Trends in Molecular Medicine - December 2013

December 2013 Volume 19, Number 12 pp. 705–754

Editorial Team Editor Christopher Pettigrew

Reviews

705

Challenges and advances towards the rational design of mRNA vaccines

Charlotte Pollard, Stefaan De Koker, Xavier Saelens, Guido Vanham, and Johan Grooten

714

Microbiota impact on the epigenetic regulation of colorectal cancer

Tao Yang, Jennifer L. Owen, Yaíma L. Lightfoot, Michael P. Kladde, and Mansour Mohamadzadeh

726

Non-canonical roles of lysyl-tRNA synthetase in health and disease

Alex Motzik, Hovav Nechushtan, Shen Yun Foo, and Ehud Razin

732

Developing epigenetic diagnostics and therapeutics for brain disorders

Irfan A. Qureshi and Mark F. Mehler

742

Common non-epigenetic drugs as epigenetic modulators

Jörn Lötsch, Gisbert Schneider, Daniel Reker, Michael J. Parnham, Petra Schneider, Gerd Geisslinger, and Alexandra Doehring

Executive Editor, Microbiology Lakshmi Goyal Journal Manager Olaf Meesters Journal Administrators Ria Otten Patrick Scheffmann Advisory Editorial Board S. Agrawal, Cambridge, USA K. Alitalo, Helsinki, Finland H. Blau, Stanford, USA K. Culver, East Hanover, USA K. Davies, Oxford, UK I. Dunham, Hinxton, UK F. Gage, La Jolla, USA D. Gordon, Manchester, UK W. Günzburg, Vienna, Austria D. Gurwitz, Tel Aviv, Israel M.A. Liu, Emeryville, USA K. Matsushima, Tokyo, Japan A. McMichael, Oxford, UK A. Papavassiliou, Athens, Greece W.A. Petri, Charlottesville, USA D. Rader, Philadelphia, USA C. Soto, Houston, USA T. Soussi, Paris, France J.S. Stamler, Cleveland, USA G.C. Tsokos, Boston, USA L-C. Tsui, Hong Kong, PR China J. Uitto, Philadelphia, USA S. Waxman, New Haven, USA Editorial Enquiries Trends in Molecular Medicine Cell Press 600 Technology Square Cambridge, MA 02139 Tel: +1 617 386-2133 Fax: +1 617 397-2810 E-mail: [email protected]

Cover: Epigenetic changes have emerged as key signatures of pathology in a range of brain disorders and can be observed both in the central nervous system and periphery. These signatures may have value both as diagnostic markers and therapeutic targets. On pages 732–741, Mehler et al. review the latest developments in epigenetic medicine and the potential to treat neurodevelopmental and neurodegenerative diseases. Cover image from iStockphoto/Eraxion.

Review

Challenges and advances towards the rational design of mRNA vaccines Charlotte Pollard1,2, Stefaan De Koker2, Xavier Saelens3,4, Guido Vanham1,5, and Johan Grooten2 1

Virology Unit, Department of Biomedical Sciences, Institute of Tropical Medicine, B-2000 Antwerp, Belgium Laboratory of Molecular Immunology, Department of Biomedical Molecular Biology, Ghent University, B-9052 Ghent, Belgium 3 Molecular Virology Unit, Inflammation Research Center, VIB, B-9052 Ghent, Belgium 4 Molecular Virology Unit, Department of Biomedical Molecular Biology, Ghent University, B-9052 Ghent, Belgium 5 Department of Biomedical Sciences, University of Antwerp, B-2000 Antwerp, Belgium 2

In recent years, mRNA vaccines have emerged as a safe and potent approach for the induction of cellular immune responses. Whereas initial studies were limited to the ex vivo loading of dendritic cells (DCs) with antigenencoding mRNA, recent progress has led to the development of improved mRNA vaccines that enable direct in vivo targeting of DCs. Although preclinical studies demonstrated their potency in inducing antitumor immunity, several bottlenecks hinder the broader application of mRNA vaccines. In this review, we discuss the challenges associated with mRNA-based vaccination strategies, the technological advances that have been made to overcome these limitations, and the hurdles that remain to be tackled for the development of an optimal mRNA vaccine. Vaccine-induced cellular immunity Over the past decades, traditional vaccines consisting of inactivated pathogens or subunit vaccines have been successfully used against infectious diseases for which humoral immunity is considered the primary correlate of protection. However, these approaches have limitations for chronic infections where cellular immune responses are required to control infection. Cellular immune responses such as cytolytic T cells (CTLs; see Glossary) have the capacity to target virally infected cells directly, as opposed to neutralizing or killing the causal agent, that is, the virus or bacteria, which is usually achieved through humoral immunity. Therefore, CTLs enable the eradication of the cellular reservoir of the pathogen and ultimately terminate pathogenic persistence. The significance of cellular immune responses has also been recognized in the field of cancer immunotherapy. A common aim in therapeutic tumor vaccination is the induction of tumor-specific

Corresponding author: Pollard, C. ([email protected]). Keywords: vaccination; cellular immunity; mRNA; dendritic cell; nucleic acid-based vaccines; cytotoxic T cells. 1471-4914/$ – see front matter ß 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.molmed.2013.09.002

effector T cells that reduce tumor mass and induce immunological memory to control relapse [1,2]. For efficient induction of CTL responses, the antigen needs to enter the endogenous route of antigen processing. This requires antigenic proteins originating from the pathogen to be delivered to the cytosol where proteasomal processing takes places. Subsequently, the resultant peptides are transported by the transporter associated with antigen processing (TAP) into the endoplasmic reticulum where the antigenic peptides bind with major histocompatibility complex (MHC) class I molecules. The MHC class I/peptide complexes are then transported to the cell surface for recognition by CD8+ T lymphocytes. Thus, the delivery of immunogens into the antigen-processing pathways promoting MHC class I-restricted CTL responses has been a major objective in current vaccine development. Nucleic acid-based vaccines have been shown to be particularly suitable for the generation of potent CTL responses as they enable expression of the encoded protein in the cytosol of the target antigen-presenting cell (APC). Although nucleic acid-based vaccine research has been mainly focused on the development of plasmid DNA and viral vectors, the limitations associated with these ‘classical’ approaches have recently led to an increased interest in mRNA-based vaccination strategies. Method of choice: mRNA vaccines A major motive for the use of mRNA vaccines is their superior safety profile as compared with plasmid DNA or viral vectors [3]. The risks of genome integration, long-term expression, and/or induction of pathogenic anti-DNA autoantibodies have hindered FDA approval of DNA-based vaccines for human use. Similarly, the human application of viral vectors is impeded by their potential reversion to pathogenicity and the presence of strong vector-specific humoral immune reactions, especially upon boosting. As these concerns are not applicable for mRNA vaccines, they are not classified by the FDA as ‘gene therapy’. In addition to its safer pharmaceutical properties, the transient expression of mRNA-encoded antigen enables a more controlled antigen exposure and minimizes the risk of tolerance induction that can be associated with long-term antigen exposure. Because additional sequences such as plasmid backbone and viral packing proteins are Trends in Molecular Medicine, December 2013, Vol. 19, No. 12

705

Review Glossary Adjuvant: a substance often used in vaccines to enhance immune responses against the antigen with which it is mixed. Apoptosis: process in which the cell activates an internal death program, characterized by DNA degradation, nuclear condensation and fragmentation, and plasma membrane blebbing, and ultimately resulting in phagocytosis of the cellular remains, without the induction of inflammatory responses. Autoantibodies: antibodies specific for self-antigens. Autoantibodies can cause damage to cells and tissues and result in systemic autoimmune diseases such as systemic lupus erythematosus. 50 Cap: the presence of a 7-methylguanosine linked to the first nucleotide of mRNA via a 50 –50 triphosphate bridge. The process of 50 capping is crucial for the development of mature mRNA and allows binding of proteins responsible for transport and translation of the mRNA molecule. Cytotoxic/cytolytic T lymphocyte (CTL): T lymphocytes whose major effector function constitutes the killing of virally infected cells. Most cytotoxic T cells are CD8+ T cells that recognize microbial peptides displayed by class I MHC molecules. Dendritic cells (DCs): bone marrow-derived immune cells that have a key function in the initiation of adaptive immune responses by presenting antigen to naı¨ve T cells. Electroporation: mechanical transfection method that uses an externally applied electrical field to permeabilize the cell plasma membrane. Electroporation is a highly efficient strategy for the introduction of exogenous DNA into mammalian cell cultures. Endocytosis: mechanism used by cells to actively internalize molecules through the formation of plasma membrane invaginations, called endosomes. Epitope: specific portion of the antigen that is recognized by an antibody or a T cell receptor. Epitopes bind to MHC molecules for recognition by the T cell receptor. Food and drug agency (FDA): US agency responsible for the regulation of food safety, tobacco products, medical drugs, and vaccines. Gene therapy: according to the FDA, gene therapy is ‘a medical intervention based on modification of the genetic material of living cells’. Both the administration of ex vivo modified cells and the direct in vivo delivery of genetic material are considered gene therapy. Immunotherapy: modification of immune responses for the treatment of disease. Active immunotherapy such as cancer vaccination involves the modification of the host immune system. Passive immunotherapy is independent of the patient’s immune system. Liposomes: artificial vesicles composed of an aqueous interior surrounded by one or more phospholipid bilayers. Liposomes can be used for the delivery of nutrients, pharmaceutical drugs, and vaccines. Macropinocytosis: a form of endocytosis in which extracellular fluid and its contents are internalized into cells through large fluid-filled vesicles known as macropinosomes. Major histocompatibility complex (MHC) molecule: set of membrane glycoproteins that present peptides for recognition by T lymphocytes. Class I MHC molecules are present on most nucleated cells and present peptides generated in the cytosol to CD8 T cells, whereas class II MHC molecules are largely restricted to professional APCs and present peptides degraded in the endosomes to CD4 T cells. Pattern recognition receptors (PRRs): receptors of the innate immune system that recognize molecular patterns on pathogenic surfaces and trigger innate immune responses against these pathogens. Polymers: natural or synthetic molecules consisting of multiple subunits known as ‘monomers’. Polymers used for the delivery of nucleic acids are mostly cationic, thus allowing them to electrostatically bind to nucleotides. Protamine: small, arginine-rich nuclear protein essential for sperm head condensation and DNA stabilization during spermatogenesis. Ribonuclease (RNAse): nuclease enzyme that catalyzes the degradation of RNA molecules. Toll-like receptor (TLR): type of PRR that shares structural homology and signal transduction pathways with the type I IL-1 receptors. TLRs are named after the Toll gene first identified in Drosophila. Transfection: process by which nucleic acids are introduced into mammalian cells. Transporter associated with antigen processing (TAP): heterodimeric molecule composed of the ATP-binding cassette molecules TAP-1 and TAP-2. TAP molecules mediate transport of peptides from the cytosol into the lumen of the endoplasmic reticulum where they associate with MHC class I molecules. Viral vectors: viruses that have been genetically modified to deliver exogenous genes into the cell (e.g., for vaccination or gene therapy). For safety reasons, virulence genes are deleted.

lacking in mRNA vaccines, immune responses are elicited only against the encoded antigen, and the pre-existence or induction of antivector antibodies is of no issue. Furthermore, RNA does not need to cross the nuclear barrier for 706

Trends in Molecular Medicine December 2013, Vol. 19, No. 12

protein expression and thus enables transfection of non or slowly dividing cells such as dendritic cells (DCs). Finally, as mRNA is transcribed from plasmid DNA templates by bacteriophage RNA polymerases (T7, T3, or S6), production and purification is easy and of low cost, and modifications can be easily introduced. Most studies employing antigen-encoding mRNA for induction of T cell immunity have used autologous DCs that were loaded ex vivo with mRNA and reintroduced into the patient. These mRNA-loaded DC vaccines have been tested both in the context of cancer immunotherapy and for therapeutic vaccination against HIV-1 [4,5]. Although clinical testing has shown safety and feasibility of this approach, the ex vivo manipulation of DCs requires a complex personalized vaccination procedure, which strongly impedes its application to larger numbers of patients. Therefore, there is a high demand for new vaccination approaches that target DCs directly in vivo. Successful induction of T cell immunity by direct in vivo mRNA-based vaccination is associated with several challenges. First, the antigen-encoding mRNA should be targeted to APCs before degradation by extracellular ribonucleases (RNAses) occurs. Following endocytosis by the target APC, the mRNA must escape the acidic endolysosomal compartment to enable antigen translation in the cytosol. Once inside the cytosol, the mRNA structure should be sufficiently stable and efficiently recruit translation initiation factors. Second, although cytosolic antigen translation will strongly stimulate the CD8+ T cell arm, potent CD4+ T cell responses should also be evoked to provide sufficient help for inducing and maintaining antigen-specific CD8+ T cells. Resolving these issues should enable a better presentation of the mRNA encoded antigen and evoke more potent T cell responses. However, in addition to an efficient presentation of the encoded antigen, priming of effector T cell responses requires the antigen to be presented by APCs in the correct activation status. Although mRNA vaccines can activate DCs by triggering several pattern recognition receptors (PRRs), DC activation is at best moderate. Developing strategies that allow a more optimal DC activation without interfering with mRNA uptake and translation represents a second major hurdle of in vivo mRNA vaccines. In the following sections, we will extensively elaborate on the bottlenecks that have hindered the broad application of mRNA-based vaccines, discuss the recent technological advances that have been made to overcome these limitations, and the challenges that remain to be tackled for the development of an optimal mRNA vaccine. Improving mRNA stability and antigen translation Several efforts have been undertaken to increase the stability and translation efficiency of mRNA vaccines (Figure 1). The extracellular half-life of mRNA can be extended by intranodal delivery of mRNA vaccines or by encapsulation of naked mRNA into particulate delivery systems to protect it from degradation by ubiquitous RNAses [6–9]. Methods to increase the intracellular stability and translation of mRNA include structural modifications affecting the 50 cap analog, the untranslated regions (UTRs), and the poly(A) tail [3].

Review


(A) Improving mRNA stability and translaon Peripheral ssue

Lymph node

Intracellular

• Parculate delivery systems

• Intranodal delivery

• • • •

Elongated polyA tail Modified 5′cap β-globine UTR Nucleode analogs Cap

(AAA)n

(B) Promong angen presentaon (I) Angen presentaon via the MHC-I processing route

CD8 T cell

Proteasomal degradaon MHC-I Protein translaon ER

Lamp1 MHC-II Late endosome/ lysosome

CD4 T cell

(II) Angen presentaon via the MHC-II processing route TRENDS in Molecular Medicine

Figure 1. Strategies to increase presentation of the mRNA-encoded antigen. Methods used to promote antigen presentation of mRNA vaccines include enhancing the extracellular half-life, increasing intracellular stability, and promoting help by CD4+ T cells through presentation via MHC-II molecules. (A) The extracellular half-life of mRNA can be extended by intranodal delivery of mRNA vaccines or by encapsulation of naked mRNA into particulate delivery systems. Methods to increase the intracellular stability and translation of mRNA include structural modifications affecting the 50 cap analog, the UTR, and the poly(A) tail. (B) Induction of CD4 T cells can be promoted by the addition of an MHC-II targeting sequence of an endosomal or lysosomal protein. Abbreviations: MHC, major histocompatibility complex; UTR, untranslated region.

Increasing the extracellular half-life: go intranodally or wrap-it-up! Barring a few exceptions, studies using extranodal delivery of naked mRNA vaccines have failed to elicit antigenspecific immune responses. By contrast, intranodal delivery of naked mRNA in mouse models has been shown to efficiently induce both CD4+ and CD8+ T cell responses. It is currently unknown how intranodal administration of antigen-encoding mRNA increases vaccine potency, but it is probable that the direct bioavailability of the mRNA molecule to a high concentration of APCs increases the likelihood that it will be taken up and translated before degradation by extracellular nucleases occurs. Three recently published studies describe the use of mRNA for

intranodal antitumor vaccination [10–12]. Kreiter et al. demonstrated that mRNA administered intranodally is selectively taken up by resident DCs, resulting in a T cell stimulatory environment and expansion of antigen-specific CD4+ and CD8+ T cells [10]. The exclusive uptake of mRNA by lymph node resident CD11c+ cells was also confirmed in a study using CD11c-DTR transgenic mice where DCs were depleted via diphtheria toxin (DT) treatment [12]. Furthermore, both prophylactic and therapeutic antitumor immunity was elicited, and the protective effect was superior over antitumor immunity induced by subcutaneous injection of tumor peptide adjuvanted with CpG. When comparing intranodal administration of naked antigen-encoding mRNA to the subcutaneous, near-nodal or intradermal 707

Review route, the authors found that only intranodal injection resulted in efficient antitumor immunity. Intranodal tumor vaccination was also superior over subcutaneous immunization with bone marrow-derived DCs electroporated with antigen-encoding mRNA, thus highlighting the potential of intranodal mRNA delivery for cancer immunotherapy [10]. A common strategy to increase the extracellular stability of extranodally delivered mRNA vaccines is to complex naked mRNA with non-viral carrier systems. In this regard, ovalbumin (OVA)-encoding mRNA complexed with the cationic liposome 1,2-dioleoyl-3-trimethylammonium-propane (DOTAP) has been shown to inhibit growth of OVA-expressing tumors both in prophylactic and therapeutic settings [13]. Addition of the helper lipid 1,2-dioleoyl-sn-glycero-3phosphoethanolamine (DOPE) in the cationic DOTAP lipids further improved the induction of CTL responses. Another strategy using non-viral carrier systems to improve mRNA vaccination is the encapsulation in histidylated lipopolyplexes [14]. Using this approach, Mockey et al. demonstrated that intravenous injection of mRNA encoding melanomaassociated antigen (MART1) complexed with histidine-rich cationic polymers and further encapsulated in histidylated cationic liposomes protected against melanoma tumor progression. When mRNA was complexed with either polymers or liposomes alone, no effect on tumor proliferation was observed. Recently, Fotin-Mleczek et al. described a two-component mRNA-based vaccine consisting of protamine complexed mRNA mixed with naked mRNA [15]. When injecting mice intradermally with naked luciferase-encoding mRNA, they observed high translation efficiency but low immunogenicity. Conversely, co-delivery of mRNA with the arginine-rich protein protamine strongly activated the innate immune system, but failed to express the encoded protein. The authors reasoned that a two-component mRNA vaccine consisting of protamine complexed and naked mRNA would stimulate the immune system while simultaneously enabling antigen expression. When immunizing mice intradermally with such a two-component mRNA vaccine encoding OVA, both prophylactic and therapeutic antitumor responses were induced against OVA-expressing tumor cells. However, the additional requirement for naked mRNA to allow for efficient protein expression may be dependent on the nature of the delivery system used to complex mRNA. Non-viral carriers with a strong affinity towards mRNA impede its release from the complexes and therefore have a negative influence on mRNA translation [16]. In addition, for efficient protein expression, the formed complexes have to escape from the endosomal compartment, a process which is dependent on the physicochemical properties (e.g., presence of helper groups that promote fusion with the cell membrane) of the carrier system [17]. Increasing intracellular stability and translation: changing the nucleotides All current mRNA vaccines contain one or more structural modifications to improve stability and/or translation efficiency. The most commonly used modifications are the ARCA cap and elongated poly(A) tails. Capping of mRNA species facilitates recognition of mature mRNA by the 708


translation initiation factor eIF4E and as a consequence improves translation initiation and RNA stability during protein synthesis [18]. Conventional cap analogs are incorporated into the RNA in both forward and reverse orientations leading to two isomeric RNA populations [19]. Because RNAs capped with reverse 50 caps cannot be translated, only 50% of the capped mRNA will result in protein expression. The antireverse cap analog (ARCA) contains a base modification that ensures that caps are only inserted in the functional, translation competent orientation [20]. This will in turn result in an increase in protein expression and enhanced priming and expansion of naı¨ve antigen-specific T cells [21]. The poly(A) tail protects mRNA from degradation and promotes subsequent binding of poly(A) binding protein. Thus, the addition of poly(A) tails to antigen-encoding mRNA has been reported to result in higher levels of protein expression [22]. Transcript stability can be further increased by the incorporation of the pyrimidine-rich stabilizing 50 or 30 UTR of b-globin. Although the exact mechanism underlying this phenomenon is unknown, RNA transcripts encoding b-globin are highly stable so that protein synthesis of globins continues after enucleation of the red blood cell. An alternative strategy to increase protein expression from in vitro transcribed mRNA is the incorporation of chemically modified nucleotide analogs. Replacement of 25% of uridine and cytidine with 2-thiouridine and 5methyl-cytidine increases transfection efficiency of human and murine epithelial cells up to seven- and ninefold, respectively. In mice, intramuscular injection of erythropoietin (EPO)-encoding mRNA containing these modified nucleotides similarly increased EPO protein levels fourfold compared with mice treated with unmodified EPO mRNA [23]. This increase in protein expression was attributed to loss of innate immune activation and its resulting inhibition of mRNA translation. Indeed, when uridine and cytidine were substituted with their modified counterparts, a decrease of mRNA binding to the PRRs Toll-like receptor (TLR)3, TLR7, TLR8, and retinoic acid-inducible gene-I (RIG-I) was observed in peripheral blood monocytes along with a significant reduction of interferon (IFN)-g, interleukin (IL)-12, and IFN-a secretion [23]. Modification of selected nucleotides is a widespread phenomenon in naturally occurring RNA. It has thus been proposed that the use of modified RNA nucleotides aids the innate immune system to discriminate between bacterial and mammalian RNA [24]. However, it cannot be excluded that additional mechanisms contribute to the increase in protein translation induced by these nucleotide analogs because they have been shown to stabilize RNA structures and alter codon specificity [25]. Yet, the lack of immune activation may counterindicate the use of these modified RNA nucleotides for vaccination purposes due to a reduced activation of DCs into efficient APCs required for naive T cell priming. Stimulating the helper T cell branch of adaptive immunity Efficient induction of immune responses by mRNA vaccination is dependent on endosomal mRNA escape and subsequent cytosolic expression of the antigen-encoding

Review mRNA. Consequently, the encoded antigens are presented mainly by MHC-I molecules resulting in T cell responses that are largely confined to the CD8+ T cell compartment [26]. Increasing evidence suggests that CD4+ T cells play a central role in the initiation of cellular and humoral immunity, the generation of effective antitumor responses and in the establishment of memory responses [27]. Thus, an increase in CD4+ responses could further potentiate mRNA-induced immune responses. We showed that administration of a local protein boost to mice primed with antigen-encoding mRNA could increase the level of effector CD4+ T cell responses up to fortyfold [26]. This effect is probably attributable to the MHC-II mediated route of antigen presentation of soluble antigens. In an alternative approach utilized by Bonehill and colleagues, mRNA encoding a tumor antigen was coupled to mRNA encoding the MHC-II targeting sequence of an endosomal or lysosomal protein [28] (Figure 1). Electroporation of human monocyte-derived DCs with mRNA encoding MAGE-A3 without added sorting signals was capable of stimulating antigen-specific CD8+ T cells but not CD4+ T cells. Coupling of the MAGE-A3 mRNA to the targeting signal of the type II transmembrane protein invariant chain (Ii), the type I transmembrane protein lysosome-associated membrane protein-1 (LAMP1), or the mature DC-specific DCLAMP efficiently stimulated MAGE-A3-specific CD4+ T cells and increased the CD8+ T cell stimulatory capacity. The LAMP1- and DC-LAMP-derived sorting signals were superior over the Ii targeting signal in their antigen-specific CD4+ T cell stimulatory capacity. Increasing mRNA immunogenicity: in search for the perfect adjuvant A common premise in vaccine research stipulates that activation of the innate immune system is crucial for the subsequent induction of adaptive immune responses. After uptake by APCs, the antigen is processed and loaded onto MHC molecules thus allowing recognition by the T cell receptor and providing the so-called signal 1. However, full activation of these epitope-specific T cells requires a second signal that is provided when co-stimulatory molecules on the APC (most notably CD80 and CD86) bind to their cognate receptor on the T cell. A third signal consisting of secreted (inflammatory) cytokines instructs T cell polarization (Figure 2). Adjuvants may increase antigen presentation (signal 1) by targeting the antigen specifically towards APCs, which constitutes a typical feature of particulate antigen delivery vehicles. In addition, adjuvants may promote signals 2 and 3 by triggering various PRRs that in turn upregulate co-stimulatory ligands on the APC, and/or by stimulating the release of cytokines. Innate immune sensing of mRNA vaccines: the almighty PRRs As RNA molecules are often crucial elements of invading pathogens, the innate immune system has evolved to recognize foreign RNA nucleotides via specific PRRs (Figure 3). These PRRs include members of the TLR family localized in the endolysosomal compartments, and cytosolic receptors capable of detecting nucleic acids in the


Signal 1 TCR

MHC

T cell

APC B7 CD28 Signal 2

Cytokines Signal 3 TRENDS in Molecular Medicine

Figure 2. Three signals required for T cell differentiation. Differentiation of naı¨ve T cells into effector T cells by APCs requires the presence of three signals. Signal 1 is delivered through the T cell receptor when it engages an appropriate peptide–MHC complex. Interaction between the co-stimulatory ligands (CD80/CD86) on the APC and their corresponding receptor on the T cell provides signal 2. Signal 3 entails the release of proinflammatory cytokines responsible for T cell polarization. Abbreviations: APC, antigen presenting cell; TCR, T cell receptor; MHC, major histocompatibility complex.

cytoplasm. In the endosomal compartment of human cells, single-stranded RNA (ssRNA) molecules can be recognized by TLR7 and TLR8 receptors with distinct sequence specificity (Box 1). Although both receptors require a uridine-rich tetramer as a minimum requisite for activation, TLR7-mediated type I IFN production in plasmacytoid DCs is preferentially induced by GU-rich sequences, whereas AU-rich sequences stimulate TLR8-mediated tumor necrosis factor (TNF) responses in monocytes [29]. In addition to activating TLR7/8, ssRNA viruses and in vitro transcribed ssRNA can activate TLR3 through double-stranded (ds) replication intermediates or through the formation of ds secondary structures, respectively [30,31]. The TLR3 receptor does not distinguish between base pair sequences, but requires a minimum length of 45 bp for binding [32]. RNA molecules present in the cytosol can be recognized by two distinct families of cytosolic sensors: the RIG-I-like receptor (RLR) family or the nuclear oligomerization domain (NOD)-like receptor (NLR) family. Within the RLR family, RIG-I recognizes viral RNA or in vitro transcribed mRNA through the presence of small regions of base pairing (10–18 bp) in ssRNA molecules [33,34]. Activation of MDA-5 by ssRNA molecules has not been reported. Because MDA-5 activation requires a much longer dsRNA structure than RIG-I (at least 2 kb), it is unlikely that the small base pairing stretches present on in vitro transcribed mRNA are long enough to initiate MDA-5 signaling. Recently, an uncapped GU-rich ssRNA sequence was reported to activate the NLR NOD2 [35]. It remains to be established whether non-GU rich mRNA can also induce NOD2-mediated signaling. The PRR-dependent immune-activating property of mRNA vaccines was confirmed in studies using non-viral carriers of antigen-encoding mRNA [15,26]. Thus, by triggering the above-mentioned PRRs, mRNA vaccines not only provide antigenic signal 1 but also supply signals 2 and 3 required for activation of naı¨ve T cells. 709

Review


Naked or parculated mRNA vaccine

Endosome ssRNA

dsRNA

TLR7

TLR3

MyD88

TRIF

dsRNA

RIG-I

NOD-2

IPS-1

Cytoplasm

IKKα

TBK-1, IKKε

NF-κB

IRF7

Proinflammatory cytokines

Nucleus

IRF3

Type I IFN (IFN-α, IFN-β)

(T cell smulaon Figure 2)

(anviral responses Figure 4)

TRENDS in Molecular Medicine

Figure 3. Signaling pathways activated by RNA-binding to pattern recognition receptors. The innate immune system has evolved to recognize foreign RNA nucleotides via specific pattern recognition receptors. In the endosomal compartment, ssRNA molecules are recognized by TLR7, whereas dsRNA binds to TLR3. TLR7 signaling is mediated by the adaptor protein MyD88, resulting in the activation of IKKa and NF-kB, which then stimulates transcription of proinflammatory genes. IKKa also activates IRF7, which leads to expression of type I IFN. Activation of TLR3 by dsRNA involves TRIF, which in turn activates IRF3 through TBK-1 and IKKe, resulting in IFN gene transcription through activation of IRF3 and IRF7. In the cytoplasm, dsRNA binds to NOD-2 and RIG-I, resulting in activation of IKKa and TBK-1 and IKKe, respectively. Abbreviations: TLR, Toll-like receptor; MyD88, myeloid differentiation primary response 88; IKKa, IkB kinase a; NF-kB, nuclear factor-kB; IRF7, interferon (IFN) regulatory factor 7; TRIF, Toll/IL-1 receptor (TIR) domain-containing adaptor inducing IFN-b; IRF3, interferon (IFN) regulatory factor 3; TBK-1, TANK-binding kinase 1; IKKe, IkB kinase e; NOD-2, nucleotide-binding oligomerization domain-containing protein 2; RIG-I, retinoic acid-inducible gene-I; ssRNA, single-stranded RNA; dsRNA, double-stranded RNA.

The type I IFN antiviral response: good or bad? In a study performed by Kim et al. it was demonstrated that the recognition of in vitro transcribed mRNA by PRRs activates the innate immune system, resulting in secretion of type I IFN [36]. Similarly, we showed that immunization of mice with a DOTAP/DOPE complexed mRNA vaccine induces detectable levels of type I IFN in serum collected 1–24 h after immunization [26]. In addition to their role in the induction of innate immune responses, type I IFNs stimulate the generation of adaptive immunity. In this regard, it has been shown that type I IFNs are essential for the induction of cellular immune responses when using the ssRNA agonist polyUs21 or the dsRNA agonist polyI:C to adjuvant protein vaccines [37,38]. Surprisingly, we found that when using mRNA to encode vaccine antigens, type I IFN negatively interfered with the strength of the evoked T cell response, with antigen-specific T cell responses being 710

dramatically increased in mice lacking the type I IFN receptor (IFNaR/) [26]. Subsequent studies clearly showed increased expression levels of mRNA-encoded protein in IFNaR/ DCs compared with wild type DCs, thus pointing to a negative role of type I IFNs in the expression of the mRNA-encoded antigen. Type I IFNs play a key role in antimicrobial host defense by initiating innate immune responses against viral RNA (Figure 4). This antiviral resistance may lead to: (i) phosphorylation of the a subunit of the translation initiator factor 2 (eIF2a), resulting in downregulation of mRNA translation; (ii) induction of the 20 50 oligoadenylate synthetase (20 50 OAS), thus promoting RNA degradation through RNAse L activity; and (iii) triggering of adenosine deaminase acting on RNA (ADAR-1) and of the apolipoprotein B mRNA-editing enzyme catalytic polypeptide-like editing complex-3 (APOBEC-3), which deaminates RNA viral

Review


Box 1. Differential expression of mRNA-binding PRRs in DC subsets TLR3 and TLR7/8 are differentially expressed on different DC subsets in a species-specific manner. In human cells, TLR7 is exclusively expressed in plasmacytoid DCs, whereas TLR8 is more broadly expressed in monocytes, macrophages, and conventional DCs [46]. Murine TLR7 is expressed on plasmacytoid DCs and CD4a+ conventional DCs, but not on CD8a+ DCs [47]. TLR3 expression is not present on plasmacytoid DCs [46,48]. Thus, mRNA vaccines can activate several TLRs, and which TLR is stimulated may depend on the specific DC subset that is targeted. This issue may be of high relevance when using a particulate delivery system to protect the vaccine mRNA from degradation. Depending on the size and the physicochemical properties of the particle, the antigen can be targeted to different APCs [49]. In a study using human peripheral blood mononuclear cells (PBMCs), it was shown that plasmacytoid DCs preferentially take up nanometric protamine RNA particles resulting in production of IFN-a, whereas microparticles are phagocytosed far less efficiently. By contrast, phagocytosis by CD11c+ monocytes was not size selective [50]. As opposed to TLR expression, expression of the cytosolic sensors of the RIG-I family is not restricted to certain cell types. It is assumed that all cells can detect cytosolic pathogens in an autonomous manner.

replication intermediates [39]. In turn, these processes negatively influence cell viability and can eventually result in apoptosis. When using mRNA as an antigen expression tool, these antiviral defense mechanisms might interfere with the efficient expression of mRNA vaccines. We hypothesized that in antigen expression systems, the negative side of type I IFN on the stability and translation of mRNA prevails over its immunostimulatory properties, resulting in an overall negative effect on mRNA-based vaccines. However, additional studies on the role of type I IFN in mRNA vaccine-induced T cell functionality are required to fully understand this love/hate relationship. In any case, it is clear that mRNA vaccines should be designed in a way that innate immune responses resulting in inhibition of antigen expression are reduced to a minimum. Adjuvanting mRNA vaccines The search for adjuvants that increase mRNA vaccine efficacy has been a daunting process. Although the use of TLR agonists has become a popular method to improve protein or DNA vaccination, combining antigen-encoding mRNA with the TLR3 ligand polyI:C or the TLR4 ligand lipopolysaccharide (LPS) has been shown to have a detrimental effect on protein expression [12,40]. The simultaneous delivery of protein-encoding mRNA and DC maturation stimuli has been hypothesized to hamper RNA internalization by abrogating macropinocytosis [40]. However, as polyI:C and LPS are strong inducers of type I IFN, it is equally possible that the inhibitory effect of these TLR ligands on protein expression is mediated by type I IFN-dependent antiviral defense mechanisms suppressing vaccine mRNA translation as described in the previous section. In addition, the specific response of DCs to inflammatory stimuli promoting DC maturation may also contribute to the negative outcome of combining antigen-encoding mRNA with TLR3 and TLR4 ligands. In response to maturation stimuli, protein synthesis of capped mRNA species in DCs is rapidly enhanced via a phosphoinositide 3-kinase (PI3K)-dependent pathway.

IFN α/β

↑ PKR

↑ 2′5′ OAS

eIF2α

RNaseL

Inhibion of translaon

Apoptosis

↑

ADAR APOBEC3

RNA deaminaon

RNA cleavage TRENDS in Molecular Medicine

Figure 4. Effects of type I IFN on mRNA stability and translation. Binding of foreign RNA molecules by pattern recognition receptors induces type I IFNs. In turn, type I IFN upregulates the expression of PKR, 20 50 OAS, ADAR, and APOBEC3. PKR phosphorylates the a subunit of eIF2a, resulting in downregulation of mRNA translation. 20 50 OAS activates RNase L, leading to degradation of RNA. eIF2a and RNase L can additionally signal for the induction of apoptosis. ADAR and APOBEC3 deaminate RNA molecules, which eventually results in RNA mutations. Cleaved RNA molecules can again bind to PRRs, thus amplifying the type I IFN loop. Abbreviations: IFN, interferon; PKR, protein kinase R; OAS, oligoadenylate synthetase; ADAR, adenosine deaminase acting on RNA; APOBEC3, apolipoprotein B mRNA-editing enzyme, catalytic polypeptide 3; eIF2a, eukaryotic translation initiation factor; RNase L, ribonuclease L.

However, this initial peak in cap-mediated translation is observed only during the first hours of maturation and is immediately followed by a marked reduction in protein synthesis. The inhibition of cap-dependent translation is correlated with eIF2a phosphorylation and an increased production and degradation of eIF4GI and the eIF4GI-like factor DAP5 [41]. These counteracting mechanisms may in fact be inherent to differences in the specific cellular processes underlying antigen presentation of mRNA-based vaccines as opposed to protein-based vaccination strategies. Whereas for protein vaccines the antigen is readily available to enter the cellular processing pathways, for mRNA vaccines the translation machinery inside the APC must first synthesize relevant epitopes from the mRNA template. As a consequence of this difference in kinetics, DC activation may interfere with antigen presentation when using mRNA vaccines in combination with classical adjuvants. In an attempt to synchronize these two events, mRNA encoding immunostimulatory molecules have been co-delivered with antigen-encoding mRNA. Van Lint et al. used a mix of mRNA encoding CD40 ligand, constitutive active TLR4 and CD70 (Trimix) to adjuvant intranodal delivery of tumor-associated antigen (TAA) mRNA. Addition of the Trimix resulted in increased CTL responses and enhanced antitumor responses as compared with the non-adjuvanted mRNA vaccine [12]. However, co-delivery of the mRNA vaccine with Trimix still resulted in decreased protein expression, indicating that downstream counteracting mechanisms such as type I IFN production are activated. Concluding remarks Over the past decades it has become clear that the development of DNA- or viral vector-based vaccines for human 711

Review


Box 2. Outstanding questions Endosomal escape of mRNA vaccines. Following endocytosis by the target APC, the mRNA must escape the acidic endolysosomal compartment to enable antigen translation in the cytosol. However, the exact mechanism responsible for this so-called endosomal escape remains to be defined. Superior efficacy of intranodal delivery of naked mRNA vaccines over other delivery routes. Most studies using subcutaneous, intravenous, or intradermal administration of naked mRNA vaccines fail to induce antigen-specific cellular immunity. By contrast, intranodally delivered mRNA evokes potent T cell responses resulting in antitumor immunity. The immunological basis for the superior efficacy of intranodally delivered mRNA over other administration routes is currently unknown. Uptake and expression of particulate mRNA vaccines. Although it has been shown that intranodally administered mRNA vaccines are taken up by lymph node resident DCs, it is currently unclear which cells are targeted by mRNA complexed with non-viral carriers. This may be dependent on the physicochemical properties of the carrier system. The formation of small mRNA/ carrier complexes could result in passive drainage to the lymph nodes, whereas larger carriers will be dependent on active uptake by local APCs and subsequent migration to the lymph nodes.

use is hindered by serious safety concerns. As a result, vaccine research has focused on alternative strategies to potently induce antigen-specific CTL responses. Although several limitations still prevent a more generalized use of mRNA for immunization purposes, intensive research efforts have resulted in stabilized mRNA constructs, optimized delivery systems, and increased immunogenicity. Preclinical studies have already demonstrated the potency of mRNA-based vaccines to induce both prophylactic and therapeutic antitumor immunity. Moreover, a recent study showed for the first time that mRNA vaccines can induce protection against viral infection in mice and pigs [42]. However, the limited clinical outcome of therapeutic antitumor mRNA vaccines in cancer patients indicates that mRNA vaccine technology is still suboptimal [43–45]. Circumventing innate immune responses repressing mRNA translation, while at the same time maintaining mRNAinduced immune activation properties, remains a major hurdle in the development of mRNA vaccines. Therefore, a better understanding of immunological mechanisms triggered by mRNA vaccines is required in order to further rationalize the design of vaccine and treatment protocols, ultimately resulting in improved clinical efficacy of mRNAbased vaccination approaches (Box 2). Acknowledgments C.P. was supported by a PhD scholarship from SOFI-B, granted by the Ministry of Economy, Science, and Innovation (EWI) to the Institute of Tropical Medicine, Antwerp, and by a grant from the Fund for Scientific Research Flanders (FWO G 0266.10N). S.D.K. acknowledges the Research Foundation Flanders for a postdoctoral fellowship. X.S. acknowledges B0F12/GOA/014 for financial support.

References 1 Palucka, K. and Banchereau, J. (2012) Cancer immunotherapy via dendritic cells. Nat. Rev. Cancer 12, 265–277 2 Mellman, I. et al. (2011) Cancer immunotherapy comes of age. Nature 480, 480–489 712

Inhibitive effect of TLR ligands on mRNA vaccine expression. Adjuvanting mRNA vaccines with the TLR3 ligand polyI:C or the TLR4 ligand LPS has been shown to negatively influence mRNA expression. The exact mechanism for this inhibitive effect is currently unknown. It has been hypothesized that DC maturation induced by TLR ligands hampers RNA internalization by abrogating macropinocytosis. Another possible explanation is the induction of type I IFN-mediated antiviral responses in the host cell. Source of type I IFN secretion. Circumventing the induction of antiviral responses responsible for the inhibition of antigen expression remains a major hurdle in the development of improved mRNA vaccines. Identification of the source of mRNA vaccine-induced type I IFN and the target cells at which these cytokines exert their effects will help understanding the mechanism behind their inhibitive effects, and therefore stimulate the design of improved mRNA vaccines. Limited clinical efficacy of mRNA vaccines. Although preclinical studies demonstrated the potency of mRNA vaccines in inducing both prophylactic and therapeutic antitumor immunity, studies in cancer patients showed limited clinical efficacy of therapeutic antitumor mRNA vaccines. The reason for the discrepancy between preclinical and clinical efficacy of mRNA vaccines remains to be elucidated and could provide new insights that would ultimately result in the design of mRNA vaccines with improved clinical efficacy.

3 Pascolo, S. (2008) Vaccination with messenger RNA (mRNA). Handb. Exp. Pharmacol. 183, 221–235 4 Boudreau, J.E. et al. (2011) Engineering dendritic cells to enhance cancer immunotherapy. Mol. Ther. 19, 841–853 5 Vanham, G. and Van Gulck, E. (2012) Can immunotherapy be useful as a ‘‘functional cure’’ for infection with Human Immunodeficiency Virus1? Retrovirology 9, 72 6 Martinon, F. et al. (1993) Induction of virus-specific cytotoxic T lymphocytes in vivo by liposome-entrapped mRNA. Eur. J. Immunol. 23, 1719–1722 7 Hoerr, I. et al. (2000) In vivo application of RNA leads to induction of specific cytotoxic T lymphocytes and antibodies. Eur. J. Immunol. 30, 1–7 8 Scheel, B. et al. (2006) Therapeutic anti-tumor immunity triggered by injections of immunostimulating single-stranded RNA. Eur. J. Immunol. 36, 2807–2816 9 Lu, D. et al. (1994) Optimization of methods to achieve mRNAmediated transfection of tumor cells in vitro and in vivo employing cationic liposome vectors. Cancer Gene Ther. 1, 245–252 10 Kreiter, S. et al. (2010) Intranodal vaccination with naked antigenencoding RNA elicits potent prophylactic and therapeutic antitumoral immunity. Cancer Res. 70, 9031–9040 11 Kreiter, S. et al. (2011) FLT3 ligand enhances the cancer therapeutic potency of naked RNA vaccines. Cancer Res. 19, 6132–6142 12 Van Lint, S. et al. (2012) Preclinical evaluation of TriMix and antigen mRNA-based antitumor therapy. Cancer Res. 7, 1661–1671 13 Hess, P.R. et al. (2006) Vaccination with mRNAs encoding tumorassociated antigens and granulocyte–macrophage colony-stimulating factor efficiently primes CTL responses, but is insufficient to overcome tolerance to a model tumor/self antigen. Cancer Immunol. Immunother. 55, 672–683 14 Mockey, M. et al. (2007) mRNA-based cancer vaccine: prevention of B16 melanoma progression and metastasis by systemic injection of MART1 mRNA histidylated lipopolyplexes. Cancer Gene Ther. 14, 802–814 15 Fotin-Mleczek, M. et al. (2011) Messenger RNA-based vaccines with dual activity induce balanced TLR-7 dependent adaptive immune responses and provide antitumor activity. J. Immunother. 34, 1–15 16 Rejman, J. et al. (2010) mRNA transfection of cervical carcinoma and mesenchymal stem cells mediated by cationic carriers. J. Control. Release 147, 385–391 17 Varkouhi, A.K. et al. (2011) Endosomal escape pathways for delivery of biologicals. J. Control. Release 151, 220–228

Review 18 Sonenberg, N. and Hinnebusch, A.G. (2009) Regulation of translation initiation in eukaryotes: mechanisms and biological targets. Cell 136, 731–745 19 Pasquinelli, A.E. et al. (1995) Reverse 50 caps in RNAs made in vitro by phage RNA polymerases. RNA 1, 957–967 20 Stepinski, J. et al. (2001) Synthesis and properties of mRNAs containing the novel ‘‘anti-reverse’’ cap analogs 7-methyl(30 -Omethyl)GpppG and 7-methyl (30 -deoxy)GpppG. RNA 7, 1486–1495 21 Kuhn, A.N. et al. (2010) Phosphorothioate cap analogs increase stability and translational efficiency of RNA vaccines in immature dendritic cells and induce superior immune responses in vivo. Gene Ther. 17, 961–971 22 Holtkamp, S. et al. (2006) Modification of antigen-encoding RNA increases stability, translational efficacy, and T-cell stimulatory capacity of dendritic cells. Blood 108, 4009–4017 23 Kormann, M.S.D. et al. (2011) Expression of therapeutic proteins after delivery of chemically modified mRNA in mice. Nat. Biotechnol. 29, 154–157 24 Kariko´, K. et al. (2005) Suppression of RNA recognition by Toll-like receptors: the impact of nucleoside modification and the evolutionary origin of RNA. Immunity 23, 165–175 25 Ge, J. and Yu, Y.-T. (2013) RNA pseudouridylation: new insights into an old modification. Trends Biochem. Sci. 4, 210–218 26 Pollard, C. et al. (2013) Type I IFN counteracts the induction of antigenspecific immune responses by lipid-based delivery of mRNA vaccines. Mol. Ther. 1, 251–259 27 Swain, S.L. et al. (2012) Expanding roles for CD4+ T cells in immunity to viruses. Nat. Rev. Immunol. 12, 136–148 28 Bonehill, A. et al. (2004) Messenger RNA-electroporated dendritic cells presenting MAGE-A3 simultaneously in HLA class I and class II molecules. J. Immunol. 172, 6649–6657 29 Forsbach, A. et al. (2012) Sequence-specific TLR8-dependent immune response. J. Immunol. 180, 3729–3738 30 Kariko´, K. et al. (2004) mRNA is an endogenous ligand for Toll-like receptor 3. J. Biol. Chem. 279, 12542–12550 31 Ceppi, M. et al. (2005) Double-stranded secondary structures on mRNA induce type I interferon (IFN a/b) production and maturation of mRNA-transfected monocyte-derived dendritic cells. J. Gene Med. 7, 452–465 32 Botos, I. et al. (2009) The toll-like receptor 3:dsRNA signaling complex. Biochim. Biophys. Acta 1789, 667–674 33 Schlee, M. et al. (2009) Recognition of 50 triphosphate by RIG-I helicase requires short blunt double-stranded RNA as contained in panhandle of negative-strand virus. Immunity 31, 25–34


34 Schmidt, A. et al. (2009) 50 -Triphosphate RNA requires base-paired structures to activate antiviral signaling via RIG-I. Proc. Natl. Acad. Sci. U.S.A. 106, 12067–12072 35 Sabbah, A. et al. (2009) Activation of innate immune antiviral response by NOD2. Nat. Immunol. 10, 1073–1080 36 Kim, D-H. et al. (2004) Interferon induction by siRNAs and ssRNAs synthesized by phage polymerase. Nat. Biotechnol. 22, 321–325 37 Rajagopal, D. et al. (2010) Plasmacytoid dendritic cell-derived type I interferon is crucial for the adjuvant activity of Toll-like receptor 7 agonists. Blood 115, 1949–1957 38 Longhi, M.P. et al. (2009) Dendritic cells require a systemic type I interferon response to mature and induce CD4+ Th1 immunity with poly IC as adjuvant. J. Exp. Med. 206, 1589–1602 39 Pichlmair, A. and Reis e Sousa, C. (2007) Innate recognition of viruses. Immunity 27, 370–383 40 Diken, M. et al. (2011) Selective uptake of naked vaccine RNA by dendritic cells is driven by macropinocytosis and abrogated upon DC maturation. Gene Ther. 18, 702–708 41 Lelouard, H. et al. (2007) Regulation of translation is required for dendritic cell function and survival during activation. J. Cell Biol. 179, 1427–1439 42 Petsch, B. et al. (2012) Protective efficacy of in vitro synthesized, specific mRNA vaccines against influenza A virus infection. Nat. Biotechnol. 30, 1210–1216 43 Weide, B. et al. (2008) Results of the first phase I/II clinical vaccination trial with direct injection of mRNA. J. Immunother. 31, 180–188 44 Weide, B. et al. (2009) Direct injection of protamine-protected mRNA: results of a phase 1/2 vaccination trial in metastatic melanoma patients. J. Immunother. 32, 498–507 45 Rittig, S.M. et al. (2011) Intradermal vaccinations with RNA coding for TAA generate CD8+ and CD4+ immune responses and induce clinical benefit in vaccinated patients. Mol. Ther. 19, 990–999 46 Kadowaki, N. et al. (2001) Subsets of human dendritic cell precursors express different toll-like receptors and respond to different microbial antigens. J. Exp. Med. 194, 863–869 47 Iwasaki, A. and Medzhitov, R. (2004) Toll-like receptor control of the adaptive immune responses. Nat. Immunol. 5, 987–995 48 Lande, R. and Gilliet, M. (2010) Plasmacytoid dendritic cells: key players in the initiation and regulation of immune responses. Ann. N. Y. Acad. Sci. 1183, 89–103 49 De Koker, S. et al. (2011) Designing polymeric particles for antigen delivery. Chem. Soc. Rev. 40, 320–339 50 Rettig, L. et al. (2010) Particle size and activation threshold: a new dimension of danger signaling. Blood 115, 4533–4541

713

Review

Microbiota impact on the epigenetic regulation of colorectal cancer Tao Yang1,2, Jennifer L. Owen1,2, Yaı´ma L. Lightfoot1,2, Michael P. Kladde3, and Mansour Mohamadzadeh1,2 1

Department of Infectious Diseases and Pathology, University of Florida, Gainesville, FL 32608 USA Division of Gastroenterology, Hepatology, and Nutrition, Department of Medicine, University of Florida, Gainesville, FL 32610 USA 3 Department of Biochemistry and Molecular Biology, University of Florida, Gainesville, FL 32610 USA 2

Mechanisms of colorectal cancer (CRC) development can be generally divided into three categories: genetic, epigenetic, and aberrant immunologic signaling pathways, all of which may be triggered by an imbalanced intestinal microbiota. Aberrant gut microbial composition, termed ‘dysbiosis’, has been reported in inflammatory bowel disease patients who are at increased risk for CRC development. Recent studies indicate that it is feasible to rescue experimental models of colonic cancer by oral treatment with genetically engineered beneficial bacteria and/or their immune-regulating gene products. Here, we review the mechanisms of epigenetic modulation implicated in the development and progression of CRC, which may be the result of dysbiosis, and therefore may be amenable to therapeutic intervention. A role for the microbiota in CRC development CRC is one of the leading contributors to cancer death and morbidity in the USA and is the fourth most commonly diagnosed cancer in the world, with more than 1 million new cases diagnosed annually [1,2]. CRC begins with the formation of polyps, masses of cells protruding from the bowel wall, which subsequently progress to dysplastic adenomas, and ultimately to colonic carcinoma [3]. The mechanisms of CRC pathogenesis have been widely investigated and current evidence indicates that genetic mutations, epigenetic changes, and aberrant immunologic signaling pathways are major contributors to CRC [4]. Mutations in tumor suppressor genes or oncogenes of colonic epithelial cells lead to dysregulated signaling networks that control cellular metabolism, differentiation, proliferation, and survival, resulting in the transformation of normal cells into cancer cells [5]. In addition to genetic mutations, epigenetic modulation of tumor suppressor genes, oncogenes, or proinflammatory mediators represent other important mechanisms whereby homeostatic Corresponding author: Mohamadzadeh, M. ([email protected]). Keywords: colorectal cancer; commensal bacteria; epigenetic regulation; inflammatory bowel disease; microbiota. 1471-4914/$ – see front matter ß 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.molmed.2013.08.005

714

Trends in Molecular Medicine, December 2013, Vol. 19, No. 12

balance is lost. Host chromatin can be altered epigenetically by covalent modification of histones (e.g., acetylation, methylation, phosphorylation, and ubiquitylation) or of DNA (e.g., cytosine methylation); additionally, chromatin can be regulated by nucleosome positioning that governs accessibility of trans-acting factors to DNA and the presence of non-coding RNAs [6]. Various epigenetic profile analyses comparing healthy and neoplastic cells indicate that epigenetic regulation plays a key role in tumor initiation and progression in CRC [5]. Cancer development and progression is a complex process involving a myriad of host–tumor interactions and countless molecular and cellular elements in the tumor microenvironment [7]. Recently, data have established a link between alterations in the commensal bacteria of the gut, termed the ‘microbiota’, and the pathogeneses of inflammatory bowel disease (IBD) and colitis-associated colorectal carcinoma (CAC) [8]. This is not surprising, as the gut microbiota is a pivotal component of the gastrointestinal ecosystem, wherein the microbes maintain complex interactions with host mucosal epithelial and immune cells. Microbial surface antigens and bacterial metabolic endproducts can trigger downstream signaling pathways resulting in the activation of immune cells and the production of cytokines. Thus, a balance between immunostimulation against potential pathogens and immunosuppression against gut commensal microbes and dietary antigens is required for intestinal homeostasis [9]. This homeostasis is lost in IBD, a chronic autoinflammatory syndrome manifesting as Crohn’s disease (CD) or ulcerative colitis (UC). CD is characterized by discontinuous, transmural lesions in the intestinal wall [10], whereas UC presents with diffuse, continuous, superficial inflammation in the colon [11]. Gradually, this chronic inflammation in the gut may provide a suitable microenvironment for aberrant molecular events [12]. For example, patients with IBD, especially UC, have an increased risk for developing CRC, and this risk is directly proportional to the extent of the colonic inflammation [9,13]. Furthermore, a pathological role for gut luminal bacteria in mucosal inflammation has been demonstrated in interleukin (IL)-2 and IL-10 gene knockout models of colitis [14]; and mice that are genetically susceptible to developing CRC have fewer tumors when living in germ-free conditions than when housed conventionally [2]. Thus, the importance of the

Review intestinal microbiota during CRC development and progression is gaining significant attention. It is thought that, in most patients, the progression of normal colonic mucosa to CRC requires a stepwise accumulation of genetic and epigenetic alterations [15]. The approximate amount of time for the malignant transformation of normal mucosa into adenomatous polyps and, finally, into invasive carcinoma is thought to be 5–15 years [16,17]. This window of time provides an opportunity for early detection and therapeutic intervention [15]. Here, we review evidence for the role of epigenetic regulation in the development and progression of CRC and consider how an altered microbiota plays a role in these processes and may even represent a platform for intervention. Intestinal microbiota are necessary for a healthy immune system As most pathogens gain entry by breaching a mucosal surface, the mucosal epithelia and resident immune cells of the gastrointestinal, respiratory, and urogenital tracts play critical roles in the protection of the host against infection [18]. Within the healthy intestine, the normal commensal microbiota and the cells of the intestinal mucosa coexist in a mutualistic relationship, whereby the commensal microorganisms help to shape the local and systemic immune response, and the intestinal mucosal cells modulate the habitat and composition of the microbiota [9]. In particular, the colon is the most heavily colonized tissue in the body, containing an estimated 1014 bacteria and over 70% of all the microorganisms that reside in the host [19]. Obligate anaerobic bacteria belonging to the two phyla Bacteroidetes and Firmicutes dominate the human digestive tract where more than 50 different phyla reside [20]. Other phyla, including Proteobacteria, Verrucomicrobia, and Actinobacteria, are present in the gut to a much lesser extent. An imbalance in this composition of commensal microbes, or ‘dysbiosis’, is characterized by a considerable decrease in the resident obligate anaerobic bacteria, and an increase in facultative anaerobes such as Enterobacteriaceae [21]. For example, adherent-invasive Escherichia coli (AIEC), which are Enterobacteriaceae, are isolated more frequently from the intestinal mucosa of patients with CD than from healthy individuals [22]. Interestingly, the endoplasmic reticulum-localized stress response chaperone, Gp96, is abnormally expressed on the apical surface of epithelial cells in the ileum of patients with CD, and acts as a receptor for the OmpA of AIEC to promote pathogenic invasion of this bacterium [23]. It is important to note that the composition of the microbiome of the murine gut is very similar to that of humans [24]; therefore, mouse experimental models of gastrointestinal disease are likely to have translational relevance [25]. The gut microbiota of both species can further be divided according to their location in the gut, namely, whether the microbes are present in the intestinal lumen or penetrate the mucus layer overlying the intestinal epithelium [19]. A thick mucin layer protects mucosal enterocytes from excessive exposure to bacteria and dietary antigens throughout the length of the intestines, particularly in the colon, thus preventing immune hypersensitivity responses [26].


Microbial colonization of the human gut is thought to begin at birth via commensal microbes from the maternal skin, vagina, and feces, as well as the external environment of the neonate [27,28]. Emerging data also suggest that the environment in utero is not sterile, as Enterococcus faecalis, Staphylococcus epidermidis, and E. coli have been isolated from the meconium of healthy neonates [29]. The establishment of a diverse and balanced microbiota plays a critical role in the development and maturation of a healthy immune system. This has been demonstrated using germ-free animals raised in bacteria-free conditions that exhibit physical and functional immune abnormalities, including defects in the formation of the spleen, lymph nodes, and Peyer’s patches [18]. Additionally, there are significant reductions in the number of IgA-producing plasma cells and CD4+ T cells, as well as decreased immunoglobulins and imbalanced cytokine levels [18]. Germfree mice also have few, if any, proinflammatory T helper (Th)17 cells in the intestinal lamina propria [30]. Thus, exposure to commensal bacteria greatly affects gut health and immunity. This microbial exposure is tightly regulated, as limited or no bacterial interaction significantly impairs development of the host immune system, whereas excessive bacterial contact can result in exaggerated immune responses and autoinflammatory disease [26]. The critical role of the microbiota in overall gut health and the extent of the deleterious consequences of dysbiosis address the importance of understanding the molecular interactions between the microbiota and the cells of the host. Control of intestinal inflammation and tolerance by the gut microbiota Although a basal level of inflammation is needed to recognize and eliminate invading enteropathogens, the gastrointestinal tract must also coexist with the diverse and abundant commensal bacteria and the daily estimated 130–190 g of ingested protein replete with dietary antigens [31]. To maintain intestinal homeostasis both locally and systemically, tolerance must be achieved by the induction of anti-inflammatory molecules. Commensal microbes are critical regulators of the immune system and can maintain homeostasis by stimulating antibody production and activating immune cells, as well as inducing anti-inflammatory cytokine responses (Figure 1A). The microbial pattern sensors, Toll-like receptors (TLRs), play particularly important roles in microbe recognition, homeostasis, and immune defense in the gut [32]. In the healthy gut, TLR3 and TLR5 seem to be constitutively expressed. By contrast, TLR2 and TLR4 are expressed at very low levels, suggesting that the expression of these two receptors is regulated to avoid autoinflammatory immune activation in response to commensal microbes [26]. However, current data reveal that microbiota-mediated TLR signal transduction is complex. For example, constitutive TLR4 signaling in intestinal epithelium reduced tumor burden by increasing apoptosis in ApcMin/+ mice, an experimental model of CRC in which Min (multiple intestinal neoplasia) is a mutant allele of the mouse ortholog of the tumor suppressor gene, APC (adenomatous polyposis coli) [33]. Another way in which intestinal bacteria affect the host is via the end-products of their unique metabolism. For 715

Review


(A)

Lumen

Bifidobacterium Streptococcus Lactobacillus Enterococcus Clostridium

Mucosa

Lactobacillus Enterococcus Clostridium

Epithelial cells

Bacterial surface angens and metabolites

-

Angens

-

Cytokines

Lamina propria

B cells

Regulatory T cells Dendric cells Macrophages

l t d iimmune cellll acvaon -R Regulated - IgA producon - Tolerance

t ki -C Cytokines - Physiological inflammaon

(B)

SFCA

p65 IκBα p50

p65 p50

SFCA

HDAC

HAT


Figure 1. Mechanisms for maintaining basal immune activation and tolerance. (A) Commensal bacteria help to preserve a healthy gut environment. Abundant beneficial microbes exist at the mucosal layer where epithelial cells can be stimulated by bacterial antigens such as surface layer proteins and metabolic products. The release of signaling molecules or the presentation of antigens by dendritic cells is able to induce the activation of T and B lymphocytes, meanwhile recruiting more dendritic cells and macrophages. (B) Bacterial fermentation products such as SFCAs inhibit the NF-kB signaling pathway and decrease HDAC activity to regulate homeostasis. HATs, acting as transcription coactivators, can open chromosome structure by adding acetyl residues to histones, whereas HDACs do the opposite. Inhibition of HDACs may lead to accessible chromatin. The transcription of cell cycle arrest and apoptosis-related genes can result in suppression of tumor development. Inhibition of the NF-kB signaling pathway also prevents its contribution to pathogenic inflammation. Abbreviations: SlpA, surface layer protein A; SFCA, short fatty chain acid; HAT, histone acetyltransferase; HDAC, histone deacetylase.

716

Review


example, the short chain fatty acids (SCFAs), acetate, proprionate, and butyrate, are products of microbial fermentation of dietary fiber in the colon and have been reported to be effective in immune regulation [10]. These metabolites are important sources of energy for colonic epithelial cells and are reportedly decreased in patients with IBD [10]. Although acetate is the most abundant SCFA found in the colon, butyrate has been the subject of intense study because of its role in the preservation of mucosal health and its anti-inflammatory and antitumorigenic properties [34]. The most widely studied function of butyrate is as a histone deacetylase (HDAC) inhibitor that results in hyperacetylation, maintenance of an open euchromatic state, and the activation of genes involved in cellular differentiation, apoptosis, and cell cycle arrest in tumor cells [35]. Butyrate has disparate effects in normal colonocytes versus malignant cells because most of this SCFA is oxidized to produce ATP in normal cells, resulting in cellular proliferation. Conversely, butyrate is inefficiently oxidized in tumor cells, which rely on anaerobic metabolism, and is able to reach the nucleus and function as a HDAC inhibitor [35]. Apart from this epigenetic regulation, butyrate is also capable of suppressing the activation of the proinflammatory transcription factor, nuclear factorkappa B (NF-kB), which plays a central role in the host immune responses to infection. Activated NF-kB translocates into the nucleus where it binds to specific sequences of target genes and promotes the expression of a series of proinflammatory factors [36]. Activation of NF-kB is one of the principal contributing factors to the development of inflammation-associated CRC and its constitutive activation was reported in 40% of excised CRC tissues and 67%

of human CRC cell lines [34,37]. These mechanisms of butyrate action are important to limit the inflammatory response and to maintain intestinal homeostasis (Figure 1B). Intestinal pathogenesis elicited by altered microbiota composition Discussions about whether the altered composition of the microbiota is a cause or consequence of IBD and CAC are ongoing. A possibility is that alteration of either the microbiome composition or the number of inflammatory stimuli, whichever occurs in excess first, will tip the balance of homeostasis (Figure 2). As a consequence, changes in barrier function of the epithelium, activation of the innate immune system via TLRs and other microbial sensors, and polarization of T lymphocytes will result in the release of inflammatory mediators that lead to pathologic autoinflammation. A decreased diversity of species in gut microbial communities is often associated with a high density of mucosal surface colonization and epithelial invasion in areas with active disease [38]. Further investigations have demonstrated that it is not only the alteration of microbial composition but also aberrant molecular signaling induced by metabolites of altered bacteria that are involved during the transformation from acute to chronic inflammation, IBD, or even colon cancer [39]. This is not surprising, as an estimated 20–30% of cancers are associated with chronic microbial infections, including gastric cancer (Helicobacter pylori), liver cancer (hepatitis B and C viruses), urinary bladder cancer (Schistosoma haematobium), cholangiocellular neoplasia (Opisthorchis viverrini), and cervical cancer (human papilloma viruses) [40].

Gut microbiota Cytokines

Surface angens

Healthy homeostasis Metabolic products

Chemokines

Immune system Imbalanced microbiota

Pathogenic inflammaon

IBD (i.e., UC)

CRC


Figure 2. Interactions between the gut microbiota and the immune system maintain healthy homeostasis. Either alteration of the gut microbiota composition or overactivation of the immune system may disrupt the balance and, therefore, induce inflammation and even IBD or CRC. Abbreviations: IBD, inflammatory bowel disease; UC, ulcerative colitis; CRC, colorectal cancer.

717

Review The connection among infection, inflammation, and carcinogenesis is explained, in part, by oxidative and nitrative stress and the resulting promutagenic DNA lesions: DNA–protein crosslinks, depurination/depyrimidination, base and sugar modifications, and single- and double-stranded breaks, all of which have been reported to play key roles in inflammation-related carcinogenesis [3,41]. For example, nitrative- and oxidative-induced DNA damage have been identified at sites of carcinogenesis in animal models of chronic inflammation induced by infection with the liver fluke, O. viverrini [42]. The reactive oxygen, nitrogen, and halogen species caused profound damage to cellular nucleic acids, proteins, lipids, and carbohydrates, and resulted in the induction of proinflammatory molecules, including IL-1, IL-6, tumor necrosis factor (TNF)-a, cyclooxygenase-2, and inducible nitric oxide synthase (iNOS) [42]. Molecular mechanisms underlying infection-induced inflammation and colon cancer were recently described using Helicobacter hepaticus-infected Rag2/ mice [43]. This bacterium colonizes the liver and colon of various mouse strains, but does not typically cause disease in immunocompetent mice. However, it is linked with hepatitis and chronic colitis in several immunodeficient murine models; the lack of regulatory T cells (Tregs) in Rag2/ mice renders them unable to achieve immune tolerance to this microbe [44]. In this study, liquid chromatography-coupled tandem mass spectrometry (LC-MS/MS) was used to quantify DNA and RNA damage products, revealing chlorinated nucleic acid (5-chlorocytosine, 5-Cl-C) in the colonic epithelium due to neutrophil-derived HOCl [43]. Lao et al. [45] reported that the 2-deoxyribo-halogenated product, 5-CldC, mimics 5-methyl-dC and induces inappropriate, increased CpG methylation that can silence tumor suppressor genes and initiate carcinogenesis [43]. Another role for inflammatory cell-derived nitrogen species in the disruption of intestinal homeostasis is the ability of certain microbes to use host inflammation-derived nitrate as a terminal electron acceptor for anaerobic respiration. This gives the facultative anaerobe, E. coli, a distinct growth advantage in the lumen of the inflamed intestine over commensal obligate anaerobes, which cannot utilize nitrate and must rely on fermentation for energy [46]. The means by which colonization with the gut commensal, E. coli NC101, promotes invasive colorectal carcinoma in germ-free, colitis-susceptible Il-10/ mice treated with the colon-specific carcinogen, azoxymethane (AOM), was also recently investigated [47]. The promotion of tumorigenesis was related to the DNA-damaging, polyketide synthase ( pks) pathogenicity island of the bacteria; and mucosal pks+ E. coli were found in a high percentage of IBD and CRC patients. These data suggest that colitis can promote tumorigenesis by altering microbial composition and inducing the expansion of microorganisms with the ability to damage DNA [47]. Additional evidence linking microbial infection and CRC includes the capability of enterotoxigenic Bacillus fragilis (ETBF) to trigger colitis and induce colonic hyperplasia and tumor initiation in Min mice via induction of signaling transducer and activator of transcription (STAT)-3 activation and a proinflammatory 718


Th17 response [48]. Conversely, although non-toxigenic B. fragilis (NTBF) also chronically colonized mice, it did not induce STAT-3 activation or colitis; importantly, these data reveal distinct roles for a commensal bacterium and adaptive immunity during CRC pathogenesis [48]. It would stand to reason that in a mutualistic relationship, the gut microbiome could impact genetic and epigenetic structures of the host. A supportive example of the role of diet and gut commensals in host mucosal epigenetic and microbiomic changes came from Richard Kellermayer’s laboratory, where they demonstrated that a maternal diet supplemented with methyl donors (e.g., folic acid, vitamin B12, and choline) resulted in offspring with increased susceptibility to dextran sodium sulfate (DSS)induced colitis due, in part, to aberrant methylation of antiinflammatory genes [49]. Genetic regulation of colorectal carcinogenesis For decades, genetic instability has been considered to underlie the multistep process of tumor growth and metastatic dissemination [50]. An array of genetic mutations is found in most, if not all, cancer subtypes (Figure 3). The most common gene mutation in CRC is the inactivating mutation of the APC gene [51]. Approximately 70–80% of sporadic colorectal adenomas and carcinomas have somatic mutations of APC, and nearly all of these mutations result in premature truncation of the APC protein [5]. Inactivation of APC leads to inappropriate and constitutive activation of the Wnt signaling pathway, a general mechanism of CRC development [52]. APC mutations are found in even the earliest lesions, such as microscopic adenomas consisting of a few dysplastic glands [5]. Several nonfunctional Apc mutant mouse strains show similar CRC phenotypes, with polyps arising within the small intestine or colon [53,54]. Multiple mechanisms of APC inactivation have been reported, including hypermethylation of CpG sites in the APC promoter and decreased translation due to inhibition by miRNAs [51]. Interestingly, it was recently demonstrated that chronic NF-kB activation within intestinal epithelial cells accelerates loss of heterozygosity at the APC locus via increased nitrosative DNA damage, leading to spontaneous adenomas within the small intestine and the colon [55]. TP53, the gene encoding the tumor suppressor protein, p53, is also mutated with high incidence in CRC [56]. p53 is a cell cycle regulator that induces cell growth arrest by holding the cell cycle at the G1/S regulation checkpoint. TP53 mutations can lead to the inactivation of p53 or a 17p chromosomal deletion that eliminates the second TP53 allele [57]. Not surprisingly, this abnormality often correlates with colorectal tumorigenesis [58]. It has also been reported that the K-ras oncogene is mutated in 30–60% of large adenomas and CRC; the mutated, constitutively active form of this protein affects signaling pathways controlling cellular differentiation, motility, proliferation, and apoptosis [17]. Epidemiological data reveal that approximately 15% of all colorectal tumors demonstrate a defect in DNA mismatch repair proteins resulting in insertions/deletions in repetitive sequences (e.g., microsatellites) and microsatellite instability [59]. The most commonly mutated DNA repair enzyme genes in CRC are MSH2 (bacterial mutS

Review


Wnt signaling pathway

Cell cycle regulaon

Genec mutaon DNA damage/mismatch repair

Cytokines receptor/downstream pathway

CRC pathogenesis miRNA regulaon

DNA methylaon

Epigenec dysregulaon Histone modificaon

Chroman remodeling TRENDS in Molecular Medicine

Figure 3. Genetic and epigenetic mechanisms of colorectal cancer (CRC) pathogenesis. Mutation of genes (e.g., APC) involved in the Wnt signaling pathway plays a dominant role in CRC pathogenesis. Genes that are related to cell cycle progression, DNA repair, and cytokine signaling have also been shown to be crucial for CRC pathogenesis. DNA hypermethylation of tumor suppressor gene promoter regions has been intensively studied to demonstrate its pivotal role in gene silencing. Histone modification includes histone methylation and deacetylation, both of which have been shown to be associated with DNA methylation.

homolog 2), MLH1 (bacterial mutL homolog 1), and MSH6 [60]. The subsequent colorectal tumors tend to occur in the proximal or ‘right-sided colon’, are less invasive, occur in younger patients, and are rather undifferentiated with fewer mutations in K-ras or TP53 [59]. Conversely, 55– 70% of colorectal tumors develop in the distal or ‘left-sided colon’ [60]. Regardless of location, most of these tumors demonstrate mutations that result in overactivation of the Wnt/b-catenin signaling pathway [51]. Epigenetic regulation of colorectal carcinogenesis The link between epigenetic modifications and cancer was first made in 1983 [61]. However, until recently, appreciation of epigenetic alterations has lagged behind genetic mutations with regard to their contributions to human cancer development. With the rapidly increasing understanding of specific epigenetic mechanisms involved in gene expression regulation, and the advent of techniques to study these changes, epigenetics has become a favored area of research in the field of cancer biology. Epigenetic modifications are heritable, potentially reversible, and regulate gene expression through post-replicative DNA modification, posttranslational histone modification, and chromatin remodeling without affecting the sequence of nucleotides [62] (Figure 3). As epigenetic changes are reversible, they are appealing targets for cancer therapy [63]. miRNAs Discovered over 20 years ago, miRNAs are small, noncoding RNAs composed of approximately 18–24 nucleotides. miRNAs post-transcriptionally regulate target gene

expression by inhibiting translation or inducing the degradation of mRNA, depending on the amount of complementarity with the target sequence [64]. Mature miRNAs are incorporated into the RNA-induced silencing complex (RISC) where they are able bind to the 30 untranslated region of the target mRNA. [65]. miRNAs are frequently dysregulated in tumors either by genetic or epigenetic factors, and are currently being investigated for their potential as cancer biomarkers [66]. For example, miR21 and miR-106 are upregulated in stool samples from patients with CRC compared with normal individuals, and higher levels of miR-144 were found in the stool of CRC patients with a sensitivity of 74% and specificity of 87% [67]. Because of their small size, miRNA levels are very stable in biological samples [15], and investigators have recently demonstrated that extracellular miRNAs are stable for at least 1 month, even in stool samples [68]. As there is currently no FDA-approved blood screening test for CRC, there is increasing interest in detectable, circulating markers of these tumors such as miRNA [15]. miR-21 is considered to be an oncogenic miRNA or ‘oncomir’ that is upregulated in many solid and hematological cancers, including CRC [67]. Recently, it has been demonstrated that miR-18a reduces the repair of DNA double-strand breaks by directly suppressing the translation of the multimeric ataxia telangiectasia mutated (ATM) repair enzyme [69]. Doublestrand DNA breaks are an example of genetic instability, are extremely cytotoxic, and result from endogenous, oxidative stress (e.g., inflammation) or exogenous sources such as ionizing radiation or genotoxic agents [69]. Once 719

Review double-strand DNA breaks are detected by the Mre11 sensor complex, ATM undergoes autophosphorylation, which promotes its monomerization and kinase activity [70]. This causes G1 cell cycle arrest by activating p53, with subsequent upregulation of the cyclin-dependent kinase inhibitor, p21 [71]. Furthermore, there was a significant inverse correlation between levels of miR-18a and ATM in 45 pairs of human rectal tumors and adjacent normal tissues; this downregulation of ATM was also observed in CRC cell lines when compared with normal colon biopsies [69]. DNA methylation DNA methylation is an important regulator of gene expression and was the first epigenetic mechanism identified in cancer development [61]. In this process, a methyl group (CH3) is covalently added by DNA methyltransferase (DNMT) enzymes to carbon 5 of the cytosine base, usually in CpG dinucleotides [72]. Aberrant DNA hypermethylation at CpG-rich regions near gene promoters has consistently been associated with the inactivation of target genes. Since its postulation by Toyota et al. in 1999 [73], current investigations have increasingly focused on what is known as the CpG island methylator phenotype (CIMP) of tumor suppressor genes involved in CRC. More recent studies have revealed that numerous genes are hypermethylated at their promoter regions and silenced in CRC [74–77]. These include regulators of DNA mismatch repair such as MLH1 and MGMT (O6-methylguanine-DNA methyltransferase), and negative regulators of Wnt signaling such as Wnt inhibitory factor 1 (WIF1) [78]. MGMT encodes a DNA repair protein that demethylates lesions that result from alkylation at the O6 position of guanine that are induced by dietary nitrosamines or chemotherapeutic agents [79]. CpG hypermethylation of DNA can also lead to the silencing of genes encoding tumor suppressors and transcription factors required for cellular homeostasis. Among these is RUNX3 that codes for runt-related transcription factor 3, which modulates both the transforming growth factor (TGF)-b [80] and the Wnt/b-catenin signaling pathways [81]. Runx3/ mice exhibit gastric epithelial proliferation due to decreased apoptosis and reduced sensitivity to the growth inhibitory cytokine, TGF-b. They also demonstrate intestinal hyperplasia due to enhanced Wnt signaling activity and upregulation of the Wnt target genes, CMYC and cyclin D1 [82]. Thus, suppression of RUNX3 expression leads to aberrant signaling of both pathways, which are commonly dysregulated in colonic carcinogenesis. Loss of function of RUNX3 has significant and constitutive effects on the expression of its target genes, which may then contribute to CRC development [75,81]. Decreased expression of RUNX3 mRNA due to epigenetic hypermethylation of the RUNX3 promoter occurs in a range of human CRC cell lines [83]. Treatment of CRC cell lines with the DNMT inhibitor, 5-aza-20 -deoxycytidine, restored the expression of RUNX3 mRNA, suggesting that epigenetic inactivation of RUNX3 is due to aberrant promoter methylation [83]. Moreover, the DNA mismatch repair enzyme gene, MLH1, has also been found to be inactivated by the 720


hypermethylation of its promoter region. Decreased expression of MLH1 results in a larger number of DNA mismatches and increased microsatellite instability and, ultimately, increased carcinogenesis [84]. As discussed above, mutations of the APC gene promote CRC onset by failing to negatively regulate the Wnt/b-catenin pathway. However, it has also been suggested that functional but reduced expression of APC caused by hypermethylation of its promoter can lead to CRC associated with overexpression of the oncogene, CMYC, which is a downstream effector of the Wnt/b-catenin pathway [85]. In the case of genes lacking CpG-rich promoter regions, hypomethylation of the gene body may contribute to cancer initiation [86]. Genome-wide DNA hypomethylation has been detected in many human cancers. Unlike inactivation of tumor suppressor genes by hypermethylation in CRC, global genomic hypomethylation plays a causal role in tumor formation through the activation of proto-oncogenes [73] or by promoting chromosomal instability [86,87]. For instance, immunohistochemistry analyses indicated that the c-myc oncoprotein was highly disseminated throughout the adenomatous polyps of the colonic mucosa compared with normal colonic mucosa [88,89], where c-myc was found to be solely localized in the nuclei. Elevated levels of the c-myc protein correlated with the loss of methylation of the third exon of the CMYC gene. In addition, parallel analyses of paired normal–CRC human tissues indicated that global DNA hypomethylation occurs early in carcinogenesis and that progression of this event is followed by several causative alterations, including increased activity of DNMTs and the appearance of hypermethylated sites [61,62]. Histone modifications Unlike the intensive studies aimed at understanding DNA methylation in tumor initiation, much less is known regarding the potential contribution of aberrant histone posttranslational modifications and nucleosome positioning to the initiation and progression of CRC. The basic unit of chromatin is the nucleosome, which comprises a histone octamer (two copies each of H2A, H2B, H3, and H4) around which 147 base pairs of DNA are wrapped [90]. Evidence suggests that accumulated loss of histone H4 lysine monoacetylation, which is associated with loss of H4K16 and H4K20 trimethylation, may be a hallmark in human tumor cells, including CRC [91]. Distinct post-translational methylation of the N termini of histones can play opposing roles in the expression of target genes. For example, H3K4 mono- and tri-methylation are generally considered to be markers for activation. However, di- or tri-methylation of H3K9 or H3K27 are associated with repression [92]. The Polycomb Repressive Complexes (PRC1 and PRC2) are histone-modifying complexes that set the repressive mark trimethylation of H3K27 [93]. Investigators have recently determined a ‘DNA methylome’ using murine ApcMin adenoma as a model for cancer initiation. In this study, a core panel of differentially methylated regions (DMRs) was conserved between mouse adenomas and excised tissues from patients with advanced CRC. These patterns were also distinct from the surrounding intestinal epithelia, including intestinal stem cells, indicating that the patterns

Review


were not pre-existing, but formed de novo [94]. Hypermethylated DMRs were prevalent at PRC2 binding sites in both mice and human tumors, suggesting that increased PRC2 activity may be critical in the intestinal tumorspecific pattern of DNA hypermethylation following the loss of Apc [94]. Unlike histone methylation, acetylation of histone tails by histone acetyltransferases (HATs) predominantly demarcates an active state. Addition of the acetyl group to lysines neutralizes the positive charge of the histone. This disrupts interactions between the histone N-terminal tails and the negatively charged DNA, and also interferes with higher order compaction of the chromatin fiber [95,96]. In addition, histone acetylation recruits chromatin remodeling activities that contain subunits with bromodomains that bind preferentially to acetylated nucleosomes [97]. As a consequence, a more open chromatin structure is formed, often accompanied by histone depletion that leads

to higher accessibility to the DNA by transcription factors [98]. HDACs are a class of enzymes that remove acetyl groups from acetylated histones (as well as from some nonhistone proteins such as the transcription factors, Sp1 and Sp3) [99], reversing the open chromatin structure, and resulting in a condensed heterochromatic state and inactivation of transcription [100]. Consequently, HATs and HDACs collaboratively maintain the balance of histone acetylation in vivo to achieve homeostasis [101]. In 2002, a link between DNA methylation and histone modification in cancer was reported [102]. Genes that become hypermethylated at their promoter CpG islands are often associated with histone hypoacetylation, loss of active histone methylation marks, and gain of repressive marks [98]. Accumulating evidence indicates that bacteria can directly interfere with DNA replication and repair, RNA splicing, transcription, and chromatin remodeling [6]. H. pylori, which was first discovered and identified in a chronic

Component proteins / metabolites

LTA- L. acidophilus

IL-10 DCs Component proteins / metabolites

– PI3K–Akt

MAPK–p38

Th17 cells

?

+

–

CD4+ T cells

MAPK–Erk

NF-KB β IL-10 TGF-β

T reg cells

Coacvator complex

TIC RNAP R II

IL-10, RUNX3, TIMP3

Th1 cells


Figure 4. Cascade of molecular reactions initiated by LTA-deficient Lactobacillus acidophilus. Either component proteins or metabolic products from LTA-deficient L. acidophilus are potential antigens that can be recognized by dendritic cells. Presentation of antigen to recipient CD4+ T cells induced the genesis of a population of Tregs that could further differentiate into Th17 or Th1 cells, depending on the microenvironment. Meanwhile, several signaling pathways were triggered in dendritic cells by the recognition of antigens. The MAPK–Erk1/2 pathway was shown to be crucial for the expression of the anti-inflammatory cytokine IL-10, and tumor suppressors, such as RUNX3 and TIMP3, were also upregulated by the treatment of LTA-deficient L. acidophilus both in vitro and in vivo. Abbreviations: LTA, lipoteichoic acid; RNAP II, RNA polymerase II; TIC, transcription initiation complex; Treg, regulatory T cell; Th, T helper; MAPK–Erk1/2, mitogen-activated protein kinase–extracellular signal-regulated kinase 1/2; IL-10, interleukin-10.

721

Review gastritis patient in 1982, was shown to be a potential causative agent of gastric carcinogenesis due to altered DNA methylation [103] and histone acetylation patterns [104]. In a recent study that examined the dynamic interplay between host and gut microbiota [105], wild type and Tlr2/ mice were employed to evaluate epigenomic and metagenomic differences in the colonic mucosa. Significantly modified DNA methylation patterns of host genes involved in immune processes were identified. Alterations in the number and composition of the mucosal microbiome were also shown with the use of next-generation pyrosequencing of bacterial 16S rRNA. Collectively, these consequences of gut microbial imbalance may result in epigenetic abnormalities that contribute to differential gene expression and potentially trigger the onset and progression of inflammation-associated carcinomas [105]. Tumor therapy and protection by genetically modified bacteria There are different causative mechanisms responsible for the generation and progression of CRC. On the basis of our current knowledge, drugs approved by the FDA for use in CRC treatment, include Bevacizumab (an angiogenesis inhibitor), Cetuximab (a monoclonal antibody against the epidermal growth factor receptor), and Capecitabine (a drug precursor of the DNA synthesis inhibitor, 5-fluorouracil). These drugs are designed for inhibiting and/or eliminating aberrant tumor cells. However, the regular metabolism of normal cells is also disrupted by these drugs, which may result in unwanted side effects. Although inhibitors with therapeutic potential are being designed based on HDACs, HATs, and DNMTs [106], there will always be a major concern of how to enhance the drug specificity in vivo while diminishing side effects. Changes in the composition of the microbiota can lead to altered immunologic signaling; thus, it is reasonable to speculate that oral administration of gut commensal bacteria could enhance anti-inflammatory responses at the genetic, epigenetic, and/or immune cytokine levels to rebalance disrupted homeostasis. Recently, our laboratory and other investigators have provided supportive evidence that the gut microbiota plays an essential role in intestinal epigenomic mechanisms of the host [107,108]. We demonstrated that deletion of lipoteichoic acid (LTA), a TLR2 ligand, normalizes innate and adaptive pathogenic immune responses and causes regression of established colonic polyps in a novel transgenic mouse model that develops multiple polyps in the colon and distal ileum as early as 5–6 months of age. This study revealed the proinflammatory role of LTA and the ability of LTA-deficient Lactobacillus acidophilus to suppress inflammation and protect against colon cancer [54]. Furthermore, this genetically modified gut commensal bacterium, LTA-deficient L. acidophilus, could prevent and/or ablate established colitis [109] and polyposis [54], downregulate downstream signaling [110], and upregulate tumor suppressor genes in CRC cell lines [111] (Figure 4). These observations suggest that overt inflammation can be reshaped by immune regulatory agents or even by natural microbial gene products, including the 722


Box 1. Outstanding questions Is dysbiosis a cause or consequence of CAC? Recent studies have demonstrated that dysbiosis induces a dysregulated microenvironment, which may in turn elicit protumorigenic inflammation. However, clinical data show changes in the composition of the gut microbiota in patients with CAC, indicating that bacterial dysbiosis may play a critical role in the developmental process of CAC. Studies are ongoing to rebalance the condition of bacterial dysbiosis in order to elicit healthy homeostasis to mitigate cancer development and/or progression. What are the potential epigenetic consequences of beneficial bacteria?The epigenetic consequences may include DNA methylation, histone modification in target gene regions (e.g., enhancer, promoter, etc.), and dynamic chromosome remodeling across all of the gene elements. Basically, the metabolites (e.g., small peptides and chemicals) and protein components of beneficial bacteria are the major functional molecules. The targets of these molecules in the gut are enzymes with epigenetic modification activities. Acting naturally, as either antagonists or agonists of those enzymatic activities, functional molecules can control the physiological dynamics in vivo. What are the advantages of the administration of engineered bacteria to CRC patients?Most commercial drugs available for CRC treatment are chemicals that were originally designed to function as inhibitors of certain normal physiological processes to inhibit tumor cell growth or to eliminate neoplastic cells. However, many of those drugs may initiate severe side effects. Therefore, the continuous use of chemical drugs raises concerns about their cell specificity in vivo. Beneficial bacteria have been widely accepted as healthy supplements. Engineered beneficial bacteria may show more anti-inflammatory properties and would restore gut homeostasis due to the removal of their proinflammatory gene products. Because these bacteria naturally reside in the gut, their potential to cause severe side effects is much less than that of drugs.

bacterial surface layer proteins upon which the focus of our laboratory is currently centered. Although the beneficial effects of gut commensal bacteria have been demonstrated in several experimental models [112,113], little is known about the effect of these bacteria on genomic stability or epigenetic regulation. Using an in vitro 3D intestinal mucosal model consisting of human intestinal HT-29/B6 or T84 cell lines and peripheral blood mononuclear cells (PBMCs) from human healthy individuals, researchers demonstrated that the commensal bacteria, Bifidobacterium breve and Lactobacillus rhamnosus GG, inhibited the production of the proinflammatory cytokines, IL-17 and IL-23. Global DNA methylation and histone acetylation were also evaluated to demonstrate that commensal microbial treatment resulted in increased host DNA methylation and decreased histone acetylation [114]. It is tempting to speculate about the status of DNA methylation and histone acetylation at the promoter regions of IL-17 and IL-23 in these experiments. Using the commensal bacteria Lactobacillus and Bifidobacterium, as examples, the potential strategy to restore the composition of the intestinal microbiota with genetically modified bacteria appears feasible and promising. To this end, monoassociation and fecal flora transfer studies using germ-free animal models of CRC may provide answers to several unresolved questions (Box 1). Concluding remarks The mutualistic relationship between the host and the gut microbiome benefits the host in many ways. With that said,

Review any perturbation of this delicate balance caused by altered microbial composition may induce aberrant intestinal signaling pathways and epigenetic modifications. The combinatorial impact of bacterial surface toxins and metabolic end-products may stimulate severe inflammation that can progress to cancer due to disrupted homeostasis. Instead of turning to drugs with obvious side effects for the treatment of inflammation-associated disease, genetically modified bacteria or bacterial products may be considered for a more specific correction of the original dysbiosis. At present, experimental data using genetically modified bacteria and their products are promising, and some beneficial commensal strains are currently being tested in clinical IBD cases. Although progress has been made in understanding and utilizing gut commensal microbes for these purposes, we are still at an early stage in the battle against CRC development. Additional research is necessary to better understand the association between microorganisms and carcinogenesis. Further investigation of the epigenetic regulation by natural and biologically engineered bacteria or their immune regulatory gene products in the gut will lead to the development of safer, targeted therapeutics for CRC. Acknowledgments This work was supported in part by National Institutes of Health (NIH) grant 1R01AI098833-01, Department of Defense (DoD) grant CA111002, Gatorade Trust Pilot Project Funding from the University of Florida, and NIH/National Center for Research Resources (NCRR) Clinical and Translational Science Award to the University of Florida (UL1 RR 029890).

References 1 Ferlay, J. et al. (2010) Estimates of worldwide burden of cancer in 2008: GLOBOCAN 2008. Int. J. Cancer 127, 2893–2917 2 Tjalsma, H. et al. (2012) A bacterial driver-passenger model for colorectal cancer: beyond the usual suspects. Nat. Rev. Microbiol. 10, 575–582 3 Tudek, B. and Speina, E. (2012) Oxidatively damaged DNA and its repair in colon carcinogenesis. Mutat. Res. 736, 82–92 4 Arends, M.J. (2013) Pathways of colorectal carcinogenesis. Appl. Immunohistochem. Mol. Morphol. 21, 97–102 5 Fearon, E.R. (2011) Molecular genetics of colorectal cancer. Annu. Rev. Pathol. 6, 479–507 6 Bierne, H. and Cossart, P. (2012) When bacteria target the nucleus: the emerging family of nucleomodulins. Cell. Microbiol. 14, 622–633 7 Tosolini, M. et al. (2011) Clinical impact of different classes of infiltrating T cytotoxic and helper cells (Th1, th2, treg, th17) in patients with colorectal cancer. Cancer Res. 71, 1263–1271 8 Couturier-Maillard, A. et al. (2013) NOD2-mediated dysbiosis predisposes mice to transmissible colitis and colorectal cancer. J. Clin. Invest. 123, 700–711 9 Cario, E. (2013) Microbiota and innate immunity in intestinal inflammation and neoplasia. Curr. Opin. Gastroenterol. 29, 85–91 10 Viladomiu, M. et al. (2013) Nutritional protective mechanisms against gut inflammation. J. Nutr. Biochem. 24, 929–939 11 Ford, A.C. et al. (2013) Ulcerative colitis. BMJ 346, f432 12 Sansone, P. and Bromberg, J. (2011) Environment, inflammation, and cancer. Curr. Opin. Genet. Dev. 21, 80–85 13 Jess, T. et al. (2012) Risk of colorectal cancer in patients with ulcerative colitis: a meta-analysis of population-based cohort studies. Clin. Gastroenterol. Hepatol. 10, 639–645 14 Okayasu, I. (2012) Development of ulcerative colitis and its associated colorectal neoplasia as a model of the organ-specific chronic inflammation-carcinoma sequence. Pathol. Int. 62, 368–380 15 Mazeh, H. et al. (2013) The diagnostic and prognostic role of microRNA in colorectal cancer – a comprehensive review. J. Cancer 4, 281–295


16 Vogelstein, B. et al. (1988) Genetic alterations during colorectal-tumor development. N. Engl. J. Med. 319, 525–532 17 Al-Sohaily, S. et al. (2012) Molecular pathways in colorectal cancer. J. Gastroenterol. Hepatol. 27, 1423–1431 18 Dongarra, M.L. et al. (2013) Mucosal immunology and probiotics. Curr. Allergy Asthma Rep. 13, 19–26 19 Sekirov, I. et al. (2010) Gut microbiota in health and disease. Physiol. Rev. 90, 859–904 20 Barengolts, E. (2013) Vitamin D and prebiotics may benefit the intestinal microbacteria and improve glucose homeostasis in prediabetes and type 2 diabetes. Endocr. Pract. 19, 497–510 http:// dx.doi.org/10.4158/EP12263.RA 21 Winter, S.E. et al. (2013) The dynamics of gut-associated microbial communities during inflammation. EMBO Rep. 14, 319–327 22 Darfeuille-Michaud, A. et al. (2004) High prevalence of adherentinvasive Escherichia coli associated with ileal mucosa in Crohn’s disease. Gastroenterology 127, 412–421 23 Rolhion, N. et al. (2010) Abnormally expressed ER stress response chaperone Gp96 in CD favours adherent-invasive Escherichia coli invasion. Gut 59, 1355–1362 24 Ley, R.E. et al. (2005) Obesity alters gut microbial ecology. Proc. Natl. Acad. Sci. U.S.A. 102, 11070–11075 25 Arthur, J.C. and Jobin, C. (2011) The struggle within: microbial influences on colorectal cancer. Inflamm. Bowel Dis. 17, 396–409 26 Kelly, D. and Mulder, I.E. (2012) Microbiome and immunological interactions. Nutr. Rev. 70 (Suppl. 1), S18–S30 27 Dominguez-Bello, M.G. et al. (2011) Development of the human gastrointestinal microbiota and insights from high-throughput sequencing. Gastroenterology 140, 1713–1719 28 Mulder, I.E. et al. (2011) Restricting microbial exposure in early life negates the immune benefits associated with gut colonization in environments of high microbial diversity. PLoS ONE 6, e28279 29 Nicholson, J.K. et al. (2012) Host–gut microbiota metabolic interactions. Science 336, 1262–1267 30 Ivanov, I.I. et al. (2008) Specific microbiota direct the differentiation of IL-17-producing T-helper cells in the mucosa of the small intestine. Cell Host Microbe 4, 337–349 31 Hadis, U. et al. (2011) Intestinal tolerance requires gut homing and expansion of FoxP3+ regulatory T cells in the lamina propria. Immunity 34, 237–246 32 Capelluto, D.G. (2012) Tollip: a multitasking protein in innate immunity and protein trafficking. Microbes Infect. 14, 140–147 33 Li, Y. et al. (2013) Constitutive TLR4 signalling in intestinal epithelium reduces tumor load by increasing apoptosis in APCMin/+ mice. Oncogene http://dx.doi.org/10.1038/onc.2012.581 34 Fung, K.Y. et al. (2012) A review of the potential mechanisms for the lowering of colorectal oncogenesis by butyrate. Br. J. Nutr. 108, 820–831 35 Leonel, A.J. and Alvarez-Leite, J.I. (2012) Butyrate: implications for intestinal function. Curr. Opin. Clin. Nutr. Metab. Care 15, 474–479 36 Iwai, K. (2012) Diverse ubiquitin signaling in NF-kB activation. Trends Cell Biol. 22, 355–364 37 Sakamoto, K. et al. (2009) Constitutive NF-kB activation in colorectal carcinoma plays a key role in angiogenesis, promoting tumor growth. Clin. Cancer Res. 15, 2248–2258 38 Manichanh, C. et al. (2012) The gut microbiota in IBD. Nat. Rev. Gastroenterol. Hepatol. 9, 599–608 39 Kipanyula, M.J. et al. (2013) Signaling pathways bridging microbialtriggered inflammation and cancer. Cell. Signal. 25, 403–416 40 Sears, C.L. and Pardoll, D.M. (2011) Perspective: alpha-bugs, their microbial partners, and the link to colon cancer. J. Infect. Dis. 203, 306–311 41 Azcarate-Peril, M.A. et al. (2011) The intestinal microbiota, gastrointestinal environment and colorectal cancer: a putative role for probiotics in prevention of colorectal cancer? Am. J. Physiol. Gastrointest. Liver Physiol. 301, G401–G424 42 Yongvanit, P. et al. (2012) Oxidative and nitrative DNA damage: key events in opisthorchiasis-induced carcinogenesis. Parasitol. Int. 61, 130–135 43 Mangerich, A. et al. (2012) Infection-induced colitis in mice causes dynamic and tissue-specific changes in stress response and DNA damage leading to colon cancer. Proc. Natl. Acad. Sci. U.S.A. 109, E1820–E1829 723

Review 44 Fox, J.G. et al. (2011) Helicobacter hepaticus infection in mice: models for understanding lower bowel inflammation and cancer. Mucosal Immunol. 4, 22–30 45 Lao, V.V. et al. (2009) Incorporation of 5-chlorocytosine into mammalian DNA results in heritable gene silencing and altered cytosine methylation patterns. Carcinogenesis 30, 886–893 46 Winter, S.E. et al. (2013) Host-derived nitrate boosts growth of E. coli in the inflamed gut. Science 339, 708–711 47 Arthur, J.C. et al. (2012) Intestinal inflammation targets cancerinducing activity of the microbiota. Science 338, 120–123 48 Wu, S. et al. (2009) A human colonic commensal promotes colon tumorigenesis via activation of T helper type 17 T cell responses. Nat. Med. 15, 1016–1022 49 Schaible, T.D. et al. (2011) Maternal methyl-donor supplementation induces prolonged murine offspring colitis susceptibility in association with mucosal epigenetic and microbiomic changes. Hum. Mol. Genet. 20, 1687–1696 50 Hanahan, D. and Weinberg, R.A. (2011) Hallmarks of cancer: the next generation. Cell 144, 646–674 51 Wu, W.K. et al. (2012) Dysregulation and crosstalk of cellular signaling pathways in colon carcinogenesis. Crit. Rev. Oncol. Hematol. 86, 251–277 52 Christie, M. et al. (2012) Different APC genotypes in proximal and distal sporadic colorectal cancers suggest distinct WNT/b-catenin signalling thresholds for tumourigenesis. Oncogene http:// dx.doi.org/10.1038/onc.2012.486 53 Robanus-Maandag, E.C. et al. (2010) A new conditional Apc-mutant mouse model for colorectal cancer. Carcinogenesis 31, 946–952 54 Khazaie, K. et al. (2012) Abating colon cancer polyposis by Lactobacillus acidophilus deficient in lipoteichoic acid. Proc. Natl. Acad. Sci. U.S.A. 109, 10462–10467 55 Shaked, H. et al. (2012) Chronic epithelial NF-kB activation accelerates APC loss and intestinal tumor initiation through iNOS up-regulation. Proc. Natl. Acad. Sci. U.S.A. 109, 14007–14012 56 Wei, J. et al. (2012) p53 family: role of protein isoforms in human cancer. J. Nucleic Acids 2012, 687359 57 Markowitz, S.D. and Bertagnolli, M.M. (2009) Molecular origins of cancer: molecular basis of colorectal cancer. N. Engl. J. Med. 361, 2449–2460 58 Bacolod, M.D. and Barany, F. (2010) Gene dysregulations driven by somatic copy number aberrations – biological and clinical implications in colon tumors: a paper from the 2009 William Beaumont Hospital Symposium on Molecular Pathology. J. Mol. Diagn. 12, 552–561 59 Boland, C.R. and Goel, A. (2010) Microsatellite instability in colorectal cancer. Gastroenterology 138, 2073–2087 60 Albuquerque, C. et al. (2011) Colorectal cancers choosing sides. Biochim. Biophys. Acta 1816, 219–231 61 Feinberg, A.P. and Vogelstein, B. (1983) Hypomethylation distinguishes genes of some human cancers from their normal counterparts. Nature 301, 89–92 62 Sharma, S. et al. (2010) Epigenetics in cancer. Carcinogenesis 31, 27–36 63 Crea, F. et al. (2012) Polycomb genes and cancer: time for clinical application? Crit. Rev. Oncol. Hematol. 83, 184–193 64 Bronevetsky, Y. and Ansel, K.M. (2013) Regulation of miRNA biogenesis and turnover in the immune system. Immunol. Rev. 253, 304–316 65 Iorio, M.V. and Croce, C.M. (2012) MicroRNA dysregulation in cancer: diagnostics, monitoring and therapeutics. A comprehensive review. EMBO Mol. Med. 4, 143–159 66 Fabbri, M. (2012) TLRs as miRNA receptors. Cancer Res. 72, 6333– 6337 67 Kong, Y.W. et al. (2012) microRNAs in cancer management. Lancet Oncol. 13, e249–e258 68 Turchinovich, A. et al. (2011) Characterization of extracellular circulating microRNA. Nucleic Acids Res. 39, 7223–7233 69 Wu, C.W. et al. (2013) MicroRNA-18a attenuates DNA damage repair through suppressing the expression of ataxia telangiectasia mutated in colorectal cancer. PLoS ONE 8, e57036 70 Stracker, T.H. et al. (2013) The ATM signaling network in development and disease. Front. Genet. 4, 37 71 Sperka, T. et al. (2012) DNA damage checkpoints in stem cells, ageing and cancer. Nat. Rev. Mol. Cell Biol. 13, 579–590 724


72 Migheli, F. et al. (2013) Comparison study of MS-HRM and pyrosequencing techniques for quantification of APC and CDKN2A gene methylation. PLoS ONE 8, e52501 73 Toyota, M. et al. (1999) CpG island methylator phenotype in colorectal cancer. Proc. Natl. Acad. Sci. U.S.A. 96, 8681–8686 74 Fan, X.Y. et al. (2011) Association between RUNX3 promoter methylation and gastric cancer: a meta-analysis. BMC Gastroenterol. 11, 92 75 Nishio, M. et al. (2010) RUNX3 promoter methylation in colorectal cancer: its relationship with microsatellite instability and its suitability as a novel serum tumor marker. Anticancer Res. 30, 2673–2682 76 Lee, B.B. et al. (2009) Aberrant methylation of APC, MGMT, RASSF2A, and Wif-1 genes in plasma as a biomarker for early detection of colorectal cancer. Clin. Cancer Res. 15, 6185–6191 77 Mirchev, M.B. et al. (2010) DNA methylation in patients with colorectal cancer – correlation with some clinical and morphological features and with local tumour invasion. Folia Med. (Plovdiv) 52, 22–30 78 Drini, M. et al. (2011) Investigating the potential role of genetic and epigenetic variation of DNA methyltransferase genes in hyperplastic polyposis syndrome. PLoS ONE 6, e16831 79 Curtin, N.J. (2012) DNA repair dysregulation from cancer driver to therapeutic target. Nat. Rev. Cancer 12, 801–817 80 Ito, Y. (2008) RUNX genes in development and cancer: regulation of viral gene expression and the discovery of RUNX family genes. Adv. Cancer Res. 99, 33–76 81 Ito, K. et al. (2008) RUNX3 attenuates b-catenin/T cell factors in intestinal tumorigenesis. Cancer Cell 14, 226–237 82 Chuang, L.S. and Ito, Y. (2010) RUNX3 is multifunctional in carcinogenesis of multiple solid tumors. Oncogene 29, 2605–2615 83 Ku, J.L. et al. (2004) Promoter hypermethylation downregulates RUNX3 gene expression in colorectal cancer cell lines. Oncogene 23, 6736–6742 84 Kim, J.H. et al. (2013) Subsets of microsatellite-unstable colorectal cancers exhibit discordance between the CpG island methylator phenotype and MLH1 methylation status. Mod. Pathol. 26, 1013–1022 85 Schwitalla, S. et al. (2013) Intestinal tumorigenesis initiated by dedifferentiation and acquisition of stem-cell-like properties. Cell 152, 25–38 86 Feinberg, A.P. et al. (2006) The epigenetic progenitor origin of human cancer. Nat. Rev. Genet. 7, 21–33 87 Baker, S.J. et al. (1990) p53 gene mutations occur in combination with 17p allelic deletions as late events in colorectal tumorigenesis. Cancer Res. 50, 7717–7722 88 Tulchin, N. et al. (1992) c-myc protein distribution. Neoplastic tissues of the human colon. Am. J. Pathol. 140, 719–729 89 Tulchin, N. et al. (1995) C-myc protein distribution – mammary adenocarcinomas of mtv/myc transgenic mice. Int. J. Oncol. 7, 5–9 90 Struhl, K. and Segal, E. (2013) Determinants of nucleosome positioning. Nat. Struct. Mol. Biol. 20, 267–273 91 Fraga, M.F. et al. (2005) Loss of acetylation at Lys16 and trimethylation at Lys20 of histone H4 is a common hallmark of human cancer. Nat. Genet. 37, 391–400 92 Barski, A. et al. (2007) High-resolution profiling of histone methylations in the human genome. Cell 129, 823–837 93 Chase, A. and Cross, N.C. (2011) Aberrations of EZH2 in cancer. Clin. Cancer Res. 17, 2613–2618 94 Grimm, C. et al. (2013) DNA-methylome analysis of mouse intestinal adenoma identifies a tumour-specific signature that is partly conserved in human colon cancer. PLoS Genet. 9, e1003250 95 Kan, P.Y. et al. (2009) The H4 tail domain participates in intra- and internucleosome interactions with protein and DNA during folding and oligomerization of nucleosome arrays. Mol. Cell. Biol. 29, 538–546 96 Shogren-Knaak, M. et al. (2006) Histone H4-K16 acetylation controls chromatin structure and protein interactions. Science 311, 844–847 97 Hassan, A.H. et al. (2001) Histone acetyltransferase complexes stabilize SWI/SNF binding to promoter nucleosomes. Cell 104, 817–827 98 Akhavan-Niaki, H. and Samadani, A.A. (2013) DNA methylation and cancer development: molecular mechanism. Cell Biochem. Biophys. (in press) 99 Wilson, A.J. et al. (2010) Apoptotic sensitivity of colon cancer cells to histone deacetylase inhibitors is mediated by an Sp1/Sp3-activated

Review

100 101 102

103

104

105

106

transcriptional program involving immediate-early gene induction. Cancer Res. 70, 609–620 Lee, Y.M. (2011) Control of RUNX3 by histone methyltransferases. J. Cell. Biochem. 112, 394–400 Icardi, L. et al. (2012) The HAT/HDAC interplay: multilevel control of STAT signaling. Cytokine Growth Factor Rev. 23, 283–291 Fahrner, J.A. et al. (2002) Dependence of histone modifications and gene expression on DNA hypermethylation in cancer. Cancer Res. 62, 7213–7218 Park, S.Y. et al. (2009) Comparison of CpG island hypermethylation and repetitive DNA hypomethylation in premalignant stages of gastric cancer, stratified for Helicobacter pylori infection. J. Pathol. 219, 410–416 Ding, S.Z. et al. (2010) Helicobacter pylori-induced histone modification, associated gene expression in gastric epithelial cells, and its implication in pathogenesis. PLoS ONE 5, e9875 Kellermayer, R. et al. (2011) Colonic mucosal DNA methylation, immune response, and microbiome patterns in Toll-like receptor 2knockout mice. FASEB J. 25, 1449–1460 Rius, M. and Lyko, F. (2012) Epigenetic cancer therapy: rationales, targets and drugs. Oncogene 31, 4257–4265


107 Licciardi, P.V. et al. (2010) Epigenome targeting by probiotic metabolites. Gut Pathog. 2, 24 108 Ritchie, M.L. and Romanuk, T.N. (2012) A meta-analysis of probiotic efficacy for gastrointestinal diseases. PLoS ONE 7, e34938 109 Mohamadzadeh, M. et al. (2011) Regulation of induced colonic inflammation by Lactobacillus acidophilus deficient in lipoteichoic acid. Proc. Natl. Acad. Sci. U.S.A. 108 (Suppl. 1), 4623–4630 110 Saber, R. et al. (2011) Lipoteichoic acid-deficient Lactobacillus acidophilus regulates downstream signals. Immunotherapy 3, 337–347 111 Lightfoot, Y.L. et al. (2013) Targeting aberrant colon cancer-specific DNA methylation with lipoteichoic acid-deficient Lactobacillus acidophilus. Gut Microbes 4, 84–88 112 Hedin, C. et al. (2007) Evidence for the use of probiotics and prebiotics in inflammatory bowel disease: a review of clinical trials. Proc. Nutr. Soc. 66, 307–315 113 Marteau, P.R. et al. (2001) Protection from gastrointestinal diseases with the use of probiotics. Am. J. Clin. Nutr. 73, 430S–436S 114 Ghadimi, D. et al. (2012) Epigenetic imprinting by commensal probiotics inhibits the IL-23/IL-17 axis in an in vitro model of the intestinal mucosal immune system. J. Leukoc. Biol. 92, 895–911

725

Review

Non-canonical roles of lysyl-tRNA synthetase in health and disease Alex Motzik1, Hovav Nechushtan2, Shen Yun Foo3, and Ehud Razin1 1

Department of Biochemistry and Molecular Biology, The Institute for Medical Research Israel-Canada, The Hebrew University–Hadassah Medical School, Jerusalem 91120, Israel 2 Oncology Department, Hadassah Hebrew University Medical Center, POB 12272, Jerusalem 91120, Israel 3 Department of Microbiology, National University of Singapore and The Hebrew University of Jerusalem (NUS–HUJ) Cellular and Molecular Mechanisms of Inflammation Programme (NRF), 1 CREATE Way, Innovation Wing, #03-09, Singapore 138602

Lysyl-tRNA synthetase (LysRS) is a highly conserved enzyme that is part of the translational machinery in all living cells. Besides its canonical role in translation, LysRS gained additional domains and functions throughout evolution. These include its essential role in HIV replication and its roles in transcriptional regulation, cytokine-like signaling, and transport of proteins to the cell membrane. These diverse processes are tightly regulated through post-transcriptional modifications, interactions with other proteins, and targeting to the various cell compartments. The emerging variety of tasks performed by LysRS may therefore be utilized by various processes and pathological conditions that are described in this review, and their ongoing investigation is of extreme importance for our understanding of basic cellular regulatory mechanisms. Rationale Lysyl-tRNA synthetase (LysRS) is an evolutionarily conserved enzyme that takes part in protein synthesis. In addition to its well-studied canonical function, this enzyme has been found to be involved in various other cellular processes. An improved understanding of these non- canonical functions and their involvement in pathological events will provide new opportunities for future therapeutic studies. Aminoacyl-tRNA synthetases (ARSs): from translation to various cellular processes ARSs are an ancient family of enzymes that are evolutionarily conserved in all life forms. Their primary, or canonical, function is to catalyze the esterification reactions that conjugate amino acids with cognate tRNAs as part of the protein translation process [1]. Aminoacylation is a twostep process in which amino acids are first activated by ATP, forming an intermediate aminoacyl adenylate, and then transferred to the 30 -end of tRNA to form the Corresponding author: Razin, E. ([email protected]). Keywords: lysyl-tRNA synthetase; HIV; cell signaling; gene regulation. 1471-4914/$ – see front matter ß 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.molmed.2013.07.011

726


aminoacyl-tRNA end-product [2,3]. All ARSs contain catalytic and anticodon recognition domains specific to their cognate amino acids. In mammalian cells, nine ARSs reside in a high molecular weight multi-tRNA synthetase complex (MSC) together with three non-enzymatic proteins, p43, p18, and p38 [4–6]. In addition to their conserved role in translation, some ARSs acquired additional domains with unique structural characteristics that are neither a part of the enzymatic core nor present in bacterial homologs. These newly evolved domains, which are generally attached to the amino or carboxy terminus, are not essential for tRNA charging, but rather are responsible for non-canonical activities unrelated to aminoacylation [6]. Evidence from recent studies suggests that defects in either canonical or non-canonical ARS functions are associated with human diseases (reviewed in [7]). For instance, potential associations of ARSs with several types of cancer through aberrant expression and interactions have been described [8]. This could involve transcriptional regulation [9,10]. Our present review focuses on the non-canonical functions of lysyl-tRNA synthetase in health and disease and is an update of a review with a more limited scope that we published several years ago [11]. The LysRS–diadenosine tetraphosphate (Ap4A) pathway The regulation of microphthalmia-associated transcription factor (MITF) transcriptional activity has been intensively investigated since the first studies in the 1990s [12,13]. We found that in several cell types, including mast cells and cardiomyocytes [14], MITF transcriptional activity is inhibited by the histidine triad nucleotide-binding protein 1 (Hint-1) through direct binding and by formation of inhibitory complexes [13,15–17], whereas transcriptional activation requires the disruption of this MITF– Hint-1 interaction [13,15,17–19]. This dissociation of Hint-1 is specifically driven by the binding of Ap4A to Hint-1, which causes its release from MITF and thus allows activation of MITF target genes. Ap4A, which is mostly hydrolyzed by the NUDT2 [nudix (nucleoside diphosphate linked moiety X)-type motif 2] gene product, was later established by our group as a bona fide second messenger [14,15,18,20–22].

Review


(A)

(C)

NUDT2 Impβ

MSC

LysRS

MSC HINT1

LysRS HINT1

MITF/USF2

Impβ NUDT2

MITF/USF2

(B) P

LysRS

MSC P

LysRS

K AP M

y wa th pa

4A

Ap

HINT1

MITF/USF2

P

LysRS

Gene transcripon


Figure 1. A model that summarizes the lysyl-tRNA synthetase (LysRS)–diadenosine tetraphosphate (Ap4A) signaling pathway. (A) In quiescent cells, transcription factors microphthalmia-associated transcription factor (MITF) and/or upstream transcription factor 2 (USF2) are inhibited by the histidine triad nucleotide-binding protein 1 (Hint-1) through direct binding and by formation of inhibitory complexes. LysRS normally resides in the high molecular weight cytoplasmic multi-tRNA synthetase complex (MSC), where it takes part in the translation process. (B) Upon extracellular stimulation, LysRS is specifically phosphorylated on Ser207 in a mitogen-activated protein kinase (MAPK)-dependent manner. This phosphorylation leads to its structural switch, causing dissociation from the MSC, nuclear translocation and, in parallel, switches the function of the protein from aminoacylation to Ap4A production. Locally synthesized in the proximity of the MITF (or USF2)–Hint-1 inhibitory complex, Ap4A binds to Hint-1 causing its dissociation from MITF (or USF2), allowing transcription of their target genes. (C) By utilizing a non-canonical nuclear translocation mechanism, Ap4A hydrolase directly associates with importin b (Impb) and translocates into the nucleus while Ap4A levels are elevated. Ap4A hydrolase then locally hydrolyzes the accumulated Ap4A into AMP + ATP in the nucleus, restoring homeostatic conditions and allowing Hint-1 to reassociate with MITF (or USF2) [20]. The nature of the downregulation of LysRS itself in terms of dephosphorylation/nuclear export or degradation is still unclear and is yet to be determined. An animated version of the LysRS–Ap4A–MITF pathway is available on the internet (http://www.youtube.com/watch?v=NGgMEz79NX4).

Although several ARSs possess the ability to produce Ap4A, it has been postulated that LysRS is a major contributor to the production of this nucleotide [23–25]. Moreover, the free form of LysRS from rat liver was reported to synthesize higher levels of Ap4A than the form associated to the MSC [26]. It is, however, important to remark that other ARS can produce dinucleotides, and all the studies above were performed in vitro before current molecular tools such as siRNA silencing and site-directed mutagenesis were available. To analyze further the mechanism underlying this induction of Ap4A production we studied LysRS in activated mast cells and cardiomyocytes (Figure 1). We found that upon IgE–antigen (Ag) binding to the high-affinity FceRI receptor, and b-adrenergic stimulation of cardiomyocytes, LysRS is specifically phosphorylated on a serine (Ser) 207 residue in a MAP kinase (MAPK)-dependent manner [14,22]. This phosphorylation leads to a structural switch in LysRS leading to its dissociation from the MSC

and nuclear translocation, and switches the function of the protein from aminoacylation to Ap4A production (Figure 1B). Locally synthesized in the proximity of the MITF–Hint-1 inhibitory complex, Ap4A binds to Hint-1 causing its dissociation from MITF, allowing MITF to transcribe its target genes. Thus, phosphorylation of a single Ser residue causes LysRS to switch its canonical function in the translation apparatus to activation of transcription [21,22]. Silencing of LysRS by siRNA led to a reduction in IgE–Ag induced MITF transcriptional activity [22]. It is important to note that treatment of cells in which LysRS had been knocked down with wild type LysRS resulted in reconstitution of the IgE–Ag-induced production of Ap4A. Furthermore, transfection of a mutant Ser207Ala LysRS was unable to reconstitute the production of IgE–Ag-induced Ap4A production. These results strongly imply that LysRS has indeed an important role in immunologically stimulated Ap4A production and that phosphorylation of Ser207 has an essential role in this 727

Review


Leu133

Thr52

Tyr173

Ser207

Ancodon-binding 1

70

Aminoacylaon 214 223

575 597 TRENDS in Molecular Medicine

Figure 2. A schematic representation of lysyl-tRNA synthetase (LysRS), its domains, and residues that affect its function. The main domains of LysRS are the N-terminal anticodon-binding domain, which recognizes and binds tRNALys, and the aminoacylation domain, which catalyzes the formation of the aminoacyl-tRNA end-product. Phosphorylation of Thr52 releases LysRS from the multi-tRNA synthetase complex (MSC) and it is translocated to the membrane. There, LysRS stabilizes the 67 kDa laminin receptor (67LR) by inhibiting its ubiquitin-dependent degradation, thereby enhancing laminin-induced cell migration of cancer cells. Leu133His and Tyr173SerfsX7 mutations inactivate the aminoacylation reaction and were found in patients with Charcot–Marie–Tooth (CMT) disease phenotypes. Phosphorylation of Ser207 leads to a structural switch in LysRS leading to its dissociation from the MSC and nuclear translocation, and switches the function of the protein from aminoacylation to Ap4A production. Ap4A binds to Hint-1 causing its dissociation from MITF, activating its transcriptional activity.

regard. More recently, structural studies have revealed the crucial role of Ser207 phosphorylation [21]. These studies are the first to demonstrate that modification of a tRNA synthetase can prevent its ability to function in the translation process, while enhancing its enzymatic activity as a producer of dinucleotides and as a transcriptional regulator. This is the result of a distinct conformational change to an open form of LysRS following phosphorylation at this particular site (Figure 2). Furthermore, chromatin immunoprecipitation analysis revealed the binding of LysRS to DNA a few minutes after IgE–Ag stimulation. This of course might result from its binding to MITF that binds to DNA. By utilizing pseudo-phosphorylated Ser207Asp LysRS mutant, we demonstrated that the binding to DNA of this ‘open form’ of LysRS is much more efficient. It is important to note that our studies also suggest that, although two stable dimers of LysRS molecules are normally present in the MSC, when one dimeric LysRS is released following Ser207 phosphorylation the other dimer can be maintained in the MSC to sustain protein synthesis [21]. Studies in transgenic mouse models are now being carried out to analyze the physiological significance of LysRS as a transcriptional regulator. An animated video which summarizes the LysRS–Ap4A–MITF pathway is available on the internet (http://www.youtube.com/watch? v=NGgMEz79NX4).

cleaved by leukocyte elastase into two different cytokines. The cleaved C-terminal polypeptide of tyrosyl-tRNA synthetase shows chemotactic activity for immune cells, and induces TNF-a and platelet tissue factor, whereas the N-terminal domain is similar to interleukin 8 [27]. Human tryptophanyl-tRNA synthetase undergoes alternative splicing to generate its N-terminal truncated form [28]. By contrast, LysRS was found to be secreted as a fulllength protein upon TNF-a treatment by several cancer cell lines. Secreted LysRS was shown to act on macrophages, inducing migration and TNF-a secretion. It appears that this extracellular LysRS signaling is mediated by G-protein activation, extracellular signal-regulated kinase (ERK), and p38 MAPK [29] (Table 1). In a recent study by Kim et al. [30] it was discovered that LysRS can induce cancer cell migration through interaction with the 67 kDa laminin receptor (67LR), a further component of the translational machinery that is converted from ribosomal subunit p40 upon laminin signaling. LysRS was shown to undergo phosphorylation at Tyr52 by p38 MAPK, causing its release from the MSC and membranal translocation. There, LysRS stabilized 67LR by inhibiting its ubiquitin-dependent degradation, thereby enhancing laminin-induced cell migration. These findings suggest an important role for LysRS in laminin-dependent cancer cell migration and metastasis.

LysRS and cell migration Several human ARSs have been shown to act as cytokinelike molecules. Tyrosyl-tRNA synthetase is secreted and

The role of LysRS in HIV infection Retroviruses, such as HIV-1, initiate the reverse transcription of their viral RNA genome from the 30 end of a host

Table 1. The various roles of lysyl-tRNA synthetase in biological processes and pathology Process Protein translation Regulation of transcription Cell migration

HIV infection

Charcot–Marie–Tooth (CMT) disease Death signaling Amyotrophic lateral sclerosis (ALS)

728

Role of LysRS Aminoacylation of tRNALys with the amino acid lysine. Production of the signaling molecule diadenosine tetraphosphate (Ap4A). Secreted as a cytokine-like molecule which induces macrophage migration and TNF-a secretion. Stabilizes the 67 kDa laminin receptor by inhibiting its ubiquitin-dependent degradation, thereby enhancing laminin-induced cell migration in cancer cells. The selective incorporation of tRNALys isoacceptors, which serve as primers for viral reverse transcription, depends on specific interaction between Gag and LysRS and its incorporation into the virions. Loss of function mutations in the catalytic domain of LysRS were found in CMT patients. Facilitates the translocation and surface exposure of calreticulin (CRT), activating the CRT pathway which leads to death of stressed cells. Mitochondrial LysRS binds to mutant superoxide dismutase 1 (SOD1), forming protein aggregates that damage mitochondrial activity and lead to disease onset.

Refs [1–6] [21,22] [29] [30] [36–43]

[45] [50] [53]

Review cellular tRNA [31]. Human tRNALys3 serves as the primer for HIV-1 reverse transcription and is selectively packaged into the virions along with the other major tRNALys isoacceptors in the cell, tRNALys1 and tRNALys2 [32]. Upon infection, the single-stranded viral RNA undergoes reverse transcription into double-stranded proviral DNA, which is integrated into host cell DNA. Before viral budding, the newly synthesized viral RNA, together with the viral precursor proteins Gag and GagPol, forms a cytoplasmic nucleoprotein complex into which tRNALys3 is recruited, and this complex also includes human LysRS [33,34]. The relative level of packaged tRNALys3 correlates with the levels of packaged LysRS and virus infectivity. Knockdown of cytoplasmic LysRS reduces the amount of tRNALys in the virus and subsequently reduces virus infectivity, whereas overexpression of LysRS causes the opposite effect [35]. The selective incorporation of tRNALys isoacceptors appears to be due to a specific interaction between Gag and LysRS [36,37]. It was recently shown that GagPol is also required for binding to tRNALys and facilitates its incorporation [38,39], and an interaction between LysRS and Pol has been proposed [40]. In a recent study by Jones and colleagues, it was shown that LysRS specifically binds to a tRNA anticodon-like element in the HIV-1 genome, which mimics the anticodon loop of tRNALys and is located proximal to the primer binding site, to which tRNALys3 binds [41]. It is, however, yet to be determined which variant of LysRS is required for HIV production. Mirande and colleagues suggested that it might be the mitochondrial LysRS [39], whereas Kleiman and colleagues suggested that it may be the newly synthesized protein [42]. A third possibility, that modified LysRS is released from the MSC, has yet to be assessed experimentally. Importantly, Musier-Forsyth and colleagues have recently reported the isolation of a cyclic peptide that can inhibit the binding of LysRS to HIV capsid protein, and thus has the potential to provide the basis of new unique anti-HIV agents [43]. Collectively, these observations provide an insight into the mechanism by which viruses such as HIV utilize the translation machinery components of the host in their replication cycle. The potential role played by LysRS in peripheral neuropathy Charcot–Marie–Tooth (CMT) disease represents a genetically and clinically heterogeneous group of peripheral neuropathies, with a prevalence of 1 in 2500 individuals. Mutations in four genes encoding ARSs have been implicated in CMT disease characterized by an axonal pathology: GlyRS (GARS), TyrRS (YARS), AlaRS (AARS), and LysRS (KARS) [44]. In the case of LysRS, a sequencingbased mutation screen in 355 patients with a CMT phenotype revealed two mutant variants in this gene leading to a Leu133His substitution and the frameshift Tyr173SerfsX7. Functional aminoacylation analyses revealed that p.Leu133His severely affects enzyme activity, whereas p.Tyr173SerfsX7 leads to deletion of the catalytic domain [45]. Although the molecular pathology of axonopathy associated with ARS mutations remains unclear, loss of enzymatic function may play a role in disease onset. Neurons with long axons are more susceptible to tRNA


charging deficits [46], suggesting that LysRS, together with several other ARSs, are candidate markers for CMT disease. This provides an opportunity to diagnose a possible disease onset by considering mutations in ARSs as possible markers. LysRS and cell death signaling In response to some anticancer therapeutics such as chemotherapies and ionizing irradiation, tumor cells can activate a complex pathway that leads to the translocation of calreticulin (CRT) from the lumen of the endoplasmic reticulum to the surface of the plasma membrane [47]. There, CRT facilitates the engulfment of cell-associated antigens by dendritic cells, which are optimal antigen presenters and are required for launching cellular immune responses [47,48]. The CRT exposure pathway can be subverted by multiple viruses [49], suggesting that infectious pathogens may suppress this pathway to evade the immune response. Interestingly, it was found that LysRS translocates to the surface of cells that have been treated with immunogenic cell death inducers. Knockdown experiments show LysRS to be indispensable for the translocation and surface exposure of CRT. Both LysRS and CRT colocalize at the surface of stressed cells within or in close proximity to microdomains that incorporate cholera toxin B, and hence constitute bona fide lipid rafts [50]. The immune response to chemotherapies may have a crucial role in mediating their long-term effects as cancer therapeutics and thus seem to be of major importance. Therefore, the possible role of LysRS in this pathway may be of great clinical interest. Further studies of the role of LysRS under these circumstances will probably demand analysis of specific post-translational changes to be able to prove its importance under these circumstances, as has been shown for the Ap4A signaling pathway. Mitochondrial LysRS in ALS Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disorder of motor neurons that results in paralysis and death within 5 years of diagnosis. Approximately 10% of ALS cases are inherited, of which 20% are associated with mutations in the Cu,Zn-superoxide dismutase, SOD1, a free radical scavenging enzyme [51]. Because many SOD1 mutations do not affect enzymatic activity and the disease has an autosomal dominant transmission, the associated pathological condition is believed to be a result of a toxic gain of function of the mutant protein. SOD1 is abundantly expressed in the cytosol, but a proportion of mutant SOD1 is also associated with mitochondria, where its aggregation could have pathological consequences [52]. Transgenic mice expressing mutant human SOD1 (hSOD1) develop mitochondrial degeneration in motor neurons [51]. A study by Kawamata and colleagues discovered that mutant hSOD1 interacts specifically with mitochondrial LysRS whereas wild type hSOD1 was unable to bind to LysRS. These interactions resulted in the formation of high molecular weight protein aggregates that correlate with impaired mtDNA-encoded protein synthesis, mitochondrial morphological abnormalities, and decreased cell survival [53]. Thus, although it is not mutated itself, mitochondrial LysRS plays a role in 729

Review ALS simply by being able to bind mutant SOD1, forming protein aggregates that damage mitochondrial activity and lead to disease onset. Concluding remarks LysRS serves as an example of what is now becoming clear: the idea of one gene–one protein–one function has been extended by the knowledge that many proteins have multiple functions that can be either vital or harmful for the cell and the organism. The fact that one protein can act in multiple distinct cellular processes and compartments such as translation, transcription, cell signaling, and apoptosis is both fascinating and challenging. The existence of such ‘moonlighting’ functions of LysRS complicates efforts to understand metabolic and regulatory networks, as well as physiological and pathological processes in organisms. The LysRS–Ap4A signaling pathway was first discovered in mast cells, but was later shown to be activated by b-adrenergic signaling in cardiomyocytes [14]. Moreover, not only is MITF transcriptional activity activated by LysRS-produced Ap4A, but another transcription factor, USF2 (upstream transcription factor 2), was found to be regulated in the same manner [18]. These findings suggest that the LysRS–Ap4A pathway is a general pathway that exists in many cell types, responds to various stimuli, and regulates the transcription of numerous genes. Knowing that LysRS is phosphorylated on Ser207 in a MAPK-dependent manner, and activates MITF, a well-known melanoma oncogene, LysRS activation may be a key process in malignant cell transformation and growth, and therefore a possible therapeutic target. Because non-canonical functions can play important roles in disease processes, an improved understanding of non-canonical functions of LysRS will provide new opportunities for therapeutic efforts to target specifically a function involved in pathology while sparing physiologically important functions. References 1 Schimmel, P. (1987) Aminoacyl tRNA synthetases: general scheme of structure–function relationships in the polypeptides and recognition of transfer RNAs. Annu. Rev. Biochem. 56, 125–158 2 Kisselev, L.L. and Favorova, O.O. (1974) Aminoacyl-tRNA synthetases: some recent results and achievements. Adv. Enzymol. Relat. Areas Mol. Biol. 40, 141–238 3 Schafer, K.P. and Soll, D. (1974) New aspects in tRNA biosynthesis. Biochimie 56, 795–804 4 Deutscher, M.P. (1984) Processing of tRNA in prokaryotes and eukaryotes. CRC Crit. Rev. Biochem. 17, 45–71 5 Kellermann, O. et al. (1982) Macromolecular complexes from sheep and rabbit containing seven aminoacyl-tRNA synthetases. I. Species specificity of the polypeptide composition. J. Biol. Chem. 257, 11041–11048 6 Guo, M. et al. (2010) New functions of aminoacyl-tRNA synthetases beyond translation. Nat. Rev. Mol. Cell Biol. 11, 668–674 7 Yao, P. and Fox, P.L. (2013) Aminoacyl-tRNA synthetases in medicine and disease. EMBO Mol. Med. 5, 332–343 8 Kim, S. et al. (2011) Aminoacyl-tRNA synthetases and tumorigenesis: more than housekeeping. Nat. Rev. Cancer 11, 708–718 9 Antonellis, A. and Green, E.D. (2008) The role of aminoacyl-tRNA synthetases in genetic diseases. Annu. Rev. Genomics Hum. Genet. 9, 87–107 10 Kim, Y.W. et al. (2012) Cancer association study of aminoacyl-tRNA synthetase signaling network in glioblastoma. PLoS ONE 7, e40960 11 Nechushtan, H. et al. (2009) Chapter 1. The physiological role of lysyl tRNA synthetase in the immune system. Adv. Immunol. 103, 1–27 730


12 Hemesath, T.J. et al. (1994) Microphthalmia, a critical factor in melanocyte development, defines a discrete transcription factor family. Genes Dev. 8, 2770–2780 13 Razin, E. et al. (1999) Suppression of microphthalmia transcriptional activity by its association with protein kinase C-interacting protein 1 in mast cells. J. Biol. Chem. 274, 34272–34276 14 Carmi-Levy, I. et al. (2008) Diadenosine tetraphosphate hydrolase is part of the transcriptional regulation network in immunologically activated mast cells. Mol. Cell. Biol. 28, 5777–5784 15 Lee, Y.N. et al. (2004) The function of lysyl-tRNA synthetase and Ap4A as signaling regulators of MITF activity in FcepsilonRI-activated mast cells. Immunity 20, 145–151 16 Nechushtan, H. and Razin, E. (2002) The function of MITF and associated proteins in mast cells. Mol. Immunol. 38, 1177–1180 17 Sonnenblick, A. et al. (2005) Immunological trigger of mast cells by monomeric IgE: effect on microphthalmia transcription factor, STAT3 network of interactions. J. Immunol. 175, 1450–1455 18 Lee, Y.N. and Razin, E. (2005) Nonconventional involvement of LysRS in the molecular mechanism of USF2 transcriptional activity in FcepsilonRI-activated mast cells. Mol. Cell. Biol. 25, 8904–8912 19 Nechushtan, H. and Razin, E. (1998) Deciphering the early-response transcription factor networks in mast cells. Immunol. Today 19, 441–444 20 Carmi-Levy, I. et al. (2011) Importin beta plays an essential role in the regulation of the LysRS-Ap(4)A pathway in immunologically activated mast cells. Mol. Cell. Biol. 31, 2111–2121 21 Ofir-Birin, Y. et al. (2013) Structural switch of lysyl-tRNA synthetase between translation and transcription. Mol. Cell 49, 30–42 22 Yannay-Cohen, N. et al. (2009) LysRS serves as a key signaling molecule in the immune response by regulating gene expression. Mol. Cell 34, 603–611 23 Randerath, K. et al. (1966) Isolation and characterization of dinucleoside tetra- and tri-phosphates formed in the presence of lysyl-sRNA synthetase. Biochem. Biophys. Res. Commun. 24, 98–105 24 Zamecnik, P.C. et al. (1966) Enzymatic synthesis of diadenosine tetraphosphate and diadenosine triphosphate with a purified lysylsRNA synthetase. Biochem. Biophys. Res. Commun. 24, 91–97 25 Yang, D.C. et al. (1985) Multienzyme complexes of mammalian aminoacyl-tRNA synthetases. Curr. Top. Cell. Regul. 26, 325–335 26 Wahab, S.Z. and Yang, D.C. (1985) Synthesis of diadenosine 50 ,5000 P1,P4-tetraphosphate by lysyl-tRNA synthetase and a multienzyme complex of aminoacyl-tRNA synthetases from rat liver. J. Biol. Chem. 260, 5286–5289 27 Wakasugi, K. and Schimmel, P. (1999) Two distinct cytokines released from a human aminoacyl-tRNA synthetase. Science 284, 147–151 28 Liu, J. et al. (2004) A new gamma-interferon-inducible promoter and splice variants of an anti-angiogenic human tRNA synthetase. Nucleic Acids Res. 32, 719–727 29 Park, S.G. et al. (2005) Human lysyl-tRNA synthetase is secreted to trigger proinflammatory response. Proc. Natl. Acad. Sci. U.S.A. 102, 6356–6361 30 Kim, D.G. et al. (2012) Interaction of two translational components, lysyl-tRNA synthetase and p40/37LRP, in plasma membrane promotes laminin-dependent cell migration. FASEB J. 26, 4142–4159 31 Bonhoeffer, S. et al. (1997) Human immunodeficiency virus drug therapy and virus load. J. Virol. 71, 3275–3278 32 Jiang, M. et al. (1993) Identification of tRNAs incorporated into wildtype and mutant human immunodeficiency virus type 1. J. Virol. 67, 3246–3253 33 Huang, M. and Martin, M.A. (1997) Incorporation of Pr160(gag–pol) into virus particles requires the presence of both the major homology region and adjacent C-terminal capsid sequences within the Gag–Pol polyprotein. J. Virol. 71, 4472–4478 34 Kleiman, L. et al. (2010) Formation of the tRNALys packaging complex in HIV-1. FEBS Lett. 584, 359–365 35 Guo, F. et al. (2005) Inhibition of cellular HIV-1 protease activity by lysyl-tRNA synthetase. J. Biol. Chem. 280, 26018–26023 36 Javanbakht, H. et al. (2003) The interaction between HIV-1 Gag and human lysyl-tRNA synthetase during viral assembly. J. Biol. Chem. 278, 27644–27651 37 Kovaleski, B.J. et al. (2007) Critical role of helix 4 of HIV-1 capsid C-terminal domain in interactions with human lysyl-tRNA synthetase. J. Biol. Chem. 282, 32274–32279

Review 38 Khorchid, A. et al. (2000) Sequences within Pr160gag–pol affecting the selective packaging of primer tRNA(Lys3) into HIV-1. J. Mol. Biol. 299, 17–26 39 Kobbi, L. et al. (2011) Association of mitochondrial Lysyl-tRNA synthetase with HIV-1 GagPol involves catalytic domain of the synthetase and transframe and integrase domains of Pol. J. Mol. Biol. 410, 875–886 40 Saadatmand, J. et al. (2008) Interactions of reverse transcriptase sequences in Pol with Gag and LysRS in the HIV-1 tRNALys3 packaging/annealing complex. Virology 380, 109–117 41 Jones, C.P. et al. (2013) Molecular mimicry of human tRNALys anticodon domain by HIV-1 RNA genome facilitates tRNA primer annealing. RNA 19, 219–229 42 Halwani, R. et al. (2004) Cellular distribution of Lysyl-tRNA synthetase and its interaction with Gag during human immunodeficiency virus type 1 assembly. J. Virol. 78, 7553–7564 43 Dewan, V. et al. (2012) Cyclic peptide inhibitors of HIV-1 capsid– human lysyl-tRNA synthetase interaction. ACS Chem. Biol. 7, 761–769 44 Antonellis, A. et al. (2003) Glycyl tRNA synthetase mutations in Charcot–Marie–Tooth disease type 2D and distal spinal muscular atrophy type V. Am. J. Hum. Genet. 72, 1293–1299 45 McLaughlin, H.M. et al. (2010) Compound heterozygosity for loss-offunction lysyl-tRNA synthetase mutations in a patient with peripheral neuropathy. Am. J. Hum. Genet. 87, 560–566


46 Antonellis, A. et al. (2006) Functional analyses of glycyl-tRNA synthetase mutations suggest a key role for tRNA-charging enzymes in peripheral axons. J. Neurosci. 26, 10397–10406 47 Obeid, M. et al. (2007) Calreticulin exposure dictates the immunogenicity of cancer cell death. Nat. Med. 13, 54–61 48 Martins, I. et al. (2010) Surface-exposed calreticulin in the interaction between dying cells and phagocytes. Ann. N. Y. Acad. Sci. 1209, 77–82 49 Kepp, O. et al. (2009) Viral subversion of immunogenic cell death. Cell Cycle 8, 860–869 50 Kepp, O. et al. (2010) Lysyl tRNA synthetase is required for the translocation of calreticulin to the cell surface in immunogenic death. Cell Cycle 9, 3072–3077 51 Wong, P.C. et al. (1995) An adverse property of a familial ALS-linked SOD1 mutation causes motor neuron disease characterized by vacuolar degeneration of mitochondria. Neuron 14, 1105–1116 52 Deng, H.X. et al. (2006) Conversion to the amyotrophic lateral sclerosis phenotype is associated with intermolecular linked insoluble aggregates of SOD1 in mitochondria. Proc. Natl. Acad. Sci. U.S.A. 103, 7142–7147 53 Kawamata, H. et al. (2008) Lysyl-tRNA synthetase is a target for mutant SOD1 toxicity in mitochondria. J. Biol. Chem. 283, 28321–28328

731

Review

Developing epigenetic diagnostics and therapeutics for brain disorders Irfan A. Qureshi1,2,3,6 and Mark F. Mehler1,2,3,4,5,6,7,8,9,10 1

Roslyn and Leslie Goldstein Laboratory for Stem Cell Biology and Regenerative Medicine, Albert Einstein College of Medicine, Bronx, NY 10461, USA 2 Institute for Brain Disorders and Neural Regeneration, Albert Einstein College of Medicine, Bronx, NY 10461, USA 3 Department of Neurology, Albert Einstein College of Medicine, Bronx, NY 10461, USA 4 Department of Neuroscience, Albert Einstein College of Medicine, Bronx, NY 10461, USA 5 Department of Psychiatry and Behavioral Sciences, Albert Einstein College of Medicine, Bronx, NY 10461, USA 6 Rose F. Kennedy Center for Research on Intellectual and Developmental Disabilities, Albert Einstein College of Medicine, Bronx, NY 10461, USA 7 Einstein Cancer Center, Albert Einstein College of Medicine, Bronx, NY 10461, USA 8 Ruth S. and David L. Gottesman Stem Cell Institute, Albert Einstein College of Medicine, Bronx, NY 10461, USA 9 Center for Epigenomics, Albert Einstein College of Medicine, Bronx, NY 10461, USA 10 Institute for Aging Research, Albert Einstein College of Medicine, Bronx, NY 10461, USA

Perturbations in epigenetic mechanisms have emerged as cardinal features in the molecular pathology of major classes of brain disorders. We therefore highlight evidence which suggests that specific epigenetic signatures measurable in central – and possibly even in peripheral tissues – have significant value as translatable biomarkers for screening, early diagnosis, and prognostication; developing molecularly targeted medicines; and monitoring disease progression and treatment responses. We also draw attention to existing and novel therapeutic approaches directed at epigenetic factors and mechanisms, including strategies for modulating enzymes that write and erase DNA methylation and histone/chromatin marks; protein–protein interactions responsible for reading epigenetic marks; and non-coding RNA pathways. Epigenetics and epigenetic medicine Although first-generation epigenetic medicines for cancer are already FDA-approved and increasingly becoming available in the marketplace, the field of epigenetics is evolving rapidly and promises to deliver advanced technologies with more widespread applications, including novel diagnostic and therapeutic modalities for brain disorders. Basic neuroscience research is revealing that epigenetic processes play leading roles in generating the extraordinary structural and functional complexity of the nervous system. Epigenetic factors and mechanisms orchestrate brain development, adult neurogenesis, synCorresponding author: Mehler, M.F. ([email protected]). Keywords: bromodomain; epigenomic; exosome; glioma; histone deacetylase; long non-coding RNA; microRNA. 1471-4914/$ – see front matter ß 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.molmed.2013.09.003

732


aptic plasticity, stress responses, and aging, and the transgenerational inheritance of cognitive and behavioral phenotypes [1]. Translational research efforts are similarly demonstrating how epigenetic mechanisms – and their deregulation – contribute to nervous system disease pathogenesis. Furthermore, targeting epigenetic processes in disease models shows the potential to dramatically reduce pathology and relieve symptoms, including mitigating neurodegeneration, promoting neural regeneration, and restoring cognitive functions [2–5]. Characterizing epigenetic and genome-wide epigenomic profiles and modulating epigenetic factors, therefore, represent novel and powerful paradigms for identifying and monitoring how

Glossary DNA methylation: the formation of 5-methylcytosine catalyzed by DNA methyltransferase enzymes that transfer methyl groups from S-adenosylmethionine to cytosine residues, which often occurs in gene regulatory regions and is associated with transcriptional repression. Exosomes: membrane-bound microvesicles secreted by various cell types into different bodily fluids that contain signaling molecules, including non-coding RNAs, involved in intercellular communication. Glioma CpG island methylator phenotype (G-CIMP): characteristic profile of DNA methylation alterations found in gene promoter regions in glioma specimens that define a subset of tumors with distinct genetic, pathological, and clinical features. Histone modifying enzymes: complementary families of enzymes, including histone acetyltransferases/deacetylases and histone methyltransferases/demethylases, responsible for ‘writing’ and ‘erasing’ histone protein posttranslational modifications. Locked nucleic acids (LNAs): chemically modified nucleic acids that are locked or restricted in terms of their conformation, enhancing their stability, target specificity, and pharmocokinetic profiles. Long ncRNAs (lncRNAs): highly versatile non-coding RNA molecules greater than 200 nucleotides in length that have a broad range of regulatory, structural, and other emerging biological roles. MicroRNAs (miRNAs): non-coding RNAs, 20–23 nucleotides in length, which participate in post-transcriptional regulation of target mRNAs through RNA interference pathways. RE1-silencing transcription factor/neuron-restrictive silencer factor (REST/ NRSF): a master transcriptional and epigenetic regulatory factor within the nervous system, implicated in the pathogenesis of many brain diseases, that binds to the RE1 motif associated with many neural genes and non-coding RNAs.

Review Table 1. Key epigenetic factors DNA methyltransferase (DNMT) enzymes Methyl-CpG-binding domain (MBD) proteins DNA excision repair enzymes Gadd45 enzymes Ten–eleven translocation enzymes Histone modifying enzymes Polycomb group (PcG) and Trithorax group (TrxG) proteins RE1-silencing transcription factor/neuron-restrictive silencer factor (REST/NRSF) MicroRNAs (miRNAs) Long non-coding RNAs (lncRNAs)

nervous system diseases unfold and for halting, preventing, or even reversing them. In this integrated overview, we highlight opportunities and challenges for developing these clinical applications. Examination of aberrant epigenetic profiles in nervous system diseases The principal epigenetic factors mediate DNA methylation (see Glossary) and hydroxymethylation, histone protein and chromatin modifications, and non-coding RNA (ncRNA) deployment (Table 1; Table S1 in the supplementary material online) [1]. These dynamic and environmentally responsive epigenetic processes collectively regulate the tissue- and cell-specific execution of genomic programs, including gene transcription; post-transcriptional RNA processing, transport, and translation; X-chromosome inactivation; genomic imprinting; gene dosage; and maintenance of genomic integrity. Accumulating evidence demonstrates that genetic variation and functional abnormalities in epigenetic enzymes and related factors modify nervous system disease risk, onset, and progression. Epigenetic mechanisms regulate (and are regulated by) disease-associated genes and pathways. Accordingly, epigenetic profiles are ubiquitously abnormal in central and peripheral nervous system disease patient-derived tissues. Here, we draw attention to salient examples linking these signatures with clinicopathological features and relevant outcomes. DNA methylation DNA methylation profiles have been interrogated most extensively in cancer, and in brain tumors more so than in other nervous system diseases. These approaches unequivocally provide diagnostic and prognostic information and can guide therapeutic decision-making. One key study utilizing Cancer Genome Atlas data revealed a DNA methylation pattern in adult glioblastoma multiforme (GBM) specimens referred to as the glioma CpG island methylator phenotype (G-CIMP) [6]. This signature defines a molecular subgroup of GBM associated with secondary or recurrent tumors, isocitrate dehydrogenase 1 (IDH1) mutations, a distinct profile of copy number variation, and a specific proneural gene expression pattern. Clinically, GBM patients with G-CIMP-positive tumors are younger at diagnosis and have a more favorable prognosis (longer median survival) than those with GCIMP-negative tumors. In addition, it was recently dem-


onstrated that mutant IDH1 produces an oncometabolite, which deregulates DNA methylation and generates the GCIMP, uncovering a causal link between genetic and epigenetic abnormalities in GBM [7]. Another study integrated global DNA methylation profiles with genomic and transcriptomic data from both pediatric and adult GBM specimens and identified six GBM subgroups associated with distinct clinical features, including patient age, tumor location, and overall survival [8]. A related study showed that 5-hydroxymethylcytosine levels in gliomas are inversely correlated with tumor grade and survival [9]. Together, these interesting observations imply that DNA (hydroxy)methylation profiles provide diagnostic and prognostic data. In terms of management, a seminal study published in 2005 reported that O6-methylguanine-DNA methyltransferase (MGMT) gene promoter methylation status in GBM specimens determines whether adding the alkylating agent, temozolomide (TMZ), is beneficial for patients undergoing radiotherapy [10]. Large randomized trials have since confirmed that MGMT methylation is associated with treatment responsiveness and longer survival [11,12]. This advantage is conferred because methylation is associated with lower levels of MGMT, a DNA repair enzyme that mitigates the effects of TMZ. Intriguingly, MGMT methylation status in tumor-derived tissue and in cell free circulating DNA in serum is highly concordant [13]. This finding suggests that serum MGMT methylation status can predict the efficacy of TMZ when tumor tissue is unavailable and, further, that there is active central–peripheral signaling in GBM coordinating DNA methylation states in the tumor and in serum. DNA methylation also provides clinically relevant data for solid tumors that metastasize to the brain. Studies of lung, breast, and renal cell carcinomas, as well as melanomas, suggest that the methylation status of MGMT and other genes influences metastatic potential, response to TMZ, and relapse rate [14–17]. Emerging data reveal that, in addition to brain tumors, every major class of nervous system disease harbors either locus-specific or genome-wide alterations in DNA methylation [1]. In some cases, these abnormalities are present in disease-specific neural cell types and tissues and are clearly linked to pathogenic processes, particularly in certain neurodevelopmental and neurodegenerative disorders. For other findings, such as DNA methylation abnormalities found in peripheral tissues, the mechanistic significance is unclear. At select loci, it appears that DNA methylation profiles in readily accessible peripheral tissues reflect those in brain directly. Alternatively, they can potentially have utility as indirect or surrogate markers. For example, monoamine oxidase A (MAOA) is an enzyme critical for neurotransmitter metabolism that exhibits variability in expression in the brain. One study found that MAOA gene promoter methylation in white blood cells predicts MAOA levels present in the brain [18]. Another study identified a robust association between aging-related DNA methylation in multiple tissues, including the blood and the brain [19]. In addition, a large case–control study demonstrated that DNA methylation profiles at the frataxin gene locus in peripheral blood mononuclear cells and buccal cells from Friedreich’s ataxia (FRDA) patients cor733

Review relate with age of onset of symptoms and clinical disease severity [20]. Also, population-based studies have reported that methylation levels of repetitive elements (i.e., Alu and LINE-1) – de facto markers of global DNA methylation – in blood correlate with the risk for Alzheimer’s disease (AD) [21] and for stroke and related diseases [22]. Histone and chromatin modifications Impaired chromatin regulation is directly responsible for the pathogenesis of a spectrum of brain diseases. Indeed, there is a rapidly increasing inventory of mutations in genes encoding histones, histone modifying enzymes, and other chromatin-related factors known to cause forms of syndromic and non-syndromic intellectual and developmental disability, primary brain tumors, and other disorders [1,8,23–27]. However, the potential array of disruptions in histone modification and chromatin states, which result from these mutations and from additional defects in chromatin regulators found in other nervous system diseases, is much more diverse and multifaceted and, thus, less well characterized than DNA methylation abnormalities. It is nevertheless clear that these aberrations are also present in both central and peripheral tissues and are associated with specific clinical phenotypes. For example, one study interrogated global levels of histone (H) 3 lysine (K) 18 (H3K18) and H3K9 acetylation, H3K4 dimethylation, and H4K20 trimethylation along with other characteristics in glioma specimens and identified subgroups in which specific histone modification patterns predict progression-free and overall survival [28]. In addition, disruptions in histone modification and chromatin states are prominent in neuropsychiatric disorders, such as depression, psychosis, and addiction [29,30]. Although these have been investigated primarily in animal models and in postmortem neuropathological specimens, preliminary data suggest that profiling peripheral tissues is informative. One study reported that levels of H3K9 dimethylation are elevated in lymphocytes of schizophrenic patients and correlate with the age of onset [31]. Also, the response to therapy can be mediated by specific histone modifications. A recent study showed that atypical antipsychotic agents decrease histone acetylation at the metabotropic glutamate 2 receptor gene promoter in the mouse and human frontal cortex [32]. Similarly, in Huntington’s disease (HD), genome-wide histone acetylation and methylation, histone variant expression, and the function of the master epigenetic regulatory factor, RE1-silencing transcription factor/neuron-restrictive silencer factor (REST/ NRSF), are all aberrant in animal models of HD. These are consistent with abnormalities present in HD patient-derived tissues, including striatal and cortical brain regions and blood, and potentially have value for diagnosis, prognostication, and monitoring therapy [33,34]. Non-coding RNAs Many investigators have characterized microRNAs (miRNAs) in brain diseases, and a small, but increasing, number has begun to interrogate the roles played by other classes of ncRNAs [35,36]. Studying model systems demonstrates that ncRNA networks are highly integrated with disease-linked genes and pathways and that ncRNAs can 734


directly modulate disease pathogenesis. For example, in Drosophila, miRNA-mediated RNA-induced silencing complex (miRNA–RISC) pathways are perturbed by Parkinson’s disease (PD)-causing mutations in leucine-rich repeat kinase 2; and resulting abnormalities in let-7 and mir-184 activity lead to degeneration of dopaminergic neurons [37]. mir-34 is expressed in adult Drosophila brain and is significantly upregulated with aging; and loss of mir-34 leads to accelerated brain aging and neurodegeneration [4]. Levels of the skeletal muscle-specific miRNA, miR-206, are increased in amyotrophic lateral sclerosis (ALS) mouse models [38]. This miRNA mediates muscle–nerve crosstalk and slows the progression of ALS by enhancing regenerative responses at neuromuscular synapses. Analyzing ncRNA profiles in tissues from patients with nearly every class of nervous system disease reveals alterations that are consistent with those in model systems, linked to pathogenesis, parallel between central and peripheral tissues, and indicative of clinicopathological phenotypes. For example, miRNAs, piwi-RNAs, and long ncRNAs (lncRNAs) are deregulated in the brain in animal models of stroke [39–42]. miRNA levels in the ischemic brain overlap with those in blood and have an evolving temporal profile. miRNA expression patterns in the blood of stroke patients correspond with these animal data, are suggestive of stroke mechanisms (i.e., large vessel, lacunar, or cardioembolic), and might provide a quantitative index for stroke severity and recovery potential. Expression patterns for other classes of ncRNAs have not been studied in stroke patients, with the exception of a single lncRNA, CDKN2B antisense RNA 1 (CDKN2B-AS1/ ANRIL). Levels of CDKN2B-AS1/ANRIL in human carotid atherosclerotic plaques and peripheral blood T lymphocytes are linked to rates of ischemic and hemorrhagic stroke. Intriguingly, polymorphisms in the genomic locus for this lncRNA (9p21) modify the risk of developing several diseases, including atherosclerosis, coronary artery disease, and intracranial aneurysms [43–45]. These observations suggest that ncRNAs might represent the long sought after peripheral biomarkers for stroke risk stratification, rapid diagnosis, mechanistic classification, and prognostication. Specific miRNA signatures predict time to relapse and overall survival in ependymomas [46]; leptomeningeal spread and responsiveness to chemotherapies in medulloblastomas [47]; event-free and overall survival and responsiveness to chemotherapies in neuroblastomas [48]; and recurrence rate in meningiomas [49]. Particular miRNA and lncRNA profiles in gliomas are associated with tumor grade, Karnofsky performance score, MGMT expression, radio- and chemo-sensitivities, comorbidities, and overall survival [50–55]. In addition to primary neuropathological specimens, cerebrospinal fluid (CSF) miRNA levels permit detection of GBM, differentiation of GBM from metastatic brain tumors, and monitoring of disease activity [56]. Particular miRNA signatures are also present in the blood of GBM patients, vary with treatment, and might be useful for monitoring recurrence [57]. In multiple sclerosis, miRNA profiles in white matter lesions discriminate between active and inactive lesions [58]. miRNA levels in blood differentiate between patients with a relapsing–re-

Review Box 1. Exosomes Exosomes are membrane-bound microvesicles derived from intracellular multivesicular bodies and the endosomal pathway that participate in local and systemic cell–cell communication. Exosomes are secreted by donor cells (e.g., neural, immune, and other cell types) into bodily fluids and release their contents into selectively targeted recipient cells [113,114]. Exosomes can transport cargo including ncRNAs and other molecules that are functional in recipient cells. For example, miRNAs transferred via exosomes repress their target genes in recipient cells [115]. In terms of pathological roles, exosomes are released by primary brain tumors, such as gliomas, internalized by normal cells, and implicated in modulating the tumor microenvironment to promote tumor invasion and angiogenesis and preventing immune responses [113,114]. Other classes of ncRNAs may similarly be found in exosomes and influence target cells. Indeed, exosomes harboring tumor-specific factors – mutant proteins, mutated and amplified oncogene sequences, and retrotransposable elements – are present in the serum, implicating them in mediating the cancer state systemically. Ongoing studies of exosome formation, contents, and delivery further suggest that the spreading of pathology in a spectrum of other nervous system diseases is associated with the release of these microvesicles. Therapeutic strategies aimed at exploiting exosomes are also being developed. Exosomes can be engineered to deliver cargo, including ncRNAs, into the brain through the periphery. Alternatively, endogenous exosomes can be harnessed to treat disease by activating endogenous immune responses. For example, one intriguing study reported the utilization of glioma-derived exosomes to generate CD8+ T cells with glioma-specific cytotoxic activity in vitro and suggested this stratagem for immunizing against gliomas [116].

mitting course and one that is progressive. miRNA expression also identifies patients treated with disease-modifying agents (e.g., glatiramer acetate and natalizumab) and those that are untreated. Similarly, miRNA expression in blood distinguishes between patients with PD and controls and medicated patients from non-medicated ones [59]. Carriers of huntingtin gene mutations exhibit high levels of miR-34b in plasma prior to the onset of symptoms [60]. Furthermore, patients with AD display increased expression of let-7 in CSF [61]. Extracellular let-7 provokes neurodegeneration by activating neuronal RNA-sensing Toll-like receptor 7 signaling and promotes the spread of pathology. Other miRNAs (i.e., miR-146a and miR-155) upregulated in AD patient CSF specifically have deleterious proinflammatory effects [62]. miRNAs, such as these, play a role in neuroimmune crosstalk locally and may do so systemically because they can be trafficked in blood and other fluids [63]. miRNAs and other ncRNAs probably mediate additional forms of central–peripheral communication, potentially accounting for their pervasive deregulation in peripheral tissues in brain diseases. Although ncRNA profiles can be sampled in readily accessible biological fluids, it is vital to understand how and why these factors might specifically reflect central pathological states. The recently recognized role of membrane-bound microvesicles, called exosomes (Box 1), in mediating intercellular communication provides some insight [64]. Not only are ncRNAs present intracellularly but, as highlighted above, they are also found extracellularly in biological fluids including CSF, blood (serum and plasma), lymphatics, urine, and saliva in health and disease [65– 70]. These extracellular ncRNAs are fairly resistant to


nucleases because they form stable complexes with proteins (i.e., Argonaute 2) or lipids (i.e., high-density lipoproteins), or alternatively they can be packaged into exosomes. Integrated epigenetic and epigenomic profiling There are still many unanswered questions regarding the causal relationship between epigenetic deregulation and nervous system disease pathogenesis and the biological significance of central and peripheral epigenetic profiles. Nevertheless, the range of observations cited above supports the advancement of epigenetic biomarker discovery, qualification, and validation for improving and personalizing nervous system disease screening, early diagnosis, and prognostication; developing molecularly targeted treatments; and monitoring progression and therapeutic responses. There exist significant challenges, however [71]. Even the most well-established epigenetic biomarker, MGMT methylation in gliomas, is not widely utilized. Persistent technical questions are one of the major obstacles that have prevented this and other epigenetic tests from being clinically adopted. Even leading next-generation sequencing platforms are relatively insensitive for measuring changes in epigenetic profiles in nervous system disorders, which are often subtle. For example, lncRNAs are low-abundance transcripts with differential expression in disease in the twofold range. However, rapid technological and methodological innovations are poised to overcome these issues by offering advantages in terms of sample preparation and quantity, speed, resolution, throughput, and cost-effectiveness. Not only do these emerging approaches permit the analysis of a single ‘candidate’ epigenetic modification but they also allow the characterization of highly integrated genome-wide epigenomic profiles utilizing single cells and molecules, by employing sophisticated technology platforms. Specific examples of such powerful tools and techniques that have recently been reported include the following: (i) a DNA array (fabricated via advanced soft lithography) for high-resolution methylation profiling of single DNA molecules [72]; (ii) a quantum dot-enabled electrophoretic mobility shift assay for high-resolution quantitative epigenetic analysis [73]; (iii) RNA aptamers – RNAs that bind with very high affinity and selectivity to a particular target – engineered to recognize molecules harboring certain epigenetic modifications [74]; (iv) a nanofluidic device for real-time multiplexed detection and automated sorting of individual molecules based on their epigenetic states [75]; and (v) a chemical sensor array that can robustly discriminate between complex profiles of histone modifications (e.g., unmethylated, mono-, di-, and tri-methylated lysines) at a single histone and across different sites [76]. In addition, novel approaches for targeting ncRNAs and microvesicles include (i) improved methods for miRNA isolation from blood [77]; (ii) single molecule nanopore platforms designed for quantitative measurement of ncRNA levels [78]; (iii) oligonucleotide molecular beacons and magnetic nanoparticles enabling the detection of ncRNAs, including those in live animals, by magnetic resonance imaging [79]; and (iv) microfluidic 735

Review


Integrated epigenomic profiling DNA methylaon/hydroxymethylaon Histone modificaons Higher-order chroman remodeling Non-coding RNA deployment RNA/DNA eding Retrotransposon regulaon Genomic imprinng X chromosome inacvaon RNA binding protein interacons RNA trafficking (exosomes/microvesicles) Cells/Tissues/Fluids

Central–peripheral Central-Peripheral communicaon Communicaon Neuroendocrine Neuroendocrine Immune surveillance surveillance Energy homeosta sis homeostasis Gut-micro biome Gut-microbiome Personalized omics profiling

Blood Cerebrospinal fluid Lymphacs Urine Saliva

Genome Transcriptome Proteome Metabolome Microbiome TRENDS in Molecular Medicine

Figure 1. Integrated epigenomic and personalized ‘omics’ profiling. Utilizing a variety of accessible body fluids and associated cell and tissue sources, combinatorial epigenomic and personalized ‘omics’ profiling are already providing physiological ‘readouts’ of dynamic multidimensional central–peripheral communications and an integrated systems biology analysis of nervous system functions in health and in various neurological disease states with an emerging spectrum of innovative diagnostic and therapeutic applications.

devices enabling the rapid isolation, sorting, and detection of microvesicles from small volume samples [80]. This integrated epigenomic profiling can be coupled with other ‘omics’ profiling approaches for the genome, transcriptome, RNA editome, proteome, metabolome, and microbiome yielding a ‘systems’ level view of pathology (along the spectrum from single cell to whole organism), and providing both mechanistic insights as well as combinatorial signatures with potential clinical applications (Figure 1) [81]. Evolving epigenetic therapeutic strategies Preclinical studies and clinical trials strongly suggest that agents targeting epigenetic factors and mechanisms, such as DNA methyltransferase (DNMT) and histone deacetylase (HDAC) enzyme inhibitors, have therapeutic efficacy in a very broad range of diseases, including brain disorders. However, the clinical utility of most first-generation epigenetic agents is limited by their lack of specificity for individual enzymes, toxicities, poor bioavailability, and other factors. As such, there is considerable enthusiasm for identifying and designing additional compounds that modulate targets across the spectrum of epigenetic writer, eraser, and reader classes and ncRNA pathways. These emerging agents have potential utility not only as biological probes for preclinical studies but also as lead compounds for optimization and clinical development. Here, 736

we highlight examples of these evolving strategies and their relevance for brain disorders. DNA methylation Compounds available for modulating DNA methylation pathways include those that increase the supply of methyl donors (e.g., folic acid, betaine, vitamin B12, and creatine), inhibit DNMT enzymes (e.g., 5-azacytidine, 5-aza-2-deoxycytidine, and zebularine), and inhibit DNA demethylation enzymes (e.g., gemcitabine). Interestingly, supplementing with methyl donors in Rett and Angelman syndromes – disease mechanism-based treatment approaches – shows trends towards clinical benefits, meriting further study [82,83]. Existing DNMT inhibitors have been evaluated in preclinical paradigms and modulate learning and memory, reward and addiction, ischemia, neurodegeneration, and epileptogenesis [84]. Thus, more efficient and specific DNMT inhibitors now being identified can potentially mitigate the pathogenesis and symptomatology in this wide range of disorders. Gemcitabine, an antimetabolite cancer therapeutic with neurotoxic effects, was recently found to inhibit Gadd45a-mediated DNA demethylation [85]. It is the first agent noted to have this ability and offers insights for developing drugs that target DNA demethylation enzymes. In addition, some commonly used drugs, such as hydralazine, procainamide, and valproic acid (VPA), also influence DNA methylation profiles, and these

Review


drugs are prime candidates for epigenome-based repurposing/repositioning [86–89]. For example, preliminary studies demonstrate that adjunctive treatment with VPA enhances the efficacy of TMZ in GBM, by influencing MGMT promoter methylation status [90]. Notably, VPA might exert these beneficial effects via other mechanisms, which include the ability to inhibit HDAC enzymes. Conversely, toxicities and adverse effects associated with these and other commonly used drugs might result from an unintended impact on epigenetic pathways.

complexes with other factors that fulfill these roles, efforts have focused on disrupting these protein–protein interactions. JQ1 is a first-in-class small molecule that selectively inhibits the interaction between bromodomain proteins and acetylated lysine residues [98] and shows promise for treating genetically diverse subgroups of GBM [99]. Emerging technologies for modulating protein–protein interactions, such as stapled peptides and other proteomimetic approaches, also hold significant promise in this regard [100].

Histone and chromatin modifications Compounds that modulate histone acetylation and methylation and other chromatin targets are increasingly becoming available. HDAC inhibitors are the most numerous among these. Their structural and mechanistic properties have been extensively studied, as have their context-specific biological functions, which include robust salutary effects in many different brain disease models. Notably, certain HDAC enzymes have roles in non-epigenetic pathways in the axon and mitochondria, and the mechanisms of action of HDAC inhibitors may include inhibiting these functions. Several HDAC inhibitors have advanced into clinical trials for nervous system disease indications, and more are likely to follow given the ongoing optimization of HDAC inhibitors, improving their selectivity for individual isoforms and pharmacokinetic and pharmacodynamic profiles. Some examples in clinical trials include (i) EVP-0334, a class I HDAC inhibitor being developed for AD (and other neurodegenerative disorders); (ii) SEN0014196 (selisistat), an HDAC inhibitor that targets SIRT1, being evaluated in HD; (iii) phenylbutyrate, a class I/II HDAC inhibitor, being studied in HD and ALS; and (iv) suberoylanilide hydroxamic acid (vorinostat) and the related compound panobinostat, class I/II HDAC inhibitors, being analyzed individually and as adjunctive agents for treating gliomas [91,92]. Of interest, emerging methodologies for analyzing histone acetylation levels in blood and other tissues are tools for monitoring and titrating therapy with these agents [93]. Efforts employing high-content and high-throughput screening of chemical libraries, structure-based drug design, and other methods are focused on identifying and designing small molecules that target additional families of epigenetic factors with ‘writer/eraser’ functions (e.g., protein methyltransferases/demethylases), as well as those with ‘reader’ domains (e.g., bromodomains) [94]. For example, one study reported the discovery of a small molecule, GSK126, which is a highly selective and potent inhibitor of the enhancer of zeste homolog 2 (EZH2) histone methyltransferase [95]. EZH2 is implicated in the development of many cancers, including lymphoma and GBM; and preliminary data suggest that GSK126 is effective in treating these disorders. A complementary study reported finding a specific small molecule inhibitor of the histone demethylases, jumonji domain containing 3, and lysinespecific demethylase 6A, which is effective in modulating acute inflammatory responses, such as those that occur after spinal cord injury and contribute to secondary damage [96,97]. In addition, because epigenetic enzymes often contain ‘reader’ domains or they are components of

Master epigenetic regulator: REST Forward-looking strategies can also be envisioned that target epigenetic factors with key roles in brain diseases, such as REST, in more complex and nuanced ways utilizing a variety of existing and novel approaches including small molecules, RNA interference (RNAi), decoy oligonucleotides, and synthetic peptide nucleic acid oligomers [101– 103]. These include modulating REST expression, REST alternative splicing, macromolecular complex assembly, REST interactions with RE1 binding sites, regulators of REST activity (i.e., double-stranded RE1/NRSE ncRNAs and the REST4 isoform) that mediate switching between REST transcriptional activator/repressor functions, interactions between REST and lncRNAs that mediate REST complex genomic site-specific deployment, REST cytoplasmic–nuclear shuttling, and REST post-transcriptional regulatory activity (Figure 2). Non-coding RNAs Considerable effort is now concentrated on developing ncRNA-based therapeutics and capitalizes on previously existing expertise in antisense oligonucleotides, RNAi, and related platforms. These approaches are most advanced for inhibiting pathogenic miRNA expression and function and employ antisense oligonucleotides, termed antimirs, engineered with various chemical modifications to increase their specificity, stability, nuclease resistance, and delivery across the blood–brain barrier [e.g., locked nucleic acids (LNAs) and antagomirs]. These molecules ameliorate nervous system pathology in preclinical models. For example, antagomir-mediated inhibition of miR-206, which is increased in human AD brains and targets brain-derived neurotrophic factor (BDNF), enhances BDNF levels, hippocampal synaptic density and neurogenesis, and memory in an AD mouse model [104]. Also, a recent study presented data suggesting that utilizing an antimiR directed at miR886-3p is a possible treatment for FRDA [105]. Alternative strategies to inhibit miRNAs include introducing oligonucleotides that interfere with miRNA–target mRNA interactions (i.e., miRNA ‘masks’ and ‘sponges’). Modulating multiple miRNAs simultaneously with a single therapeutic oligonucleotide has also been proposed. Related strategies for lncRNAs are also being developed [106]. Specifically, antagoNATs are oligonucleotides designed to target natural antisense transcripts (NATs) – lncRNAs in antisense genomic configurations relative to proteincoding genes – including those that regulate genes with key roles in the nervous system, such as BDNF, glial-derived neurotrophic factor, and ephrin receptor B2 [107]. Strategies to inhibit lncRNA functions with oligonucleotides 737

Review


(A) REST expression levels

(E) REST repressor/acvator switching REST4 dsNRSE

+ –

↓ TRF2, SCFβ-TrCP

+ –

↑ HAUSP, HIPPI

RE1

(F) REST–LncRNA interacons: genomic deployment

(B) REST alternave splicing RD1

DNA binding

RD2

N

C

IncRNA

↑ ↑

↑ REST4

SRRM4

Δ REST RE1

(C) REST complex assembly HDACs mSin3

MeCP2 LSD1 G9a Brg1 CoREST

(G) REST cytoplasmic–nuclear shuling

mH

– HAP1 RILP RE1

Dynacn (p150Glued)

(D) RE1 binding site interacons

(H) REST-mediated post-transcriponal regulaon MOR1 RE1 MOR1 mRNA RE1 decoy

RE1

↑ eIF4G-P

AAA...AAA

MOR1 Protein TRENDS in Molecular Medicine

Figure 2. Novel epigenetic therapeutic targeting strategies. Using a variety of existing and emerging approaches, it is possible to alter the functional roles of REST, a key transcriptional and epigenetic mediator of nervous system function in health and various disease states, by differentially modulating: (A) REST expression levels; (B) REST alternative splicing profiles; (C) REST macromolecular complex assembly; (D) REST/RE1 binding site interactions; (E) REST transcriptional activator/repressor activity and switching; (F) REST interactions with long non-coding RNAs that mediate REST genome-wide deployment; (G) REST cytoplasmic–nuclear shuttling; and (H) REST posttranscriptional regulatory processes. Abbreviations: Brg1, Brahma-related gene 1; CoREST, corepressor for element-1-silencing transcription factor; dsNRSE, doublestranded non-coding RNAs encoding the RE1 sequence; elF4G-P, elongation factor-4G, phosphorylated; G9a, histone H3-lysine 9 methyltransferase; HAP1, Huntingtinassociated protein 1; HAUSP, herpesvirus-associated ubiquitin specific protease; HDACs, histone deacetylases; HIPPI, Huntingtin interacting protein 1 (HIP1) protein interactor; mHtt, mutant Huntingtin; lncRNA, long non-coding RNA; LSD1, lysine-specific demethylase-1; MeCP2, methyl-CpG binding protein 2; MOR1, mu opioid receptor1; RD, repressor domain; RE1, repressor element-1; REST, RE1-silencing transcription factor; REST4, REST/NRSF truncated activator isoform 4; RILP, REST/NRSE-interacting LIM domain protein; SCF(bTrCP), Skp1–Cullin1–F-box b-transducin repeat-containing protein; mSin3, mammalian component of histone deacetylase complex; SRRM4, splicing factor-encoding gene Ser/Arg repetitive matrix 4; TRF2, telomeric repeat-binding protein 2.

(or small molecules) that interfere with sequence-specific and structural interactions between lncRNAs with other molecules (i.e., DNA, RNA, and proteins; e.g., chromatin remodeling complexes) are being considered. Although ncRNA-based therapeutics for brain disorders are in early stages of development, proof-of-principle is illustrated by the advancement of miravirsen, a LNA antisense oligonucleotide targeting miR-122, into Phase II clinical trials for hepatitis C (NCT01727934). Methods for introducing ncRNAs with salubrious effects are also being investigated. These include mimics or replacements for endogenous ncRNAs. The p137 RNA derived from cytomegalovirus is one particularly interesting example [108]. This ncRNA protects dopaminergic neurons from 738

cell death provoked by mitochondrial dysfunction in PD models. Intriguingly, fusing this ncRNA with rabies virus glycoprotein peptide, a powerful technique for drug delivery to the brain, permits peripheral intravenous administration of this agent overcoming one of the greatest challenges for any of these potential therapies [109]. Another approach for restoring beneficial ncRNAs, such as those with tumor suppressor properties, involves using other epigenetic agents (e.g., DNMT and HDAC inhibitors) that upregulate ncRNA expression. Alternative strategies for modulating ncRNA functions include the identification of small molecule agents that influence ncRNA biogenesis and effector pathways (e.g., inhibitors of miRNA precursor processing and RISC complex loading) [110–112].

Review Box 2. Outstanding questions Why do patients with central pathologies that are often brain region- and cell type-selective exhibit abnormal epigenetic profiles in peripheral tissues? Are these peripheral epigenetic signatures directly connected to disease mechanisms, indirect but potentially valuable markers thereof, or simply downstream phenomena that are neither sensitive nor specific? In addition to exosome-mediated communication, what (if any) other mechanisms might link epigenetic states in the brain to those in the periphery? Can these also be exploited for diagnostic or therapeutic applications? How can emerging technologies enabling high-resolution realtime in vivo interrogation of epigenetic and epigenomic states be used to (i) differentiate between pathogenic processes and those that are protective responses and (ii) develop corresponding clinically relevant and cost-effective assays? Can epigenetic therapeutics be designed, developed, and delivered into the brain and specific neural cell types not only in animal models but also in patients? What are the explicit barriers that must be overcome? Is the development of epigenetic drugs and associated companion diagnostics a biologically and commercially viable therapeutic strategy? What brain diseases and epigenetic targets might be most amenable to providing proof-of-concept for this approach? Can epigenetic agents be combined with each other or with other classes of drugs to more effectively address the multiple layers of molecular pathology present in complex brain disorders?

Concluding remarks We are at the vanguard of the era of epigenetic and epigenomic medicine, which is poised to revolutionize the diagnosis and treatment of nervous system diseases (Box 2). This paradigm shift is being propelled by basic and translational discovery efforts that have revealed how epigenetic changes mediate central pathology or indirectly reflect it in the periphery. Epigenetic diagnostic and therapeutic innovations are, in turn, being driven by ongoing technological progress (i) allowing higher resolution interrogation of epigenetic profiles and molecular imaging and (ii) promoting the discovery, design, and optimization of novel compounds that can modulate epigenetic pathways and their delivery into the nervous system. We foresee these epigenetic clinical applications evolving in concert with complementary diagnostic and therapeutic platforms (e.g., microfluidics, RNA aptamers, nanotechnologies, oligonucleotide-based strategies, immunotherapies, and cellular reprogramming and regenerative medicine) that are revolutionary in their own right. Acknowledgments We regret that space constraints have prevented the citation of many relevant and important references. M.F.M. is supported by grants from the National Institutes of Health (NS071571, HD071593, MH66290), as well as by the F.M. Kirby, Alpern Family, Mildred and Bernard H. Kayden and Roslyn and Leslie Goldstein Foundations.

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.molmed.2013.09.003. References 1 Mehler, M.F. (2008) Epigenetic principles and mechanisms underlying nervous system functions in health and disease. Prog. Neurobiol. 86, 305–341


2 Graff, J. et al. (2012) An epigenetic blockade of cognitive functions in the neurodegenerating brain. Nature 483, 222–226 3 Oliveira, A.M. et al. (2012) Rescue of aging-associated decline in Dnmt3a2 expression restores cognitive abilities. Nat. Neurosci. 15, 1111–1113 4 Liu, N. et al. (2012) The microRNA miR-34 modulates ageing and neurodegeneration in Drosophila. Nature 482, 519–523 5 Rivieccio, M.A. et al. (2009) HDAC6 is a target for protection and regeneration following injury in the nervous system. Proc. Natl. Acad. Sci. U.S.A. 106, 19599–19604 6 Noushmehr, H. et al. (2010) Identification of a CpG island methylator phenotype that defines a distinct subgroup of glioma. Cancer Cell 17, 510–522 7 Turcan, S. et al. (2012) IDH1 mutation is sufficient to establish the glioma hypermethylator phenotype. Nature 483, 479–483 8 Sturm, D. et al. (2012) Hotspot mutations in H3F3A and IDH1 define distinct epigenetic and biological subgroups of glioblastoma. Cancer Cell 22, 425–437 9 Orr, B.A. et al. (2012) Decreased 5-hydroxymethylcytosine is associated with neural progenitor phenotype in normal brain and shorter survival in malignant glioma. PLoS ONE 7, e41036 10 Hegi, M.E. et al. (2005) MGMT gene silencing and benefit from temozolomide in glioblastoma. N. Engl. J. Med. 352, 997–1003 11 Wick, W. et al. (2012) Temozolomide chemotherapy alone versus radiotherapy alone for malignant astrocytoma in the elderly: the NOA-08 randomised, phase 3 trial. Lancet Oncol. 13, 707–715 12 Malmstrom, A. et al. (2012) Temozolomide versus standard 6-week radiotherapy versus hypofractionated radiotherapy in patients older than 60 years with glioblastoma: the Nordic randomised, phase 3 trial. Lancet Oncol. 13, 916–926 13 Balana, C. et al. (2011) Tumour and serum MGMT promoter methylation and protein expression in glioblastoma patients. Clin. Transl. Oncol. 13, 677–685 14 Ingold, B. et al. (2009) Homogeneous MGMT immunoreactivity correlates with an unmethylated MGMT promoter status in brain metastases of various solid tumors. PLoS ONE 4, e4775 15 Hashimoto, K. et al. (2012) Methylation status of O6-methylguanineDNA-methyl transferase promoter region in non-small-cell lung cancer patients with brain metastasis. Clin. Transl. Oncol. 14, 31–35 16 Wu, P.F. et al. (2010) O(6)-Methylguanine-DNA methyltransferase expression and prognostic value in brain metastases of lung cancers. Lung Cancer 68, 484–490 17 Zhang, Y. et al. (2012) Loss of expression of the differentially expressed in adenocarcinoma of the lung (DAL-1) protein is associated with metastasis of non-small cell lung carcinoma cells. Tumour Biol. 33, 1915–1925 18 Shumay, E. et al. (2012) Evidence that the methylation state of the monoamine oxidase A (MAOA) gene predicts brain activity of MAO A enzyme in healthy men. Epigenetics 7, 1151–1160 19 Horvath, S. et al. (2012) Aging effects on DNA methylation modules in human brain and blood tissue. Genome Biol. 13, R97 20 Evans-Galea, M.V. et al. (2012) FXN methylation predicts expression and clinical outcome in Friedreich ataxia. Ann. Neurol. 71, 487–497 21 Bollati, V. et al. (2011) DNA methylation in repetitive elements and Alzheimer disease. Brain Behav. Immun. 25, 1078–1083 22 Baccarelli, A. et al. (2010) Ischemic heart disease and stroke in relation to blood DNA methylation. Epidemiology 21, 819–828 23 Robinson, G. et al. (2012) Novel mutations target distinct subgroups of medulloblastoma. Nature 488, 43–48 24 Sanchez-Mut, J.V. et al. (2012) Aberrant epigenetic landscape in intellectual disability. Prog. Brain Res. 197, 53–71 25 Wu, G. et al. (2012) Somatic histone H3 alterations in pediatric diffuse intrinsic pontine gliomas and non-brainstem glioblastomas. Nat. Genet. 44, 251–253 26 Schwartzentruber, J. et al. (2012) Driver mutations in histone H3.3 and chromatin remodelling genes in paediatric glioblastoma. Nature 482, 226–231 27 Chan, K.M. et al. (2013) The histone H3.3K27M mutation in pediatric glioma reprograms H3K27 methylation and gene expression. Genes Dev. 27, 985–990 28 Liu, B.L. et al. (2010) Global histone modification patterns as prognostic markers to classify glioma patients. Cancer Epidemiol. Biomarkers Prev. 19, 2888–2896 739

Review 29 Robison, A.J. and Nestler, E.J. (2011) Transcriptional and epigenetic mechanisms of addiction. Nat. Rev. Neurosci. 12, 623–637 30 Labrie, V. et al. (2012) Epigenetics of major psychosis: progress, problems and perspectives. Trends Genet. 28, 427–435 31 Gavin, D.P. et al. (2009) Dimethylated lysine 9 of histone 3 is elevated in schizophrenia and exhibits a divergent response to histone deacetylase inhibitors in lymphocyte cultures. J. Psychiatry Neurosci. 34, 232–237 32 Kurita, M. et al. (2012) HDAC2 regulates atypical antipsychotic responses through the modulation of mGlu2 promoter activity. Nat. Neurosci. 15, 1245–1254 33 Zuccato, C. et al. (2003) Huntingtin interacts with REST/NRSF to modulate the transcription of NRSE-controlled neuronal genes. Nat. Genet. 35, 76–83 34 Marullo, M. et al. (2010) Analysis of the repressor element-1 silencing transcription factor/neuron-restrictive silencer factor occupancy of non-neuronal genes in peripheral lymphocytes from patients with Huntington’s disease. Brain Pathol. 20, 96–105 35 Esteller, M. (2011) Non-coding RNAs in human disease. Nat. Rev. Genet. 12, 861–874 36 Faghihi, M.A. et al. (2008) Expression of a noncoding RNA is elevated in Alzheimer’s disease and drives rapid feed-forward regulation of bsecretase. Nat. Med. 14, 723–730 37 Gehrke, S. et al. (2010) Pathogenic LRRK2 negatively regulates microRNA-mediated translational repression. Nature 466, 637–641 38 Williams, A.H. et al. (2009) MicroRNA-206 delays ALS progression and promotes regeneration of neuromuscular synapses in mice. Science 326, 1549–1554 39 Liu, D.Z. et al. (2010) Brain and blood microRNA expression profiling of ischemic stroke, intracerebral hemorrhage, and kainate seizures. J. Cereb. Blood Flow Metab. 30, 92–101 40 Jeyaseelan, K. et al. (2008) MicroRNA expression in the blood and brain of rats subjected to transient focal ischemia by middle cerebral artery occlusion. Stroke 39, 959–966 41 Dharap, A. et al. (2011) Altered expression of PIWI RNA in the rat brain after transient focal ischemia. Stroke 42, 1105–1109 42 Dharap, A. et al. (2012) Effect of focal ischemia on long noncoding RNAs. Stroke 43, 2800–2802 43 Zhang, W. et al. (2012) Variants on chromosome 9p21.3 correlated with ANRIL expression contribute to stroke risk and recurrence in a large prospective stroke population. Stroke 43, 14–21 44 Cunnington, M.S. et al. (2010) Chromosome 9p21 SNPs associated with multiple disease phenotypes correlate with ANRIL expression. PLoS Genet. 6, e1000899 45 Foroud, T. et al. (2012) Genome-wide association study of intracranial aneurysms confirms role of ANRIL and SOX17 in disease risk. Stroke 43, 2846–2852 46 Costa, F.F. et al. (2011) Identification of microRNAs as potential prognostic markers in ependymoma. PLoS ONE 6, e25114 47 Weeraratne, S.D. et al. (2011) miR-34a confers chemosensitivity through modulation of MAGE-A and p53 in medulloblastoma. Neuro Oncol. 13, 165–175 48 Ryan, J. et al. (2012) MicroRNA-204 increases sensitivity of neuroblastoma cells to cisplatin and is associated with a favourable clinical outcome. Br. J. Cancer 107, 967–976 49 Zhi, F. et al. (2013) A microRNA expression signature predicts meningioma recurrence. Int. J. Cancer 132, 128–136 50 You, G. et al. (2012) Significance of miR-196b in tumor-related epilepsy of patients with gliomas. PLoS ONE 7, e46218 51 Chistiakov, D.A. and Chekhonin, V.P. (2012) Contribution of microRNAs to radio- and chemoresistance of brain tumors and their therapeutic potential. Eur. J. Pharmacol. 684, 8–18 52 Han, L. et al. (2012) LncRNA profile of glioblastoma reveals the potential role of lncRNAs in contributing to glioblastoma pathogenesis. Int. J. Oncol. 40, 2004–2012 53 Zhang, X. et al. (2012) Long non-coding RNA expression profiles predict clinical phenotypes in glioma. Neurobiol. Dis. 48, 1–8 54 Chang, C. et al. (2012) Correlation of microRNA-375 downregulation with unfavorable clinical outcome of patients with glioma. Neurosci. Lett. 531, 204–208 55 Zhang, W. et al. (2012) miR-181d: a predictive glioblastoma biomarker that downregulates MGMT expression. Neuro Oncol. 14, 712–719 740


56 Teplyuk, N.M. et al. (2012) MicroRNAs in cerebrospinal fluid identify glioblastoma and metastatic brain cancers and reflect disease activity. Neuro Oncol. 14, 689–700 57 Roth, P. et al. (2011) A specific miRNA signature in the peripheral blood of glioblastoma patients. J. Neurochem. 118, 449–457 58 Thamilarasan, M. et al. (2012) MicroRNAs in multiple sclerosis and experimental autoimmune encephalomyelitis. Autoimmun. Rev. 11, 174–179 59 Margis, R. and Rieder, C.R. (2011) Identification of blood microRNAs associated to Parkinsons disease. J. Biotechnol. 152, 96–101 60 Gaughwin, P.M. et al. (2011) Hsa-miR-34b is a plasma-stable microRNA that is elevated in pre-manifest Huntington’s disease. Hum. Mol. Genet. 20, 2225–2237 61 Lehmann, S.M. et al. (2012) An unconventional role for miRNA: let-7 activates Toll-like receptor 7 and causes neurodegeneration. Nat. Neurosci. 15, 827–835 62 Lukiw, W.J. et al. (2012) Spreading of Alzheimer’s disease inflammatory signaling through soluble micro-RNA. Neuroreport 23, 621–626 63 Soreq, H. and Wolf, Y. (2011) NeurimmiRs: microRNAs in the neuroimmune interface. Trends Mol. Med. 17, 548–555 64 Gutierrez-Vazquez, C. et al. (2013) Transfer of extracellular vesicles during immune cell–cell interactions. Immunol. Rev. 251, 125–142 65 Bryant, R.J. et al. (2012) Changes in circulating microRNA levels associated with prostate cancer. Br. J. Cancer 106, 768–774 66 Gallo, A. et al. (2012) The majority of microRNAs detectable in serum and saliva is concentrated in exosomes. PLoS ONE 7, e30679 67 Kucharzewska, P. et al. (2013) Exosomes reflect the hypoxic status of glioma cells and mediate hypoxia-dependent activation of vascular cells during tumor development. Proc. Natl. Acad. Sci. U.S.A. 110, 7312–7317 68 Peinado, H. et al. (2012) Melanoma exosomes educate bone marrow progenitor cells toward a pro-metastatic phenotype through MET. Nat. Med. 18, 883–891 69 Rana, S. et al. (2013) Exosomal tumor microRNA modulates premetastatic organ cells. Neoplasia 15, 281–295 70 Duttagupta, R. et al. (2011) Impact of cellular miRNAs on circulating miRNA biomarker signatures. PLoS ONE 6, e20769 71 Heyn, H. and Esteller, M. (2012) DNA methylation profiling in the clinic: applications and challenges. Nat. Rev. Genet. 13, 679–692 72 Cerf, A. et al. (2011) Single DNA molecule patterning for highthroughput epigenetic mapping. Anal. Chem. 83, 8073–8077 73 Zhang, Y. et al. (2012) Mapping DNA quantity into electrophoretic mobility through quantum dot nanotethers for high-resolution genetic and epigenetic analysis. ACS Nano 6, 858–864 74 Hyun, S. et al. (2011) An RNA aptamer that selectively recognizes symmetric dimethylation of arginine 8 in the histone H3 N-terminal peptide. Nucleic Acid Ther. 21, 157–163 75 Cipriany, B.R. et al. (2012) Real-time analysis and selection of methylated DNA by fluorescence-activated single molecule sorting in a nanofluidic channel. Proc. Natl. Acad. Sci. U.S.A. 109, 8477–8482 76 Minaker, S.A. et al. (2012) Antibody-free reading of the histone code using a simple chemical sensor array. J. Am. Chem. Soc. 134, 11674– 11680 77 Li, Y. and Kowdley, K.V. (2012) Method for microRNA isolation from clinical serum samples. Anal. Biochem. 431, 69–75 78 Gu, L.Q. et al. (2012) Detection of miRNAs with a nanopore singlemolecule counter. Expert Rev. Mol. Diagn. 12, 573–584 79 Hwang do, W. et al. (2010) Smart magnetic fluorescent nanoparticle imaging probes to monitor microRNAs. Small 6, 81–88 80 Chen, C. et al. (2010) Microfluidic isolation and transcriptome analysis of serum microvesicles. Lab Chip 10, 505–511 81 Sun, J. et al. (2010) A microfluidic platform for systems pathology: multiparameter single-cell signaling measurements of clinical brain tumor specimens. Cancer Res. 70, 6128–6138 82 Freilinger, M. et al. (2011) Effects of creatine supplementation in Rett syndrome: a randomized, placebo-controlled trial. J. Dev. Behav. Pediatr. 32, 454–460 83 Peters, S.U. et al. (2010) Double-blind therapeutic trial in Angelman syndrome using betaine and folic acid. Am. J. Med. Genet. A 152A, 1994–2001 84 Chestnut, B.A. et al. (2011) Epigenetic regulation of motor neuron cell death through DNA methylation. J. Neurosci. 31, 16619–16636

Review 85 Schafer, A. et al. (2010) Gemcitabine functions epigenetically by inhibiting repair mediated DNA demethylation. PLoS ONE 5, e14060 86 Csoka, A.B. and Szyf, M. (2009) Epigenetic side-effects of common pharmaceuticals: a potential new field in medicine and pharmacology. Med. Hypotheses 73, 770–780 87 Coronel, J. et al. (2011) A double-blind, placebo-controlled, randomized phase III trial of chemotherapy plus epigenetic therapy with hydralazine valproate for advanced cervical cancer. Preliminary results. Med. Oncol. 28 (Suppl. 1), S540–S546 88 Lee, B.H. et al. (2005) Procainamide is a specific inhibitor of DNA methyltransferase 1. J. Biol. Chem. 280, 40749–40756 89 Dong, E. et al. (2010) Valproate induces DNA demethylation in nuclear extracts from adult mouse brain. Epigenetics 5, 730–735 90 Berendsen, S. et al. (2012) Valproic acid for the treatment of malignant gliomas: review of the preclinical rationale and published clinical results. Expert Opin. Investig. Drugs 21, 1391–1415 91 Lee, E.Q. et al. (2012) Phase I study of vorinostat in combination with temozolomide in patients with high-grade gliomas: North American Brain Tumor Consortium Study 04-03. Clin. Cancer Res. 18, 6032–6039 92 Drappatz, J. et al. (2012) Phase I study of panobinostat in combination with bevacizumab for recurrent high-grade glioma. J. Neurooncol. 107, 133–138 93 Rigby, L. et al. (2012) Methods for the analysis of histone H3 and H4 acetylation in blood. Epigenetics 7, 875–882 94 Arrowsmith, C.H. et al. (2012) Epigenetic protein families: a new frontier for drug discovery. Nat. Rev. Drug Discov. 11, 384–400 95 McCabe, M.T. et al. (2012) EZH2 inhibition as a therapeutic strategy for lymphoma with EZH2-activating mutations. Nature 492, 108–112 96 Kruidenier, L. et al. (2012) A selective jumonji H3K27 demethylase inhibitor modulates the proinflammatory macrophage response. Nature 488, 404–408 97 Lee, K. et al. (2012) Molecular mechanism of Jmjd3-mediated interleukin-6 gene regulation in endothelial cells underlying spinal cord injury. J. Neurochem. 122, 272–282 98 Filippakopoulos, P. et al. (2010) Selective inhibition of BET bromodomains. Nature 468, 1067–1073 99 Cheng, Z. et al. (2013) Inhibition of BET bromodomain targets genetically diverse glioblastoma. Clin. Cancer Res. 19, 1748–1759 100 Jubb, H. et al. (2012) Structural biology and drug discovery for protein–protein interactions. Trends Pharmacol. Sci. 33, 241–248 101 Rigamonti, D. et al. (2009) Turning REST/NRSF dysfunction in Huntington’s disease into a pharmaceutical target. Curr. Pharm. Des. 15, 3958–3967


102 Leone, S. et al. (2008) SAR and QSAR study on 2-aminothiazole derivatives, modulators of transcriptional repression in Huntington’s disease. Bioorg. Med. Chem. 16, 5695–5703 103 Rigamonti, D. et al. (2007) Loss of huntingtin function complemented by small molecules acting as repressor element 1/neuron restrictive silencer element silencer modulators. J. Biol. Chem. 282, 24554– 24562 104 Lee, S.T. et al. (2012) miR-206 regulates brain-derived neurotrophic factor in Alzheimer disease model. Ann. Neurol. 72, 269–277 105 Mahishi, L.H. et al. (2012) miR-886-3p levels are elevated in Friedreich ataxia. J. Neurosci. 32, 9369–9373 106 Qureshi, I.A. and Mehler, M.F. (2013) Long non-coding RNAs: novel targets for nervous system disease diagnosis and therapy. Neurotherapeutics http://dx.doi.org/10.1007/s13311-013-0199-0 107 Modarresi, F. et al. (2012) Inhibition of natural antisense transcripts in vivo results in gene-specific transcriptional upregulation. Nat. Biotechnol. 30, 453–459 108 Kuan, W.L. et al. (2012) A novel neuroprotective therapy for Parkinson’s disease using a viral noncoding RNA that protects mitochondrial complex I activity. J. Exp. Med. 209, 1–10 109 Kumar, P. et al. (2007) Transvascular delivery of small interfering RNA to the central nervous system. Nature 448, 39–43 110 Tan, G.S. et al. (2012) Small molecule inhibition of RISC loading. ACS Chem. Biol. 7, 403–410 111 Shan, G. et al. (2008) A small molecule enhances RNA interference and promotes microRNA processing. Nat. Biotechnol. 26, 933–940 112 Melo, S. et al. (2011) Small molecule enoxacin is a cancer-specific growth inhibitor that acts by enhancing TAR RNA-binding protein 2mediated microRNA processing. Proc. Natl. Acad. Sci. U.S.A. 108, 4394–4399 113 Balaj, L. et al. (2011) Tumour microvesicles contain retrotransposon elements and amplified oncogene sequences. Nat. Commun. 2, 180 114 Skog, J. et al. (2008) Glioblastoma microvesicles transport RNA and proteins that promote tumour growth and provide diagnostic biomarkers. Nat. Cell Biol. 10, 1470–1476 115 Valadi, H. et al. (2007) Exosome-mediated transfer of mRNAs and microRNAs is a novel mechanism of genetic exchange between cells. Nat. Cell Biol. 9, 654–659 116 Bu, N. et al. (2011) Exosome-loaded dendritic cells elicit tumorspecific CD8+ cytotoxic T cells in patients with glioma. J. Neurooncol. 104, 659–667

741

Review

Common non-epigenetic drugs as epigenetic modulators Jo¨rn Lo¨tsch1,2, Gisbert Schneider3, Daniel Reker3, Michael J. Parnham2, Petra Schneider3, Gerd Geisslinger1,2, and Alexandra Doehring1 1

Institute of Clinical Pharmacology, Goethe University, Theodor-Stern-Kai 7, D-60590 Frankfurt am Main, Germany Fraunhofer Institute of Molecular Biology and Applied Ecology – Project Group Translational Medicine and Pharmacology (IME-TMP), Theodor-Stern-Kai 7, D-60590 Frankfurt am Main, Germany 3 Institute of Pharmaceutical Sciences, Eidgeno¨ssische Technische Hochschule (ETH), Wolfgang-Pauli-Strasse 10, 8093 Zu¨rich, Switzerland 2

Epigenetic effects are exerted by a variety of factors and evidence increases that common drugs such as opioids, cannabinoids, valproic acid, or cytostatics may induce alterations in DNA methylation patterns or histone conformations. These effects occur via chemical structural interactions with epigenetic enzymes, through interactions with DNA repair mechanisms. Computational predictions indicate that one-twentieth of all drugs might potentially interact with human histone deacetylase, which was prospectively experimentally verified for the compound with the highest predicted interaction probability. These epigenetic effects add to wanted and unwanted drug effects, contributing to mechanisms of drug resistance or disease-related and unrelated phenotypes. Because epigenetic changes might be transmitted to offspring, the need for reliable and cost-effective epigenetic screening tools becomes acute. Introduction Epigenetic effects are exerted by a variety of factors including early social experiences [1–3], physical training [4], age [5], nutritional and chemical factors such as royal jelly [6], benzene [7], asbestos, or smoking [5]. It is, therefore, to be expected that common drugs are a further factor modulating epigenetic mechanisms. Indeed, epigenetic regulation of DNA transcription, in the absence of changes in the DNA sequence (‘epi’ from Greek: over, above; http://www.epigenome.eu/en/1,1,0), has gained increasing attention as a starting point for new drug targets and as a further source of variability in drug effects [8,9]. Classical epigenetic mechanisms such as histone modifications provide direct pharmacological control of protein expression and novel classes of histone modifiers have received clinical approval for cancer therapy with broadening indications towards rheumatic diseases [10], chronic pain [11,12], and other non-malignant diseases. However, epigenetic effects are not limited to Corresponding author: Lo¨tsch, J. ([email protected]). 1471-4914/$ – see front matter ß 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.molmed.2013.08.006

742


epigenetic therapeutics but increasing evidence suggests that pharmacological treatments with common drugs may affect the patient’s epigenome [13]. Examples include the antiepileptic drug valproate, which modifies histones via direct chemical interaction with the histone deacetylase (HDAC) [14], as well as cannabinoids [15] and opioids [16] that both trigger DNA hypermethylation indirectly via branches of their downstream signaling pathways [15]. As epigenetic traits may be inheritable, producing stable phenotypes from changes in the chromosome without alterations in the DNA sequence [17], inheritable drug effects become a possibility. In addition, epigenetic drug effects interfering with protein expression in pathophysiological processes may cause unwanted effects such as opioid-induced hyperalgesia [16] or drug resistance to cytostatics [18]. Therefore, the epigenetic effects reviewed here of common drugs not intended as epigenetic therapeutics and referred to as ‘non-epigenetic drugs’ cannot be ignored, and epigenetic screening becomes a necessity in drug development [19]. The expected frequency of epigenetic drug effects will be exemplified by computational predictions and subsequent experimental verification of drug interactions with human HDACs as important players in epigenetic mechanisms. Mechanisms of epigenetic regulation addressed by drugs Classical epigenetic mechanisms (Figure 1) include covalent modifications of the DNA (methylation) and of the DNA-packaging histones (e.g., changes in the state of histone acetylation, methylation, phosphorylation, and others). Post-transcriptional regulation by miRNAs [20], which is not addressed in this review, can also be added to the list of modifications. DNA methylation and demethylation Among classical epigenetic mechanisms in vertebrates, DNA methylation has received by far the most scientific interest. DNA is covalently modified by the transfer of a methyl group from S-adenosyl methionine (SAM) to the 5position of the pyrimidine ring in the cytosine residues of CpG dinucleotides [21]. This process is catalyzed by the DNA (cytosine-5) methyltransferases DNMT1, 3A, and 3B and involves methyl-binding proteins such as Mbd1–4 and

Review


2 46 89

3 1

16 27

20 36

Morphine Cocaine

HD

HM

M

T

79

Ac

Ph

Valproate

se

Kin a

Kin a

HDA C

HAT

Me

2 4 6 8 9 27 36

20 16

se

3 1

79

Celecoxib Difluoromethylornithine SSRI Anandamide 1 tor CB cep re MAPK (p38, p42/44)

1,3

Imanib Tamoxifen

A/

,3

MT

T1

( +)D N

( -) D N M

µre Opio ce pt id or

A/

B

Fluoxene

TF

Opioid

Hydralazine

B

TF

HDAC

HD M

Mbd1-4 MeCP2

Nuk

lea

se

Gemcitabine

DN

TF

se mera oly Ap

TF


Figure 1. Epigenetic mechanisms addressed by pleiotropic epigenetic effects of non-epigenetic drugs. (Top) Enlarged scheme of human chromatin and its basic modules, the nucleosomes, which are composed of a DNA double strand wrapped around octameric histone protein complexes (top right). There are two states of chromatin, the densely packed, transcriptionally repressive, heterochromatin (top) and the more open, transcriptionally active, euchromatin (bottom). Both states differ in the sites and types of N-terminal histone modifications, in that heterochromatin is characterized by methylation (yellow dots) of lysine residues (catalyzed by HMTs), for instance, at positions 9 and 27 in histone H3, whereas euchromatin is characterized by additional modifications such as acetylation (red dots; catalyzed by HATs and their counteractive HDACs) at lysines 4, 8, and 27 in histone H3 or phosphorylation (blue dots; catalyzed by kinases) at threonine 6 in histone H3. (Bottom) Further enlarged view of the DNA strand with its different methylation states. The binding of transcription factors (TFs) to unmethylated CpG-containing binding sites (left) is inhibited by methylation of the cytosines (catalyzed by DNMTs) in the promoters and the first exons of many genes. However, methyl-binding proteins bind to methylated cytosines and form complexes (See figure legend on the bottom of the next page.)

743

Review MeCP2 [22]. In mammals, DNA methylation patterns are established during embryonic development and maintained by a copying mechanism when cells divide. After DNA replication the parental strand maintains its pattern of modified cytosines, whereas the newly synthesized daughter strand remains unmodified but is soon methylated via DNMTs [23], probably by DNMT1, preferring hemimethylated DNA as a substrate in vitro [24]. DNMT3A and DNMT3B are the so-called de novo methyltransferases, mainly responsible for introducing cytosine methylation at previously unmethylated CpG loci. Methylation of cytosines that precede a guanosine in the DNA sequence (the CpG dinucleotide) is a key epigenetic modification, and hypermethylation of gene promoters is associated with transcriptional silencing [25]. That is, functionally, DNA methylation is mostly associated with transcriptional repression. It prevents the docking of transcription factors and thus the reading of the gene. Most CpGs in the genome are methylated [26], a state associated with maintenance of chromosomal stability, prevention of translocation and endoparasitic sequence silencing. CpG dinucleotides within so-called ‘CpG islands’ [27], that is, CpG-rich regions mainly localized in promoters or first exons of approximately 60% of human genes, are typically unmethylated during development and in normal tissue. However, some physiological processes require DNMT3induced DNA methylation of CpG islands such as silencing of imprinted genes and X chromosome inactivation in women [23], serving to balance the X-chromosome-linked gene dosage between females and males. Besides methylation in germ cells or the early embryo, de novo methylation also occurs in adult somatic cells during disease development such as cancer [28] and neurological disorders [29]. In contrast to the well-defined methylation processes, the various pathways of DNA demethylation are still only rudimentarily understood. Demethylation can occur either via passive or active processes. Reduction or inhibition of DNMTs leads to a passive demethylation, whereas a more complex process of active demethylation is proposed to include various modifications such as hydroxylation by the TET family, deamination by the AID/APOBEC family, and DNA repair mechanism procedures carried out by the BER glycosylases [30,31]. Histone modifications The second and more widely therapeutically employed classical epigenetic mechanism includes modifications of the DNA-packaging histone proteins. Histones are composed of two H2A–H2B dimers and one H3–H4 tetramer and form, together with the wrapped-around 146 bp of DNA, the so-called nucleosome, which is the basic structure of chromatin. Histone proteins are post-translationally modified at specific sites on the N terminus, including lysine


acetylation and ubiquitination, methylation of lysines and arginines, serine phosphorylation, sumoylation, and others [32]. This ‘histone code’ alters the histone–DNA interaction and thus the chromatin structure, which can be favorable or unfavorable for the docking of transcription factors, depending on the type and site of the modification. Acetylation and deacetylation are regulated by histone acetyltransferases (HATs) and HDACs. HATs catalyze the transfer of the acetyl moiety from acetyl-CoA to the eamino group of histone lysine residues, resulting in acetylated lysine and CoA. Subsequently, the positively charged histones are neutralized and their interaction with the negatively charged DNA is decreased. Moreover, acetylation can regulate DNA replication, histone deposition, and DNA repair by recruiting proteins that possess the acetyllysine-binding domain bromodomain [33]. Lysine acetylation results in the more opened euchromatin state which is more accessible to transcription factors [34]. Moieties can also be removed from the histones, with opposite effects for DNA readability. HDACs catalyze the removal of acetyl groups from lysine residues, resulting in a more compacted and thus transcriptionally repressed heterochromatin structure [35]. In contrast to acetylation, methylation which is regulated by histone methyltransferases and demethylases, respectively, does not alter the overall charge of the histone tails; however, it modifies their basicity, hydrophobicity, and the affinity for anionic molecules such as DNA. Evidence for epigenetic effects of non-epigenetic drugs Besides drugs developed to modulate epigenetics on purpose (Table 1), accumulating evidence shows that other compounds (Table 2) can also interfere with epigenetic mechanisms (Box 1). Induction of epigenetic changes by chemicals has gained widespread attention as a result of the discovery that the nutritional effect of royal jelly from the honey bee Apis mellifera, which controls the development of genetically identical larvae into either queens (royal jelly fed) or worker bees (royal jelly not fed), represents an epigenetic mechanism. Royal jelly reduces DNA methylation by silencing DNMT3 expression, leading to the increased transcription of developmental and reproduction genes in newborn honey bees [36] via its major component royalactin, a 57-kDa protein [6]. Small interfering RNA (siRNA)-induced knockdown of Dnmt3 levels in bee larvae resulted in the development of queen bees [36]. However, HDAC-inhibiting activity of crude royal jelly was additionally shown [37]. The active (E)-10-hydroxy-2-decenoic acid (10HDA), a fatty acid accounting for up to 5% of royal jelly, is structurally similar to the HDAC inhibitors valproic acid and suberoylanilide hydroxamic acid (SAHA), and the epigenetic regulation of queen bee development is probably driven, in part, also by the HDACinhibiting activity in royal jelly [37].

with HDACs and HDMs, thus further influencing histone modifications (middle). In addition, methylated cytosines in the DNA strand are recognized as failures and are replaced by unmethylated cytosines during DNA repair mechanisms (right). Pleiotropic epigenetic effects of non-epigenetic drugs are shown in green with direct effects (continuous lines) caused by valproate (which binds to the catalytic center of HDACs) and by hydralazine (which stably interacts with DNMTs). Indirect effects (dashed lines) indicate those of gemcitabine (which disturbs the DNA repair mechanism) and various other drugs which alter the DNA methylation pattern by influencing DNMTs (e.g., imatinib, celecoxib, and cannabinoids/opioids) or the methyl-binding proteins Mbd and MeCP2 (fluoxetine) or by changing the histone code (morphine/cocaine), respectively. Abbreviations: HMT, histone methyltransferase; HDM, histone demethylase; HAT, histone acetyltransferase; HDAC, histone deacetylase; DNMT, DNA methyltransferase; MDB, methyl-binding proteins.

744

Review


Table 1. Epigenetic therapeutics that have been developed with the purpose to exert epigenetic effects Substance Epigenetic therapeutics Suberoylanilide hydroxamic acid (SAHA) (vorinostat) PXD101 (belinostat) ITF2375 (givinostat)

Chemical class

Target

Epigenetic consequences

Indications

Refs

Hydroxamic acids

HDACs

Histone acetylation

Advanced cutaneous T cell lymphoma (CTCL) (FDA-approved) Hepatocellular carcinoma (Phase I/II) Hematological malignancies (Phase II) Systemic-onset juvenile idiopathic arthritis (SOJIA) (Phase I) Advanced cutaneous T cell lymphoma (CTCL) and peripheral T cell lymphoma (FDA-approved) Solid tumors (Phase I/II) Myeloid malignancies (FDA-approved)

[101]

Depsipeptide (Romidepsin)

Cyclic peptides

MS-275 (entinostat) 5-Azacytidine (5-Aza-CR; Vidaza) 5-Aza-20 -deoxycytidine (5-Aza-CdR; Decitabine)

Benzamides Nucleoside analogs

DNMTs

DNA hypomethylation

Epigenetic effects of drugs with structural similarities to epigenetic factors are highly likely, as shown for valproic acid which, besides exerting channel blocking actions, is also a HDAC inhibitor [14]. Several mechanisms have been described by which drugs, so far not known to exert epigenetic effects, may interfere with DNA methylation or histone modification. We identified these by a PubMed (http:// www.ncbi.nlm.nih.gov/pubmed?db=pubmed) search for ‘epigenetic’, combined with various drug classes, including ‘antidepressants’ or ‘antihypertensive’ or ‘anticancer drugs’. Only studies in which actual evidence shows epigenetic changes, that is, DNA (de-)methylation or histone modifications, respectively, were included in the analysis and not reports with shown changes in protein expression for which an epigenetic background is a mere hypothesis [13]. Direct interaction with primary enzymes of epigenetic processes The antiepileptic drug, valproate, acts as a HDAC inhibitor to inhibit sodium channels and to increase the synaptic availability of GABA in the brain [38]. It causes hyperacetylation of the N-terminal tails of histones H3 and H4 and directly inhibits HDAC activity, probably by binding to its catalytic center thus blocking substrate access [14]. In addition, exposure of breast cancer cells to valproate led to a downregulation of genes that regulate the structure of chromatin with subsequent chromatin decondensation and enhanced DNA sensitivity to nucleases [39]. The antihypertensive drug, hydralazine, causes adverse immunologic reactions such as drug-induced lupus-like syndrome, which was found to be associated with DNA hypomethylation [40]. Chemical structure analyses showed a high-affinity and energetically stable interaction between hydralazine and the DNMTs (Figure 2) resulting in direct inhibition of methyltransferase activity [41,42]. Preliminary results of a Phase II study showed that the combination of hydralazine and valproate could provide an effective therapy for patients with myelodysplastic syndromes [43]. Interference with DNA repair mechanisms The cytidine analog, gemcitabine, an anticancer agent which, among other actions, inhibits DNA polymerase

[102] [103] [104] [105]

[106] [107]

activity, was also shown to function as an epigenetic drug [44]. In this respect, the growth arrest and DNA damage inducible protein, 45 alpha (Gadd45a), was shown to directly interact with the endonuclease XPG involved in the nucleotide excision repair mechanism [45]. As a consequence, the DNA repair machinery is activated, causing the excision of methylated cytosines and their replacement by unmethylated cytosines and thus a net DNA demethylation [46]. Gemcitabine inhibits this Gadd45a-mediated DNA repair process and the associated net demethylation finally resulting in a gene-specific hypermethylation of the promoter region of the tumor suppressor gene MLH1 accompanied by reduced MLH1 expression in cell models [44]. Interaction with epigenetic mechanisms via drug signaling pathways Epigenetic changes following drug exposure can be triggered via downstream signaling pathways modified by the drug. This includes branches of pathways targeting epigenetic mechanisms or effects that are the result of modifications of cellular homeostasis originally intended to be produced when administering drugs of a particular class. These epigenetic side effects have been described for several pathways, on which many drugs, such as G-proteincoupled receptor signaling or cytokine pathways, exert their effects. G-protein-coupled receptor pathways Stimulation of the Gi-protein-coupled cannabinoid receptor CB1 involves activation of the p38 and p42/44 mitogenactivated protein (MAP) kinase-dependent pathways [15] and the natural CB1 ligand, anandamide, increases the overall methylation status in differentiating human keratinocytes [15] via induction of DNMT. A similar effect has been described for opioids. Thus, the chronic exposure in addicts and pain patients also increased DNA methylation at specific sites in a CpG-rich island in the OPRM1 gene, coding for the m-opioid receptor [16,47], and at the global DNA methylation site LINE-1 [16]. However, it is unclear whether the same mechanism is involved as that in CB1 signaling. But the methylation in LINE-1, a member of the retrotransposons, mobile genetic elements that use a germline copy mechanism to spread 745

Review


Table 2. Non-epigenetic drugs, that is, drugs that have been discovered with other mechanisms of action but for which epigenetic effects have been successfully discovered after drug development Substance Chemical class Target Non-epigenetic drugs with successively discovered epigenetic effects Arachidonoyl CB1/2 cannabinoid Anandamide receptors ethanolamine

a-Difluoromethylornithine

Difluoromethylornithine

Celecoxib

COX-2 inhibitor

Cocaine

Tropane alkaloid

Escitaloprame

Benzofurans

Fluoxetine

Phenylpropylamines

Gemcitabine

Cytidine analog

DNA repair machinery

Hydralazine

Phthalazines

Ca2+ balance in the vascular smooth muscle

Imatinib

Benzamides

Tyrosine kinases abl, c-kit, and PDGF-R

Opioids

Opioid

m-Opioid receptor

Tamoxifen

Estrogen receptor modulator

Estrogen receptor

Valproate

Short-chain fatty acids

HDACs

Ornithine decarboxylase COX-2

Serotonin reuptake pump of neuronal membranes

throughout the genome [48] and that are involved in cerebral gene expression [49], correlated positively with the pain scores in opioid-treated patients providing a new putative epigenetic mechanism of opioid-induced hyperalgesia [16]. Cytokine pathways Modifications in the expression of the tumor necrosis factor (TNF)-a have been repeatedly reported to explain the changes in DNA or histone methylation following drug exposure. In mice with genetic deletion of receptors for the cytokines interferon (IFN)-g and TNF-a [50,51], and in the presence of nonsteroidal anti-inflammatory drugs (NSAIDs), such as ibuprofen or acetylsalicylic acid, which indirectly reduce TNF expression, an increase in brain levels of P11 (S100A10, S100 calcium-binding protein 746

Epigenetic consequences

Indications

Refs

DNA hypermethylation

AIDS-associated wasting syndrome, Multiple sclerosisassociated spastic symptoms Facial hirsutism, sleeping sickness Inflammation, pain

[15]

Reversal of the global DNA hypomethylation and the specific hypermethylation of the ER-a gene in rats with induced colon tumors Decreased expression of histone methyltransferase G9a and subsequent lower methylation levels at H3K9 Reduced mRNA expression for DNMTs and subsequent decreased gene-specific methylation levels Induction of methyl-CpGbinding proteins Inhibition of DNA repair process and the associated demethylation process ! DNA hypermethylation Hypomethylation by a stable interaction with DNMTs causing inhibition of methyltransferase activity Increase in DNTM3A and EZH2 expression associated with promoter hypermethylation and downregulation of the tumor suppressor PTEN DNA hypermethylation Decreased expression of histone methyltransferase G9a and subsequent lower methylation levels at H3K9 Downregulation of estrogen receptor responsive genes pS2 and progesterone receptor due to promoter hypermethylation Hyperacetylation of the Nterminal tails of histones H3 and H4

[59]

[108]

Major depression

[57]

[72,73] Cancer

[44]

Hypertension, vasodilation

[41,42]

Leukemia

[69]

Pain, substitution therapy of opiate addiction

[16] [77]

Breast cancer

[71]

Epilepsy, bipolar disorder, diabetic peripheral neuropathy

[14]

A10) triggered by antidepressants such as S-citalopram [52] or fluoxetine [53] was abolished. The P11 protein interacts with several cellular components, including ion channels and the serotonin 5-HT1B receptor [54] and has been associated with depression [55,56]. Decreased p11 levels were associated with a p11 promoter DNA hypermethylation in the Flinders sensitive line (FSL) genetic rodent model of depression [57]. S-Citalopram reduced the mRNA expression for the DNA methyltransferases, Dnmt1 and Dnmt3a and reversed hypermethylation, leading to increased p11 expression [57]. NSAIDs antagonized the selective serotonin reuptake inhibitor (SSRI)-induced increase in frontal cortical levels of IFN-g and TNF-a, and patients receiving NSAIDs displayed impaired therapeutic responses to the SSRI citalopram [51]. A pathway was suggested whereby SSRIs increase cytokines, which in

Review

Box 1. Possible epigenetic mechanisms of action: nonepigenetic drugs Direct interaction with epigenetic-modulating enzymes DNA methylation may be inhibited by a high affinity to and stable ‘interaction with DNMTs’ by hydralazine. HDAC may be inhibited by drug ‘binding to the catalytic center of HDACs’, as shown for valproate. Interaction with DNA repair mechanisms Replacement of methylated cytosines by unmethylated cytosines during ‘DNA repair’ is inhibited by gemcitabine, with the net result of a more methylated DNA. Alterations of epigenetics via pathways that project or branch towards epigenetic processes ‘G-protein-coupled receptor pathways’ may branch to p38 and p42/44 MAPK-dependent pathways that, in turn, actively project to DNMTs via a yet unknown pathway. This is the putative mechanism of DNA hypermethylation associated with cannabinoids (anandamide) and possibly opioids. ‘Cytokine pathways’ may modulate the expression of DNA methyltransferases. Increased levels of IFN-g and TNF-a have been associated with reduced DNMT expression and thus hypomethylation. This mechanism may stay behind the epigenetic effects of SSRIs (S-citalopram, fluoxetine) and possibly celecoxib and other NSAIDs (ibuprofen, acetyl salicylic acid). Many additional pharmacological mechanisms seem to project to epigenomic regulatory processes without detailed elucidation of the pathways.

turn stimulate the formation of p11 [50]. A link to DNMT downregulation fits with the observation of global DNA hypomethylation in blood cells from systemic lupus erythematosus patients [58]. Whether cytokine-mediated mechanisms are also involved in the reversal of the methylation pattern in rats with induced colon tumors, following administration of the cyclooxygenase-2 (COX-2) inhibitor celecoxib [59], is uncertain. The tumor-induced general DNA hypomethylation and the specific hypermethylation of the estrogen receptor (ER)-a gene were both reversed and, additionally, the gene-specific reduced methylation was accompanied by an increase in the amount of ER-a mRNA [59]. Interestingly, it has been shown that celecoxib inhibits tumor growth COX-2 independently at clinically relevant doses [60], although the rather low plasma levels cannot explain these effects. One possible mechanism for this pleiotropic effect discussed in the literature may be intercalation and accumulation of the drug into cell membranes leading to inhibition of proteins and enzymes directly linked to tumor cell growth and proliferation [61]. Although not explicitly described so far, celecoxib, when intracellularly enriched, may also interfere with epigenetically important proteins. TNF-mediated phosphorylation of S276 on RelA/p65 is required for RelA/p65–DNMT1 interactions and for chromatin loading of DNMT1 with subsequent transcriptional repression [62]. Celecoxib inhibits the TNF-a-induced nuclear accumulation of the nuclear factor (NF)-kB p65 subunit resulting in reduced expression of TNF-induced cytokines [63]. At higher concentrations, however, celecoxib shows opposite effects, meaning that it activates NF-kB and thereby increases TNF-a levels [64]. Taken together, epigenetic mechanisms may provide


explanations for the various biochemical effects shown so far for celecoxib. Further evidence of epigenetic drug effects Many epigenetic changes observed following drug exposure have been described without investigation of the mechanisms involved. Their epigenetic effects seem to be triggered indirectly as a consequence of structural or functional alterations in cellular components or signaling pathways, which are primarily addressed by the drug and have been reported to be responsible for many of the known epigenetic mechanisms. Imatinib, a specific inhibitor of tyrosine kinases, abl (the Abelson proto-oncogene), c-kit, and PDGF-R (platelet-derived growth factor receptor) [65], prolongs the survival of patients with various forms of leukemia [66]. However, imatinib resistance has emerged as a clinical problem in leukemia patients [67]. Although this is often attributable to acquired mutations in the target kinases [67], an additional epigenetic mechanism has also been proposed. Accordingly, in leukemia cells, a 4-month exposure to imatinib caused a promoter hypermethylation and therefore downregulation of the tumor suppressor gene PTEN (phosphatase and tensin homolog deleted on chromosome 10) [68]. This effect seemed to be triggered by induced expression of DNMT3A and of the polycomb group protein EZH2 (Enhancer of Zeste homolog 2) [69], both previously shown to interact with each other during methylation of EZH2 targets [70], such as MYT1 (Myelin transcription factor 1) or the proto-oncogene Wnt-1 [70]. Recently, a combination therapy of hydralazine and magnesium valproate was shown to reverse imatinib resistance in patients with chronic myeloid leukemia [18]. The development of drug resistance to the ER antagonist tamoxifen involves epigenetic silencing of estrogenregulated genes [71]. MCF-7 breast cancer cells permanently treated with tamoxifen showed suppression of estrogen receptor mRNA and protein, and downregulation of the ER responsive gene pS2 and the progesterone receptor. Although tamoxifen withdrawal restored ER expression and function, it failed to restore the suppression of ER responsive genes. These downregulations were caused by gene promoter hypermethylations [71] by, as yet, unspecified mechanisms. PS2 and progesterone receptor expression were restored following cotreatment with the demethylating agent, 5-azacytidine [71]. The inhibitor of ornithine decarboxylase, a-difluoromethylornithine led to a reversal of the methylation pattern in rats with induced colon tumors [59]. The tumorinduced general DNA hypomethylation and the specific hypermethylation of the ER-a gene were reversed and, additionally, the gene-specific reduced methylation was accompanied by an increase in the amount of ER-a mRNA. However, although the effects were similar to those of celecoxib [59], their mechanism is unclear. Chronic administration of fluoxetine in traumatic brain injury models induced hippocampal neurogenesis and altered epigenetic signaling parameters [72]. The ratio of acetylated histone 3 normalized to total histone 3, as well as the expression of the methyl-CpG-binding protein 1 (MBD1) were increased following fluoxetine treatment 747

Review


Cys70

HN

NH2 Glu128

Phe204 His321

N N Val95, Phe130 Asn441

Thr109 Arg174 TRENDS in Molecular Medicine

Figure 2. Comparison of observed and modeled interaction patterns of hydralazine with N-acetyltransferase and DNA methyltransferase. We compared potential ligand–protein interactions of hydralazine with Mycobacterium marinum arylamine N-acetyltransferase (MMNAT; PDB ID: 3LTW, X-ray structure resolution of 2.1 A˚; green residues [109]) and a comparative protein model (‘homology model’) of human DNA methyltransferase (blue residues [42]). The hydralazine– MMNAT complex is currently the only available crystal structure containing hydralazine as a ligand. DNA methyltransferase is a potential epigenetic ‘off’-target of hydralazine [110]. Hydrogen bonds are depicted as arrows pointing from hydrogen bond donor to acceptor. Filled arrowheads represent structure-derived interactions observed in the hydralazine–MMNAT complex. Open arrowheads indicate computationally estimated hydrogen bonds. The computer model suggests the His321 in DNA methyltransferase might be involved in arene–arene interactions with hydralazine, thus assuming the role of MMNAT Phe204. Accordingly, Asn441 in DNA methyltransferase might undergo hydrophobic interactions that are observed for MMNAT Val95 and Phe130 in the crystal complex.

[72]. Similarly, in normal adult rat brains, MBD1 and methyl-CpG-binding protein 2 (MeCP2) were found to be induced after repeated fluoxetine injections [73]. This apparently contradicts previous in vitro studies associating MeCP2 with reduced histone H3 acetylation by binding to methylated CpG sites and thereby recruiting a HDAC [74,75]. This mechanism seems to suggest an additional epigenetic response to fluoxetine, in addition to its abovementioned epigenetic effects via brain cytokines. Repeated exposure to cocaine [76] or morphine [77] was associated with decreased lysine methylation at position 9 of histone H3 (H3K9me2/3) and with a decreased expression of the respective histone methyltransferase, namely G9a, in mouse nucleus accumbens. As G9a levels in the nucleus accumbens involved in reward were shown to influence various behavioral responses to repeated morphine exposure, such as locomotor sensitization, analgesic tolerance, and physical withdrawal [77], this may provide an epigenetic-based mechanism for opiate addiction. Frequency of HDAC-inhibiting potential among nonepigenetic drugs One of the major mechanisms by which drugs interact with epigenetic processes is interference with histone deacetylation. This is the target of the largest group of current epigenetic therapeutics [78] and is also modified by valproate among non-epigenetic drugs. To estimate whether or not such interaction with HDAC is a frequent although largely unexplored phenomenon among nonepigenetic drugs, a computational estimate of the expected fraction of common drugs that may potentially interact with HDAC was performed in silico and 748

predictions were subsequently verified in vitro for the compound with the highest probability (lowest statistical P-value) of HDAC binding. To this end, we investigated the Library of Pharmacologically Active Compounds (LOPAC1280, Sigma-Aldrich) for structural and pharmacophoric similarities to known HDAC inhibitors. The molecules were represented by a topological pharmacophore descriptor (CATS) [79] that accounts for patterns of potential ligand–receptor interaction sites in the drug compounds (Box 2). In brief, molecules are treated as 2D (planar) molecular graphs, where each vertex corresponds to an atom. Then, pharmacophore feature types (hydrogen bond donor or acceptor, positively or negative ionizable, aromatic) are assigned to the graph vertices, and a histogram is generated that lists the occurrence frequencies of pairs of features spaced a certain number of bonds apart. Here, we counted pairs separated up to nine bonds in the molecular graph. Finally, the histogram values were scaled by the relative feature type occurrence in each compound (for details, see [80]). This resulted in a 210-dimensional vector representing the topological pharmacophoric features of each molecular structure. These molecular representations were compared to a set of carefully annotated lead compounds that were extracted from the recent literature (COBRA collection; 12 642 bioactive reference compounds known to bind to a total of 980 target protein subtypes [81]). This set of reference compounds contained 51 known HDAC inhibitors. Then, we employed a self-organizing map (SOM, Kohonen network [82]) to perform a nonlinear clustering of these data, and for each of the approved drugs inferred potential HDAC inhibition by similarity to co-clustered known HDAC inhibitors. The SOM technique has already been successfully used for target prediction, de novo drug design, and screening library profiling [83,84]. Owing to its focus on pharmacophore-based similarities, the approach complements the similarity ensemble approach (SEA) developed by the Shoichet group [85,86], and related tools like PASS [87] or the techniques implemented by Mestres and coworkers [88], to name some of the prominent examples. To infer the significance of acquired predictions, we computed background similarity score distributions of allagainst-all comparisons. This allowed us to estimate the Pvalue of a prediction as the probability to expect at least such a similarity score between a test molecule and the set of known ligands at random [89]. Applying a permissive threshold of P < 0.05 we identified 68 (5%) of the LOPAC compounds as potential HDAC inhibitors. To investigate whether these potential HDAC inhibitors have already been validated, we looked up their target annotations in the ChEMBL database [90]. We found 48 of the candidate compounds, yet none were annotated to inhibit any of the HDAC subtypes. Our top HDAC target prediction (P = 0.007) was for Compound 1, a low-nanomolar type C inhibitor of human matrix metalloproteinases MMP2 and MMP9 [91]. This antiangiogenic agent shares a few key features with known HDAC inhibitors, which motivate the SOM-based prediction (Box 2). Of note, the SEA method, which uses ChEMBL as a knowledge base, did not predict HDAC as a potential target of Compound 1. Also, HDAC is not reported as a

Review


Box 2. SOM-based prediction of potential HDAC inhibitors A self-organizing map (SOM) allows for the unsupervised clustering of molecular data and the prediction of macromolecular targets for bioactive substances. The method originates from the field of neural networks and has been successfully employed for property and activity prediction in drug discovery [80]. The SOM algorithm generates a nonlinear, neighborhood-preserving projection of high-dimensional molecular data (e.g., compounds represented by

numerical descriptors) [111] (Figure I). By using P-value statistics, the SOM clusters (here: 15 10 = 150 clusters) were assessed for the presence of statistically significant pairs of test compounds and reference ligands with known macromolecular targets. As a result, target panels are predicted for test compounds that are statistically significantly similar in their properties (here: pharmacophore features, CATS descriptor [100]).

Self-organizing map (SOM) (A)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Reference compounds (known HDAC inhibitors) from SOM cluster (6/6) O

1 2 3 4

HO

5 6

N H

7

O O

8

HO

9 10

1 2 3

S

N H

Cl

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

(B)

N

O HO

O

N H

N

4 5

SOM predicon P = 0.007

6 7 8 9 10

O HO–NH

O O N S O Compound 1 TRENDS in Molecular Medicine

Figure I. Self-organizing map (SOM)-based prediction of potential histone deacetylase (HDAC) inhibitors. The 2D map representations resulting from SOM-based clustering display the overall compound density of the reference compounds (Collection of Bioactive Reference Analogs, COBRA) together with 1219 drugs (Library of Pharmacologically Active Compounds, LOPAC) (A), and the set of known HDAC inhibitors used in this study (B). The color code ranges from white (no compounds) through blue (few) to red (many compounds). The best-scoring candidate drug, Compound 1, was colocated in cluster (6/6) on the SOM, which also contained several known HDAC reference inhibitors (the two structures shown at the top).

target of Compound 1 in CAS (American Chemical Society, Columbus, OH, USA; https://scifinder.cas.org/). We therefore tested Compound 1 for HDAC inhibition in vitro. It turned out that this drug potently inhibits HDAC8 with an IC50 of 11 mM (HDAC1 to 11 inhibition assays were performed by Cerep SA, Celle l’Evescault, France; http:// www.cerep.fr). HDAC8 is involved in neuroblastomal differentiation [92], skull morphogenesis [93], and transcriptional regulation of estrogen-related receptor a [94]. Considering the low micromolar antiangiogenic cellular activity of Compound 1 [91], the observed HDAC8 inhibition might be pharmacologically relevant [95]. This novel finding corroborates the applicability of our SOM approach and demonstrates the usefulness but also the need for sophisticated computational methods and meticulously annotated and curated reference data for reliable target prediction. It goes without saying that every computational target prediction is error-prone and requires careful experimental assessment. Keeping this inherent caveat in mind, robust prediction tools can help unravel drugs with likely ‘off’-target liabilities that would otherwise have gone unnoticed.

Consequences of pleiotropic epigenetic drug effects A measureable affinity of a drug to an epigenetic target might not necessarily constitute a therapeutic limitation. However, evidence is accumulating that the modification of epigenetics, as a major regulatory mechanism of gene expression, by drugs may alter the patient’s phenotype. For example, chronic opioid treatment was associated with increased methylation at the LINE-1 site; however, this impacted on the patient’s phenotype as it correlated with increased pain [16], suggesting that hypermethylation of unspecified pain-relevant genes was triggering hyperalgesia. LINE-1 is a retrotransposon [48] that may reshape the genetic circuitry of neurobiological processes by mobilizing protein-coding genes that are differentially expressed and active in the brain [49] and drug-induced changes in its methylation status may have unexpected effects on central nervous system (CNS) functions. The hereditary transmission of epigenetic changes is still under discussion [96]. Clear evidence has been found in plants, such as rice treated with heavy metals [97]. However, equivalent data for animals are still sparse, although evidence supporting such mechanisms is accumulating. For 749

Review example, mental stress in mice induced by postnatal separation of the pups from the mother induced depressive behavior and altered the DNA methylation patterns in the brains of the pups as well as in sperm, and this was associated with changes in methylation status and gene expression in both generations [98]. Moreover, a predisposition for the development of nonpolyposis colorectal cancer has been attributed to a cytosine methylation of the mutL homolog 1, colon cancer, nonpolyposis type 2 gene (MLH1) [99]. Epigenetic silencing of this DNA mismatch repair gene was transmitted from a mother, in whom endometrial cancer had developed at the age of 45 years, to her son [99]. Future directions The growing concern that epigenetic events may be involved in chemically and/or nutritionally mediated adverse health outcomes has driven demands for epigenetic screening in product safety assessment [19]. Present evidence clearly indicates that this includes drugs. The strategies by which epigenetic effects can be identified and possibly avoided during drug development, however, vary according to their different mechanisms (Box 1). Direct chemical interactions between a drug molecule and an epigenetic enzyme (i) can be predicted by contemporary computational methods and subsequently experimentally verified, as shown above. As these interactions usually do not represent drug class effects, chemical members of the same class can be chosen during drug discovery which do not exhibit epigenetic effects. Because they represent a class effect, but are not easily avoidable, epigenetic effects (ii) of cytostatic drugs interfering with DNA repair mechanisms are also likely to be predictable. However, if epigenetic processes (iii) are addressed via signaling pathways through which the drug exerts its therapeutic effect, they may only be predictable if the effect has been described, although avoidance of epigenetic effects depends on the ability to select a target downstream of the point at which the pathway branches towards the epigenetic mechanism. As these signaling-mediated epigenetic mechanisms, such as the CB1-mediated effect of anandamide via mitogen-activated protein kinase (MAPK), are only partly understood, their prediction is rarely possible. This applies particularly to the many epigenetic drug effects for which the mechanisms are largely unknown. These effects have often been discovered accidently, a causal relationship between drug presence and epigenetic changes is not always proven and their prediction probably requires direct testing. Epigenetic screens for identification of these effects still need to be refined and extended. This includes reproducible identification of causal links between drugs and epigenetic changes and the development of reporter assay tools to monitor such changes [19]. As for general product safety [19], in vitro high-throughput screens could assess all three branches of epigenome control, namely, DNA methylation, histone modifications, and the noncoding RNAs; the latter have not been reviewed here. The identification of epigenetic interactions with epigenetic enzymes and the resulting modifications of the epigenome should be followed by an experiment-based assessment of the unwanted phenotypic effects resulting from these 750


interactions to correctly judge their clinical impact and ponder it against the benefits of drug administration. Concluding remarks In this review, we have shown that there is reason for growing concern that epigenetic, possibly heritable, changes in gene expression, superimposed on DNA nucleotide sequences, may be caused by drugs that have not been developed to produce these effects (Box 3). An increasing number of reports show that epigenetic modifications are associated with drug administration, and in many cases this association is likely to be causal. Drugs can interfere at several stages of epigenome control, including direct chemical interactions with epigenetic key enzymes, such as the interaction of valproate with HDAC [14] or of hydralazine with DNMTs [41,42], interference via DNA repair [46] or with branching pathways to DNMTs that are activated following drug coupling at its G-protein-coupled target receptors [15]. Although current evidence for these effects is still limited to a few drugs and mechanisms, the computer predictions clearly show that the phenomenon is probably frequent. A single mechanism, such as the interaction of drug molecules with a specific epigenetic enzyme (HDAC), probably holds true for approximately 5% of all currently approved drugs. We derived this number from a computational study that predicted HDAC inhibition and has been validated for the statistically most significant prediction by in vitro binding experiments. We thereby have proven a known pharmacologically active compound to possess a previously unknown potential for epigenetic effects. This further stresses the importance of investigating such effects in drug discovery programs. Nevertheless, unintended, epigenetic effects of non-epigenetic drugs are not necessarily unwanted. They may contribute to the clinical efficacy of the drug, such as the interaction of valproic acid with HDAC, that probably adds to its effects on GABA and calcium channels and thus contributes to its effectiveness in diabetic neuropathic pain Box 3. The importance, frequency, and consequences of epigenetic drug effects Epigenetic effects, that is, altering transcription via changes in DNA methylation or histone conformation, can be exerted by a multitude of environmental factors including chemicals and drugs. While epigenetically acting therapeutics form a new class of drugs, increasing evidence suggests that common drugs are also among epigenetic modulators. These effects can be exerted via several molecular pathways interfering with most of the known epigenetic mechanisms. This phenomenon seems to be frequent as computational prediction of the interaction with only one of the epigenetic enzymes, HDAC, revealed that 5% of all known drugs interact with this enzyme. Epigenetic changes may alter the patient’s phenotype at several levels from the therapeutic drug effects via the disease to be treated up to pathophysiological changes unrelated to the primary therapy goal. Epigenetic changes may be inheritable. Therefore, drug-induced epigenetic changes need to be reliably identified by computational prediction combined with prospective laboratory tests and predictive screening models, which should be added to pharmaceutical drug development requirements to take a major step towards comprehensive drug safety assessment strategies.

Review therapy [100]. However, epigenetic effects may also reduce the effectiveness of the drug, such as putative opioidinduced hyperalgesia caused by hypermethylation of global methylation sites at LINE-1 [16], or the resistance to imatinib due to tumor suppressor gene PTEN promoter hypermethylation as a result of imatinib-mediated induction of DNMT3A [68,69]. Moreover, when considering their likely hereditary transmission [96], drug-induced epigenetic changes need to be reliably identified. The demonstration of the utility of computational predictions in this context suggests that approaches at epigenetic drug effects should embrace both in silico and in vitro assays in concert to increase the predictive performance while keeping the experimental costs at bay by preselecting candidate drugs that might interact with epigenetic mechanisms. Broad testing will reveal the realistic potential of modern computational prediction methods that are applied in concert with molecular biology techniques. The identification of putative transgenerational epigenetic drug effects of human relevance should be added to pharmaceutical drug development requirements to take a major step towards comprehensive drug safety assessment strategies. Acknowledgments This work has been financially supported by the ‘Landesoffensive zur Entwicklung wissenschaftlich-o¨ konomischer Exzellenz’: ‘LOEWESchwerpunkt: Anwendungsorientierte Arzneimittelforschung’ and TRIP (G.G., J.L.). We thank Wolf von Waldow for creating Figure 1.

Disclaimer statement G.S. is a scientific consultant to pharmaceutical industry and shareholder of inSili.com LLC, Zu¨rich, Switzerland and AlloCyte Pharmaceuticals AG, Basel, Switzerland. The authors have declared that no other competing interests exist.

References 1 Liu, D. et al. (1997) Maternal care, hippocampal glucocorticoid receptors, and hypothalamic–pituitary–adrenal responses to stress. Science 277, 1659–1662 2 Diorio, J. and Meaney, M.J. (2007) Maternal programming of defensive responses through sustained effects on gene expression. J. Psychiatry Neurosci. 32, 275–284 3 Meaney, M.J. (2001) Maternal care, gene expression, and the transmission of individual differences in stress reactivity across generations. Annu. Rev. Neurosci. 24, 1161–1192 4 Zhang, F.F. et al. (2011) Physical activity and global genomic DNA methylation in a cancer-free population. Epigenetics 6, 293–299 5 Christensen, B.C. et al. (2009) Aging and environmental exposures alter tissue-specific DNA methylation dependent upon CpG island context. PLoS Genet. 5, e1000602 6 Kamakura, M. (2011) Royalactin induces queen differentiation in honeybees. Nature 473, 478–483 7 Bollati, V. et al. (2007) Changes in DNA methylation patterns in subjects exposed to low-dose benzene. Cancer Res. 67, 876–880 8 Yoo, C.B. and Jones, P.A. (2006) Epigenetic therapy of cancer: past, present and future. Nat. Rev. Drug Discov. 5, 37–50 9 Minucci, S. and Pelicci, P.G. (2006) Histone deacetylase inhibitors and the promise of epigenetic (and more) treatments for cancer. Nat. Rev. Cancer 6, 38–51 10 Viatte, S. et al. (2013) Genetics and epigenetics of rheumatoid arthritis. Nat. Rev. Rheumatol. 9, 141–153 11 Doehring, A. et al. (2011) Epigenetics in pain and analgesia: an imminent research field. Eur. J. Pain 15, 11–16 12 Denk, F. and McMahon, S.B. (2012) Chronic pain: emerging evidence for the involvement of epigenetics. Neuron 73, 435–444 13 Csoka, A.B. and Szyf, M. (2009) Epigenetic side-effects of common pharmaceuticals: a potential new field in medicine and pharmacology. Med. Hypotheses 73, 770–780


14 Gottlicher, M. et al. (2001) Valproic acid defines a novel class of HDAC inhibitors inducing differentiation of transformed cells. EMBO J. 20, 6969–6978 15 Paradisi, A. et al. (2008) Anandamide regulates keratinocyte differentiation by inducing DNA methylation in a CB1 receptordependent manner. J. Biol. Chem. 283, 6005–6012 16 Doehring, A. et al. (2013) Chronic opioid use is associated with increased DNA methylation correlating with increased clinical pain. Pain 154, 15–23 17 Berger, S.L. et al. (2009) An operational definition of epigenetics. Genes Dev. 23, 781–783 18 Cervera, E. et al. (2012) Epigenetic therapy with hydralazine and magnesium valproate reverses imatinib resistance in patients with chronic myeloid leukemia. Clin. Lymphoma Myeloma Leuk. 12, 207–212 19 Rasoulpour, R.J. et al. (2011) Epigenetic screening in product safety assessment: are we there yet? Toxicol. Mech. Methods 21, 298–311 20 Lee, R.C. et al. (1993) The C. elegans heterochronic gene lin-4 encodes small RNAs with antisense complementarity to lin-14. Cell 75, 843– 854 21 Patra, S.K. et al. (2008) Demethylation of (cytosine-5-C-methyl) DNA and regulation of transcription in the epigenetic pathways of cancer development. Cancer Metastasis Rev. 27, 315–334 22 Joulie, M. et al. (2010) Mammalian methyl-binding proteins: what might they do? Bioessays 32, 1025–1032 23 Kim, J.K. et al. (2009) Epigenetic mechanisms in mammals. Cell. Mol. Life Sci. 66, 596–612 24 Pradhan, S. et al. (1999) Recombinant human DNA (cytosine-5) methyltransferase. I. Expression, purification, and comparison of de novo and maintenance methylation. J. Biol. Chem. 274, 33002– 33010 25 Herman, J.G. and Baylin, S.B. (2003) Gene silencing in cancer in association with promoter hypermethylation. N. Engl. J. Med. 349, 2042–2054 26 Ehrlich, M. et al. (1982) Amount and distribution of 5-methylcytosine in human DNA from different types of tissues of cells. Nucleic Acids Res. 10, 2709–2721 27 Bird, A.P. (1986) CpG-rich islands and the function of DNA methylation. Nature 321, 209–213 28 Sharma, S. et al. (2010) Epigenetics in cancer. Carcinogenesis 31, 27–36 29 Sananbenesi, F. and Fischer, A. (2009) The epigenetic bottleneck of neurodegenerative and psychiatric diseases. Biol. Chem. 390, 1145– 1153 30 Franchini, D.M. et al. (2012) 5-Methylcytosine DNA demethylation: more than losing a methyl group. Annu. Rev. Genet. 46, 419–441 31o Bhutani, N. et al. (2011) DNA demethylation dynamics. Cell 146, 866–872 32 Kouzarides, T. (2007) Chromatin modifications and their function. Cell 128, 693–705 33 Mujtaba, S. et al. (2007) Structure and acetyl-lysine recognition of the bromodomain. Oncogene 26, 5521–5527 34 Roth, S.Y. et al. (2001) Histone acetyltransferases. Annu. Rev. Biochem. 70, 81–120 35 Kuo, M.H. and Allis, C.D. (1998) Roles of histone acetyltransferases and deacetylases in gene regulation. Bioessays 20, 615–626 36 Kucharski, R. et al. (2008) Nutritional control of reproductive status in honeybees via DNA methylation. Science 319, 1827–1830 37 Spannhoff, A. et al. (2011) Histone deacetylase inhibitor activity in royal jelly might facilitate caste switching in bees. EMBO Rep. 12, 238–243 38 Rosenberg, G. (2007) The mechanisms of action of valproate in neuropsychiatric disorders: can we see the forest for the trees? Cell. Mol. Life Sci. 64, 2090–2103 39 Marchion, D.C. et al. (2005) Valproic acid alters chromatin structure by regulation of chromatin modulation proteins. Cancer Res. 65, 3815–3822 40 Cornacchia, E. et al. (1988) Hydralazine and procainamide inhibit T cell DNA methylation and induce autoreactivity. J. Immunol. 140, 2197–2200 41 Arce, C. et al. (2006) Hydralazine target: from blood vessels to the epigenome. J. Transl. Med. 4, 10 42 Singh, N. et al. (2009) Molecular modeling and molecular dynamics studies of hydralazine with human DNA methyltransferase 1. ChemMedChem 4, 792–799 751

Review 43 Candelaria, M. et al. (2011) Hydralazine and magnesium valproate as epigenetic treatment for myelodysplastic syndrome. Preliminary results of a phase-II trial. Ann. Hematol. 90, 379–387 44 Schafer, A. et al. (2010) Gemcitabine functions epigenetically by inhibiting repair mediated DNA demethylation. PLoS ONE 5, e14060 45 Fagbemi, A.F. et al. (2011) Regulation of endonuclease activity in human nucleotide excision repair. DNA Repair (Amst.) 10, 722–729 46 Barreto, G. et al. (2007) Gadd45a promotes epigenetic gene activation by repair-mediated DNA demethylation. Nature 445, 671–675 47 Nielsen, D.A. et al. (2009) Increased OPRM1 DNA methylation in lymphocytes of methadone-maintained former heroin addicts. Neuropsychopharmacology 34, 867–873 48 Kazazian, H.H., Jr (2004) Mobile elements: drivers of genome evolution. Science 303, 1626–1632 49 Baillie, J.K. et al. (2011) Somatic retrotransposition alters the genetic landscape of the human brain. Nature 479, 534–537 50 Snyder, S.H. (2011) Serotonin, cytokines, p11, and depression. Proc. Natl. Acad. Sci. U.S.A. 108, 8923–8924 51 Warner-Schmidt, J.L. et al. (2011) Antidepressant effects of selective serotonin reuptake inhibitors (SSRIs) are attenuated by antiinflammatory drugs in mice and humans. Proc. Natl. Acad. Sci. U.S.A. 108, 9262–9267 52 Dell’aversano, C. et al. (2009) Antipsychotic and antidepressant cotreatment: effects on transcripts of inducible postsynaptic density genes possibly implicated in behavioural disorders. Brain Res. Bull. 79, 123–129 53 Manev, R. et al. (2001) Fluoxetine increases the content of neurotrophic protein S100b in the rat hippocampus. Eur. J. Pharmacol. 420, R1–R2 54 Svenningsson, P. and Greengard, P. (2007) p11 (S100A10) – an inducible adaptor protein that modulates neuronal functions. Curr. Opin. Pharmacol. 7, 27–32 55 Svenningsson, P. et al. (2006) Alterations in 5-HT1B receptor function by p11 in depression-like states. Science 311, 77–80 56 Tzang, R.F. et al. (2008) Association study of p11 gene with major depressive disorder, suicidal behaviors and treatment response. Neurosci. Lett. 447, 92–95 57 Melas, P.A. et al. (2012) Antidepressant treatment is associated with epigenetic alterations in the promoter of P11 in a genetic model of depression. Int. J. Neuropsychopharmacol. 15, 669–679 58 Javierre, B.M. and Richardson, B. (2011) A new epigenetic challenge: systemic lupus erythematosus. Adv. Exp. Med. Biol. 711, 117–136 59 Pereira, M.A. et al. (2004) Modulation by celecoxib and difluoromethylornithine of the methylation of DNA and the estrogen receptor-a gene in rat colon tumors. Carcinogenesis 25, 1917–1923 60 Grosch, S. et al. (2006) Cyclooxygenase-2 (COX-2)-independent anticarcinogenic effects of selective COX-2 inhibitors. J. Natl. Cancer Inst. 98, 736–747 61 Maier, T.J. et al. (2009) Cellular membranes function as a storage compartment for celecoxib. J. Mol. Med. (Berl.) 87, 981–993 62 Liu, Y. et al. (2012) Phosphorylation of RelA/p65 promotes DNMT-1 recruitment to chromatin and represses transcription of the tumor metastasis suppressor gene BRMS1. Oncogene 31, 1143–1154 63 Funakoshi-Tago, M. et al. (2008) Celecoxib potently inhibits TNFainduced nuclear translocation and activation of NF-kB. Biochem. Pharmacol. 76, 662–671 64 Niederberger, E. et al. (2001) Celecoxib loses its anti-inflammatory efficacy at high doses through activation of NF-kB. FASEB J. 15, 1622–1624 65 Pytel, D. et al. (2009) Tyrosine kinase blockers: new hope for successful cancer therapy. Anticancer Agents Med. Chem. 9, 66–76 66 Kantarjian, H. et al. (2002) Hematologic and cytogenetic responses to imatinib mesylate in chronic myelogenous leukemia. N. Engl. J. Med. 346, 645–652 67 Cools, J. et al. (2005) Resistance to tyrosine kinase inhibitors: calling on extra forces. Drug Resist. Updat. 8, 119–129 68 Nishioka, C. et al. (2010) Long-term exposure of leukemia cells to multi-targeted tyrosine kinase inhibitor induces activations of AKT, ERK and STAT5 signaling via epigenetic silencing of the PTEN gene. Leukemia 24, 1631–1640 69 Nishioka, C. et al. (2011) Imatinib causes epigenetic alterations of PTEN gene via upregulation of DNA methyltransferases and polycomb group proteins. Blood Cancer J. 1, e48 752


70 Vire, E. et al. (2006) The polycomb group protein EZH2 directly controls DNA methylation. Nature 439, 871–874 71 Stone, A. et al. (2012) Tamoxifen-induced epigenetic silencing of oestrogen-regulated genes in anti-hormone resistant breast cancer. PLoS ONE 7, e40466 72 Wang, Y. et al. (2011) Fluoxetine increases hippocampal neurogenesis and induces epigenetic factors but does not improve functional recovery after traumatic brain injury. J. Neurotrauma 28, 259–268 73 Cassel, S. et al. (2006) Fluoxetine and cocaine induce the epigenetic factors MeCP2 and MBD1 in adult rat brain. Mol. Pharmacol. 70, 487–492 74 Nan, X. et al. (1998) Transcriptional repression by the methyl-CpGbinding protein MeCP2 involves a histone deacetylase complex. Nature 393, 386–389 75 Jones, P.L. et al. (1998) Methylated DNA and MeCP2 recruit histone deacetylase to repress transcription. Nat. Genet. 19, 187–191 76 Maze, I. et al. (2010) Essential role of the histone methyltransferase G9a in cocaine-induced plasticity. Science 327, 213–216 77 Sun, H. et al. (2012) Morphine epigenomically regulates behavior through alterations in histone H3 lysine 9 dimethylation in the nucleus accumbens. J. Neurosci. 32, 17454–17464 78 New, M. et al. (2012) HDAC inhibitor-based therapies: Can we interpret the code? Mol. Oncol. 6, 637–656 79 Schneider, G. et al. (1999) ‘Scaffold hopping’ by topological pharmacophore search: a contribution to virtual screening. Angew. Chem. Int. Ed. Engl. 38, 2894–2896 80 Reutlinger, M. et al. (2013) Chemically advanced template search (CATS) for scaffold-hopping and prospective target prediction for ‘orphan’ molecules. Mol. Inform. 32, 133–138 81 Schneider, P. and Schneider, G. (2003) Collection of bioactive reference compounds for focused library design. QSAR Comb. Sci. 22, 713–718 82 Kohonen, T. (1982) Self-organized formation of topologically correct feature maps. Biol. Cybernet. 43, 59–69 83 Schneider, G. et al. (2009) Self-organizing molecular fingerprints: a ligand-based view on drug-like chemical space and off-target prediction. Future Med. Chem. 1, 213–218 84 Schneider, P. et al. (2009) Self-organizing maps in drug discovery: compound library design, scaffold-hopping, repurposing. Curr. Med. Chem. 16, 258–266 85 Keiser, M.J. et al. (2007) Relating protein pharmacology by ligand chemistry. Nat. Biotechnol. 25, 197–206 86 Lounkine, E. et al. (2012) Large-scale prediction and testing of drug activity on side-effect targets. Nature 486, 361–367 87 Poroikov, V. et al. (2007) PASS: identification of probable targets and mechanisms of toxicity. SAR QSAR Environ. Res. 18, 101–110 88 Gregori-Puigjane, E. and Mestres, J. (2008) Coverage and bias in chemical library design. Curr. Opin. Chem. Biol. 12, 359–365 89 Baldi, P. and Nasr, R. (2010) When is chemical similarity significant? The statistical distribution of chemical similarity scores and its extreme values. J. Chem. Inf. Model. 50, 1205–1222 90 Gaulton, A. et al. (2012) ChEMBL: a large-scale bioactivity database for drug discovery. Nucleic Acids Res. 40, D1100–D1107 91 Rossello, A. et al. (2005) N-i-Propoxy-N-biphenylsulfonylaminobutylhydroxamic acids as potent and selective inhibitors of MMP-2 and MT1-MMP. Bioorg. Med. Chem. Lett. 15, 1321–1326 92 Oehme, I. et al. (2009) Targeting of HDAC8 and investigational inhibitors in neuroblastoma. Expert Opin. Investig. Drugs 18, 1605–1617 93 Haberland, M. et al. (2009) Epigenetic control of skull morphogenesis by histone deacetylase 8. Genes Dev. 23, 1625–1630 94 Wilson, B.J. et al. (2010) An acetylation switch modulates the transcriptional activity of estrogen-related receptor a. Mol. Endocrinol. 24, 1349–1358 95 Wolfson, N.A. et al. (2013) HDAC8 substrates: histones and beyond. Biopolymers 99, 112–126 96 Guerrero-Bosagna, C. and Skinner, M.K. (2012) Environmentally induced epigenetic transgenerational inheritance of phenotype and disease. Mol. Cell. Endocrinol. 354, 3–8 97 Ou, X. et al. (2012) Transgenerational inheritance of modified DNA methylation patterns and enhanced tolerance induced by heavy metal stress in rice (Oryza sativa L.). PLoS ONE 7, e41143 98 Franklin, T.B. et al. (2010) Epigenetic transmission of the impact of early stress across generations. Biol. Psychiatry 68, 408–415

Review 99 Hitchins, M.P. et al. (2007) Inheritance of a cancer-associated MLH1 germ-line epimutation. N. Engl. J. Med. 356, 697–705 100 Agrawal, R.P. et al. (2009) Management of diabetic neuropathy by sodium valproate and glyceryl trinitrate spray: a prospective doubleblind randomized placebo-controlled study. Diabetes Res. Clin. Pract. 83, 371–378 101 Duvic, M. and Vu, J. (2007) Vorinostat: a new oral histone deacetylase inhibitor approved for cutaneous T-cell lymphoma. Expert Opin. Investig. Drugs 16, 1111–1120 102 Yeo, W. et al. (2012) Epigenetic therapy using belinostat for patients with unresectable hepatocellular carcinoma: a multicenter phase I/II study with biomarker and pharmacokinetic analysis of tumors from patients in the Mayo Phase II Consortium and the Cancer Therapeutics Research Group. J. Clin. Oncol. 30, 3361–3367 103 Rambaldi, A. et al. (2010) A pilot study of the histone-deacetylase inhibitor givinostat in patients with JAK2V617F positive chronic myeloproliferative neoplasms. Br. J. Haematol. 150, 446–455 104 Vojinovic, J. et al. (2011) Safety and efficacy of an oral histone deacetylase inhibitor in systemic-onset juvenile idiopathic arthritis. Arthritis Rheum. 63, 1452–1458


105 Grant, C. et al. (2010) Romidepsin: a new therapy for cutaneous T-cell lymphoma and a potential therapy for solid tumors. Expert Rev. Anticancer Ther. 10, 997–1008 106 Pili, R. et al. (2012) Phase I study of the histone deacetylase inhibitor entinostat in combination with 13-cis retinoic acid in patients with solid tumours. Br. J. Cancer 106, 77–84 107 Ghoshal, K. and Bai, S. (2007) DNA methyltransferases as targets for cancer therapy. Drugs Today (Barc.) 43, 395–422 108 Maze, I. et al. (2011) Cocaine dynamically regulates heterochromatin and repetitive element unsilencing in nucleus accumbens. Proc. Natl. Acad. Sci. U.S.A. 108, 3035–3040 109 Abuhammad, A.M. et al. (2010) Probing the architecture of the Mycobacterium marinum arylamine N-acetyltransferase active site. Protein Cell 1, 384–392 110 Kuck, D. et al. (2010) Novel and selective DNA methyltransferase inhibitors: docking-based virtual screening and experimental evaluation. Bioorg. Med. Chem. 18, 822–829 111 Reutlinger, M. and Schneider, G. (2012) Nonlinear dimensionality reduction and mapping of compound libraries for drug discovery. J. Mol. Graph. Model. 34, 108–117

753

Trends in Molecular Medicine - December 2013

Recommend Documents