrosine kinase receptors signaling cascade are currently being used. These proteins constitute a group of enzymes involved in the control of mitogen signals, energy status, cell survival, and angiogenesis (144, 145). Several tyrosine kinase inhibitors (TKIs) have been developed for thyroid carcinoma treatment, but not all of them have received approval from international health agencies (145). Many TKIs are still in the initial clinical phase of study. Sorafenib, lenvatinib, vandetanib, and cabozantinib are multikinase inhibitors approved by the Food and Drug Administration and the European Medical Agency (EMA) for use in patients with advanced thyroid carcinomas. In fact, these patients had significantly increased progression-free survival rates with the use of these agents (145).

In addition to TKIs, other downstream targets can regulate metabolic pathways (**Figure 4**). PI3K was described as a component of the insulin receptor intracellular signaling pathway and its main substrate is AKT. Once activated, AKT increases phosphorylation events in both the cytoplasm and the nucleus, stimulating glucose uptake; glycolytic flux; inhibition of apoptosis; and activation of mTOR, an important regulator of metabolism and cancer growth (145, 146). PI3K/AKT signaling is inhibited by the tumor suppressor phosphatase and tensin homolog (PTEN), which is often mutated in many types of tumors. Therefore, the loss of PTEN activity allows PI3K/AKT constitutive activation (145, 146). LY294002 is a classic molecule that inhibits PI3K and has already been tested in some carcinomas, including thyroid cancer. Currently, other new drugs targeting PI3K/AKT have been tested in several carcinomas and are in different clinical trial phases of study (144, 145).

Mammalian target of rapamycin protein is the main downstream effector of PI3K/AKT, and its upregulation is involved in tumorigenesis (144–146). Our group demonstrated that the activation of PI3K/AKT/mTOR inhibits iodide uptake by diminishing sodium/iodide symporter transporter (NIS) expression in non-tumor cells (149). Loss of iodide uptake is a negative process in the course of disease evolution, since thyroid cancer treatment depends on the administration of radioactive iodine (145). Several mTOR inhibitors (rapamycin analog), such as everolimus, have been studied in preclinical and clinical trials, including thyroid carcinomas (144, 145). Recently, phase II studies using everolimus in patients with advanced thyroid cancer have reported a partial response and sustained stable disease in a small proportion of patients (5–45%, respectively), with progression-free survival rates of approximately 11–16 months (150). The use of everolimus as monotherapy shows moderate effects, but its clinical relevance mainly derives from its relatively low toxicity profile (144, 151). The combination of another mTOR inhibitor (temsirolimus) and other drugs (MEK inhibitors) has shown synergic effects *in vitro* and is being tested in clinical trials (152, 153).

AMP kinase is a potent physiological mTOR inhibitor. Under low ATP/AMP ratio conditions, AMPK is activated, leading to metabolic adaptations, such as increased catabolism and decreased anabolism that are partially mediated by mTOR inhibition. The crosstalk between mTOR and AMPK has been extensively studied (137, 138, 140). The first evidence associating AMPK with cancer development was the discovery of liver kinase B1 (LKB1). LKB1 is a serine/threonine kinase and the major upstream kinase responsible for AMPK activation through phosphorylation (138). LKB1 is recognized as a tumor suppressor that associates bioenergetics with cell growth control and downregulation of mTOR activity through AMPK activation (138, 142). We have studied the effects of a pharmacological AMPK activator (AICAR) on PTC cell lines and observed decreased cell proliferation and the induction of apoptosis (139). These results suggest that AMPK may be a good target for thyroid cancer therapy (**Figure 4**). Epidemiological studies reported that in thyroid cancer patients who are also diabetic, metformin, an oral anti-diabetic drug, can activate AMPK, resulting in a reduced tumor size and higher remission rates (154). Interestingly, our group showed that the expression and activity of AMPK are increased in human PTC and in PTC cell lines (BCPAP and TPC1) compared to those in non-tumor tissues and non-tumor cell lines (139, 143). As PTCs are well-differentiated and slow-growing carcinomas, it is believed that increased AMPK activation could impair tumor growth.

Although tumor energy metabolism has common characteristics, most molecular targets that may be used for tumor treatment are ubiquitously expressed and function in the entire body. Therefore, it is difficult to produce specific effects only in tumor cells. Glycolytic inhibitors, such as 2-deoxyglucose (2DOG) and 3-bromopyruvate (3-BP), can be used as adjuvant agents to sensitize tumors. Unlike 2DOG, 3-BP acts in many targets and inhibits HK, GAPDH, and MCT activities, thus leading to decreased aerobic glycolysis (155). Furthermore, inhibition of PFK1 activity, a rate-limiting step of glycolysis, is also an interesting strategy in cancer therapy. This protein is activated by fructose-2,6-bisphosphate (F2,6BP), which is produced by PFKFBs. A small molecule, 3-(3-pyridinyl)-1-(4-pyridinyl)-2-propen-1-one,

and MAPK signaling pathways are shown in red. Metabolic alterations in thyroid cancer cells can also be inhibited, such as glutamine uptake (phenylacetate) and glucose metabolism (2DOG, 3-BP, DCA). Finally, other drugs can activate molecules like AMPK, such as metformin and AICAR, which are shown in green. Abbreviations: 2DOG, 2-deoxyglucose; 3-BP, 3-bromopyruvate; DCA, dichloroacetate, AICAR, 5-aminoimidazole-4-carboxamide-ribonucleoside, NOX, NADPH oxidade; MAPK, mitogen-activated protein kinase; AMPK, AMP kinase.

has been found to inhibit PFKFB3, leading to decreased glycolytic flux and slower tumor growth (**Figure 4**) (156).

Other attractive targets for cancer therapy include the inhibition of lactate production. Some previous studies have shown that loss of LDHA function by dichloroacetate (DCA) results in dramatically diminished cellular transformation or xenograft tumor growth in breast cancer (157). Glutaminolysis and amino acid metabolism are very important for tumors. Glutamine is the most abundant amino acid in the plasma and is heavily consumed by tumor cells. Therapies that decrease plasma glutamine concentrations induce tumor regression and prevent muscular catabolism, an endogenous source of glutamine. Phenyl acetate is a promising drug that can reduce the availability of glutamine in the blood and shows low toxicity (158).

Finally, another strategy targeting metabolism in tumors is diet restriction, fasting, or a ketogenic diet (a low-carbohydrate and high-fat diet) (159–162). The hypothesis is based on the glucose dependency of many tumor types. Interestingly, these diets do not increase plasma glucose levels, but produce ketone bodies that can be used as a carbon source for energy production in oxidative processes, altering the Warburg phenotype (159–162).

## CONCLUSION

In summary, although thyroid cancer studies are emerging, the mechanism of tumor progression remains unclear. As described previously, metabolic reprogramming is the hallmark of cancer cells. Recent molecular studies in thyroid cancer revealed that oncogenes and tumor suppressor genes not only control growth and apoptotic phenotypes of thyroid carcinomas, but also directly affect cellular energy metabolism and are implicated in the Warburg phenotype. The higher glucose and glutamine consumption associated with the disruption of mitochondrial OXPHOS create a favorable environment for tumor progression.

#### AUTHOR CONTRIBUTIONS

RC—writing and elaborating the figures. RF—writing and reviewing the manuscript. DC—writing and reviewing the final version.

#### REFERENCES


### ACKNOWLEDGMENTS

This work was supported by grants from Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq, 472630/2012- 1), Fundação de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ, E-26/111.447/2012), Programa de Oncobiologia, and Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES).

61 with follicular variant. *Hum Pathol* (2006) 37:824–30. doi:10.1016/j. humpath.2006.01.030


thyroid cancer through MYC up-regulation. *Neoplasia* (2014) 16:757–69. doi:10.1016/j.neo.2014.08.003


tomography uptake in well-differentiated thyroid carcinoma. *J Clin Endocrinol Metab* (2004) 89:91–5. doi:10.1210/jc.2003-031027


case reports. *J Am Coll Nutr* (1995) 14:202–8. doi:10.1080/07315724.1995. 10718495


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

# Metabolic Symbiosis and immunomodulation: How Tumor Cell-Derived Lactate May Disturb innate and Adaptive immune Responses

*Alexandre Morrot1,2\*† , Leonardo Marques da Fonseca3†, Eduardo J. Salustiano3†, Luciana Boffoni Gentile3 , Luciana Conde4 , Alessandra Almeida Filardy4 , Tatiany Nunes Franklim3 , Kelli Monteiro da Costa3 , Celio Geraldo Freire-de-Lima3 and Leonardo Freire-de-Lima3 \**

#### *Edited by:*

*Stephen G. Maher, Trinity College, Dublin, Ireland*

#### *Reviewed by:*

*Paolo E. Porporato, Università degli Studi di Torino, Italy Prashant Trikha, Nationwide Children's Hospital, United States*

#### *\*Correspondence:*

*Alexandre Morrot alexandre.morrot@ioc.fiocruz.br, morrot@micro.ufrj.br; Leonardo Freire-de-Lima leolima@biof.ufrj.br*

*† These authors have contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology*

*Received: 15 December 2017 Accepted: 09 March 2018 Published: 23 March 2018*

#### *Citation:*

*Morrot A, Fonseca LM, Salustiano EJ, Gentile LB, Conde L, Filardy AA, Franklim TN, da Costa KM, Freire-de-Lima CG and Freire-de-Lima L (2018) Metabolic Symbiosis and Immunomodulation: How Tumor Cell-Derived Lactate May Disturb Innate and Adaptive Immune Responses. Front. Oncol. 8:81. doi: 10.3389/fonc.2018.00081*

*<sup>1</sup> Faculdade de Medicina, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil, 2 Laboratório de Imunoparasitologia, Instituto Oswaldo Cruz, Fundação Oswaldo Cruz (Fiocruz), Rio de Janeiro, Brazil, 3 Instituto de Biofísica Carlos Chagas Filho, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil, 4 Instituto de Microbiologia, Departamento de Imunologia, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil*

The tumor microenvironment (TME) is composed by cellular and non-cellular components. Examples include the following: (i) bone marrow-derived inflammatory cells, (ii) fibroblasts, (iii) blood vessels, (iv) immune cells, and (v) extracellular matrix components. In most cases, this combination of components may result in an inhospitable environment, in which a significant retrenchment in nutrients and oxygen considerably disturbs cell metabolism. Cancer cells are characterized by an enhanced uptake and utilization of glucose, a phenomenon described by Otto Warburg over 90 years ago. One of the main products of this reprogrammed cell metabolism is lactate. "Lactagenic" or lactate-producing cancer cells are characterized by their immunomodulatory properties, since lactate, the end product of the aerobic glycolysis, besides acting as an inducer of cellular signaling phenomena to influence cellular fate, might also play a role as an immunosuppressive metabolite. Over the last 10 years, it has been well accepted that in the TME, the lactate secreted by transformed cells is able to compromise the function and/or assembly of an effective immune response against tumors. Herein, we will discuss recent advances regarding the deleterious effect of high concentrations of lactate on the tumor-infiltrating immune cells, which might characterize an innovative way of understanding the tumor-immune privilege.

#### Keywords: cancer, metabolism, lactate, immune evasion, cytokines

## INTRODUCTION

#### Cancer as a Metabolic Disease

Historically, cancer has been considered a product of multiple pathologies. In second century AD, the philosopher and physician Claudius Galenus was the first to employ the Greek word *onco* (swelling) for all types of tumors, leaving Hippocrates' term *karkinos* exclusively for malignant tumors. During his time, Galenus asserted that tumors were the result of "black bile" accumulation. It was only during the nineteenth century that this theory was revisited and cancer begun to be perceived as the result of acquired metabolic abnormalities (1). Nowadays, it is well accepted that cancer development and progression is modulated by the disordered growth of cells featuring self-sufficiency of

**39**

growth signals, evasion of apoptosis, angiogenesis, invasiveness, and metastasis (2, 3). When cells break free from the restraints on cell division, they start assuming inappropriate proliferation rates and distinct metabolic profiles, becoming abnormal in their own way (4, 5). Cells originating from solid tumors may gain the ability to dissolve the extracellular matrix (ECM), invade nearby tissues, reaching the bloodstream or lymphatic vessels, or remain within the boundaries of the original tissue, being characterized as malignant or benign tumors, respectively. Several genomic changes lead normal cells through malignant transformation. These changes can be anything from point mutations and deletions to chromosome rearrangements, as long as they result in irreversible changes affecting cell cycle (6). Any individual suffers several mutations in various cell types during its lifetime, due to diverse exogenous or endogenous factors. Most of these mutations are promptly corrected or lead to apoptosis. The accumulation of uncorrected mutations leads to the development of benign or malignant tumors (3). Loss of tumor suppressor factors, germline mutations, and overexpression of oncogenes are some of the changes that may collaborate for the occurrence of somatic mutations that escape DNA repair processes (7, 8).

The tumor microenvironment (TME) comprises both cellular and non-cellular components (9–11). The acellular components include the ECM, as well as soluble signals secreted by transformed and tumor-associated cancer cells. Several cell types associate with tumors, including fibroblasts, endothelial cells, and immune cells. Together, all components form an organ-like structure capable of interacting with the organism as a whole (12–14). To maintain tumor growth, several adaptations may be driven by neoplastic cells. A well-known mechanism is the formation of immature and abnormal vessels, in a phenomenon named neoangiogenesis (15, 16). In this context, both the platelet-derived growth factor and the vascular endothelial growth factor (VEGF) are recognized as the main proangiogenic signals upregulated by cancer cells during tumor growth (15, 17, 18).

The incredible proliferation rate of tumor cells can make a single mutated cell generate a tumor of ≈1 cm in diameter containing over 109 cells. Such a high proliferation ratio demands effective metabolic pathways, capable of meeting the steep energy requirements while supplying the biosynthetic precursors that maintain cell anabolism and redox balance in the neoplastic cell (19). Reprogramming of cellular metabolism has been observed in several types of cancer and is considered a hallmark of this disease (3, 20). The elucidation of these atypical metabolic activities is a lively field in the study of cancer biology, showing great potential for the development of novel therapeutic approaches. Several studies have shown that inhibition of some metabolic pathways of cancer cells is able to prevent tumor growth and metastasis (21, 22).

### METABOLIC SYMBIOSIS: A PROPOSED CONCEPT OF ENERGY MANAGEMENT BETWEEN CANCER CELLS IN THE TME

The impact of the acidosis induced by lactate and protons in the TME is a hot topic in cancer research (23–25). It is a wellestablished fact that a high enough lactate production can overcome the cellular proton buffering capability, resulting in a decrease of the cellular pH. Such condition, besides influencing the dynamics of waste and reuse of energy by cancer cells, modulates the function of distinct tumor-associated cell types as well (26–29). Several papers published over the last 10 years, demonstrated that when cancer cells experience low tension of oxygen, the hypoxia-inducible factor-1α (HIF-1α) transcription factor is stabilized, increasing glucose (Glc) uptake and secretion of substantial levels of lactate and protons out of cytoplasm by the monocarboxylated transporter 4 (MCT4) (**Figure 1B**), promoting a biochemical event termed lactic acidosis. By contrast, when cancer cells are adjacent to blood vessels and oxygen availability is sufficient, the transformed cells preferably use lactate as energy source (29–33). For this reason, lactate should not be considered a waste metabolite. In fact, it is reused by different cell subpopulations in a tumor (28, 29). Recently, Lee and colleagues (34) showed that an oxygen-regulated protein (NDGR3), which is usually degraded under normoxia *via* the prolyl hydroxylase 2/Von Hippel–Lindau (PDH2/VHL) system, becomes protected from degradation when bound to lactate. The authors demonstrated that when stable, NDRG3 is able to bind the proto-oncogene c-Raf, a serine/threonine-protein kinase, and induce activation of the Raf–ERK pathway, thus promoting cell growth and angiogenesis. However, inhibition of lactate production compromises NDRG3-mediated hypoxia responses (34).

This metabolic model of lactate reuse in the TME has been described as a metabolic symbiosis, where lactate works as a medium to convey energy from highly glycolytic/hypoxic transformed cells to more oxidative cancer cells (35, 36). In the TME, the uptake of lactate and protons by oxygenated cancer cells occurs in a dynamic way through the monocarboxylated transporter 1 (MCT1) (**Figure 1B**), which has been previously identified as gatekeeper of this metabolic symbiosis (29). In the same study, the authors demonstrated that cells with inhibited or silenced MCT1 became more sensitive to cell death, which may support lactate management within the TME as a valid therapeutic strategy. Therefore, it would plausible to speculate that the high-lactate concentration at the intercellular space might affect the functionality of diverse tumor-associated cells, including those of the innate and adaptive immune system (see sections below).

## THE EFFECT OF LACTATE IN THE TME

In normal cell metabolism, the consumed Glc is catabolized into pyruvate, which is then transported to the mitochondria to fuel the tricarboxylic acid cycle in a series of redox reactions. The resulting free electrons go through the electron transport chain (ETC), beginning the oxidative phosphorylation (OXPHOS) and ultimately leading to a high production of ATP (37) (**Figure 1A**). In the early 1920s, Otto Warburg observed that tumor cells remain in glycolytic state, constitutively absorbing Glc and converting pyruvate to lactate, in the presence of oxygen (**Figure 1A**). Lactate production is 40-fold greater in tumor cells, so the transport of lactate to the ECM by MCTs (38–41) is essential for the glycolytic switch. This metabolic behavior is named "Warburg Effect," or aerobic glycolysis, one of the main characteristics studied in

Figure 1 | Scheme summarizing metabolic differences between normal and cancer cells and metabolic symbiosis. In normal cells, glucose (Glc) is initially metabolized to pyruvate and further to carbon dioxide (CO2) through tricarboxylic acid cycle and oxidative phosphorylation (OXPHOS) processes in the mitochondria, generating 36 ATP molecules per Glc molecule consumed (A). In this process, O2 is indispensable, since it is used as the final electron acceptor (A). However, in cancer cells undergoing aerobic glycolysis (Warburg effect), Glc is broken down into pyruvate and finally converted into lactate, deviating Glc metabolites from energy production to anabolic process. This event generates two ATP molecules per Glc molecule. The panel (B) illustrates an event named metabolic symbiosis. It has been well documented that when cancer cells are near or distant of blood vessels, they are well or poorly oxygenated, respectively. It is also known that when cancer cells are subject to low oxygen tension (↓O2) hypoxia-inducible factor-1α (HIF-1α) is stabilized, increasing the transcriptional activation of genes encoding glucose transporters (GLUTs), lactate dehydrogenase A (LDHA), as well as the uptake of Glc and secretion of lactate and protons out of cytoplasm through the monocarboxylated transporter 4 (MCT4). However, when transformed cells are close to blood vessels and the availability of O2 is enough, lactate is taken by monocarboxylated transporter 1 (MCT1) and utilized as energy source after conversion into pyruvate by lactate dehydrogenase B (LDHB). In this way, lactate may not be pointed out as a waste metabolite, since it is reused by different cell subpopulations in a tumor.

cancer metabolism (19) (**Figure 1B**). Glycolysis produces ATP faster yet less efficiently than OXPHOS, forcing the tumor cell to consume much more Glc than a normal cell to produce enough energy to maintain its high proliferative status. Therefore, the glycolysis is an advantage for the tumor cell only when Glc supply is not limited. The importance of Glc for the metabolism of cancer cells is so evident that low-carbohydrate diet as a therapeutic approach for cancer patients, aiming to starve tumor cells, was described to limit the growth of incurable cancers in a pilot trial with 10 patients (42). In that regard, the uptake of a radioactive Glc analog, [18F]fluorodeoxyGlc, is used as a diagnostic tool for the positron emission tomography (FDG-PET) imaging of highly proliferative tumor regions (43, 44). Currently we know that tumor cells primarily fulfill their energetic needs by the oxidation of Glc, glutamine and other nutrients coupled to the ETC, using oxygen as the final acceptor of electrons (45, 46). In cancer cells, the anaerobic respiration is optional, and there is no mitochondrial defect (47); in fact, tumor cells still retain OXPHOS and mitochondrial activity (39). The reduced mitochondrial activity is a direct result either of oxygen deprivation or activation of HIF-1α (48, 49), which is able to promote the transcriptional activation of genes encoding glucose transporters (GLUTs), as well as glycolytic enzymes, such as lactate dehydrogenase A (LDHA) (50). When the supply of oxygen is low, LDHA is essential to sustain glycolysis and the production of ATP by regenerating NAD<sup>+</sup> form NADH. In this way, HIF-1α regulates the production of lactate, the end by-product of this reaction, which consumes two ATP but generates four ATP, generating two net ATP per Glc molecule as seen in (**Figures 1A,B**) (50). Upstream of HIF-1α and the previously discussed Raf–ERK, the Ras oncogenic pathway seems to be critical for the metabolic reprogramming observed during carcinogenesis. Overexpression of oncogenic H-RasV12 was able to drive immortalized fibroblasts to consume more Glc and to release more lactate (51). Conversely, Ras inhibition in a model of glioblastoma (GBM) effectively shuts down Glc uptake and glycolysis itself, leading to the downregulation of 12 genes from the glycolytic pathway and increased extracellular pH due to reduced lactate efflux (52). The role of PDH2 is prominent in this, since oncogenic H-Ras signaling, as well as hypoxia, triggers oxidative stress, PDH2 dimerization, and inactivation, leading to HIF-1α stabilization and ultimately the OXPHOS to glycolysis shift (53).

An immediate consequence of the Warburg effect is the accumulation of lactate and protons in the TME (23, 54). It has been shown that in patients diagnosed with different stages of cervical cancer, primary tumors exhibiting high-lactate levels often lead to the manifestation of metastatic foci (55). The same group, using human larynx squamous carcinoma cells, showed that increased lactate concentration can augment cell motility and migration, corroborating the data observed in patients (56). In addition, the use of siRNA to inhibit the expression of LDHA, whose expression can be induced by lactate itself, is able to inhibit the migration of glioma cells as well as downregulate active matrix metalloproteinase-2 (57). Increased lactic acid is also responsible for the overexpression of factors related to tumor progression, such as CD44, hyaluronic acid and transforming growth factorbeta (TGF-β) (58) (**Figure 2**), a pro-carcinogenic cytokine able to activate the epithelial–mesenchymal transition process, an event that permits dissemination of tumor cells from the primary site into the surrounding stroma, setting the stage for metastatic spread (59–61). In addition, due to its antioxidant properties, increased lactate concentrations may offer a degree or resistance against any therapy relying on the production of oxygen reactive species, such as radiotherapy (62). As it stands, further studies on the production of lactate by solid tumors represent an important step toward the understanding of tumor progression and malignancy, as well as for therapy development.

The TME is characterized by acidity and low oxygen tension (63, 64), events capable of modulating not only the growth and survival of tumor cells but also the recruitment of inflammatory cells that are reeducated in the microenvironment to favor tumor spread and metastasis. In this scenario, various inflammatory cells, such as T lymphocytes, dendritic cells (DCs), natural killer (NK) cells, and macrophages (Mϕ), acquire pro-carcinogenic

Figure 2 | Overview of immunosuppressive effects of lactate in the tumor microenvironment (TME). In a hypoxic environment, Glc enters the cell *via* glucose transporter (GLUT) 1 and is broken down into pyruvate and then in lactate, which is transported out of the cell *via* monocarboxylated transporter 4 (MCT4). The lactate produced by transformed cells culminates in an acidified TME. This phenomenon is able to suppress the anticancer immune responses, particularly through impaired T and natural killer (NK) cells activation, reduced antigen presentation, compromised dendritic cell (DC) differentiation and maturation. It also promotes the emergence of the M2 Mϕ, which secretes high levels of pro-carcinogenic cytokines, such as transforming growth factor-beta (TGF-β) and vascular endothelial growth factor (VEGF), involved in processes such as epithelial–mesenchymal transition (EMT) and angiogenesis, events implicated in metastasis and cancer progression.

properties (**Figure 2**) (65–69). Recruitment and accumulation of those cells in the TME is an essential phenomenon to sustain the tumor growth (70). The immune system's role in the first phases of tumor establishment is well described, being capable of detecting and destroying cancer cells, halting their growth and spread, in a phenomenon termed immunosurveillance (71, 72), which was initially proposed by Paul Ehrlich and later developed by Sir Frank Macfarlane Burnet and Lewis Thomas (13, 73). Defects in this event might favor tumor progression and, consequently, the acquisition of a malignant phenotype. Any cancer cell that manages to escape death triggered by immune response could still have their proliferation hindered by immune mechanisms, reaching an equilibrium. On the other hand, the immunogenicity is molded through selective pressure exerted by the immune system, in an event termed immunoediting (74–77). Consequently, novel tumor variants emerge, bearing more mutations, making them more likely to evade detection and elimination by immune effector cells like NK and CD8<sup>+</sup> T cells (74, 78). The immunoediting stage is the longest phase, and it is characterized by the dynamic coevolution of cancer and immune cells (78–80). Ultimately, cells reach an escape phase, where the accumulation of edited cells drives tumor growth and the manifestation of clinical symptoms (74). The presence of the immune system in the TME undoubtedly compromises tumor growth and, in fact, correlates with favorable prognosis in some cancer types, such as renal, ovarian, colorectal, and breast cancers (74, 81). The expression of molecules able to compromise cell-to-cell interaction (3, 82–84), as well as soluble factors such as VEGF (85), cytokines (86–88), prostaglandin E2 (PGE2) (89), soluble Fas and FasL (90), and soluble MICA (91), all contribute to the appearance of multifaceted local and regional immunosuppressive networks (92–94). For example, the occurrence of the IL-4, TGF-β, IL-13, and IL-10 cytokines in the TME induces the emergence of M2 instead of M1 Mϕ (86, 87). In addition, secretion of nitric oxide, IL-10, arginase-1, IL-6, and VEGF promotes cell death and avoids the antitumor function of immune cells (95–99).

As stated earlier, the TME is rich in lactate (38–41), an immunosuppressive soluble factor that promotes cancer development (54, 74, 100). Particularly, several studies have shown that tumorderived lactate is capable of inhibiting the activation of immune cells such as monocytes, Mϕ and T lymphocytes (101–103). It has been demonstrated that high LDHA levels are deeply correlated with tumor size, as well as with the clinical stages of the disease (104, 105). Accordingly, the infiltration of immune cells in the TME correlates with high LDHA expression and/or activity (106). Besides lactate accumulation in the primary tumor site, its immunosuppressive properties can outspread to distant sites, thus stimulating invasion and metastasis in a paracrine fashion (107) (**Figure 2**).

Despite being mainly produced by skeletal muscle, various tissues generate lactate, and its elimination in healthy conditions is handled primarily through the liver and secondarily through the kidneys (108). The citric acid cycle is also a source of lactate, as pyruvate can be diverted to lactate and NAD<sup>+</sup> generation through LDH activity (109, 110). The high amounts of lactate in the extracellular microenvironment contribute to lowering the extracellular pH, which can be as low as 6.0–6.5 (111), producing acidosis and inducing angiogenesis and a reduction in efficacy of the immune system (101, 112, 113). Tumor-associated immune cells from myeloid and lymphoid origin can be modulated by hypoxic conditions as well as high levels of lactate, then favoring the acquisition of malignant phenotypes (23, 64, 114–116).

#### LACTATE AND MYELOID CELLS

Over the last 15 years, several studies demonstrated that tumorderived lactate modulates the functionality of immune cells, contributing to the establishment of an immunosuppressive microenvironment, which favors the developing of tumors (106, 117–119). Lactate promotes the development of myeloid-derived suppressor cells (MDSCs), the most prominent bone marrowderived cell population that exerts broadly immunosuppressive functions (106). In the TME, MDSCs potently suppress both innate and adaptive immunity by preventing the maturation of DCs, suppressing NK-cell cytotoxicity, inhibiting T-cell activation, and favoring the differentiation of regulatory T cells (**Figure 2**) (117).

In addition, lactate suppresses monocytes' LPS-induced activation by influencing their gene expression. Particularly, the expression of most LPS-induced genes was significantly delayed in the presence of lactate, including TNF, IL-23, CCL2, and CCL17. These effects are mediated by delayed LPS-induced phosphorylation of protein kinase B (AKT) and degradation of IkB, with reduced nuclear accumulation of NFκB (119). Furthermore, lactate stabilizes the transcription factor HIF-1α in monocytes, which ultimately promote the expression of PGE2 and the growth of human colon cancer HCT116 cells (120).

Another suppressive function of lactate is to impair the differentiation of monocytes into Mϕ or DCs in the TME (118, 121–123) and in non-tumor conditions (103, 124) (**Figure 2**). It was reported that lactate blocks LPS activation of bone marrowderived Mϕ (BMDMs) (123), and also, in hypoxia or normoxia, lactate drives tumor-associated Mϕ polarization to the "tumor friendly" M2 profile (**Figure 2**) (125, 126). Mechanistically, lactate stabilizes HIF-1α, which leads to the transcription of a broader set of M2-associated genes, including VEGF, TGF-β and arginase-1, as well as Fizz1, Mgl1, and Mgl2 (**Figure 2**) (98, 125–128). The role of PDH2 as a regulator of the metabolic reprogramming in Mϕ was observed in both RAW264 cells and in primary BMDM, since transfection with shRNA targeting PDH2 or conditional PDH2 knocking out led to decreased ATP levels along and increased lactate release into the medium (129). M2 Mϕ and their products favor tumor growth and metastasis by suppressing antitumor immune responses, activating and enhancing angiogenesis. Particularly, VEGF triggers the development of neovascularization of the tumor. Similarly, arginase-1 plays an indirect role in angiogenesis through reorganization of the tumor ECM and contributes for the generation of essential metabolites during cell division, such as polyamines, supporting cancer cells growth (130–133). The importance of arginase-1 in tumor development was demonstrated by the use of arginase-1 KO mice, which presented tumors 50% smaller than tumors from wild type mice (125, 134). Finally, distinct studies have shown that when present in high levels, lactate inhibits antigen presentation and IL-12 production by DCs (54, 135) (**Figure 2**) and enhances IL-10 production as well, generating an immunosuppressive profile in the TME (136, 137).

## LACTATE AND LYMPHOID CELLS

The immunobiological effects of lactate on immune cells from lymphoid origin have been mainly investigated in NK and T cells. The cytotoxic effect mediated by both cell types is of fundamental importance in immunological surveillance against the emergence and spread of malignant disease. NK cells represent large granular lymphocytes that induce their antitumor responses through the ligation of particular cell-surface receptors (138), such as the natural killer group 2, member D receptor (NKG2D), which induces the release of cytotoxic granules that promote the lysis of cancer cells (139, 140).

An elegant study developed by Husain and colleagues (106) revealed that the low production of lactate in LDHA-depleted tumors was able to improve the cytolytic functions of NK cells. However, when NK cells where pretreated with lactate *in vitro*, its cytolytic property was compromised and/or abrogated. The molecular mechanism responsible for such effect was investigated, and the authors demonstrated that the decline of NK cytotoxic activity was promoted by the lower expression of granzyme and perforin in lactate-treated cells (106). Furthermore, it was described that lactate works as a potent inhibitor of histone deacetylases, suggesting that lactate might be able to regulate (at transcriptional level) several genes involved not only in cell metabolism but also in immune responses, such as NCR1/ NKp46, an activating NK cell receptor (141, 142).

In 2014, Crane and coworkers demonstrated that GBM cells secrete LDH-5, an enzymatically active isoform of LDH (143), that besides catalyzing the conversion of pyruvate to lactate in an efficient way (144), is also capable to upregulate HIF-related pathways (145) and induce the expression of NKG2D ligands on healthy monocytes, thus subverting antitumor immune responses (143). In a previous study realized by the same group, it was demonstrated that TGF-β downregulates NKG2D expression in NK cells *in vitro* (146), supporting the idea that the elevated production/concentration of TGF-β in acidic TME is one of the main evasion mechanisms adopted by cancer cells (146).

It is well accepted that a robust presence of T cells in the TME is associated with good clinical outcome in distinct types of cancers (147–149). It is important to notice that new progresses in cancer immunotherapy are allied to the use of monoclonal antibodies directed against T cell-immune checkpoints. Examples include CTLA4 (149–151) and PD-1 (152, 153). These outstanding findings undoubtedly confirm the necessity of an effective T cell activation to control tumor growth and spread (147). As with other types of immune cells, cancer cells limit T cell immunity by distinct ways. In this regard, the acidification of the TME is a clear example, and several papers have demonstrated that lactate plays a pivotal role in this process (54, 154–156). As with transformed cells, activated T cells may generate energy through aerobic glycolysis (157–159). The upregulation of glycolytic enzymes intensifies the uptake of Glc and glycolytic rate, favoring the secretion of lactate into the microenvironment (157). It is possible to imagine that when together in the TME, both cancer cells and activated T lymphocytes significantly increase the production/secretion of lactate. As it stands, it has been demonstrated that the very acidic microenvironment suppresses the proliferation and cytokine production by activated T cells (101). A recent study developed by Brand and colleagues revealed that pathophysiological concentrations of lactic acid repeal the upregulation of the nuclear factor of activated T cells in both NK and T cells, which significantly reduced the production of IFN-γ (160). These results corroborate previous findings that lactic acid is able to downmodulate the function of cells from lymphoid origin, then contributing to tumor escape from immune attack.

### CONCLUDING REMARKS AND PERSPECTIVES

This review presents a snapshot of metabolic changes in cancer cells, describing how, even in aerobic conditions, transformed cells opt for glycolysis instead of OXPHOS to sustain their energy demand, high proliferation and biosynthesis rates, a process named "Warburg effect." This metabolic reprogramming culminates in a high-lactate and protons output, which is also exported to the extracellular environment by MCT4, generating acidosis, neoangiogenesis, and immunosuppression, directly modulating the TME. Although several genetic, biochemical, and pathophysiologic mechanisms have been identified as causes of malignancy in high-lactate tumors, it remains unclear why seemingly identical tumors may exhibit extreme differences in their lactate levels. Although it is certainly another challenge for future research in this field, several reports point out that high-lactate amounts help in generating a hostile microenvironment for normal cells, affecting the activation and differentiation of effector immune cells as well as antigen presentation and the production of cytokines. Future studies, particularly in solid tumors characterized by highly acidic environments, are needed to better understand the effect of lactate and other "waste" metabolites on cancer progression. The participation of lactate in TME and its immunosuppressive actions not only make it crucial for tumor survival and growth but also turns it into an interesting and promising candidate to therapeutic target in cancer chemotherapy. Following this reasoning, classic and novel drugs that modulate TME pH might be useful as potential immunomodulatory tools in cancer patients, particularly in combination with immunotherapeutic strategies.

## AUTHOR CONTRIBUTIONS

Wrote the paper: AM, LF, LG, LC, AF, TF, KC, ES, CF-d-L, and LF-d-L. All the authors read and approved the final version of the manuscript.

## FUNDING

This work was supported by grants from Fundação do Câncer, Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ), Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES), and Conselho Nacional de Desenvolvimento Cientifico e Tecnológico (CNPq).

### REFERENCES


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

*Copyright © 2018 Morrot, Fonseca, Salustiano, Gentile, Conde, Filardy, Franklim, da Costa, Freire-de-Lima and Freire-de-Lima. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Posttranslational Modifications of Pyruvate Kinase M2: Tweaks that Benefit Cancer

*Gopinath Prakasam1†, Mohammad Askandar Iqbal 2†, Rameshwar N. K. Bamezai1 \* and Sybille Mazurek <sup>3</sup> \**

*1School of Life Sciences, Jawaharlal Nehru University, New Delhi, India, 2Department of Biotechnology, Faculty of Natural Sciences, Jamia Millia Islamia, New Delhi, India, 3 Institute of Veterinary Physiology and Biochemistry, University of Giessen, Giessen, Germany*

#### *Edited by:*

*Leonardo Freire-de-Lima, Universidade Federal do Rio de Janeiro, Brazil*

#### *Reviewed by:*

*Daniele Vergara, University of Salento, Italy Andrew M. Chan, The Chinese University of Hong Kong, Hong Kong Pankaj Seth, Harvard University, United States Leonardo Marques da Fonseca, Universidade Federal do Rio de Janeiro, Brazil*

#### *\*Correspondence:*

*Rameshwar N. K. Bamezai bamezai@hotmail.com; Sybille Mazurek sybille.mazurek@vetmed. uni-giessen.de*

*† These authors have contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology*

*Received: 29 November 2017 Accepted: 22 January 2018 Published: 07 February 2018*

#### *Citation:*

*Prakasam G, Iqbal MA, Bamezai RNK and Mazurek S (2018) Posttranslational Modifications of Pyruvate Kinase M2: Tweaks that Benefit Cancer. Front. Oncol. 8:22. doi: 10.3389/fonc.2018.00022*

Cancer cells rewire metabolism to meet biosynthetic and energetic demands. The characteristic increase in glycolysis, i.e., Warburg effect, now considered as a hallmark, supports cancer in various ways. To attain such metabolic reshuffle, cancer cells preferentially re-express the M2 isoform of pyruvate kinase (PKM2, M2-PK) and alter its quaternary structure to generate less-active PKM2 dimers. The relatively inactive dimers cause the accumulation of glycolytic intermediates that are redirected into anabolic pathways. In addition, dimeric PKM2 also benefits cancer cells through various non-glycolytic moonlight functions, such as gene transcription, protein kinase activity, and redox balance. A large body of data have shown that several distinct posttranslation modifications (PTMs) regulate PKM2 in a way that benefits cancer growth, e.g., formation of PKM2 dimers. This review discusses the recent advancements in our understanding of various PTMs and the benefits they impart to the sustenance of cancer. Understanding the PTMs in PKM2 is crucial to assess their therapeutic potential and to design novel anticancer strategies.

Keywords: M2 isoform of pyruvate kinase, posttranslational modifications, cancer metabolism, glycolysis, Warburg effect, cancer therapy

### INTRODUCTION

The key metabolic anomaly associated with tumor cells in contrast to their non-transformed counterparts is rewiring of metabolism to escalate glucose uptake rate and to break down glucose largely into lactate, regardless of the presence of oxygen, a phenomenon known as aerobic glycolysis or Warburg effect (1). The preferential expression of M2 isoform of pyruvate kinase (PKM2) by the cancer cells is required for metabolic reprogramming, a nearly universal phenomenon in cancer (2–6). To further ripe benefits out of the PKM2 expression, cancer cells posttranslationally modify it into an enzymatically less active dimeric state. Thus, retarding the PKM2 catalyzed conversion of phosphoenol pyruvate (PEP) into pyruvate, causing piling up of the glycolytic intermediates and their channeling into pentose phosphate pathway (PPP) as well as phospholipid and amino acid synthesis to facilitate biosynthesis of macromolecules and reducing equivalent [nicotinamide adenine dinucleotide phosphate (NADPH + H+)], eventually supporting cancer cell proliferation (4). The dimer and tetramer ratio of PKM2 is regulated by numerous factors including metabolic intermediates [fructose 1, 6-bisphosphate (FBP), succinylaminoimidazolecarboxamide ribose 5-phosphate (SAICAR) and serine], physical interaction with oncoproteins, and importantly by various posttranslation modifications (PTMs) (7–9). Notably, the PTMs contribute to expansion of functional proteome in the mammalian cells (10). Recent advancements in PTMs research have uncovered many intricate regulatory mechanisms through which PTMs regulate the cell signaling and metabolic pathways. PKM2 protein sequence consists of many conserved amino acid sites that are modified covalently by PTMs such as phosphorylation, acetylation, hydroxylation, oxidation, ubiquitination, and sumoylation. This review deals with investigating the extrinsic and intrinsic stimuli responsible for PTMs of PKM2 and scrutinizes the impact on PKM2 biology along with the associated canonical and non-canonical functional implications that benefit cancer.

### MAMMALIAN PK GENE, TRANSCRIPT VARIANTS, AND PROTEINS

Pyruvate kinase (ATP-pyruvate 2-O-phosphotransferase, EC 2.7.1.40), a terminal glycolytic pathway enzyme, catalyzes the irreversible transphosphorylation reaction between PEP and ADP to yield a molecule of pyruvate and ATP in an oxygenindependent manner. All living entities ubiquitously express enzyme PK, thereby conserving at least one form of PK. The human genome consists of two PK genes; the *PKLR* gene on chromosome 1 and the *PKM* gene on chromosome 15, which code for four PK isoforms designated as PKL, PKR, PKM1, and PKM2 (11). The expression of PK isoforms is highly regulated and is tissue-specific, suggesting that different kinetic properties of PK isoforms satisfy different metabolic needs of different tissues. For instance, PKM2 is predominantly expressed in the tissues with high anabolic biosynthesis such as proliferating embryonic cells and tumor cells. Likewise, tissues bearing a high rate of gluconeogenesis, such as liver, possess accordingly compatible PKL isoform. The *PKLR* gene, by using tissue-specific alternate promoters, codes for the full-length PKR isoform in erythrocytes and a shorter variant PKL isoform (lacking 1 exon) in the liver as well as to a smaller extent in intestine and kidney (12). The *PKM* gene codes for PKM1 and PKM2 isoforms, through alternative splicing of the mutually exclusive exons 9 and 10. Out of 12 exons of PKM pre-mRNA, the mature transcript that includes exon 9 and excludes exon 10 is designated as PKM1. The mature mRNA that includes exon 10 and excludes exon 9 is termed PKM2 (M2- PK). The proteins coded by these mature mRNAs' are of identical size (531 amino acids); however, out of 56 amino acids that are coded by the exon 9 and exon 10, 22 amino acids are different (13). The expression of the constitutively active PKM1 isoform is ideal for adult differentiated tissues that demand a large supply of ATP such as brain, adult skeletal muscle, and heart. The PKM2 isoform is mainly expressed in the dividing cells with growing anabolic demands, e.g., embryonic cells and cancer cells (5, 13).

The prototypic PKM2 isoform is highly expressed in embryonic tissues and is steadily replaced by other isoforms during differentiation. There are several studies substantiating the isoform shift back to PKM2 during tumorigenesis. For instance, L-M2 shift in hepatocarcinogenesis (14, 15), M1–M2 shift during metabolic transformation, i.e., aerobic glycolysis (16). Although M2 is tightly associated with tumor growth and metabolism (as suggested by the reduction in tumor growth on replacing M2 by M1), it is also present in some differentiated tissues such as lung, adipose, pancreatic islets, retina, distal renal tubules, and so on (17, 18).

### REGULATION OF PKM2 RE-EXPRESSION

The molecular mechanism that regulates the alternative splicing of PKM isoforms remained ambiguous until recently. The study by David et al. unraveled a potential mechanistic link by which cancer cells could re-express PKM2. In this study, the authors revealed the role of c-Myc (the most frequent deregulated oncogene)-mediated transcriptional activation of heterogeneous nuclear ribonucleoproteins [hnRNPA1, hnRNPA2, and polypyrimidine tract binding protein (PTB)] in regulating alternative splicing of the PKM isoforms (19). The hnRNPs' selectively bind the sequence flanking exon 9 in PKM pre-mRNA to repress its inclusion to the mature mRNA, and thus, indirectly facilitating the inclusion of exon 10 into the mature mRNA, resulting in the expression of PKM2 mRNA in cancer cells.

The expression of PKM2 has been shown to be influenced by a spectrum of signaling pathways that are stimulated by tumor microenvironment (hypoxia and nutrient status), aberrant oncogenic mutations, growth factors, and hormones. These signaling pathways, through the network of transcription factors, regulate PKM2 expression. Transcription factors that directly bind to the consensus sequence in the *PKM* gene include hypoxia-inducible factor-1α (HIF-1α), nuclear factor kappa B (NF-κB), peroxisome proliferator-activated receptor gamma (PPARγ), and specificity protein 1 (Sp1) (8, 20). Under hypoxic conditions, HIF-1 stimulates the expression of PKM2 isoform by binding to the hypoxic response element present in the first intron of PKM gene. Notably, PKM2 interacts with HIF-1 transcription complex to co-regulate its own transcription through a positive feedback loop (21). In addition to hypoxia, aberrant oncogenic signaling in cancer cells has been shown to facilitate the normoxic stabilization of HIF-1α to rewire the metabolism by inducing the expression of PKM2. For example, mammalian target of rapamycin (mTOR), downstream to the axis of receptor tyrosine kinase/PI3K/protein kinase B (AKT), which is frequently deregulated in cancer cells, is known to stimulate the expression of c-Myc and normoxic stabilization of HIF-1. Both transcription factors increase the transcriptional activation of many glycolytic enzymes including PKM2 to stimulate aerobic glycolysis (22). Accordingly, the stimulation of the PI3K/AKT/mTOR pathway by insulin facilitated the normoxic stabilization of hypoxic inducible factor 1 (HIF1), thereby increasing PKM2 expression to promote cancer type metabolism (23). Following EGFR stimulation, NF-κB (p65) induces the expression of PKM2 by recruiting HIF-1α to co-regulate the transcriptional activation of the *PKM* gene (24). In PTEN null mouse liver cells, hyperactivated AKT2 induced the expression of peroxisome proliferator-activated receptor gamma (PPARγ), a nuclear hormone receptor and a transcription factor. PPARγ specifically binds to the promoter of the *PKM* and *HK* (hexokinase) genes to augment their expression, thus causing the metabolic reprogramming which contributes to the liver pathophysiology (25). Hormones, such as insulin (23, 26, 27), triiodothyronine-T3 (28), and glucocorticoid (29) also regulate PKM2 gene expression. The Sp1 transcription factor constitutively activates the transcription of the *PKM* gene by binding to the consensus DNA binding sites (GC boxes) in the PKM gene promoter. The Sp3 transcription factor synergies Sp1 to enhance the abovementioned function (30). Glucose stimulation facilitates dephosphorylation of Sp1 by protein phosphatase 1, which enhances its DNA binding activity (GC boxes). As a result demonstrates an increased PKM transcription (31).

In addition to the abovementioned mechanisms, PKM2 gene expression is also enhanced by epigenetic hypomethylation of intron 1 within the PKM gene (32). miRNAs including miR-122, miR-133a, miR-133b, and miR-326 also affect the expression of PKM2 and thus metabolic reprogramming in cancer (33–36). In conclusion, the plasticity in the regulation of the PKM gene expression is exploited by cancer cells to promote their growth and metabolism.

### STRUCTURAL BIOLOGY OF PKM2

Human PKM2 protein comprises 531 amino acids and each monomeric subunit is subdivided as A-, B- and C domains based on their characteristic functional features (**Figure 1A**). The adjoining region between the A- and B domains together forms the catalytic active site. The C domain near the carboxyl terminus harbors the binding site for the allosteric activator (FBP) as well as the intersubunit contact domain (ISCD) and the nuclear localization signal sequence. Notably, the 22 amino acids that differ between PKM1 and PKM2 (coded by the alternatively spliced exons) span across the ISCD domain (37). This difference allows PKM2 to

FIGURE 1 | Schematic illustration of the PKM2 domain structures and major sites of posttranslational modifications (PTMs). (A) The interface between the A1 and B domains of PKM2 forms the catalytic active site (ACS); the flanking region between the A2 and C domains, known as intersubunit contact domain (ISCD), is a key component that enables the formation of tetrameric oligomers; the C domain accommodates the allosteric activator (FBP) binding site and a nuclear localization signal sequence (NLS). (B) The major sites of PKM2 that are subjected to phosphorylation, acetylation, oxidation, hydroxylation, and ubiquitination are marked and the modifying enzymes or factors that dictate the PTMs are added on the right. Abbreviations: PK, pyruvate kinase; p, phosphorylation; Ac, acetylation; Ox, oxidation; OH, hydroxylation; Ub, ubiquitination; FGFR1, fibroblast growth factor receptor-1; ERK1/2, extracellular signal-regulated kinase 1/2; AKT, protein kinase B; p300, histone acetyltransferase; ROS, reactive oxygen species; PHD3, prolyl hydroxylase 3; PKM2, M2 isoform of pyruvate kinase; FBP, fructose 1, 6-bisphosphate.

have unique functional properties such as allosteric regulation by FBP and the ability to exist in tetrameric and dimeric forms. In contrast, PKM1 is not allosterically regulated and exists only in a tetrameric form.

The dimeric PKM2 emerges as the result of an interaction between two monomers at their A domain. Two PKM2 dimers associate at their C domain (ISCD) to form the tetrameric PKM2. In addition, PKM2 can form hybrids with PKL in liver, jejunum, colon, and rectum. Likewise, heterotetrameric hybrids between PKM2 and PKM1 were reported in esophagus, stomach, and also recently in lung and breast cancer cell lines (29, 38–40).

### DYNAMIC PKM2 ACTIVITY REGULATION AND CANCER METABOLISM

Although the different isozymes of PK catalyze the identical biochemical reaction, they differ in their kinetic properties such as varied affinity to its substrate PEP, allosteric regulation, and ability to exist in more than one oligomeric state. Unlike PKM1 isoform, a constitutively active tetrameric enzyme, the activity of PKM2 is influenced by its oligomeric state, i.e., dimer or tetramer. Whereas the tetrameric form of PKM2 has a high affinity for PEP and is highly active the dimeric form is characterized by a low PEP affinity and is nearly inactive under physiological conditions (3). The less active dimeric form of PKM2 supports the channeling of glucose carbons into synthetic processes which debranch from glycolytic intermediates, while the active tetrameric PKM2 is relatively less supportive to anabolic synthesis. The tetramer:dimer ratio of PKM2 is regulated by different mechanisms, i.e., metabolic intermediates, binding of oncoproteins, and posttranslational modifications (5).

The glycolytic intermediate FBP and the amino acid serine synthesized from glycolytic glycerate 3-P induce the subunit association from dimeric PKM2 to the enzymatically highly active tetramer (41, 42). In the same way, binding of SAICAR (an intermediate from the *de novo* purine biosynthesis pathway) to PKM2 stimulates PK activity under glucose-deprived conditions (43). PKM2 activity is allosterically inhibited by the amino acids—alanine, tryptophan, and phenylalanine (6, 8, 9). Besides the metabolic intermediates, numerous proteins are known to physically interact with PKM2 to modulate its enzymatic activity. The E7 oncoprotein encoded by the high-risk type human papillomavirus 16 binds to PKM2 to facilitate its subunit dissociation from tetramer to inactive dimer even in the presence of the allosteric activator FBP (44). Interaction of PKM2 with cytosolic promyelocytic leukemia (cPML), Jumonji C domain containing dioxygenase (JMJD5), and Tyr-46-phosphorylated MUC1-C (mediated by EGFR) inhibit PK activity through different mechanisms (45–47). cPML selectively binds the tetrameric PKM2 and inhibits its activity without dissociating it into PKM2 dimers. JMJD5 decreases PKM2 activity by inhibiting the formation of active tetramers. However, the underlying mechanism of inhibition of PK activity by Tyr-46-phosphorylated MUC1-C is not known. Interestingly, the interaction between PKM2 and death-associated protein kinase, a serine/threonine kinase enhances the enzyme activity of PKM2 (48). Thus, PKM2 acts as a molecular switch in regulating the metabolic fate of glycolytic intermediates (energy generation or anabolism) *via* modulation of its activity and structural configuration by the abovementioned mechanisms.

### POSTTRANSLATIONAL MODIFICATIONS OF PKM2 AND IMPLICATIONS IN CANCER

Unlike other isoforms of PK, PKM2 harbors numerous conserved PTM sites, often modified by phosphorylation, acetylation, prolylhydroxylation, oxidation, ubiquitination, and sumoylation in response to various stimuli in cancer cells (**Figure 1B**). These PTMs modulate structural and functional properties of PKM2, such as oligomeric state, catalytic activity, binding of allosteric activators, protein stability, conditional protein interaction, and subcellular localization. Eventually, the posttranslationally modified PKM2 provides metabolic and non-metabolic benefits to cancer cells.

## PHOSPHORYLATION

### Tyrosine Phosphorylation

The pioneering study from the laboratory of Eigenbrodt et al. demonstrated that Src—non-receptor tyrosine kinase—phosphorylates PKM2 to inhibit its catalytic activity (49). Further studies by his team suggested that tyrosine phosphorylation results in an inhibition of PKM2 thereby supporting the diversion of the glycolytic flux into synthetic routes of cell building blocks (2, 50). Nearly two decades later, Hitosugi et al. confirmed the findings and demonstrated that aberrant oncogenic tyrosine kinases (e.g., FGFR1, BCR-ABL, and JAK2) phosphorylate tyrosine (Y) 105 residue of PKM2 to inhibit its activity. They further showed that Y105 phosphorylation hinders the allosteric activation of PKM2 by FBP, resulting in PKM2 activity inhibition and channeling of glycolytic intermediates into biosynthetic metabolism to support tumor growth (4) (**Figure 2**). Hitosugi et al., using a phosphoproteomic screen, also identified additional phosphorylated tyrosine residues in PKM2 including, Y83, Y148, Y175, Y370, and Y390. However, subsequent characterization using mutation analysis revealed that only Y105 phosphorylation is significant in inhibiting PKM2 activity and in supporting tumor metabolism and growth. The dynamic regulation of PKM2 Y105 phosphorylation/dephosphorylation may result in a fine-tuning between aerobic glycolysis and oxidative phosphorylation to support distinct metabolic needs of the proliferating and quiescent cells. The role of tyrosine phosphatases that dephosphorylate PKM2 to rescue its activity remained unclear. Recent studies on adipocytes, using substrate trapping and mutagenesis assay, revealed an interaction between protein-tyrosine phosphatase 1B and the tyrosine 105 domain of PKM2 thereby inducing its dephosphorylation (51) (**Figure 2**).

Apart from tyrosine phosphorylation, PKM2 has been shown to specifically bind tyrosine-phosphorylated peptides or proteins, which in turn force PKM2 to release its allosteric activator FBP, leading to an inhibition of PKM2 activity, and thus, supporting anabolic metabolism in dividing tumor cells (52). Yang et al. reported that the epidermal growth factor (EGF) stimulated nuclear translocation and association of PKM2 with Y333

FIGURE 2 | Regulation of M2 isoform of pyruvate kinase (PKM2) enzyme activity and non-glycolytic nuclear functions by phosphorylation. Aberrant oncogenic FGFR1-mediated tyrosine (Y) phosphorylation of PKM2 at Y105 residue facilitates the formation of the enzymatically inactive dimeric form of PKM2 which promotes aerobic glycolysis and biosynthesis of macromolecules by causing accumulation and diversion of the glycolytic intermediates into synthetic pathways of cell building blocks. Conversely, protein-tyrosine phosphatase 1B (PTP1B), a tyrosine protein phosphatase, dephosphorylates tyrosine phosphorylated PKM2 and reverses the above-stated features to establish metabolic homeostasis under physiological conditions. Unlike the Y105 phosphorylation, serine (S) or threonine (T) phosphorylation of PKM2 at S37, S202, and T454 by growth factor-stimulated serine/threonine protein kinases [extracellular signal-regulated kinase (ERK) 1/2, protein kinase B (AKT), and proviral insertion in murine lymphomas 2 (PIM2)], direct the nuclear translocation of PKM2 with the aid of nuclear importin α5. The nuclear-localized PKM2 assists the transcriptional activation of β-catenin and signal transducer and activator of transcription 5 (STAT5) to facilitate the expression of different genes such as cyclin D1, c-Myc, glucose transporter 1 (GLUT1), lactate dehydrogenase A (LDHA), and PKM2, all essential for tumor cell metabolic reprogramming and division. Green arrows in the illustration indicate the augmented glycolytic flux. Red arrows indicate the restricted flux of glycolysis and the fate of the glycolytic intermediates as a result of the dimeric state of PKM2.

phosphorylated β-catenin. This interaction mediates the binding of the β-catenin/TCF transcription complex to the CCDN1 promoter responsible for the co-activation of cyclin D1 transcription, thus, promoting brain tumor development (53).

### Serine/Threonine Phosphorylation

Unlike tyrosine phosphorylation of PKM2 that principally modulates its glycolytic function, serine/threonine phosphorylation of PKM2 supports its nuclear localization and opens up new vistas of non-glycolytic moonlight functions, i.e., gene transcriptional regulation and protein kinase activity (**Figure 2**).

Cancer cells stimulated with EGF triggers extracellular signalregulated kinase 2 (ERK2), serine/threonine protein kinase, to phosphorylate PKM2 at the serine (S) 37 residue. PIN and importin α5 selectively bind the S37-phosphorylated PKM2 and induce its translocation into the nucleus where PKM2 serves as a transcriptional co-activator of β-catenin to express its target gene, c-Myc. The transcription factor c-Myc, in turn, enhances the expression of glucose transporter 1 (GLUT1), lactate dehydrogenase A, and remarkably PKM2, through a positive feedback loop. Together, EGF-stimulated S37 phosphorylation of PKM2 facilitates metabolic reprogramming in cancer cells (54). Likewise, an extracellular matrix protein 1 (a secretory glycoprotein) has been shown to stimulate PKM2 S37 phosphorylation, to enhance the transcriptional activation of glycolytic enzymes that support aerobic glycolysis in tumor cells (55). AKT stimulated by insulinlike growth factor (IGF-1) in cancer cells has been shown to interact and phosphorylate PKM2 at the S202 residue to promote its entry into the nucleus to facilitate the transcriptional activation of signal transducer and activator of transcription (STAT)5A targets gene, cyclin D1 (56). This study proposes PKM2 as a coactivator of the transcription factor, STAT5A, which functions downstream to the IGF/PI3K/AKT signaling pathway to promote tumor growth (56). The effect of PKM2 serine phosphorylation by A-Raf is dependent upon the metabolism of serine, alanine, and glutamine. For instance, primary fibroblasts which are characterized by serine consumption and glutamine production, expression of wild-type A-Raf induces PKM2 dimerization which leads to inhibition of the metabolism of glucose to lactate. On the other hand, NIH 3T3 cells, characterized by glutamine consumption and serine production, expression of gag-A-Raf favors the metabolism of glucose to lactate by increasing highly active tetrameric form of PKM2 (57). Oncogenic proviral insertion in murine lymphomas 2 (a serine/threonine protein kinase) has been shown to phosphorylate the threonine (T) 454 residue of PKM2 to stimulate its non-glycolytic nuclear function. The non-glycolytic function of T454 phosphorylated PKM2, primarily involves the transcriptional co-activation of HIF-1α and β-catenin transcription factors to support cancer growth and drug resistance (58).

## ACETYLATION

Acetylation is another important PTM where the class of enzymes termed N-terminal- or lysine-acetyltransferase catalyzes the transfer of the acetyl group from acetyl-CoA to the alpha- or epsilon-amino group of the lysine residue. Many basic cellular functions, such as chromatin remodeling, gene expression as well as several protein functions, and to maintain cellular homeostasis, depend upon an exact regulation of the dynamic process between acetylation and deacetylation.

Acetylation of PKM2 lysine residues 305 and 433 alter its activity, protein stability, and non-glycolytic protein kinase function (**Figure 3**). High glucose concentration induces lysine (K) 305 acetylation, mediated by the PCAF (P300/CBP-associated factor) acetyltransferase, in cancer cells to reduce PK activity. Lysine 305 acetylation targets PKM2 for lysosomal degradation through a chaperon-mediated autophagy. Hence, the state of reduced PKM2 expression and activity stockpiles the upper glycolytic intermediates and redirects the flux through the PPP shunt for biosynthesis of building blocks (nucleotides and amino acids), which eventually promotes tumor growth (59). Unlike the molecular mechanism described earlier, a recent study by Lv et al. revealed a remarkable non-glycolytic protein kinase function of PKM2 following its K433 acetylation mediated by p300 acetyltransferase in response to a spectrum of mitogenic and oncogenic stimuli (60). In detail, K433 acetylation of PKM2 promotes nuclear localization and protein kinase activity of the enzymatically inactive dimeric PKM2 to phosphorylate STAT3 at the Y705 residue and histone H3 at the T11 residue. The latter activates a transcriptional program that supports cell proliferation and tumorigenesis (60). Different deacetylases specifically remove the acetylation mark from PKM2 and limit the ability of PKM2 to promote tumorigenesis. For example, deacetylation of nuclear PKM2 by SIRT6 (sirtuin 6) at the K433 residue results in the export of PKM2 to the cytoplasm, thereby restricting its non-glycolytic functions (i.e., transcriptional co-activator and protein kinase) that support cancer (61). SIRT2 averts the metabolic benefits of PKM2 to cancer by deacetylation of PKM2 at the K305 residue which prevents the lysosomal degradation of PKM2 (62).

## HYDROXYLATION

In hydroxylation, proline or lysine amino acid residues of mammalian proteins are covalently modified by the addition of a hydroxyl group. Proline hydroxylation of mammalian proteins is primarily catalyzed by the prolyl hydroxylases family of enzymes in the presence of molecular oxygen, Fe2<sup>+</sup> (iron), 2-oxoglutarate, and ascorbate. Proline hydroxylation has an important role in numerous cellular mechanisms, including cancer. A well-known example is the regulatory role of proline hydroxylation of the transcription factor HIF1 alpha depending upon oxygen supply.

Proline (p) hydroxylation of PKM2 at 403 and 408 by the PHD3 enzyme (**Figure 4**) favors the interaction of PKM2 with the HIF1 transcription complex, which results in recruitment of p300-acetyltransferase to facilitate the transactivation of HIF target genes (21). Another study by Wang et al. revealed that under hypoxic conditions, JMJD5 interacts with PKM2 and facilitates its nuclear localization to promote HIF1 mediated transcriptional activation of glycolytic enzymes that support cancer cell metabolism and proliferation (47) (**Figure 4**).

## OXIDATION

High reactive oxygen species (ROS) concentrations induce an oxidation of PKM2 at Cysteine 358 which results in a dissociation of PKM2 to the less active dimeric form. The functional characterization of PKM2 oxidation revealed an essential role in the conservation of the redox homeostasis in cancer cells. The enzymatically inert C358 oxidized PKM2 causes an accumulation of glycolytic intermediates which are required to channel glycolytic carbons into PPP thereby generating NADPH + H<sup>+</sup> for detoxification of ROS (63) (**Figure 5**). Besides hypoxia and addition of either hydrogen peroxide or thiol oxidizing diamide into the cultivation medium of tumor cells also insulin is described to increase intracellular ROS production and cysteine oxidation of PKM2 (23).

## UBIQUITINATION AND SUMOYLATION

M2 isoform of pyruvate kinase has been shown to interact with the HECT domain (a protein domain found in ubiquitin-protein ligases) of HERC1, a giant protein with a molecular weight of 532 kDa, which acts as both an E3 ubiquitin ligase and a guanine nucleotide exchange factor. The interaction between PKM2 and HERC1 neither altered PKM2 activity nor induced its proteasomal degradation (64). Since PKM2 also phosphorylates GDP, a possible role of PKM2 as a local GTP producer for the guanine nucleotide exchange function of the RLD1 (RCC1-like domain) of HERC 1 is discussed. Another E3-ligase which interacts with PKM2 is Parkin. Monoubiquitination of PKM2 and PKM1 at lysine (K) 186 and 206 residues by Parkin resulted in a reduction of PKM activity but did not affect its stability. The interaction between PKM2 and Parkin was favored by glucose starvation. Parkin silencing rescued the activity of PK and supported tumor growth (65).

FIGURE 3 | The impact of PKM2 acetylation on its enzymatic activity, protein stability, nuclear localization, and non-glycolytic functions. High glucose concentration facilitates the acetylation of PKM2 lysine (K) 305 residue by PCAF acetyltransferase. K305-acetylated PKM2 exhibits reduced enzyme activity and undergoes lysosomal degradation through chaperon-mediated autophagy. Low PKM2 expression and reduced enzyme activity support the anabolic metabolism and rapid division of tumor cells. Mitogens and oncogenic signals elicit acetyltransferase-p300 to acetylate K433 residues of PKM2, resulting in the release of the allosteric activator FBP from the PKM2 protein, translocation of PKM2 into the nucleus, and activation of the protein kinase activity of PKM2. Inside the nucleus, K433-acetylated PKM2 protein kinase phosphorylates STAT3 at the Y705 residue and histone H3 at T11, thereby activating the transcriptional program that the supports cell proliferation and tumorigenesis. SIRT2 deacetylates K305 residue, and SIRT6 deacetylates K433 residue of PKM2, thereby impairing the tumor-promoting property of PKM2. Abbreviations: EGF, epidermal growth factor; PCAF, P300/CBP-associated factor; p300, histone acetyltransferase; Ac, acetylation; SIRT2, NAD-dependent deacetylase sirtuin-2; SIRT6, NAD-dependent deacetylase sirtuin-6; FBP, fructose 1, 6-bisphosphate; STAT3, signal transducer and activator of transcription 3; LAMP1, lysosomal-associated membrane protein 1; PKM2, M2 isoform of pyruvate kinase.

prolyl hydroxylase 3; JMJD5, Jumonji C domain-containing dioxygenase; p300, histone acetyltransferase; HRE, hypoxic response element; GLUT1, glucose

In addition, the laforin/malin E3-ubiquitin ligase complex, which is mutated in neurodegenerative Lafora disease, interacts and polyubiquitinates both M-isoforms of PK. When accessed for the biochemical properties, the enzyme activity of the PKM isoforms remained unaffected. Interestingly, this posttranslational modification impairs the nuclear localization of PKM2 following UV treatment. In the case of PKM1 neither UV treatment nor laforin/ malin-mediated polyubiquitination affects its nuclear localization (66). Apart from interacting with E3 ubiquitin ligases, PKM2 does also bind the deubiquitin enzyme USP20 (ubiquitin-specific protease, a deubiquitin enzyme). USP20 interaction with PKM2 was suggested to apparently enhance its protein stability (67). The interaction of PKM2 with SUMO-E3 ligase PIAS3 (protein inhibitor of activated STAT3) induces sumoylation of PKM2 and translocation of the PKM2–PIAS3 complex into the nucleus. The functional significance of the interaction between PKM2

transporter 1; LDHA, lactate dehydrogenase A; PKM2, M2 isoform of pyruvate kinase.

and PIAS3 as well as of the nuclear localization of the complex remains ambiguous (68). In summary, the functional relevance of the interaction of PKM2 with E3-ligases and deubiquitin enzymes as well as with SUMO E3-ligases is still not completely understood.

## GLYCOSYLATION

Threonine 405 and serine 406 residues of PKM2 were recently identified to be O-GlcNAcylated (*O*-linked β-*N*-acetylglucosamine) by O-GlcNAc transferase and this modification is dynamically regulated by the fluctuations in nutrient supply. Investigation of various human cancer cells and sporadic breast tumor tissue samples revealed PKM2 *O*-GlcNAcylation across the tumor cell types and patient samples. PKM2 T405 and S406 O-GlcNAc modification reduces PK activity by acting as a barrier that

FIGURE 5 | Effect of ROS-mediated oxidization of PKM2 on cellular metabolism and redox balance. Cancer cells produce large amounts of ROS that oxidize the cysteine (C) 358 residue of PKM2, which leads to the inhibition of PKM2 as well as a piling up of glycolytic intermediate glucose 6-phosphate which is then diverted into the PPP. The oxidative PPP facilitates the synthesis of NADPH + H+ which regenerates GSH thereby minimizing ROS-induced damages of the cancer cells. Abbreviations: ROS, reactive oxygen species; PPP, pentose phosphate pathway; NADPH, nicotinamide adenine dinucleotide phosphate; GSH, glutathione; GSSG, glutathione disulfide; H2O2, hydrogen peroxide; ETC, electron transport chain; PKM2, M2 isoform of pyruvate kinase; PPP, pentose phosphate pathway.

impedes the formation of stabilizing H-bonds at the C–C dimerization interface, thereby, destabilizing PKM2 active tetramer into enzymatically less active dimer. The state of reduced PK activity redirects the glycolytic metabolic intermediates to the anabolic metabolic pathway to support the synthesis of biomass and tumor proliferation. In addition, *O*-GlcNAcylation of PKM2 facilitates its translocation into the nucleus and the latter requires additional Ser37 phosphorylation of PKM2, mediated by EGF-stimulated ERK1/2 protein kinase and importin α5. O-GlcNAcylated, nuclear PKM2 activates transcription factor c-Myc to express GLUT1 and LHDA to further fine-tune aerobic glycolysis. The expression of *O*-GlcNAcylation-ablated PKM2 (T405A/S406A) reverses the pro-cancer metabolism and abrogates the tumor progression (69). Notably, this study highlights the crosstalk between PKM2 PTMs Thr405/Ser406 *O*-GlcNAcylation, and Ser37 phosphorylation is mutually inclusive to translocate PKM2 into the nucleus for executing non-glycolytic nuclear function (69).

## METHYLATION

Co-activator-associated arginine methyltransferase 1 also known as PRMT4 methylates specifically the dimeric form of PKM2 at Arg445/447/455 residues in the C domain. TEPP-46 and FBP, the allosteric activators that induce PKM2 tetramerization limits PKM2 methylation. Importantly, PKM2 activity remains unaltered by methylation; however, methylated PKM2 reprograms the metabolic phenotype toward aerobic glycolysis from oxidative phosphorylation to support tumor cell proliferation, migration, and metastasis. Mechanistically, methylated PKM2 localizes to the mitochondrialassociated endoplasmic reticulum membrane to interact with inositol 1, 4, 5-trisphosphate receptors (InsP3Rs) and to reduce InsP3R expression, thus, decreasing the mitochondrial membrane potential and Ca2<sup>+</sup> uptake. This is essential to support TCA cycle and oxidative phosphorylation by activating the pyruvate dehydrogenase in Ca2<sup>+</sup> dependent processes. The tumor supportive features of methylated TABLE 1 | The effect of various posttranslation modifications (PTM) of M2 isoform of pyruvate kinase (PKM2) on tumor metabolism and growth.


PKM2 were averted by PKM2 knockdown or by introducing PKM2 methylation defective mutant or by delivering competitive nonmethylated PKM2 peptide using nanoparticles (70).

### CONCLUDING REMARKS

The broad spectrum of PTMs in PKM2 which impart both metabolic and non-metabolic benefits to cancer underlines the significance of this isoenzyme in cancer cells. While PKM2 is expressed not only in tumor tissues but also in some differentiated cells and tissues, the posttranslational modifications of PKM2 are very specific for tumor cells. Accordingly, targeting the different

#### REFERENCES


posttranslational modifications of PKM2 is a promising strategy in cancer treatment (**Table 1**). Owing to the heterogeneous nature of tumors, it is important to expand our understanding of the roles of PTMs in PKM2, warranting further investigation and high-throughput PTM data from different tumors in order to refine the clinical targeting of PKM2 PTMs.

### AUTHOR CONTRIBUTIONS

GP, MI, RB, and SM equally participated in drafting the review article; RB and SM critically reviewed for important intellectual content; and GP designed all the graphics.


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

The reviewer LF and the handling editor declared their shared affiliation.

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

# DNA Damage, Repair, and Cancer Metabolism

#### *Marc-Olivier Turgeon1 , Nicholas J. S. Perry1 and George Poulogiannis1,2\**

*1Department of Cancer Biology, Institute of Cancer Research, London, United Kingdom, 2Division of Computational and Systems Medicine, Department of Surgery and Cancer, Imperial College London, London, United Kingdom*

Although there has been a renewed interest in the field of cancer metabolism in the last decade, the link between metabolism and DNA damage/DNA repair in cancer has yet to be appreciably explored. In this review, we examine the evidence connecting DNA damage and repair mechanisms with cell metabolism through three principal links. (1) Regulation of methyl- and acetyl-group donors through different metabolic pathways can impact DNA folding and remodeling, an essential part of accurate double strand break repair. (2) Glutamine, aspartate, and other nutrients are essential for *de novo* nucleotide synthesis, which dictates the availability of the nucleotide pool, and thereby influences DNA repair and replication. (3) Reactive oxygen species, which can increase oxidative DNA damage and hence the load of the DNA-repair machinery, are regulated through different metabolic pathways. Interestingly, while metabolism affects DNA repair, DNA damage can also induce metabolic rewiring. Activation of the DNA damage response (DDR) triggers an increase in nucleotide synthesis and anabolic glucose metabolism, while also reducing glutamine anaplerosis. Furthermore, mutations in genes involved in the DDR and DNA repair also lead to metabolic rewiring. Links between cancer metabolism and DNA damage/DNA repair are increasingly apparent, yielding opportunities to investigate the mechanistic basis behind potential metabolic vulnerabilities of a substantial fraction of tumors.

#### Keywords: metabolism, DNA repair, DNA damage, cancer, reactive oxygen species

### INTRODUCTION

Over the past decade, a renewed interest in cancer metabolism has emerged. The idea originated from the work of Otto Warburg and colleagues—first published in the 1920s—who noticed the propensity for cancer cells to consume increased quantities of glucose (1, 2). We now understand that there is extensive metabolic rewiring in cancer cells, and we are starting to decipher how cancer cells reprogram their metabolism to adapt to changes in their microenvironment and support their high metabolic needs (3–5). Cancer metabolism is a broad topic and our current understanding of how metabolic reprogramming is linked to malignant transformation has been extensively reviewed elsewhere (6–8). Therefore, in this review, we decided to focus on the connections between cell metabolism and another major aspect of cancer biology, DNA-repair/DNA-damage pathways.

The major hallmarks, as described by Pavlova and Thompson (7), include: deregulated glucose and amino acid uptake, opportunistic ways of acquiring nutrients, use of metabolic intermediates for biomass and nicotinamide adenine dinucleotide phosphate (NADPH) synthesis, increased demand for nitrogen, and alterations in metabolite-driven gene expression and interaction with the microenvironment (7, 9). Importantly, the wide differences in metabolic dependencies across cancer types, or stage of progression, as well as the dynamic shifts between metabolic pathways, make the study of cancer metabolism and the development of new therapies targeting these pathways very challenging.

#### *Edited by:*

*Leonardo Freire-de-Lima, Universidade Federal do Rio de Janeiro, Brazil*

#### *Reviewed by:*

*Daniele Vergara, University of Salento, Italy Bernd Groner, Georg Speyer Haus, Germany Vandna Kukshal, Washington University in St. Louis, United States*

#### *\*Correspondence:*

*George Poulogiannis george.poulogiannis@icr.ac.uk*

#### *Specialty section:*

*This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology*

*Received: 15 December 2017 Accepted: 17 January 2018 Published: 05 February 2018*

#### *Citation:*

*Turgeon MO, Perry NJ and Poulogiannis G (2018) DNA Damage, Repair, and Cancer Metabolism. Front. Oncol. 8:15. doi: 10.3389/fonc.2018.00015*

**61**

detoxification. A principal ROS detoxification mechanism involves reduced glutathione (GSH), determined by glutamine and cysteine availability, as well as

NADPH levels. High ROS-induced DNA damage leads to excessive burden on the DNA repair machinery.

In addition to the accepted role for cell metabolism in cancer, it is well established that DNA-repair/DNA-damage pathways are important in cancer progression because dysregulation leads to higher levels of genomic instability, increased mutation rate, and enhanced intra-tumor heterogeneity (10–13). There are currently three principal mechanisms through which changes in cell metabolic status are thought to have an influence on DNAdamage/DNA-repair pathways: chromatin remodeling, doublestrand break (DSB) repair, and redox homeostasis (**Figure 1**). Uncovering new links between these important aspects of cancer biology might lead to the development of new targeted therapies in DNA-repair deficient cancers or even improving the efficacy of existing therapies, such as PARP inhibitors, anthracyclines, and platinum salts. In this review, we examine work seeking to uncover the links between DNA-repair/DNA-damage and cell metabolism.

### Metabolic Status Affects DNA Folding and Repair Pathways

The first link between DNA repair and cell metabolism involves DNA folding and organization. Chromatin packaging and remodeling, through different histone posttranslational modifications—including acetylation, methylation, phosphorylation, and ubiquitination, as well as through DNA modifications such as methylation—can regulate gene expression levels by modulating the access to DNA of different protein complexes (14, 15). Interestingly, similar mechanisms can also regulate access of DNA-repair proteins to the DNA double-helix (16). When repairing DSBs, the first step involves unfolding DNA to allow access of the repair complexes to the DSB (17). There is a growing body of evidence suggesting that these mechanisms can in fact regulate the choice of DNA-repair pathway (i.e., homologous recombination), or non-homologous end-joining used to repair the DSB (16). Substrates added to histones or DNA are derived from metabolic intermediates (18–20) (**Figure 2**). For example, methyl-group donors mostly come from the S-adenosylmethionine (SAM) pathway (18), while the sole acetyl-group donor is acetyl coenzyme A (acetyl-CoA), known for the transfer of its acetyl group to oxaloacetate to form citrate and start the tricarboxylic acid (TCA) cycle (21). It has been shown that expression levels of the enzyme, ATP citrate lyase, can regulate the availability of acetyl-CoA in the cell (22). Restricting the amount of acetyl-group donors can disrupt proper DNA organization and have an impact on DNA folding and DNA remodeling essential to successful DNA DSB repair (22, 23). Acetyl-CoA can also be generated from acetate and CoA from the acetyl-CoA synthetase 2 enzyme. In fact, under metabolic stress conditions in which oxygen and lipids are deprived, acetate-generated acetyl-CoA is enhanced and acetatederived carbons are incorporated in lipid synthesis (24, 25). Acetate-derived acetyl-CoA has also been shown to be involved in histone acetylation, suggesting that acetate availability could influence histone acetylation in cancer cells (26).

Additionally, changes in the activity of the SAM pathway due to availability of necessary nutrients or substrates also influence DNA or histone methylation by regulating the pool of methyldonors (27). The SAM pathway is interconnected with methionine, tetrahydrofolate (THF), one-carbon metabolism, and choline pathways. These pathways tend to compensate for one another; however, depletion of choline, folate, or methionine can still have an impact on the final SAM concentration (18). Such changes in the SAM pathway not only influence gene expression of cancer-associated genes through epigenetic modifications but can also impact DNA folding during DNA repair processes. The latter can also be influenced directly by metabolic intermediates such as fumarate. DNA-PK dependent activation of fumarase contributes to enhanced local generation of fumarate, which in turn promotes DNA repair *via* inhibition of KDM2B-mediated histone demethylase activity (28). Accumulation of fumarate can also lead to epithelial-to-mesenchymal transition and poor clinical outcome (29). Since modifications in histone and DNA are important for proper repair of DNA DSBs (30), changes in nutrient and substrate levels that disrupt these modifications are bound to affect DNA-repair pathways.

### Nucleotide Levels Affect DNA-Repair Potential of Cancer Cells

Another major mechanism upon which cell metabolism can regulate DNA-repair/DNA-damage is through the regulation of the pool of nucleotides used for DNA replication and repair (31, 32). Many different metabolic pathways are involved in *de novo* nucleotide synthesis and can have an impact on the levels of intracellular nucleotides available (**Figure 2**) (31, 33, 34). An important precursor in the synthesis of the ribose backbone, essential for both purines and pyrimidines, comes from ribose-5-phosphate, an intermediate of the pentose-phosphate pathway (PPP) (31). Briefly, the PPP utilizes glucose-6-phosphate (G6P), an intermediate of glycolysis, which can be redirected to generate metabolic intermediates necessary for nucleotide and protein synthesis, as well as generation of NADPH. The PPP is a major focus in cancer metabolism, and it has been carefully reviewed elsewhere (31, 35, 36). The increased glucose consumption observed in cancer cells has been shown, at least in some cancers, to be used to fuel the PPP, where it can potentially be used to generate reducing power in the form of NADPH and nucleotide precursors (32). Additionally, other metabolic pathways are involved more specifically in purine or pyrimidine ring synthesis. The amide group of glutamine is essential in two steps of inosine monophosphate synthesis, an intermediate in *de novo* purine synthesis, and in one step of uridine monophosphate synthesis, an intermediate in *de novo* pyrimidine synthesis (37, 38). In glioblastoma, it has been shown that α-ketoglutarate is diverted out of the TCA cycle to synthesize more glutamine through the glutamine synthetase enzyme. The resulting glutamine is then used for *de novo* purine synthesis (39). Furthermore, aspartate is essential for the synthesis of pyrimidines, and glycine is also essential for purine synthesis (38). Loss-of-function mutations in arginosuccinate synthase 1, which leads to reduced arginine production in the urea cycle, promotes the use of aspartate for pyrimidine synthesis (33). Aspartate is mostly synthesized from glutamate and oxaloacetate. Therefore, the availability of glutamine and/ or other amino acids/metabolic substrates can influence the amount and ratio of nucleotides produced in a cell and might provide a regulatory mechanism for DNA-repair.

respectively are synthesized from glutamine, and aspartate and carbamoyl phosphate, respectively. Redox homeostasis: nicotinamide adenine dinucleotide phosphate (NADPH) is generated from the PPP and the serine synthesis pathway. Glutathione (GSH) is generated from cysteine and glutamate. Chromatin remodeling: methyl-group donors (CH3) are generated from the S-adenosyl methionine (SAM) pathway and acetyl-group donors (acetyl coenzyme A) are generated from the citrate cycle or from acetate. GLUT-1, glucose transporter 1; SCL1A5, alanine, serine, cysteine-preferring transporter 2 (ASCT2); G6P, glucose 6-phosphate; GA3P, glyceraldehyde 3-phosphate; SAH, S-adenosyl homocysteine; xCT, cystine-glutamate antiporter; MCT, monocarboxylate transporter family.

### Deregulated Redox Homeostasis Promotes DNA Damage

The metabolic regulation of reactive oxygen species (ROS) levels is the third and final link that we address here between DNA-repair/DNA-damage and cell metabolism. High ROS levels affect many aspects of tumor biology and, here, we focus on their role in inducing DNA damage and genomic instability. Most of the DNA lesions that are formed by ROS-induced damage are single strand breaks (SSBs) that can be repaired through nucleotide or base excision repair (NER/BER) (40). However, these SSBs can lead to stalling of the replication fork or error in replication, ultimately leading to DSBs (40). The accumulation of oxidative DNA damage increases the burden on the DNA-repair machinery. Tight regulation of cellular redox stress is essential, since high ROS levels can lead to oxidative stress and oxidative damage of proteins, DNA, and lipids, while a certain level of ROS is essential for activating signaling pathways involved in multiple biological processes (41–43). Cells have evolved a number of ways to balance ROS levels (**Figure 2**). Glutathione (GSH) is one of the major ROS-scavenging molecules (44), and it occurs in two versions; the reduced form, sulfhydryl GSH, and the oxidized, glutathione disulfide (GSSG). The enzyme GSH peroxidase catalyzes the reduction of H2O2 to water and lipid hydroperoxides to their corresponding lipid alcohols and in turn oxidizes GSH to GSSG. The reverse reaction is catalyzed by the enzyme glutathione reductase, also known as glutathione-disulfide reductase using the reducing potential of NADPH (45).

Glutathione is synthesized in the cytoplasm from cysteine and glutamate, and cysteine availability is the rate-limiting step of the synthesis reaction (46, 47). Cysteine can be synthesized from serine through the transsulfuration pathway (47), or it can be transported into the cells as cystine in exchange for glutamate using the cystine-glutamate antiporter xCT (39). Interestingly, mTORC2 can phosphorylate and inhibit the xCT antiporter, therefore providing a mechanism to regulate intracellular cysteine levels (48). Furthermore, it has been shown that, in some cancer types, growing cells in DMEM containing cystine leads to xCT dependent glutamine dependency *in vitro*, providing evidence for context dependent changes in metabolism (5). The main source of NADPH comes from the PPP and a significant amount also comes from the serine synthesis pathway through the THF pathway. Both harness the energy from the catabolism of metabolic intermediates to generate NADPH (31, 47). Interestingly, an important regulator of ROS levels is the transcription factor nuclear factor E2-related factor 2 (NRF2) that has been shown to directly regulate the key serine synthesis enzymes (PHGDH, PSAT1, ATF4) (49). This provides a potential mechanism through which NRF2 can regulate ROS levels by influencing NADPH and GSH levels. Importantly, the DNA-repair protein, BRCA1, regulates NRF2, providing an additional link between DNA repair pathways and ROS levels (50). Overall, there are still many open questions with regards to how cancer cells can rewire their metabolism based on different nutrient availability, but there is growing evidence to suggest that maintaining cellular redox homeostasis plays a major role on DNA-damage/DNA-repair pathways.

### DNA Damage Response (DDR) Triggers Metabolic Rewiring

While DNA-repair pathways can be influenced by cellular metabolic status and nutrient availability in the tumor microenvironment, accumulation of DNA damage due to extrinsic and intrinsic genotoxic stress, or deficient DNA repair can also cause abrupt rewiring of cell metabolism. Cells have evolved the DDR pathway (DDR) to monitor this genotoxic stress and maintain accurate transmission of genetic information to subsequent generations. The DDR can, therefore, halt cell-cycle progression, induce DNArepair mechanisms, or trigger programmed cell death when DNA damage is irreparable (51). Ataxia telangiectasia mutated (ATM) and ataxia telangiectasia and Rad3-related (ATR) kinases are two key enzymes in the recognition of DNA damage and implementation of the DDR (51, 52). Upon activation by DNA-damage, ATM and ATR generate a second wave of phosphorylation that impacts many downstream effector proteins. An important aspect of DDR, driven by ATM and ATR activation, is the induction of metabolic rewiring to promote the resolution of genotoxic stress. ATM has been shown to activate the PPP through induction of the rate-limiting enzyme glucose-6-phosphate dehydrogenase (G6PD), to support the synthesis of reducing power in the form of NADPH, and generate ribose-5-phosphate for nucleotide synthesis (53, 54).

Increased carbon flux to the PPP from glucose derivatives is driven not only by increased glucose consumption often observed in cancer cells (55, 56) but also through ROS-mediated inactivation of many glycolytic enzymes, notably PKM2 that promotes flux into the oxidative arm of the PPP as well as serine biosynthesis (57). Many studies have shown increased glucose dependency following DDR (58), and these data demonstrate the challenge posed to cancer cells under genotoxic stress. They need to contain ROS levels, enhance nucleotide synthesis to repair DNA damage, and inhibit cell-cycle progression until resolution. To this end, SIRT4 activation upon DDR has been shown to inhibit glutamine consumption in HepG2, HeLa, HEK293T cells, and in lung tissue (58). The reduced glutamine consumption or, more specifically, reduced anaplerosis of glutamine into the TCA cycle, confers improved cell survival compared to SIRT4 knockout cells that cannot activate this response (58). While this study suggests that DDR undoubtedly causes changes in metabolism, conflict in the literature indicates that the nature of these changes is context- and cell type-dependent. For example, the concept that cells undergoing genotoxic stress need to synthesize more nucleotides to repair the DNA damage contradicts the reduced glutamine consumption. As mentioned previously, the amide group of glutamine is essential for purine and pyrimidine *de novo* synthesis. Other interesting points to note are that in non-cancerous tissues/cells, there are increases in fatty acid oxidation (FAO) and oxidative phosphorylation (OXPHOS) in response to acute or chronic genotoxic stress (59). Increased FAO/OXPHOS is driven by AMP kinase (AMPK) activation resulting from the depletion of ATP. In the context of that study, cells are depleted of ATP because of the utilization of NAD+ by the DNA repair enzyme PARP-1. PARP-1 generates PAR chains from NAD+ and PARylates DNA to induce repair. This leads to depletion of NAD+, reduction of ATP synthesis, and activation of AMPK (59). This study further supports that DNA damage impacts cell metabolism. However, given the current literature, it is still unclear how a cell's nutrient availability can dictate the nature of its metabolic rewiring.

### Metabolic Changes Driven by DDR Gene Mutations

The p53 protein encoded by the *TP53* gene is another key player in DDR through its role in DNA repair and cell-cycle regulation (60–62). While p53 levels are kept low under normal conditions, upon DNA damage or other cellular stress, p53 is activated by phosphorylation events that prevent its degradation (60). Wildtype p53 activation regulates cell-cycle entry by transcriptionally repressing cyclin B and CDC25B, and it also plays an important role in DNA-repair. More specifically, it determines whether or not to repair the DNA or trigger senescence or apoptosis. p53 also plays many modulatory roles in the metabolic rewiring of cells. Among these, is its regulation of glycolysis through transcriptional activation of the *TP53*-induced glycolysis and apoptosis regulator (TIGAR) protein (63). TIGAR inhibits glycolysis by dephosphorylating fructose-2,6-biphosphate (F2,6B) leading to degradation of F2,6P and inhibition of phosphofructo kinase 1, an enzyme integral to glycolysis. p53 has also been shown to inhibit other enzymes of glycolysis as well as directly repress the expression of glucose transporters, GLUT1 and GLUT4 (61, 64). One might surmise that this impediment in glycolysis potentially leads to a surplus in glycolytic intermediates, leading to redirection toward the PPP. However, p53 has also been shown to inhibit the flux of glucose to the PPP by directly binding to G6PD and preventing its active dimer formation (61).

In addition, p53 has been shown to transcriptionally regulate proteins involved in the electron transport chain and mitochondrial stability, therefore, providing a mechanism to regulate OXPHOS (61, 65). Together, these data suggest that the role of p53 in metabolic regulation is context and tissue dependent. Importantly, *TP53* is one of the most mutated genes in cancer and many gain-of-function mutations have been described, leading to different changes in p53 depending on the mutation and the context. For example, mutant p533KR cannot induce cell-cycle-arrest, senescence, or apoptosis, but it can still regulate metabolic target genes resulting in decreased ROS levels and reduced glycolytic flux (61, 66). Therefore, mutations in DDR or DNA-repair genes may result from the application of selective pressures established by the metabolic niches of cancer cells.

Other genes involved in DNA-repair are also commonly mutated in cancer (13). Some of the most well-known DNArepair genes associated with cancer are *BRCA1* and *BRCA2* but mutations in other genes such as *ATM*, *ATR*, and *CDK12* [for a more extensive list, see Ref. (13)] also lead to DNA-repair deficient cancers (13, 67–70). Some cancers with a similar phenotype to DNA-repair mutant cancer do not harbor any mutations in those pathways. Together, these DNA-repair deficient cancers have been described as having a "BRCAness" phenotype (13, 71). Some metabolic changes have been described in the context of *BRCA*1 or *BRCA2* mutations (50, 72–74); however, it is still unclear whether these are due to their DNA-repair function or some alternative role. Since connections have been drawn between metabolism and DNA-repair, it seems likely that the "BRCAness" phenotype and *BRCA1/2* mutations would be associated with changes in metabolism. Again, as with most other mutations found in cancer,

#### REFERENCES


the metabolic changes associated with these DNA repair defects might reflect the context and tissue of origin.

### CONCLUSION

Historically, cancer has been thought of as a genetic disease driven by the accumulation of multiple mutation "hits" (75–77). However, in recent years, this paradigm has begun to shift, and cancer is often regarded as a "metabolic disease" (78, 79), which is influenced by complex interactions between the tumor and its microenvironment. Therefore, when examining data suggesting that mutations lead to the rewiring of cancer metabolism, in fact, it may be more pertinent to question whether these mutations arise in the first place from selective pressures applied by extrinsic metabolic factors in the tumor microenvironment. In a noteworthy example, while KRAS-mutant cancers exhibit resistance to low-glucose growth conditions, glucose deprivation has been shown to drive the accumulation of novel KRAS mutations in KRAS wild-type cancers (80). This indicates that extrinsic factors, such as the availability of essential nutrients, directly influence the fate of resulting genetic alterations that confer growth and survival advantages of cancer cells in their given microenvironment.

With regards to mutations in DNA-repair genes, it is possible that the highly dynamic and fluctuating conditions encountered in the tumor microenvironment provide an evolutionary pressure to select for cancer cells that have heightened genomic instability and are more proficient at adapting to these changes as a result. Consistently, it has been shown that tumors with increased heterogeneity are less responsive to therapy and behave more aggressively than tumors arising from single clonal populations (81–83). As we have discussed, a multitude of extrinsic metabolic factors affects DNA repair and, subsequently, genomic stability (**Figures 1** and **2**). Therefore, when considering future strategies to refine existing or develop novel cancer therapies, it will be essential at all stages of research and drug design to take account of the dynamic interplay between microenvironment and metabolic factors that are proving to influence treatment efficacy so substantially.

#### AUTHOR CONTRIBUTIONS

MOT and GP wrote the manuscript. MOT and NP designed the figure. NP drew the figure and contributed to the revision of the manuscript, and bibliography.

#### FUNDING

MOT and NP are supported by Cancer Research UK Ph.D. studentships. Work in the GP lab is supported by the Institute of Cancer Research and a Cancer Research UK Grand Challenge award (C59824/A25044).


of clinical radiosensitivity. *Int J Radiat Oncol Biol Phys* (2006) 66(5):1498–505. doi:10.1016/j.ijrobp.2006.08.064


BRCA-positive and BRCA-negative breast cancer. *J Clin Oncol* (2008) 26(26):4282–8. doi:10.1200/JCO.2008.16.6231


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

# Metabolic Profile of Oral squamous carcinoma cell lines relies on a higher Demand of lipid Metabolism in Metastatic cells

*Ana Carolina B. Sant'Anna-Silva1 , Gilson C. Santos1,2, Samir P. Costa Campos1 , André Marco Oliveira Gomes1 , Juan Alberto Pérez-Valencia1 and Franklin David Rumjanek1 \**

*<sup>1</sup> Instituto de Bioquímica Médica Leopoldo de Meis, Centro de Ciências da Saúde, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil, 2Centro Nacional de Biologia Estrutural e Bioimagem I (CENABIO I)/Centro Nacional de Ressonância Magnética Nuclear (CNRMN), Laboratório de Ressonância Magnética Nuclear de Biomoléculas (bioNMR), Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil*

#### *Edited by:*

*Simona Pisanti, Università degli Studi di Salerno, Italy*

#### *Reviewed by:*

*Daniele Vergara, University of Salento, Italy Tone Frost Bathen, Norwegian University of Science and Technology, Norway*

> *\*Correspondence: Franklin David Rumjanek franklin@bioqmed.ufrj.br*

#### *Specialty section:*

*This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology*

*Received: 30 October 2017 Accepted: 16 January 2018 Published: 02 February 2018*

#### *Citation:*

*Sant'Anna-Silva ACB, Santos GC, Campos SPC, Oliveira Gomes AM, Pérez-Valencia JA and Rumjanek FD (2018) Metabolic Profile of Oral Squamous Carcinoma Cell Lines Relies on a Higher Demand of Lipid Metabolism in Metastatic Cells. Front. Oncol. 8:13. doi: 10.3389/fonc.2018.00013*

Tumor cells are subjected to a broad range of selective pressures. As a result of the imposed stress, subpopulations of surviving cells exhibit individual biochemical phenotypes that reflect metabolic reprograming. The present work aimed at investigating metabolic parameters of cells displaying increasing degrees of metastatic potential. The metabolites present in cell extracts fraction of tongue fibroblasts and of cell lines derived from human tongue squamous cell carcinoma lineages displaying increasing metastatic potential (SCC9 ZsG, LN1 and LN2) were analyzed by 1 H NMR (nuclear magnetic resonance) spectroscopy. Living, intact cells were also examined by the non-invasive method of fluorescence lifetime imaging microscopy (FLIM) based on the auto fluorescence of endogenous NADH. The cell lines reproducibly exhibited distinct metabolic profiles confirmed by Partial Least-Square Discriminant Analysis (PLS-DA) of the spectra. Measurement of endogenous free and bound NAD(P)H relative concentrations in the intact cell lines showed that ZsG and LN1 cells displayed high heterogeneity in the energy metabolism, indicating that the cells would oscillate between glycolysis and oxidative metabolism depending on the microenvironment's composition. However, LN2 cells appeared to have more contributions to the oxidative status, displaying a lower NAD(P)H free/bound ratio. Functional experiments of energy metabolism, mitochondrial physiology, and proliferation assays revealed that all lineages exhibited similar energy features, although resorting to different bioenergetics strategies to face metabolic demands. These differentiated functions may also promote metastasis. We propose that lipid metabolism is related to the increased invasiveness as a result of the accumulation of malonate, methyl malonic acid, n-acetyl and unsaturated fatty acids (CH2)n in parallel with the metastatic potential progression, thus suggesting that the NAD(P)H reflected the lipid catabolic/anabolic pathways.

Keywords: oral squamous cell carcinoma, metastasis, cancer progression, nuclear magnetic resonance, metabolomics, metabolic reprogramming

### INTRODUCTION

Oral squamous cell carcinoma (OSCC) is the most prevalent type of head and neck squamous cell carcinoma frequently leading to metastasis to the neck lymph nodes (1). OSCC evolves from high proliferation rates of the oral squamous epithelium ultimately forming an *in situ* carcinoma (2, 3) followed by metastasis and a high lethality rate (4, 5).

Compared to normal cells, cancer cells have been shown to display a reprogrammed metabolism resulting from the specific energy demands imposed by growth factor signaling (6, 7). Furthermore, in the case of metastatic cells, migration and colonization of distant tissues also contribute to the extra energy burden. Thus, we envision metastatic cells as a subpopulation of cells that were selected in terms of a fine-tuned coordination that integrates nutrient uptake, anabolic, and catabolic processes. In addition, the microenvironment is variable insofar as the tumor anatomy is concerned. Whereas glucose, glutamine, and oxygen are freely available for those cells located on the surface of the tumor mass, the inner layers of cells are confronted by a radically different milieu characterized by paucity of nutrients and by hypoxia (8, 9). Consequently, these constraints introduce selective pressures that will reward metabolic plasticity. Those cells that can adjust to the different environments in the tumors will either thrive locally or eventually become detached and give rise to potentially metastatic cells. Successful adjustment can be achieved by gain of function through the concerted activation of expression of key enzymes that affect the metabolic flux and proliferative pathways as well as genes involved in the acquisition of resistance to anoikis through suppression of apoptotic programs. However, it is important to bear in mind that the metastatic phenotype probably results from non-adaptive innovation, that is, through the integration of pre-existing signaling pathways. By becoming manifest, these pathways confer different properties that enable cells to survive in an otherwise incompatible microenvironment (10–12).

Recently, the metabolomic approach using nuclear magnetic resonance (NMR) has become increasingly more informative. The availability of metabolomic data has been very useful for unraveling the metabolic pathways of several types of cancer as well as the biochemical features pertaining to metastasis (13–15). The main advantage of metabolomics rests on its ability to instantly and globally analyze metabolites quantitatively and qualitatively so that not only the involved pathways can be highlighted, but also their fluxes could be deduced (16, 17). Likewise, two-photon fluorescence lifetime imaging microscopy (FLIM), a non-invasive technique, has been successfully used to probe intact living cells in order to investigate their metabolism, thus affording a snapshot of their energy status. Experimentally, the auto fluorescence generated by both NADH and NADPH has been used to investigate the mitochondrial redox state and hence the energy producing pathways (18–20).

In the present study, we performed 1 H NMR and FLIM determinations combined with functional experiments in order to evaluate the metabolic alterations that may be relevant to the metastatic phenotypes of tongue squamous cells carcinoma (SCC) cells.

#### MATERIAL AND METHODS

#### Cell Lines

In the present study, cell lines developed and isolated from squamous cellular carcinoma SCC-9 (ATCC CRL-1629) by Agostini et al. (21) were used. The first cell line produced named SCC-9 ZsGreen stably expresses a green fluorescent zebrafish plasmid (ZsG). The paper describes how SCC-9 cells were inoculated into the footpads of BALB/c nude mice and were recovered as LN1 cells, the first metastatic generation. Another round of inoculation of LN1 cells produced LN2 cells, the second metastatic generation. Normal fibroblasts isolated from biopsies (3) were a gift by Dr. Ricardo Colleta from Department of Oral Diagnosis (School of Dentistry of Piracicaba, State University of Campinas, Brazil).

### Cell Culture

For SCC-9 derived cells, Dulbecco's Modified Eagle Medium: nutrient Mixture F12 (DMEM/F12; Gibco®, Life Technologies™, USA) was used. Media were supplemented with 10% fetal bovine serum (FBS—Cultilab, Brasil) and hydrocortisone 400 ng/ ml (Sigma-Aldrich, USA). For fibroblasts, we used Dulbecco's Modified Eagle Medium (DMEM *low*; Gibco®, Life Technologies™, USA) supplemented with 10% *donor bovine serum* (DBS; Gibco®, Life Technologies™, USA) and 1% penicillin–streptomycin. 1.1 × 106 cells of ZsG, LN1, LN2, and fibroblasts were transferred to 60.1 cm2 Petri dishes and maintained for 48 h in an incubator series 8000 water-jacketed CO2 (Thermo Scientific), with 5% of CO2 humidity atmosphere. At least four independent biological replicates of each cell line were used for experimental analysis. All cell lines were genotyped and tested free for *Mycoplasma* sp. infection using polymerase chain reaction.

#### NMR Metabolomics

ZsG, LN1, and LN2 cells were cultivated in DMEM/F12 and fibroblasts in DMEM low until 80% confluence. Then, ~1.2 × 107 cells from each replicate were trypsinized, centrifuged, and all the cell pellets were normalized to ~30 mg for metabolite extraction. We adapted the protocol for polar phase extraction according to adaptations of the method of Bligh and Dyer biphasic extraction previously described (22). Briefly, cells were extracted with methanol/chloroform/water (2:1:0.8), vortexed for 2 min after each addition, and homogenized during 30 min in a shaker in an ice bath. The polar aqueous phase (supernatant) was centrifuged at 4,600 *g* for 20 min at 15°C, the supernatant was dried overnight under vacuum (SpeedVac) and stored at −80°C until used. The samples were suspended in 50 mM phosphate buffer, pH 7.4, containing 10% of D2O and 0.1 mM of 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS) for chemical shift reference.

All spectra were acquired in a Bruker Avance III HD spectrometer running at 500.13 MHz for 1 H at 298 K. One dimensional spectra were acquired with excitation sculpting for water saturation, 20 ppm spectral width, 1.74 s relaxation delay, 64 K points, and 3 K accumulations. For assignments, we used 1 H-1 H TOCSY and pJRES. Spectra were aligned and processed in TOPSPIN 3.5 (Bruker-Biospin) and exported to AMIX (Bruker-Biospin), normalized by DSS intensity as internal reference, 0.02 ppm binned, and finally water and DSS signal were excluded. The peak intensities were used for the relative metabolite quantification in each sample.

For statistical analysis, datasets of at least four independent samples of each ZsG, LN1, LN2, and fibroblasts cell culture were collected. For the univariate analysis, we used multiple *t*-test with two-stage setup method of Benjamini–Krieger–Yekutieli, with false discovery rate (FDR) of 5%, with approach, assuming consistent SD, on GraphPad Prism 6.0. For multivariate analysis, we used Metaboanalyst 3.0 (23) with default practice (skip missing value imputation), filtered data with median intensity value (due to large amount of data), and normalized by sum of intensities. In addition, we applied the pareto scaling to reduce the impact of the buckets with high prevalence in each comparison [signaled as threonine, threitol, and (CH2)n] in the statistical analysis and to keep the data structure intact. Multivariate methods were used to compare all clusters (four groups) or paired groups (fibroblasts vs. ZsG; ZsG vs. LN1; LN1 vs. LN2) by means of principal component analysis and partial least-square discriminant analysis (PLS-DA). To validate class discrimination and avoid overfitting, we used a permutation test (1,000 permutations) based on separation distance, B/W-ratio, and cross-validation by the leave-one-out method (24, 25). We also calculated the variable importance in the projection (VIPscore) to analyze the ranking of the most important metabolites in separation groups and the PLS-regression coefficients for components 1, 2, and 3.

REACTOME free software (26) was used to determine the related metabolic pathways to metabolites accumulated in each comparison. Briefly, we listed all the related enzymes that produce each metabolite by manual curation, and it was analyzed into REACTOME using *Homo sapiens* as background, showing the corresponding metabolic pathways. FDR correction was used with significance of 0.05.

#### Fluorescence Lifetime Imaging

The autofluorescence lifetime images of isolate cultured cells were acquired using a laser scanning confocal fluorescence microscope Nikon Eclipse TE2000-U equipped with a Spectra-Physics MaiTai HP Laser (Spectra-Physics). The microscope is coupled to a fluorescence correlation spectrometer Alba Flim (ISS, Inc.) and data were collected by VistaVision software (ISS, Inc.) and analyzed by SimFCS 4 software (27). The samples were excited by two photons at 740 nm using a 60×/1.20 Plan-Apochromat water immersion objective lens (Nikon). The emission was detected on an Avalanche Photodiode detector, through a bypass filter 450 nm. A total of 100 frames from five different fields of each sample were collected. The image size acquired was 256 × 256 and the pixel dwell time was 40 µs. The calibration was done using a solution of 250 µM of NADH in water. The fluorescence lifetime was calculated by the phasor approach. Briefly, through Fast Fourier Transformation, the fluorescence decay in each pixel is plotted on the phasor plot by the coordinates G and S. The distribution of phasor points reveals areas of different lifetimes within the universal semicircle (28). Free and bound NADH have lifetimes of 0.38 and 3.4 ns, respectively, and their fluorescence intensity decays are plotted in the universal semicircle (29, 30).

#### High-Resolution Respirometry

High-resolution respirometry (Oroboros Oxygraph-2k) was performed to evaluate the oxygen consumption of intact cells, as previously described (31). 106 cells/ml of ZsG, LN1 and LN2 cells were cultivated separately and suspended in culture medium (DMEM/F12) without fetal bovine serum and phenol red. Then, cells were placed into a respiration chamber until the steadystate respiratory flux was attained (~10 min). Subsequently, the following parameters were measured: ATP synthase independent respiration in the presence of oligomycin (oligo) allowing the observation of oxygen consumption uncoupled to ATP synthesis ("leak"); content of oxygen consumption inhibited by the presence of oligomycin ("coupled"); titration with carbonyl cyanide-*4*-(trifluoromethoxy) phenylhydrazone (FCCP) to assess the maximal oxygen consumption rates once it stimulates the electron transfer system efficiency ("ETS"); residual oxygen consumption ("ROX") in the presence of rotenone (NADH dehydrogenase inhibitor) and antimycin A (AA) (cytochrome *bc1* inhibitor). Data acquisition and analysis were carried out with DatLab 5.1 software (Oroboros Instruments, Innsbruck, Austria).

#### Lactate Release in Culture Media

After 24 h of incubation with completed DMEM/F12 + 10% FBS and hydrocortisone 400 ng/ml, the culture medium was replaced by fresh DMEM/F12 without phenol red and FBS. Aliquots from the culture medium were collected at 0 and 60 min of incubation to evaluate lactate release through enzymatic assay. The lactate assay was performed in hydrazine/glycine buffer, pH 9.2, containing 5 mg/ml β-NAD<sup>+</sup> and 15 U/ml lactate dehydrogenase (LDH; Sigma-Aldrich). NADH absorbance was monitored in a microplate reader (SpectraMax M5, Molecular Devices) at 340 nm.

#### Proliferation Assay

Proliferation assays were performed using Sulforhodamine B (SRB) colorimetric assays, as previously described (32). Briefly, cells were placed into 96-well plates. After each endpoint (0, 24, 48, and 72 h), the culture medium was removed and cells were fixed with 10% trichloroacetic acid for 1 h at 25°C. The plates were then washed with distilled water and incubated with SRB solution (1% in acetic acid) for 15 min. After rinsing with 1% acetic acid, the cell monolayers were dried and the proteins were solubilized with Tris pH 10.4. Absorbance was measured at 490 nm using a spectrophotometer (SpectraMax Plus 384, Molecular Devices).

#### Statistical Analysis

Besides metabolomics, all the experiments were plotted as means ± SD for *n* independent experiments. Statistical significance to evaluate two groups was determined by the unpaired *t*-test, one-way ANOVA, and Dunnett posttest. Univariate analysis was done by multiple *t*-test, two-way ANOVA, and posttest Holm–Sidak. All set at alpha = 0.05. The graphs were generated by GraphPad Prism version 6.0 for Windows (GraphPad Software, La Jolla, CA, USA) and Excel (Microsoft Corporation*®*) with 95% of confidence level.

### RESULTS

#### Metabolic Profiles of Normal, Tumor, and Metastatic Cells

In order to gather insights into the intracellular metabolism of cells representing different stages of tumor progression, 1 H NMR spectra of cellular extracts were compared. Tongue normal fibroblasts were also investigated. High quality spectra of at least four replicates from each cell line (ZsG, LN1, and LN2) as well as fibroblasts samples were obtained. All determinations revealed similar spectral characteristics within each cluster of cell line, indicating that metabolic profiles of individual biological replicates were reproducible. Representative 1D 1 H NMR of the polar phase of the cellular extracts are shown in **Figures 1A,B** within the range of 0.8 and 8.0 ppm. Assignments of previously identified metabolites were obtained by comparing chemical shifts and spectral peak multiplicities with data from the literature and BMRB (33), HMDB (34), and COLMAR (35) data bases.

The major differences between each step of the metastasis progression, i.e., fibroblasts vs. ZsG, ZsG vs. LN1, and LN1 vs. LN2 were associated to the accumulation or decrease of threonine, threitol, n-acetyl glutamate, methyl malonate, malonate, n-acetyl, creatine, ribose, lactate, ethanolamine, phosphoethanolamine, and unsaturated lipids (**Table 1**).

Metabolomic profiles comparing the isolated groups of cells showed clearly the discrimination between normal and cancer cells by multivariate analysis PLS-DA score plot and VIP scores (**Figures 1C,D**). These analyses were validated using cross validation by the leave-one-out method (23–25), revealing an accuracy of −0.58 *R*<sup>2</sup> (clear variation) of 0.90 and *Q*<sup>2</sup> (predictive capability) of 0.57 for two components, representing a reliable classification model. In both measures, the value 1 indicates absolute fitting and high predictive power (36).

Several metabolic changes accompany tumorigenesis progression toward metastasis. To evaluate these alterations during the transformation and over tumor development, we previously defined paired comparisons replicating parental and derived cells. Then, we compared the metabolic profiling between each group and according to multivariate analysis PLS-DA, many metabolites were informative to distinguish the clusters. Among these values, we gathered the buckets with the highest VIP scores (>1; *p* < 0.05 from Student's *t*-test and FDR correction *q* < 0.05) and loading factors, considering the most discriminating power in each the comparison (**Table 1**; Figure S1 in Supplementary Material).

#### Metabolite Accumulation Is due Mainly to Lipid and Nitrogen Metabolism, through Aminoacid Transformation

In order to understand the origin and the biochemical implications of those metabolites, all the enzymes related to their formation were analyzed using the free software REACTOME (26), thus revealing the most representative biochemical pathways involved. The summary of the major findings is shown in **Table 2** and Table S1 in Supplementary Material. It is interesting to mention that there is only one metabolic pathway, related to metabolism of amino acids through aldehyde dehydrogenase 6 family member A1 (ALDH6A1), which use NAD<sup>+</sup> as cofactor, in fibroblasts and ZsG comparison. ZsG vs. LN1 comparison shows that β-oxidation and lipid metabolism are the most important processes in LN1 cells. In the same way, lipid metabolism is highly dependent on



*Table shows the metabolites identified, 1 H chemical shift values in ppm scale (*δ*H) and status represented by arrows* ↑↓ *indicating increased or decreased metabolites, respectively, based on VIP scores (*>*1; p* < *0.05 from Student's t-test and FDR correction q* < *0.05).*

NAD+, through hydroxyacyl-CoA dehydrogenase/3-ketoacyl-CoA thiolase/enoyl-CoA hydratase [trifunctional protein], alpha subunit (HADHA). A similar pattern is observed in LN1 vs. LN2 comparison. However, the fatty acid synthase (FASN) uses NADPH as cofactor. Also, interactomes of the related enzymes were performed using STRING software (37), showing clusters of acetylase activity and amino acid metabolism for fibroblast vs. ZsG and ZsG vs. LN1 comparisons, connecting lipid metabolism to acetylation in this last comparison. Between LN1 and LN2, there is a single cluster which is related to beta-oxidation (Figure S2 in Supplementary Material). Lipid metabolism appears to be prominently linked to metastatic phenotypes, while cell cycle regulation seems to be a feature of the less invasive cells.

### NAD(P)H FLIM Reveals that LN2 Cells Displays Longer Lifetimes

Most of biochemical pathways depend on reactions promoted by the availability of NAD(P)H, including fatty acids metabolism. Thus, the measurements of NAD(P)H free/bound ratios afford clues about enzymatic activities and enhanced processes (38). In this regard, we performed FLIM to evaluate the NADH free to bound ratio in ZsG, LN1 and LN2 cells. To simplify the analysis, we used the phasor plot approach, in which the fluorescence decay in each pixel is plotted on the phasor plot universe (29).


*Table shows the enzymes related to the accumulation of metabolites and the corresponding cell comparisons.*

With this approach, it is possible to observe that when cells are undergoing glycolysis there is a higher free/bound NAD(P) H ratio. In contrast, when cells have an increment in oxidative phosphorylation (OXPHOS), a lower free/bound NAD(P)H ratio is observed (39).

In **Figure 2A,** the intensity map of NADH fluorescence shows a different distribution of these molecules. To analyze the classical metabolic states (glycolytic and oxidative states), we divided the lifetime distribution of the cells in two cursors, named short and long lifetime cursor (**Figures 2B,C**). The short lifetime represents a higher NAD(P)H free/bound ratio (colored in red), which means lower binding to dehydrogenases, which is classically associated to a glycolytic profile. The long lifetime represents a lower NAD(P)H free/bound ratio (colored in green). This is related to an increased proportion of binding of NAD(P)H molecules to dehydrogenases, hence associated to OXPHOS (39) (**Figure 2E**). Moreover, we quantified the percentage of pixels in each image within the cursors and observed that ZsG and LN1 cell lines showed the same pattern of NAD(P)H free/bound ratio whereas LN2 has a lower ratio, indicating that these cell line exhibited a more oxidative profile (**Figure 2D**).

To better understand and compare the cell lines regarding the lifetime distribution, we analyzed the FLIM maps of each replicate pictured (40). Because the cell lines were shown to be heterogeneous and their phasor plots covered a wide range of different lifetimes, we evaluated the mean phasor plot of each image acquired taking into account all cells present in a certain field in order to obtain the phasor fingerprint of each cell population. **Figure 3** shows the distribution regarding means of G and S coordinates of ZsG, LN1, and LN2 cells. Analyzing the metabolic trajectory, we did not observe significant differences between the cell lines regarding the color-coded scale (**Figure 3B**), the phasor fingerprint (**Figure 3C**), and also the comparison of the mean G position of each cell line (**Figure 3D**, left). Notwithstanding, LN1 cells seem to have a dislocated dispersion of lifetimes when compared to ZsG and LN2. This can be seen by analyzing the mean S position of each cell line (**Figure 3D**, right), in which the pixels of LN1 images have a lower position on the phasor plot. Thus, we observed that although there were no dramatic differences among the cell lines regarding the metabolic pathways (glycolytic vs. oxidative states), they were different when the lifetime distributions were considered.

#### Mitochondrial Respiration and Lactate Exportation Does Not Impair the Malignant Transformation in SCC Derived Cell Lines

To further analyze the contributions of oxidative metabolism accompanying the FLIM experiments, we performed

Figure 2 | NADH fluorescence lifetime distribution in ZsG, LN1, and LN2 cells. (A) Color-coded images generated by the cursor selection according to the phasor plot (colored squares). (B) Two-photon fluorescence microscopy intensity images showing NADH fluorescence of ZsG, LN1 and LN2 cells excited at 740 nm. (C) Phasor plot illustrating the metabolic trajectories or the fraction of free/bound NADH in FLIM images. (D) Fraction of pixels in the phasor plot. (E) Metabolic trajectory on the phasor plot. All the images were obtained on a two-photon microscope (Nikon Eclipse TE2000-U) and Plan-Apochromat 60×/1.20 water immersion objective lens (Nikon). Data were acquired on VistaVision software and Alba FLIM instrument (ISS) and analyzed on SimFCS software. Values represent mean ± SD; *N* = 4; \*\*\**P* < 0.0001; ns compared within each group.

high-resolution respirometry assays using OXPHOS modulators in intact cells. **Figure 4A** displays a representative trace of oxygen consumption rate in each isolated cell line in the presence of culture medium containing the essential substrates (glucose and glutamine). In order to evaluate the dependence of respiratory complexes on oxygen consumption, we used membrane permeable compounds that affect mitochondrial function. **Figure 4B** shows the absolute values of oxygen consumption showing that it is not possible to distinguish ZsG, LN1, and LN2 by examining their oxidative status.

Metabolomic data showed that cells accumulate lactate during their malignant process (**Figure 1D**). Moreover, the proliferative phenotype is associated to higher lactate production and transportation to extracellular matrix in keeping with an enhanced glycolytic activity. In order to investigate this feature, we carried out a dynamic lactate liberation assay in presence or absence of AA, which stimulates glycolytic activity. Our results showed that even in the presence of AA there were no differences between all three cell lines (**Figure 4C**). Additionally, proliferation essays showed that LN1 and LN2 are less proliferative than ZsG cell line (**Figure 4D**).

#### DISCUSSION

The ability of cancer cells to detach from the primary tumor migrate through the blood/lymphatic system and eventually colonize distant tissues must be accompanied by metabolic reprograming. Probing those changes can help to understand the mechanisms of the metastatic processes. Accordingly, we first identified the major metabolites accumulated as the product of metabolic pathways in tongue SCC progression model used here. Taking a holistic approach, we performed 1 H RMN spectroscopy of ZsG, LN1, and LN2 cell lines as well as normal fibroblasts. We found that the major variations that could be associated to metastasis involved pathways of amino acids and fatty acid metabolism. These were revealed by alterations in the amounts of threonine, ribose, n-acetyl, malonate, methylmalonate, malonate, threitol, n-acetyl glutamate, ethanolamine, phosphoetanolamine, unsaturated lipids (CH2)n, and lactate (**Figures 1A–D**). In order to understand the metabolic pathways related to the principal metabolites accumulated in cells, we carried out paired comparisons considering parental (or one step before in malignant process) and derived cell lines having observed that therein lay the highest differences. Thus, the fibroblasts vs. ZsG, ZsG vs. LN1, and LN1 vs. LN2 cell lines were compared (**Table 1**; Figure S1 in Supplementary Material). The results in **Table 1** show the main differences encountered for metabolites present in the cell lines. The individual occurrence and concentrations of the compounds listed at any given moment could arise from either upstream activation of enzymes participating in the catabolism of various precursors, or from inhibition of downstream enzymes participating in their breakdown. To make a mechanistic sense of these findings, one would have to take into account the fluxes of all involved pathways and then propose models that would fit into coherent biochemical patterns compatible with the phenotypes of the cells displaying different metastatic potentials. Despite the unknown kinetics following the dynamic interplay that contributed toward the pattern observed, the results reflect in reproducible manner instant snapshots of the metabolic status of each cell line. That is, the differences detected are definitely informative in terms of the metabolic reprograming that took place for each cell line. In other words, the results emphasize that

in intact cells (A) and absolute results of O2 flow per cell (pmol × s−<sup>1</sup> × 10−<sup>6</sup> cells) (B). During the assay, cells were maintained in DMEM F12 medium without FBS. Oxygen consumption was evaluated after the addition of modulators of mitochondrial activity [oligomycin, FCCP, rotenone and antimycin A (AA)]. Routine respiration corresponds to endogenous oxygen consumption of cells. All parameters were corrected to ROX values. (C) Lactate release after 60 min of incubation with fresh DMEM F12 medium without FBS in the absence (Ctrl) or presence (+AA) of AA. (D) Proliferation using sulforhodamine B (SRB) colorimetric assays. Values represent means ± SD; *n* = 4; \*\**P* < 0.005 comparing among cell lines (oxygen consumption and proliferation) or within cell line (lactate release).

the cell lines have individual profiles that can suggest specific ontologies and functional meanings. Nevertheless, it is possible to speculate that some of the compounds listed in **Table 1** may actually have roles as intermediaries. This means that in their own right the metabolites could act as effectors participating in regulatory networks, altering the expression patterns and, therefore, impacting cellular plasticity. Among the compounds listed in **Table 1**, some of the molecules, such as creatine, threitol, threonine, phosphoethanolamine, ethanolamine, and ribose, irrespective of their potential additional roles in other metabolic pathways, could be grouped under the functional class of osmolytes. These are compounds that interchangeably stabilize the intracellular milieu by buffering osmotic perturbations, for instance. The notion that the osmolytes and compatible solutes listed in **Table 1** may play such a role opens up the possibility that they could act as effectors of a much broader regulatory network, given their ability to affect flexible and random coil conformations of transcription factors (41–43). In this way, the compounds detected in our metabolomic analyses may have a dual role. On the one hand, they may arise from differentially activated metabolic pathways responding to energy demands and supplying precursors such as building blocks for membranes, DNA, etc. Alternatively, the metabolites could act as bona fide regulators by altering the conformations of members of the signaling pathways and thus generate phenotypes such as the metastatic ones. This model is attractive since one could invoke new levels of regulation that are independent of gain or loss of function induced by those regulations and to grow and survive, cancer cells display alternate metabolic pathways toward tumorigenesis and progression. This hypothesis is currently being investigated in our laboratory.

Further information extracted from the metabolome involved the manual search for enzymes related to metabolite accumulation in their synthesis/degradation using biochemical pathways found in REACTOME free software (Table S1 in Supplementary Material). The results obtained pointed clearly to the involvement of the pathways connected to lipid metabolism. The importance of lipid metabolism becomes apparent when considering that lipid metabolism byproducts can influence growth promoting activity and hence tumorigenesis (44). However, the less invasive cells exhibit enzymes, which although could be broadly included in the lipid metabolism category are associated to specialized reactions pertaining to lipid biochemistry, or perhaps more to the point, catabolism of hydrophobic compounds.

Upon comparison between the metabolomic data between fibroblasts and ZsG, we found that threonine accumulated in fibroblasts. The metabolism of threonine involves the reaction catalyzed by threonine synthase (THNSL1 and THNSL2) forming l-threonine from O-phospho-l-homoserine (45). Threonine biosynthesis initiates from the precursor aspartate leading to a chain of reactions dependent of ATP and NADPH (46). In contrast, ZsG cells accumulated metabolites that promote biosynthesis relying in the reduction of NADP<sup>+</sup> producing NADPH, such as malonate and ribose. It is worth mentioning that methylmalonate (accumulated in LN1 cells) and malonate are both regulated by ALDH6A1, and that this enzyme is related to fatty acid degradation and amino acid metabolism (**Table 2**) (47).

Creatine pathway is not related to oxidation/reduction of NAD(P)H despite playing important roles in maintaining storage of ATP in skeletal muscle cells. During muscular contraction, the creatine-phosphate molecules can be decomposed forming ATP before fermentation reestablishes ATP content (48). Creatine may thus be adjuvant in maintaining the demand of ATP (49) in ZsG cell lines. Our data also shows enhanced activity of cell cycle regulation in ZsG and LN1 cells. This has been observed in several cancer cell lines with chromatin cohesion defects, which lead to a poor response to paclitaxel (50). This could mean that ZsG and LN1 cells may be more sensitive to spindle poisons than LN2 cells. Again, the catabolism of some fatty acids such as ω-fatty acids play a major role in the metabolism and biotransformation of exogenous compounds such as xenobiotics. This is achieved for example, by P450 (CYP) mixed function oxidase system that mediates by hydroxylation reactions that in general increase the solubility of these compounds, thus facilitating their excretion. The cytochrome P450 mixed function oxidase enzymes are also known to be involved in the biosynthesis of endogenous substrates such as cholesterol (51).

Phosphoethanolamine and ethanolamine are accumulated in LN1 cells when compared to ZsG. Phospholipids are required to build bilayer membranes in cells, and many studies show that its content increases as long as malignant process arises (52, 53). The membrane components including lipids and proteins allow the formation of ion channels, receptors, and signal transducers important to the interaction of extracellular matrix and cytosol (50), suggesting that those metabolites play important roles in proliferation and growth signaling in LN1 cells.

Comparison between LN1 and LN2 cell lines show that the unsaturated lipids (CH2)n are the only class accumulated in LN2 cells. (CH2)n metabolism shares pathways with valine, leucine, and isoleucine degradation (amino acid metabolism) and fatty acid metabolism. All of them are related to NAD(P)H dynamics. FLIM data confirm these observations (**Figure 2D**), showing that LN2 cells displays lower NAD(P)H free/bound ratio.

Ribose metabolism is associated to the pentose phosphate pathway (PPP), an important pathway that provides cancer cells with NADPH, ribose-5-phosphate by oxidative pathway. NADPH is essential in the antioxidant defense by glutathione production, while ribose-5-phosphate is an important element for nucleotide biosynthesis. Upregulation of PPP promotes cancer cell survival, proliferation, angiogenesis, invasion and metastasis, and resistance to radiotherapy and chemotherapy. On the other hand, the non-oxidative moiety of PPP reenters fructose-6-phosphate and glyceraldehyde-3-phosphate of the glycolytic pathway, fueling proliferation (54, 55). In the context of our model, the accumulation of ribose was observed in ZsG and LN1 cells. Interestingly, respirometry assays did not show any differences between the cell lines. Plausibly, this could be due to the non-oxidative PPP, fueling the bioenergetic pathways. Furthermore, *N*-acetylglutamate is an obligatory allosteric activator of carbamoyl phosphate synthetase I (CPS-1) (56). CPS-1 is related to cell growth and metabolite levels associated with nucleic acid biosynthesis pathway, as shown by CPS-1 knockdown in lung adenocarcinoma (57). Our analysis revealed that *N*-acetylglutamate is increased in ZsG and LN1 cells, which is compatible with amino acid metabolism and its contributions to nucleotide biosynthesis.

We did not detect differences in lactate exported from cells (**Figure 4C**) when the three cell lines were compared. The balance between accumulation and exportation lactate in tumor cells has been described as an important factor in regulating glucose metabolism and NAD<sup>+</sup>/NADH availability (58, 59). Much is known about lactate regulation in cancer cells and metabolic switch bearing glycolysis and tumorigenesis (60). Indeed, ZsG cells displaying higher proliferative rates were shown to incorporate less lactate when compared to LN1 and LN2 cells.

The interaction processes between NAD(P)H and proteins was also investigated here. Because these coenzymes play important roles in energy metabolism and, therefore, impact on tumor transformation and progression (20, 38, 61), their participation in the establishment of the metastatic phenotypes was determined, reflecting the importance of measuring NAD(P)H lifetime as an informative biomarker for understanding metabolic reprograming, mitochondrial physiology, oxidative stress, and apoptosis (39). Real-time NAD(P) H lifetime imaging by two-photon fluorescence microscopy allowed prospection into the dynamics of NAD(P)H, by analyzing its binding to specific dehydrogenases, representing an increased NAD(P)H free/bound ratio in glycolytic cells (39). As observed in FLIM experiments, LN2 cells exhibit a relatively smaller free/bound NAD(P)H ratio. However, there were no significant differences when comparing free/bound ratios of ZsG to LN1 cells.

Taken together, the results of the metabolomic analysis and FLIM analysis allowed us to conclude that several pathways connected to lipid metabolism appear to be prominently linked to metastatic phenotypes, while cell cycle regulation and amino acid metabolism are most related to less invasive cells.

### AUTHOR CONTRIBUTIONS

AS-S and GS contributed to N.M.R. metabolomics performance and analysis; AS-S, AG, and SC contributed to FLIM microscopy performance and interpretation of data; AS-S contributed to respirometry, lactate release, and proliferation assays; AS-S and JP-V contributed to the analyses of metabolic pathways; AS-S, JP-V, GS, SC, and FR discussed the data; FR coordinated the project. All authors participated in the writing and revision process.

#### ACKNOWLEDGMENTS

We would like to thank to Prof. Dr. Fabio Almeida for kindly providing support with N.M.R. operation, data acquisition, and brilliantly contributions to scientific discussions.

#### FUNDING

This project was supported by Fundação de Amparo à Pesquisa do Estado do Rio de Janeiro—FAPERJ (E-26/201.251/2014),

#### REFERENCES


Coordenação de Aperfeiçoamento de Pessoal de Nível Superior— CAPES (AUXPE-COFFECUB 2528/2014 Process 23038.001758/ 2014-47), and Conselho Nacional de Desenvolvimento Científico e Tecnológico—CNPq (403310/2016-4).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fonc.2018.00013/ full#supplementary-material.


dataset rows on the {K}-fold cross-validation quality parameters. *Mol Biosyst* (2014) 11(1):13–9. doi:10.1039/C4MB00414K


**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 Sant'Anna-Silva, Santos, Campos, Oliveira Gomes, Pérez-Valencia and Rumjanek. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Targeting Ongoing Dna Damage in Multiple Myeloma: effects of Dna Damage response inhibitors on Plasma cell survival

#### *Ana Belén Herrero1,2 and Norma Carmen Gutiérrez1,2,3\**

*1Cancer Research Center-IBMCC (USAL-CSIC), Salamanca, Spain, 2 Institute of Biomedical Research of Salamanca (IBSAL), Salamanca, Spain, 3Hematology Department, University Hospital of Salamanca, Salamanca, Spain*

#### *Edited by:*

*Luciana Boffoni Gentile, Federal University of Rio de Janeiro, Brazil*

#### *Reviewed by:*

*Valerio Donato, New York University, USA Gavin P. McStay, New York Institute of Technology, USA Eduardo J. Salustiano, Federal University of Rio de Janeiro, Brazil*

*\*Correspondence:*

*Norma Carmen Gutiérrez normagu@usal.es*

#### *Specialty section:*

*This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology*

> *Received: 08 March 2017 Accepted: 01 May 2017 Published: 19 May 2017*

#### *Citation:*

*Herrero AB and Gutiérrez NC (2017) Targeting Ongoing DNA Damage in Multiple Myeloma: Effects of DNA Damage Response Inhibitors on Plasma Cell Survival. Front. Oncol. 7:98. doi: 10.3389/fonc.2017.00098*

Human myeloma cell lines (HMCLs) and a subset of myeloma patients with poor prognosis exhibit high levels of replication stress (RS), leading to DNA damage. In this study, we confirmed the presence of DNA double-strand breaks (DSBs) in several HMCLs by measuring γH2AX and RAD51 foci and analyzed the effect of various inhibitors of the DNA damage response on MM cell survival. Inhibition of ataxia telangiectasia and Rad3-related protein (ATR), the main kinase mediating the response to RS, using the specific inhibitor VE-821 induced more cell death in HMCLs than in control lymphoblastoid cells and U266, an HMCL with a low level of DNA damage. The absence of ATR was partially compensated by ataxia telangiectasia-mutated protein (ATM), since chemical inhibition of both kinases using VE-821 and KU-55933 significantly increased the death of MM cells with DNA damage. We found that ATM and ATR are involved in DSB repair by homologous recombination (HR) in MM. Inhibition of both kinases resulted in a stronger inhibition that may underlie cell death induction, since abolition of HR using two different inhibitors severely reduced survival of HMCLs that exhibit DNA damage. On the other hand, inhibition of the other route involved in DSB repair, nonhomologous end joining (NHEJ), using the DNA-PK inhibitor NU7441, did not affect MM cell viability. Interestingly, we found that NHEJ inhibition did not increase cell death when HR was simultaneously inhibited with the RAD51 inhibitor B02, but it clearly increased the level of cell death when HR was inhibited with the MRE11 inhibitor mirin, which interferes with recombination before DNA resection takes place. Taken together, our results demonstrate for the first time that MM cells with ongoing DNA damage rely on an intact HR pathway, which thereby suggests therapeutic opportunities. We also show that inhibition of HR after the initial step of end resection might be more appropriate for inducing MM cell death, since it prevents the occurrence of a compensatory NHEJ repair mechanism. These preclinical observations provide the rationale for its clinical evaluation.

Keywords: DNA damage response, DNA damage, ataxia telangiectasia-mutated protein, ataxia telangiectasia and Rad3-related protein, double-strand break repair, homologous recombination, non-homologous end joining

## INTRODUCTION

Multiple myeloma (MM), the second most common hematological malignancy, arises from the abnormal clonal proliferation of malignant plasma cells (1, 2). Current therapies have significantly improved survival of MM patients. On the one hand, high-dose melphalan followed by autologous hematopoietic cell transplant has become a standard of care for young patients after bortezomibbased induction regimens. On the other hand, the introduction of novel agents, particularly bortezomib combined with alkylating drugs and prednisone, or lenalidomide plus dexamethasone has also improved the outcome of patients who are ineligible for high-dose therapy (3, 4). Nevertheless, MM remains an incurable disease, and new therapeutic strategies are still needed. A prominent feature of MM cells is their genome instability, whose underlying molecular basis is not fully understood. Recently, it has been proposed that high levels of ongoing intrinsic DNA damage and deregulated double-strand break (DSB) repair influence the acquisition of genomic changes (5–8).

Double-strand breaks pose a serious threat to cell viability and genome stability if left unrepaired or if they are repaired incorrectly (9). These DNA lesions can be generated exogenously, by exposure to a variety of genotoxic agents, or endogenously, due to various factors that slow down or stall replication forks, a phenomenon known as replication stress (RS). Sources of RS include fragile sites, replication–transcription complex collision, secondary DNA structures, and oncogenic stress (10, 11). To limit the impact of DSBs, cells have evolved the DNA damage response (DDR), a signal transduction cascade that coordinates the signaling and repair of these genomic lesions (9). Ataxia telangiectasia-mutated protein (ATM) and ataxia telangiectasia and Rad3-related protein (ATR) are central components of the DDR. ATM is a serine–threonine kinase that phosphorylates several key proteins, leading to cell cycle arrest through phosphorylation of CHK2, DNA repair, or apoptosis (12–14). It is mainly activated by DSBs, such as those induced by ionizing radiation (IR). In contrast, ATR is the key kinase in signaling the response to single-strand DNA (ssDNA), which can occur at persistent DSBs and on stalled replication forks. It is considered to be the main kinase mediating the response to RS (10).

Once DNA damage has been detected, DSBs are mainly repaired by two pathways: non-homologous end joining (NHEJ) and homologous recombination (HR). NHEJ is based on a direct ligation of the two ends of damaged DNA molecules and repairs DSBs mainly in G1 phase, although it is active throughout the cell cycle (15–17). HR promotes the recovery or repair of lesions that arise during replication and has a less important role in the repair of non-replication-associated DSBs (18). The pathway starts with the 5′-to-3′ resection of DNA ends, which is initiated by the endonuclease MRE11. The generated 3′-ssDNA overhangs bind to replication protein A, which is then exchanged for RAD51, the recombinase involved in the search for homology in the sister chromatid, thereby allowing subsequent repair (19).

High levels of RS leading to DNA damage have recently been described in a subset of MM (8). These patients, who overexpressed genes belonging to a chromosomal instability and DNA damage signature, displayed poor prognosis. On the other hand, several inhibitors of proteins involved in the DDR have been developed and some of them are either close to or are already being clinically trialed (20, 21). However, the effect of these types of compounds on MM cell viability has not been investigated in depth. In this study, we confirmed the presence of DSBs in several human myeloma cell lines (HMCLs) and evaluated cell survival after inhibition of the DDR. Our results show that MM with ongoing endogenous DNA damage depends on ATR over ATM and on HR over NHEJ, providing evidence of the presence of RS in these cells. However, the absence of ATR was partially compensated by ATM, and NHEJ was activated when HR was inhibited before DNA resection. These findings suggest that multiple inhibition of the DDR or inhibition of HR after DNA resection could extend the therapeutic opportunities. The implications of our findings for the treatment of the disease are discussed.

### MATERIALS AND METHODS

#### Cells and Culture Conditions

The HMCL MM1S was acquired from American Type Culture Collection, and JJN3, RPMI-8226, U266, and OPM2 were obtained from Deutsche Sammlung von Mikroorganismen und Zellkulturen. LINF903, an Epstein–Barr virus-transformed B-cell line established from a healthy individual, was obtained from the National DNA Bank of the University of Salamanca, Spain. MM and LINF903 cell lines were cultured in RPMI 1640–l-glutamine medium (Sigma-Aldrich, St. Louis, MO, USA) supplemented with 10% of fetal bovine serum (Sigma-Aldrich) and antibiotics (Gibco Life Technologies, Grand Island, NY, USA). All cells were incubated at 37°C in a 5% CO2 atmosphere. The presence of mycoplasma was routinely checked with a MycoAlert kit (Lonza, Basel, Switzerland) and only mycoplasma-free cells were used in the experiments.

#### Immunofluorescence Staining

Cells (50,000) were mounted on glass slides by cytospinning for 10 min at 1,000 rpm*.* Cells were fixed in 2% paraformaldehyde for 20 min, permeabilized in 0.2% Triton X-100 in PBS for 10 min, blocked in 3% BSA in PBS for 30 min and incubated with antiγ-H2AX (mouse, clone JBW301, Millipore) or anti-Rad51 (Ab-1, rabbit, Millipore) at 1:1,000 dilution for 2 h. After washing, slides were incubated with fluorescent secondary antibodies (1:1,000, Alexa Fluor 488 goat anti-mouse IgG or Alexa Fluor 568 antirabbit) for 1 h. Slides were mounted with ProLong Gold antifade reagent (Invitrogen, Carlsbad, CA, USA), and images were acquired using a DeltaVision system made up of a Olympus IX71 microscope, a Photometrics Coolsnap camera, and SOFWORX software. A 60× oil immersion objective was used.

#### Immunoblotting

Cells were washed with PBS and resuspended in RIPA lysis buffer (Santa Cruz Biotechnology) containing protease inhibitors (Complete, Roche Applied Science, Indianapolis, IN, USA) and phosphatase inhibitors (Santa Cruz Biotechnology). Protein concentration was measured using the Bradford assay (BioRad, Hercules, CA, USA). Protein samples (20 μg/lane) were subjected to SDS-PAGE and transferred to PVDF membrane (BioRad). After blocking, membranes were incubated with anti-human antibodies. The following primary antibodies were used: anti-p-ATM (pSer1981, mouse, clone 10H11.E12, Santa Cruz Biotechnology), anti-ATM (rabbit, clone D2E2, Cell Signaling, Danvers, MA, USA), anti-p-ATR (Ser428, rabbit, Cell Signaling), anti-ATR (goat, Santa Cruz Biotechnology) anti-tubulin (rabbit, Abcam, Cambridge, UK). Horseradish peroxidase-linked donkey anti-rabbit, anti-mouse, or anti-goat antibodies (Santa Cruz Biotechnology) were used as secondary antibodies at 1:10,000 dilution. Immunoblots were incubated for 1 h at RT and developed using enhanced chemiluminescence Western blotting detection reagents (Amersham Biosciences, Piscataway, NJ, USA).

#### Reagents

KU-55933 and VE-821 were obtained from MedChemtronica (Stockholm, Sweden). Mirin was purchased from Sigma and NU7471 from Santa Cruz Biotechnology.

### Cell Apoptosis Assays

Apoptosis was measured using Annexin V–fluorescein isothiocyanate/propidium iodide (Annexin V-FITC/PI) double staining (Immunostep, Spain) according to the manufacturer's guidelines.

### Cell Cycle Analysis

Cells were washed in PBS and fixed in 70% ethanol for later use. Cells were rehydrated with PBS, resuspended in 500 µl of PI/ RNase staining solution (Immunostep), and incubated for 20 min at RT in the dark. Samples were analyzed using a FACSCalibur flow cytometer.

#### HR Functional Assay

To determine the *in vivo* levels of HR, a reporter plasmid (22) was integrated into the genome of JJN3 and U266 cell lines as follows: cell lines were transfected with 1 µg of the HR reporter plasmid linearized by digestion with *Nhe*I. Amaxa Cell Line Nucleofector Kit V and an Amaxa Nucleofector device (Lonza, Allendale, NJ, USA) were used with programs X-005, for the U266 cell line, and T-016 for the JJN3 cell line. Three days after transfection G418 was added at 500 µg/ml. Medium containing G418 was changed every 3 days. Stable pools were obtained after 3 weeks of selection and were named U266-HR and JJN3-HR. In the chromosomally integrated reporter cassette, a unique DSB can be introduced by the rare-cutting endonuclease *I-Sce*I. Upon induction of a DSB, a functional GFP gene can be reconstituted by gene conversion, the predominant HR repair pathway in mammalian cells (22). To evaluate HR efficiency, 106 cells per transfection were cotransfected with 5 µg of an *I-Sce*I-expressing plasmid together with 0.5 µg of pDsRed-N1

in the indicated cell lines. Data are the mean of three independent experiments. At least 75 cells per experiment were counted. Error bars correspond to the SD (\*\*\**p* < 0.001 and \*\**p* < 0.01 compared with LINFB cells, Student's *t*-test). (B) Representative immunofluorescence images of cells stained with anti-γH2AX or anti-RAD51 antibodies. (C) Western blot of kinase proteins involved in the DNA damage checkpoint.

#### Figure 2 | Continued

Multiple myeloma cells with ongoing DNA damage are hypersensitive to the simultaneous inhibition of ataxia telangiectasia-mutated protein and ataxia telangiectasia and Rad3-related protein kinases. (A) Cell survival after 48 h of treatment with 4 mM caffeine determined by Annexin V/propidium iodide (Annexin V/PI) staining. (B) Cell survival after 72 h in the absence (untreated = UT) or the presence of 10 µM KU-55933, 5 µM VE-821, or both. Cell viability in the absence of treatment (only DMSO) was taken as 100%, and values obtained for all other conditions were normalized with respect to the UT value. (C) Examples of flow cytometry dot plots of cells stained with Annexin V/PI after 72 h of treatment with the indicated chemical inhibitors. Data are the mean of at least three independent experiments (\*\*\**p* < 0.001, \*\**p* < 0.01, and \**p* < 0.05).

to normalize measurements with respect to the transfection efficiency and were incubated in the presence or absence of various DDR inhibitors. Live cells were selected by FSC/SSC gating, and live GFP+ and DsRed+ cells were quantified by flow cytometry. HR efficiency was calculated as the ratio of GFP+ to DsRed+ cells.

#### RESULTS

#### HMCLs Exhibiting DNA Damage Are Hypersensitive to a Combination of ATM and ATR Inhibitors

Recent reports have shown ongoing constitutive DNA damage in several HMCLs and in plasma cells isolated from patients (6, 8). To corroborate these findings we quantified γ-H2AX and Rad51 foci, markers of DSBs, in HMCLs and in LINF903, a lymphoblastoid B cell line obtained from normal lymphocytes (7). HMCLs exhibited a higher percentage of cells with γ-H2AX and Rad51 foci than control lymphoblastoid cells, with the exception of U266 (**Figures 1A,B**), in agreement with recent published data (8). Activation of the DDR was detected in HMCLs with ongoing DNA damage by the presence of p-ATM and p-ATR (**Figure 1C**). The confirmation of high levels of DNA damage in most of the HMCLs prompted us to investigate the effect of various inhibitors of proteins involved in the DDR on MM cell viability. First, we analyzed the sensitivity of MM cell lines exhibiting low (U266) or high levels of DNA damage to caffeine, a well-known inhibitor of both ATM and ATR kinases. MM1S, RPMI-8226, and OPM2 were found to be more sensitive to the drug than U266, as revealed by Annexin V-FITC/ PI staining (**Figure 2A**). Specific inhibitors were then used to determine whether cell death induced by caffeine in HMCLs was due to inhibition of ATM, ATR, or both kinases. Inhibition of ATM with 10 µM KU-55933 resulted in a small reduction in cell survival in U266 and MM1S cells relative to untreated cells, whereas no significant effect was detected in any other HMCLs (**Figure 2B**). On the other hand, pharmacological inhibition of ATR with 5 µM of the specific inhibitor VE-821 (23) triggered a stronger apoptotic response in MM1S, RPMI-8226, and OPM2 than in LINF903 and U266 cell lines, in agreement with a recent report (8). Interestingly, inhibition of both kinases using a combination of the two drugs significantly increased cell death caused by ATR inhibition only in cells with high endogenous DNA damage (**Figures 2B,C**). These results indicated that ATM compensates for the absence of ATR, since the level of cell death caused by both inhibitors was higher than that observed with the individual inhibitors.

#### ATM and ATR Participate in DSB Repair by HR in MM, and Inhibition of Both Kinases Produces a Stronger Inhibitory Effect

Ataxia telangiectasia-mutated protein and ATR kinases are mainly known for phosphorylating their substrates CHK2 and CHK1 leading to cell cycle arrest in G1 or G2 (12, 24–26). However, even though many HMCLs exhibited endogenous DNA damage we found that cell cycle profiles did not change significantly after treatment with the combination of ATM and ATR checkpoint inhibitors (**Figures 3A,B**). This result suggests that ATM and ATR may carry out other functions in cells with DNA damage that make them dependent on the presence of both kinases. It has previously been shown that unrepaired DSBs can lead to cell death (27) and also that ATM is important for DSB repair by HR, since its downregulation reduces HR efficiency (14). Based on these findings, we hypothesized that inhibition of both kinases might strongly affect DSB repair by HR, leading to the death of MM cells with accumulated DNA damage. To analyze whether inhibition of ATM, ATR, or both caused defects in recombination, we took advantage of an HR substrate that was integrated within the chromatin of the MM cell lines JJN3 and U266. DSBs were then generated by transfection of the resulting JJN3-HR and U266-HR cells with an I-*Sce*I endonuclease-expressing plasmid, and HR efficiency in the presence or absence of the kinase inhibitors was estimated. We found that ATM inhibition by treatment with KU-55933 (10 µM) strongly reduced HR efficiency in JJN3-HR and U266-HR (**Figures 3C,D**), although cells were still able to perform HR to some extent. ATR inhibition by VE-821 also resulted in a notable reduction in HR that was lower than that induced by ATM inhibition in JJN3-HR, but higher in U266-HR. A stronger suppressive effect on HR was observed when both kinases were simultaneously inactivated, indicating that ATM partially compensates for the lack of ATR and *vice versa*.

#### Total Inhibition of Recombination Severely Affects Survival of MM Cells with Ongoing DNA Damage

Based on the results reported above we speculated that sensitivity to ATR inhibition is caused by an additional role of this protein to that performed in DSB repair. Cell death induced by the chemical inhibition of both kinases could be caused by the individual effect of ATR inhibition along with the more marked decrease in HR capability. It has previously been shown that inhibition of ATR in conditions of RS leads to replication fork collapse and the accumulation of DSBs (28, 29), so we hypothesized that such accumulation, together with defects in DSB repair, might underlie hypersensitivity to ATR inhibition in MM. To

of multiple myeloma cells but greatly reduces the efficiency of homologous recombination (HR) by gene conversion. (A) Cell cycle profiles of the indicated human myeloma cell lines untreated (UT) or treated with 10 µM KU-55933 and 5 µM VE-821. Cells were collected after 30 h of treatment and stained with propidium iodide. (B) Distribution of cells in G1, S, and G2/M phases of the cell cycle. (C) GFP+ and DsRed+ cells after 30 h of transfection with 5 µg of I-*Sce*I endonuclease-expressing plasmid and 0.5 µg pf pDSRed2-N1. (D) HR efficiency was calculated as the ratio of GFP+ to DsRed+ cells. The ratio in UT cells was taken to be 100% and in all other situations calculated as the percentage of UT controls. Data are the mean of three independent experiments. Error bars represent the SD (\*\*\**p* < 0.001, \*\**p* < 0.01, and \* *p* < 0.05).

explore this possibility, cells were treated with VE-821, and the presence of DSBs was analyzed by monitoring γH2AX foci. As shown in **Figures 4A,B**, treatment of MM cells with VE-821 clearly increased the number of cells with γH2AX foci relative to untreated controls. Number of γH2AX foci per cell also increased and in several cells became uncountable.

To explore whether the strong inhibition of HR causes the death of MM cells with ongoing DNA damage we used mirin, an MRE11 inhibitor that abolishes HR by impeding DSB resection (30). We also explored the contribution of the other main pathway involved in DSB repair, NHEJ, to the survival of MM cells exhibiting DNA damage. For this purpose, we used NU7441, a NHEJ inhibitor that targets DNA-PK (31). We found that treatment for 72 h with mirin reduced survival of MM1S, RPMI-8226, and JJN3 in a dose-dependent manner, whereas survival of LIN903 and U266 cells was higher, as expected, and remained similar at the different doses of mirin employed (**Figures 4A,B**). On the other hand, inhibition of NHEJ with concentrations of NU7441 up to 40 µM did not affect survival of any of the cell lines analyzed, but clearly decreased survival of HMCLs with a high level of ongoing DNA damage when treated with 25 and 50 µM mirin. Based on these results we concluded that mirin at 25 or 50 µM might not completely inhibit HR, and, if so, cells might survive HR inhibition by the compensatory NHEJ pathway. To explore this hypothesis, JJN3-HR and U266-HR cells carrying the HR reported cassette were treated with different doses of mirin, and HR efficiency was analyzed after 30 h of culture (**Figure 5C**). We found that the recombination efficiency of live cells decreased with increasing doses of mirin, from a partial loss of efficiency at 25 µM to a nearly total loss at 100 µM. These results indicated that HMCLs with ongoing DNA damage are hypersensitive to HR inhibition and also that defects in DNA resection induced by mirin can be compensated by NHEJ, a mechanism able to rejoin broken molecules before total resection occurs (32). Importantly, the reduction in HR was very similar in JJN3-HR and U266-HR (**Figure 5C**), which demonstrated that U266 was more resistant to mirin because it suffered less endogenous DNA damage, and not because mirin was not active in these cells.

To confirm the dependency of MM cells on HR we used B02, a recently discovered inhibitor of Rad51 (33). HMCLs were treated with different doses of B02, and HR efficiency and cell viability were analyzed. As expected, B02 clearly inhibited HR in both U266-HR and JJN3-HR reporter cell lines (**Figure 6A**), although inhibition of HR resulted in more cell death in JJN3 and RPMI-8226 than in U266 (**Figures 6B,C**), further confirming that MM cells with ongoing DNA damage are hypersensitive to HR inhibition. We found that simultaneous inhibition of HR using B02 and NHEJ with NU7441 did not increase cell death induced by B02 alone (**Figure 6B**). This result clearly indicated that endogenous DNA damage in MM is mainly repaired by HR and that if this pathway is inhibited after DNA resection occurs the break can no longer be repaired by a compensatory NHEJ mechanism. An illustration summarizing all the results is shown in **Figure 7**.

#### DISCUSSION

High levels of DNA damage can make tumor cells dependent on a proper DDR and represent a vulnerability of cancer cells that could be targeted therapeutically (34). In this study, we show that inhibition of the DDR is particularly toxic for MM cells exhibiting high levels of DSBs. Specifically, MM cells were found to be hypersensitive to agents that inhibit DSB repair by HR, a pathway upregulated in MM and important for the resolution of breaks associated with RS.

Recently, Cottini et al. showed that several HMCLs suffer RS, which is induced, at least in part, by the overexpression of MYC. In fact, silencing this oncogene in HMCLs exhibiting DNA damage and MYC overexpression reduced the number of lesions. On the other hand, its overexpression in U266, an HMCL with low levels of DNA damage and no MYC overexpression, triggered an increase in DSBs (8). The authors also found that ATR, the main kinase involved in the RS response, promoted survival of MYC-overexpressing cells. Here, we confirm the presence of high levels of DNA damage in several HMCLs, with the exception of U266, whose level of endogenous DNA damage was similar to that found in normal lymphoblastoid B cells, which were used in this work as healthy controls. We also confirmed that HMCLs

independent experiments. Error bars correspond to the SD (\*\*\**p* < 0.001 and \**p* < 0.05).

exhibiting DNA damage rely on activated ATR for survival. Moreover, we observed that these cells were more sensitive to ATR than to ATM inhibition, probably because the first induces replication fork collapse (28, 29), leading to the accumulation of DSBs (**Figure 4**). However, we found that the absence of ATR was compensated by ATM, since cell death induced by the inhibition of both kinases was more extensive than that obtained using single kinase inhibitors. It is especially noteworthy that the lethal effects of ATM and ATR inhibitors were found only in cells with extensive endogenous DNA damage, which bolsters the therapeutic opportunities. ATR and ATM transduction pathways were initially considered as two parallel pathways: IR induced the ATM pathway, whereas RS activated the ATR pathway. However, the regulation of ATM and ATR was later

shown to be mutually dependent in response to various DNAdamaging conditions like UV light, IR, and replication stalling (35–39). Thus, ATM has been demonstrated to play a role in maintaining fragile site stability, which is revealed only in the absence of ATR (40). Our results showing an increase in MM cell death with the combination of the two kinase inhibitors indicate the importance of both proteins for coping with endogenous DNA damage in MM and suggest that a double therapy could be more efficient for killing malignant cells. In this regard, it has been reported that in *ATM*-defective chronic lymphocytic leukemia cells, inhibition of ATR signaling by AZD6738 leads to an accumulation of unrepaired DNA damage that results in cell death by mitotic catastrophe (41). Dual inhibition of ATR and ATM has also been shown to potentiate the activity of the DSBinducing drugs trabectedin and lurbinectedin by perturbing the DDR and HR repair (42).

We show for the first time that ATM and ATR are involved in DSB repair by HR in MM and also that inhibition of both kinases completely abolishes HR efficiency. These results, obtained using an integrated-GFP-based DNA repair reporter substrate, support a previously described model in which ATM and ATR collaborate to maintain the activity of CtIP for efficient DNA end resection during HR (43). We found that ATM/ATR inhibition did not affect cell cycle distribution of HMCLs with high endogenous DNA damage (**Figures 3A,B**), which led us to hypothesize that the effect on HR, together with the accumulation of DSBs due to ATR inhibition, might underlie cell death induction. This hypothesis was supported by the finding that abolition of HR using the MRE11 inhibitor mirin, or the RAD51 inhibitor B02, severely decreased survival of HMCLs exhibiting high levels of endogenous DNA damage. Here, we show that inhibition of the other route involved in DSB repair, NHEJ, using the DNA-PK inhibitor NU7441 did not affect MM cell viability, which clearly demonstrated that endogenous DNA damage in MM is repaired by HR. We found that NHEJ inhibition did not increase cell death when HR was simultaneously inhibited with B02, but did clearly enhance cell death when HR was inhibited with mirin, which interferes with recombination before DNA resection occurs. These results are consistent with previous observations by Shibata et al., who showed that inhibition of the endonuclease activity of MRE11 promoted NHEJ (32). They are also consistent with the notion that 3′-ssDNA ends generated after DNA resection as substrates for HR can no longer be channeled toward NHEJ (18). Importantly, our results indicate that inhibition of HR after the initial step of end resection might be more appropriate for inducing MM cell death, since it prevents the occurrence of a compensatory NHEJ repair mechanism.

Many studies have shown that combinations of genotoxic agents with inhibitors of the DDR produce greater cell death of tumor cells compared with single agents (21, 44–46). In MM, the cytotoxic effects of bendamustine, melphalan, and doxorubicin on p53-deficient cells are enhanced by AZD7762, a CHK1/CHK2 inhibitor (47). In addition, inhibition of RAD51 has been found to sensitize MM cells to IR and to the DSB-inducing drug doxorubicin (48). However, the present study is the first to indicate that the single-agent activity of HR inhibitors induces apoptosis of MM cells with intrinsic DNA damage, providing a therapeutic window. It is important to note that partial inhibition of HR did not result in complete myeloma cell death at the times assayed, probably because residual traces of activity may be enough to ensure repair of endogenous DSBs. However, doses that completely abolished HR, such as the combination of ATM and ATR inhibitors, or concentrations of B02 greater than 15 µM drove most of the MM cells with intrinsic DNA damage to apoptosis. It is possible that the high doses of HR inhibitors needed to induce complete MM cell death could have a toxic effect, especially over prolonged periods, but maybe one that is less severe than that induced by combinations of DDR inhibitors with genotoxic agents, since the latter induce DNA damage and are also toxic to non-tumor cells. Nevertheless, specific (or temporary) inhibition of HR could be efficient not only for reducing MM cell survival but also for mitigating genomic instability and disease progression, given that we and others have demonstrated a high level of RAD51 expression and increased HR that lead to the genome instability characteristic of the disease (5, 7). Taken together, our findings suggest that HR inhibition could be a promising target for the treatment of MM. Therapeutic outcomes and toxicity profiles of the different inhibitors used as single agents or in combinations need to be tested in clinical trials.

### AUTHOR CONTRIBUTIONS

AH conceived and designed the work, performed the experiments, wrote the manuscript, and approved the final version of the manuscript. NG contributed to design the work, revised it critically for important intellectual content, ensured that any part of the work was appropriately investigated and resolved, corrected the manuscript, and approved the final version to be published.

#### REFERENCES


#### ACKNOWLEDGMENTS

We thank Dr. V. Gorbunova for the HR reporter plasmid and Dr. Wiesmueller for pCMV-I-SceI plasmid. We are grateful to A. L. Prieto for his help with the Deltavision microscope.

### FUNDING

This work was supported by the Spanish Association for Cancer Research (AECC, GCB120981SAN), the INNOCAMPUS program (CEI10-1-0010), and the Samuel Solórzano Barruso foundation (FS/25-2015).


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

The reviewer, ES, and handling editor declared their shared affiliation, and the handling editor states that the process nevertheless met the standards of a fair and objective review.

*Copyright © 2017 Herrero and Gutiérrez. 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) or licensor 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.*