RNA editing, epitranscriptomics, and processing in cancer progression

Abstract

The transcriptome is extensively and dynamically regulated by a network of RNA modifying factors. RNA editing enzymes APOBEC (apolipoprotein B mRNA editing enzyme, catalytic polypeptide-like) and ADAR (adenosine deaminase, RNA-specific) irreversibly recode primary RNA sequences, whereas newly described methylases (writers) and de-methylases (erasers) dynamically alter RNA molecules in response to environmental conditions. RNA modifications can affect RNA splicing, nuclear-cytoplasmic transport, translation, and regulation of gene expression by RNA interference. In addition, tRNA base modifications, processing, and regulated cleavage have been shown to alter global patterns of mRNA translation in response to cellular stress pathways. Recent studies, some of which were discussed at this workshop, have rekindled interest in the emerging roles of RNA modifications in health and disease. On September 10th, 2014, the Division of Cancer Biology, NCI sponsored a workshop to explore the role of epitranscriptomic RNA modifications and tRNA processing in cancer progression. The workshop attendees spanned a scientific range including chemists, virologists, and RNA and cancer biologists. The goal of the workshop was to explore the interrelationships between RNA editing, epitranscriptomics, and RNA processing and the enzymatic pathways that regulate these activities in cancer initiation and progression. At the conclusion of the workshop, a general discussion focused on defining the major challenges and opportunities in this field, as well as identifying the tools, technologies, resources and community efforts required to accelerate research in this emerging area.

Keywords:

• ADAR
• APOBEC
• editing
• epitranscriptomic
• DNA mutations
• methylation
• RNA modification
• tRNA

Introduction

Chemical modifications of proteins and DNA have recognized roles in regulating gene expression in normal development and disease. In particular, recent advances in characterizing epigenetic modifications to histones and DNA have transformed our understanding of how gene expression is regulated. However, until recently, little attention has been given to understanding the role of RNA modifications in regulating gene expression and associated cellular functions. Ironically, the presence of RNA modifications, such as the 5′-trimethylguanosine cap of nascent mRNA, mRNA adenosine methylation, and various tRNA chemical modifications have been known for decades. Beyond these well-known examples, at least 110 additional RNA modifications have been described in the literature.¹ Recent advances in high-throughput analyses have demonstrated that the transcriptome is extensively and dynamically altered by a variety of mechanisms. To assess the impact of RNA processing and modifications on cancer, the Division of Cancer Biology (DCB) at the National Cancer Institute (NCI) recently organized a workshop to explore the role(s) of RNA editing, epitranscriptomics, and tRNA modifications and the enzymatic pathways that regulate these activities in cancer progression. A panel of experts, chaired by John Coffin (Tufts), Samie Jaffrey (Weill-Cornell), and Peter Dedon (MIT), summarized the current state of the field and highlighted promising research directions and opportunities that could enhance our understanding of how these modifications affect cancer initiation and progression. The meeting summary is presented here to highlight research priorities and to stimulate basic and translational research to generate a better understanding of the molecular mechanisms regulating epitranscriptomic processes and their involvement in carcinogenesis.

APOBEC and ADAR RNA Editing

RNA editing enzymes APOBEC and ADAR irreversibly recode RNA sequences by catalyzing cytidine-to-uridine (C-to-U) and adenosine-to-inosine (A-to-I) editing, respectively. The human APOBEC family of polynucleotide cytosine deaminases comprises 9 active enzymes: APOBEC1, AID, and APOBEC3 A/B/C/D/F/G/H. Although the founding member of this family was initially implicated in RNA editing, these enzymes also play fundamental roles in DNA editing processes central for antibody diversification and innate immunity to a wide variety of DNA-based parasites including exogenous viruses and endogenous transposons. Moreover, recent reports have also revealed that APOBEC mediated editing is responsible for characteristic DNA mutations in multiple cancer types including bladder, cervical, breast, head and neck, and lung cancers.² The ADAR family is comprised of 3 members, ADAR1 (which include 2 isoforms ADAR1p150 and ADAR1p110), ADAR 2, and ADAR 3. ADARs primarily edit adenosine residues in 3′ UTR and intronic regions of both coding and non-coding RNAs, suggesting a prominent regulatory role in mRNA translation and regulation of the RNAi pathway.³

John Coffin (Tufts University) opened the workshop with an overview of the APOBEC and ADAR editing enzyme families, their evolution, normal cellular functions, and potential roles in cancer initiation and progression. APOBECs have a strong preference for ssDNA substrates as part of their adaptive and innate immune functions. For example, family member Activation-Induced Deaminase (AID) is essential for antibody gene diversity and class switching via somatic hypermutation. AID is a likely contributor to mutagenesis and driver events in B cell cancers and may also play a role in some forms of gastric and liver cancer. APOBEC1 has both RNA and DNA editing functions, targeting both host and viral sequences, and contributing to viral hypermutation. C-to-U transition mutations in mature mRNAs, catalyzed by APOBEC1, can introduce stop codons or missense mutations potentially producing truncated proteins that have lost regulatory regions or have attained new carcinogenic functions. ADAR mutations are part of the genetically heterogeneous spectrum of molecular lesions in Aicardi–Goutières syndrome (AGS), which leads to cellular nucleic acid accumulation disorders and the aberrant activation of the type-1 interferon inflammatory immune responses. Dr. Coffin noted that many viruses have evolved a variety of evasion strategies to deal with host APOBEC immune activities and these evasion strategies may have carcinogenic consequences for infected hosts. For example, Vif is an HIV-encoded protein that recruits a cellular E3 ubiquitin ligase that targets APOBEC3s for proteasomal degradation. Indeed, APOBECs most likely play a key role in protecting humans against oncogenic effects of infection by viruses, such as the murine and feline leukemia viruses found in a number of animals with which we are in close contact. In contrast, a lack of APOBEC function can leave affected hosts susceptible to retrovirus expansion and integration into the germline genome (endogenization), such as has been observed in the recent Koala retrovirus-induced (KoRV) lymphoma outbreak in Australia. Similarly, functional deficiencies in the ADAR enzyme family have been observed in pathogenic syndromes at higher risk for the development of cancer.

The role of editing enzyme substrate specificity, in particular the ability of APOBECs to edit DNA, was discussed by Reuben Harris (University of Minnesota). Dr. Harris reviewed the historic milestones that have guided our understanding of APOBEC cytosine to uracil conversions in DNA, including hierarchical details of the family tree, which was followed by observations that APOBEC directed cytosine mutations have been identified in many tumor genomes.⁴ The 7 APOBEC3 enzyme subfamily members have strong biochemical preferences for single-stranded DNA substrates and have been broadly implicated in providing innate immunity to parasitic DNA elements including naked DNA, viruses, and transposons. Recent studies from Dr. Harris’ lab and others have described APOBEC3B mutagenesis in a variety of human cancers including breast, head/neck, cervical, bladder, and lung). Key supporting evidence included expression data showing APOBEC3B upregulation in tumors but not matched normal tissues, correlations with APOBEC3B expression levels and tumor mutation loads, biochemical deamination preferences resembling the observed cytosine mutation spectra in tumors, and higher expression levels being predictive for poor clinical outcomes for breast cancer. Moreover, APOBEC3 deamination also explains the phenomenon of kataegis, or clustered mutations, recently reported in whole breast cancer genomic DNA sequences.⁵ Dr. Harris also presented a molecular link between viral infection and APOBEC3B upregulation in head/neck and cervical cancers, including the first data to show that the E6 oncoprotein from high-risk HPV isolates is alone sufficient to induce APOBEC3B upregulation. In addition to biological functions in RNA editing, Dr. Harris also noted that APOBEC1 also has robust DNA cytosine editing activity and has been implicated recently in esophageal adenocarcinomas. Overall, Dr. Harris emphasized an emerging theme in which APOBEC-mediated DNA cytosine to uracil deamination underlies a mutator phenotype in many different cancers.

Gordon Carmichael (University of Connecticut) followed with a presentation on ADAR editing induced by virus infection. ADAR functions can be either anti-viral or proviral, the latter providing an advantage to the virus. Many viruses express dsRNA during their life cycles and supply substrates for RNA editing by ADAR.⁶ However, the vast majority of viruses have not been studied for their interaction with the ADARs. Current examples of ADAR editing of viral RNA include measles virus, hepatitis delta virus (HDV), and polyoma virus. Measles virus studies revealed the importance of ADAR1p150 (the large interferon-inducible isoform) acting as an antiviral host factor. ADAR1p150 hypermutates the measles genome especially within the matrix gene, which results in reduced measles matrix protein translation and lower infectivity. Proviral ADAR1 activities include utilization of the A-to-I editing function by HDV to produce essential capsid proteins and A-to-I editing of poly (A) sites by oncogenic polyomaviruses to provide the switch from early to late transcription. Dr. Carmichael concluded with an overview of how edited sites can be identified and which approaches might be the most useful for genome-wide detection of ADAR edited sites.

Kazuko Nishikura (Wistar Institute) presented recent data from her lab describing the role of ADAR adenosine deaminase regulation of miRNA biogenesis and function. Preferred ADAR substrates contain repetitive adenosine-rich sequences located within untranslated regions of mRNAs (i.e. introns, and 5′ and 3′ UTRs), which generate stem-loop structures similar to those in primary miRNA transcripts. Dr. Nishikura's group previously demonstrated that direct interactions between RNAi and RNA editing mechanisms control miRNA biogenesis and alter miRNA target sequence. More recently they have shown that the loading of the EBV (Epstein-Barr-virus) encoded miRNA, miR-BART6-5p, onto the host RISC processing complex is inhibited by A-to-I editing of the primary transcript (pri-miR-BART6) by ADAR1. Moreover, 4 binding sites of miR-BART6-5p were identified within the human Dicer mRNA 3′UTR, revealing a unique strategy for EBV to manipulate the host RNAi machinery.⁷ Since Dicer silencing has been linked to tumor epithelial-to-mesenchymal transition (EMT) and metastasis, Dr. Nishikura is investigating the role of miR-BART6-5p in the regulation of EMT in EBV-positive breast cancer as well as the inhibitory role of ADAR1 in the control of miR-BART6-5p functions. The study is anticipated to reveal viral miRNA-mediated mechanisms that affect the carcinogenic outcomes of EBV infections associated with a variety of human cancers such as Burkett's lymphoma, Hodgkin's disease, and nasopharyngeal carcinoma.

Epitranscriptomics

In addition to RNA editing, the transcriptome is known to be extensively and dynamically altered by a variety of reversible RNA modifications including N⁶-methyladenosine (m⁶A) and 5-methlycytosine (m⁵C).^8,9 The biological significance of m⁶A and m⁵C has only recently emerged due to the identification of the writers, erasers, and readers that regulate the transcriptome through these modifications. For example, the human fat mass and obesity associated protein (FTO) is an m⁶A demethylase (eraser) that links reversible RNA epigenetic marks with nutrient sensing and energy homeostasis, and the YTH domain family 2 (YTHDF2) was recently identified as an m⁶A reader involved in regulating mRNA stability.^10,11 However, molecular characterization of the epitranscriptomic landscape and the enzyme systems that regulate the various reversible RNA modifications has only just begun.

Samie Jaffrey (Weill Cornell Medical College) opened the epitranscriptomics session by noting that internal methylated adenosines in RNA molecules (in contrast to the 5′ methyl cap structure) had been suspected since the early 1970s, but that interest waned due to technical challenges. However, recent advances have stimulated resurgence of studies of RNA modifications. In particular, the development of specific antibodies to N⁶-methyladenosine (m6A) followed by next generation sequencing (MeRIP-seq) has allowed mapping of transcriptome-wide distributions of m6A modifications. Dr. Jaffrey presented work from his lab, in collaboration with Chris Mason, in which thousands of m6A peaks were identified in both coding and non-coding RNAs. He further described the distribution of m6A across genes, in particular noting enrichment of m6A in both the 5′ untranslated regions (UTRs) and near mRNA stop codons. In addition, a consensus sequence for m6A modifications was mapped to purine-purine-adenosine-cytosine-uracil (RRA*CU) sites. Switching gears, Dr. Jaffrey described the roles of the methyltransferase like 3 (MTTL3) and WTAP components of the multi-protein methyltransferase complex required for introducing the m6A modification. Dr. Jaffrey also discussed in vitro evidence from his lab and others showing that adenosine methylation is reversible and that FTO and its homolog ALKBH5 can demethylate RNA.¹² Next, Jaffrey presented some of the proposed functional roles for m6A. Knockout studies have implicated proteins associated with regulating m6A modifications in stem cell pluripotency, gametogenesis, spermatogenesis, and other processes. Further, FTO knockout mice have altered neurotransmission as evidenced by the fact that they do not respond as expected to dopamine surges.¹³ Lastly, Dr. Jaffrey described potential roles for m6A modifications in regulating mRNA translation. Dr. Jaffrey ended his presentation by proposing that cancer-specific translation may occur through cancer-induced methylation pathways that influence the translation of specific cohorts of mRNAs.

In the second talk, Jing Crystal Zhao (Sanford Burnham Medical Research Institute) described her lab's efforts to understand the functional mechanisms of m6A RNA modification in mouse embryonic stem cells. As a first step, Dr. Zhao focused on the enzymes that write and erase the m6A modifications. While FTO and ALKBH5 are known to function as m6A demethylases and METTL3 is considered a potential m6A methyltransferase, no in vitro methylation assays have confirmed METTL3 RNA methyltransferase activity and no studies have shown that knockdown of METTL3 reduces m6A levels. Additionally, METTL3 is only one member of the methyltransferase like (METTL) protein family and it is possible that other family members could serve as the m6A methyltransferase. Using mouse embryonic stem cells (mESCs) for her studies, Dr. Zhao demonstrated that knockdown of METTL3 or METTL14 reduced m6A levels by 40–60%, that METTL3 and METTL14 can methylate RNA in vitro, and that there is significant overlap in genome-wide METTL3 and METTL14 m6A targets indicating that METTL3 and METTL14 are mammalian m6A methyltransferases. Next, Dr. Zhao presented data showing that knockdown of METTL3 or METTL14 resulted in the loss of self-renewal capability, the downregulation of pluripotency genes, and the upregulation of developmental and differentiation genes. Finally, Dr. Zhao presented data showing that m6A decreased the stability of mRNA by diminishing interactions with the RNA stabilizing protein HuR. Taken together, the data suggest a model whereby METTL3 and METTL14 methylate mRNA by generating m6A which inhibits HuR binding and results in increased mRNA decay.¹⁴ Zhao finished her presentation by suggesting future areas of research including investigating m6A target specificity and elucidating m6A functional pathways.

Chris Mason (Weill Cornell Medical College) concluded the Epitranscriptomics session by examining the role of methylation in cancer. He described how the central dogma of molecular biology (DNA to RNA to protein) has been complicated by the discovery and investigation of the epigenome and now the epitranscriptome, which add additional layers of modulation. Epigenetic modifiers are commonly and recurrently mutated in leukemia, glioblastoma, and colon cancer. Current experimental and analytical methodologies can map DNA methylation profiles at base-pair resolution. Dr. Mason noted that the epitranscriptome may be even more expansive than the epigenome, since more than 100 RNA modifications have already been cataloged in the RNA Modifications Database,¹⁵ and mapping the epitranscriptome has only just started. Dr. Mason suggested that this vast array of modifications likely has the capacity to tightly modulate the functions of RNA. MeRIPPER, an open source tool for analyzing MeRIP-seq data developed in the Mason lab, was used to identify enriched peaks of m6A modifications.¹⁶ Metagene analysis showed multiple peaks of m6A enrichment, including peaks near mRNA stop codons. Interestingly, DNA sequences under the stop codon peaks are highly conserved, indicating selective constraints for m6A modifications at these locations. Integrative analysis of m6A locations and sites of RNA editing revealed that these 2 levels of RNA regulation are mutually exclusive, leading Mason to posit that m6A may attenuate RNA editing. Next Dr. Mason looked at the role of m6A and the epitranscriptome in cancer. He used the CBioPortal ¹⁷ to demonstrate that genomic alterations of readers, writers, and erasers of RNA modifications are prevalent in a number of cancers. The genotypes and phenotypes of cancers are known to evolve over the time from original diagnosis to relapse, and Mason showed preliminary data demonstrating that the epitranscriptome likewise evolves in cancer. Mason finished his presentation by highlighting next generation single-molecule and direct RNA sequencing tools that have the potential to move the field forward.¹⁸

tRNA Modifications and Processing

tRNAs have long been known to be the most extensively modified RNA species.¹⁹ However, recent studies suggesting that tRNA modifications regulate global translational responses to cellular stress have provoked a reexamination of a decades-old finding that tRNAs are altered in cancer cells relative to normal tissues.^20,21

In addition, recent advances in high-throughput analyses of tRNA modifications have sparked a resurgent interest in the role of tRNA modifications and processing in normal developmental processes and cancer.

Peter Dedon (Massachusetts Institute of Technology) kicked off the tRNA Modifications and Processing session with a tour of the tRNA molecule, highlighting key steps in tRNA biogenesis and the myriad chemical modifications that influence tRNA structure and function. He described the elegant choreography of the tRNA maturation process and the disease states that result from missteps in this process. After introducing the session's speakers, Dr. Dedon shared data resulting from collaborative work between his laboratory and Thomas Begley's (University at Albany – SUNY), revealing how environmental stress leads to the reprogramming of dozens of modified ribonucleosides in tRNA in order to enhance translation of transcripts encoding stress response proteins. Encouraged by data from Begley's laboratory that oxidative stress leads to increased 5-methoxycarbonylmethyluridine (mcm⁵U) modification of tRNA, they took advantage of innovations in mass spectrometry-based technology to look both globally and quantitatively at specific base and ribose modifications in yeast tRNAs in response to stress.²⁰ Dr. Dedon described an intricate system in which the levels of different tRNA modifications are dynamically “reprogrammed” in response to exposure to distinct stressors, resulting in an altered ability to decode particular degenerate codons that appear more frequently in the genes that encode stress response proteins. While the talk focused on a single wobble modification in tRNA^LEU that is enhanced following exposure to oxidative stress, Dr. Dedon noted other examples of this mechanism in studies with Dr. Begley, and they speculate that this is a general system whereby the genetic code and specific tRNA modifications cooperate to ensure efficient translation of proteins needed to survive many different types of environmental stress.

Thomas Begley (University at Albany – SUNY) continued on the stress response theme, but moved the discussion into mammalian systems and cancer models. Enzymes required for the response to reactive oxygen species (ROS) frequently contain the unusual amino acid selenocysteine, which is translated from an internal UGA stop codon by a specialized tRNA and accessory proteins. This tRNA^SEC contains a mcm⁵U modification at the wobble position, which is added by the ALKBH8 enzyme, one of 2 human homologs (the other being hTRM9L) of the yeast Trm9 tRNA methyltransferase. Interestingly, these 2 homologs appear to have opposing roles in tumorigenesis, as ALKBH8 activity is required for growth of bladder cancer tumors, whereas hTRM9L is a tumor growth suppressor. Begley showed that ALKBH8-deficient mice lack the mcm⁵Um modification in their tRNA, and consequently have a compromised ROS response, which leads to elevated ROS levels and increased accumulation of DNA damage. Over-expression of ALKBH8 may lead to over-active ROS detoxification, which would allow for adaptation to higher ROS levels and ultimately a cancer-supporting state. Normally, ALKBH8 and hTRM9L levels are kept in tight balance. However, hTRM9L is silenced in colon cancer models, and over expression of hTRM9L expression in colorectal cancer cells prevents tumor growth.²² Putting these data together, Dr. Begley presented a model whereby ALKBH8 and hTRM9L may work together to control ROS, DNA damage, and tumor growth. ALKBH8 turns on translation of the selenocysteine-containing proteins needed for ROS detoxification and DNA repair, while hTRM9L may prevent their translation or lead to their degradation by an unknown mechanism. Skewing the ratio of these 2 homologs in either direction affects ROS management and the stress response, ultimately having the potential to affect tumor growth characteristics.

The process of tRNA maturation and trafficking was presented by Anita Hopper (The Ohio State University), who focused on the subset of 10 yeast tRNA families that contain an intron. After extensive processing and modification in the nucleus, these tRNAs get exported to the cytoplasm for splicing, additional chemical modification, and aminoacylation. Surprisingly, these mature tRNAs are imported back into the nucleus by a process called retrograde transport before being re-exported to the cytoplasm for use in protein synthesis.²³ This retrograde transport process is a form of quality control, monitoring mature tRNAs to ensure that they are correctly processed and modified before they enter the pool available for protein synthesis. By monitoring end-processing as well as tRNA modification state, this process enhances surveillance by the previously identified yeast rapid tRNA decay pathway. These 2 processes work together to assure that tRNAs present in the cytoplasm can translate with fidelity. While work from Dr. Hopper and others has greatly increased our knowledge of tRNA processing enzymes, there are still many open questions and many enzymes waiting to be identified. For this reason, Dr. Hopper and colleagues embarked on a genome-wide search for mutations that affect tRNA biology. Taking advantage of yeast knock-out collections and 2 different libraries containing temperature-sensitive forms of nearly every essential gene, they used Northern blot analysis to examine the effect of individually disrupting every yeast gene on tRNA biogenesis. 85 novel non-essential genes and 83 essential genes were identified that when disrupted affect tRNA processing, stability, or modification. Many of the genes discovered in this screen have known human homologs. Following up on one intriguing phenotype from the screen, Dr. Hopper and her student, J. Wu, discovered that intron degradation in yeast requires that the intron first be phosphorylated by the enzyme with both tRNA ligase and kinase activities that ligates exons after splicing.²⁴ Notably, knocking down the human kinase Clp1, also involved in splicing tRNA introns, leads to the accumulation of free introns in human cells. Point mutations in this gene have been found in patients with neurological disorders, raising the intriguing question of whether aberrant accumulation of tRNA introns could be responsible for neurological phenotypes.

Paul Anderson (Harvard Medical School) described stress-induced cleavage of tRNAs by the secreted ribonuclease angiogenin, which targets the anticodon loop to form 2 tRNA halves, called tiRNAs. Although only 1–5% of the total tRNAs are cleaved, tiRNA formation significantly decreases total cellular translational activity by partnering with the known translational inhibitor YB-1 to displace the active translational initiation factor complex required for 7-Methylguanosine (m7G) cap-dependent translation. tiRNAs also induce assembly of cytoplasmic stress granules, which contain stalled translation pre-initiation complexes. While angiogenin cleavage creates both 5′ and 3′ tRNA halves, only 5′ tiRNAs blocks translation and form stress granules. Interestingly, tRNA^ALA- and tRNA^CYS- derived 5′ fragments are the most active; both contain a conserved motif at their 5′ ends, which recruits YB-1.²⁵ However, tiRNAs do not affect translation of proteins that circumvent the m7G cap to initiate protein synthesis, for instance by using an internal ribosome entry site (IRES) or upstream open reading frame (uORF). In this way, tiRNA production enhances cell survival by allowing preferential translation of IRES- or uORF-containing proteins, including many proteins required for the cellular stress response. Indeed, Anderson found that DNA oligos mimicking the 5′ tiRNA^ALA sequence caused a modest but significant cytoprotective response in motor neurons and osteosarcoma cells exposed to toxic conditions. Point mutations in angiogenin have been found in the neurodegenerative disease Amyotrophic lateral sclerosis (ALS), and tiRNAs are over-expressed in prostate cancer where they enhance cell survival and support progression toward a metastatic state.

Workshop Summary

At the conclusion of the workshop, a round table discussion led by Joan Steitz (Yale) focused on defining the major challenges and opportunities in this emerging research area. The discussion focused on resources, tools and technologies, knowledge gaps, and community efforts that will be required to accelerate research. Key points are summarized below.

Resource Building:

• One overarching concern identified by workshop participants was the need to establish a comprehensive catalog of all possible RNA modifications organized by RNA type (mRNA, tRNA, rRNA, etc.), cell or tissue type in which specific RNA modifications occur, and how different environmental conditions (i.e., healthy versus cancer) affect RNA modification status.

• In addition to surveying known RNA modifications, discovery of additional unknown RNA modifications is also an important goal.

• Annotating this collection by complementing the RNA base modifications with associated orthogonal genetic, epigenetic, and transcriptional data would enhance the power and utility of this resource.

• Developing specific reagents, such as monoclonal antibodies, to detect, measure, quantify, and manipulate RNA (and DNA) editing enzymes and RNA epitranscriptomic modifying enzymes will be essential for rapid progress.

  Overcoming Knowledge Gaps:

• Many fundamental questions concerning the regulation of ADAR/APOBEC expression, activity, and potential carcinogenic role(s) through effects on the genome, transcriptome, and/or proteome need to be better understood.

• Identification of the specific substrates for RNA modifying enzymes and the precise target sites that direct their activities are needed. Such data would determine whether the substrates are RNA or DNA (or both), coding or noncoding RNA, whether the modifications are reversible or non-reversible, and what precise sequence features are required for target elements.

• There needs to be a comprehensive accounting of the enzyme systems that regulate reversible RNA modifications to fully characterize the role of epitranscriptomics in normal development and disease.

• Overcoming these barriers will require developing new methodologies in addition to computational or bioinformatic approaches linked to the development of innovative tools (see below).

  Tool Development:

• New tools for high-resolution (base-pair resolution) mapping of m6A in the mammalian transcriptome are needed to manipulate and follow the fate of RNA modifications to functionally characterize these pathways. Antibody libraries specific for each RNA modification, as well as creative approaches to map specific sites where RNA modifications occur in the transcriptome and which modified bases are bound by RNA modifying proteins, need to be developed to allow capture of all modified RNA bases.

• In addition to discovery of additional RNA modifications and the enzyme systems that orchestrate their deposition in the transcriptome, we need more tools to manipulate RNA modifying enzymes or agents (or the modifications themselves), chemical probes to detect and track the fate of RNA modified bases, and pharmacologic inhibitors to allow RNA modifications to be experimentally manipulated in vivo and in vitro.

• Lastly, improved animal model systems are needed to better understand the consequences of RNA and DNA editing processes in vivo. As one example, mice – in contrast to humans - express only one APOBEC enzyme, and it is poorly induced by interferon, indicating it is regulated differently in mice than in humans. Transgenic mice expressing some or the entire repertoire of human APOBEC proteins would help enhance research.

  Comparative Studies:

• It will be critical to understand how appropriate studies using lower organisms will inform our understanding of the extent and role of RNA modifications in humans. For example, high-throughput assays in lower organisms could facilitate the discovery of the RNA modifying enzyme systems that could be validated in more appropriate mammalian models.

  Team Building:

• These studies will require the efforts of collaborative teams containing complementary skillsets such as chemists, bioinformaticians, nucleic acid biologists, virologists, and cancer researchers. The workshop participants recognized the precedent that formation of highly collaborative consortiums has fostered more rapid progress than could be attained by traditional investigator-initiated single grant mechanisms.

Addressing these fundamental needs will enable rigorous hypothesis-driven studies to fully exploit this emerging research area and elucidate central questions such as what cellular and environmental signals regulate RNA modification and processing events, how RNA modification and processing affect cancer processes, and whether these systems can be targeted therapeutically.

Attendees: Paul Anderson (Brigham and Women's Hospital; Boston, MA), Tom Begley (University at Albany, SUNY; Albany, NY), Gordon Carmichael (University of Connecticut Health Center; Farmington, CT), John Coffin (Tufts University; Boston, MA), Phil Daschner (National Cancer Institute, NIH; Bethesda, MD), Peter Dedon (Massachusetts Institute of Technology; Cambridge, MA), Dan Gallahan (National Cancer Institute, NIH; Bethesda, MD), Sean Hanlon (National Cancer Institute, NIH; Bethesda, MD), Reuben Harris (University of Minnesota; Minneapolis, MN), Anita Hopper (The Ohio State University; Columbus, OH), Samie Jaffrey (Weill Medical College, Cornell University; New York, NY), Christopher Mason (Weill Medical College, Cornell University; New York, NY), Kazuko Nishikura (The Wistar Institute; Philadelphia, PA), Betsy Read-Connole (National Cancer Institute, NIH; Bethesda, MD), Dinah Singer (National Cancer Institute, NIH; Bethesda, MD), Joan Steitz (HHMI, Yale University; New Haven, CT), Jenny Strasburger (National Cancer Institute, NIH; Bethesda, MD), Keren Witkin (National Cancer Institute, NIH; Bethesda, MD), Jing Crystal Zhao (Sanford Burnham Medical Research Institute; La Jolla, California.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.