ARTICLES NATURE METHODS | VOL.9 NO.2 | FEBRUARY 2012 | 189 Informational recoding by adenosine-to-inosine RNA editing diversifies neuronal proteomes by chemically modifying structured mRNAs. However, techniques for analyzing editing activity on substrates in defined neurons in vivo are lacking. Guided by comparative genomics, here we reverse-engineered a fluorescent reporter sensitive to Drosophila melanogaster adenosine deaminase that acts on RNA (dADAR) activity and alterations in dADAR autoregulation. Using this artificial dADAR substrate, we visualized variable patterns of RNA-editing activity in the Drosophila nervous system between individuals. Our results demonstrate the feasibility of structurally mimicking ADAR substrates as a method to regulate protein expression and, potentially, therapeutically repair mutant mRNAs. Our data suggest variable RNA editing as a credible molecular mechanism for mediating individual-to-individual variation in neuronal physiology and behavior. Adenosine-to-inosine (A-to-I) RNA editing results when adeno- sine is deaminated to inosine in double-stranded (ds)RNA sub- strates by ADARs 1 . In animals, a variety of mRNAs encoding proteins involved in fast electrical and chemical neurotransmis- sion are subject to A-to-I editing 2,3 , and ADAR activity is cru- cial for neuronal integrity and behavior in several genetic model organisms 4–6 . Editing of mRNA substrates by ADARs involves the formation of imperfect, structurally diverse dsRNA duplexes in exonic templates. These form in the presence of editing site complementary sequence elements, which are cis-acting comple- mentary regions often found in neighboring intronic regions 7,8 . Editing is generally studied as a bulk property via reverse transcriptase–PCR followed by analysis of cDNA clones or electropherograms. ADAR activity in small subpopulations of neurons is rarely studied in this way because of the technical demands of analyzing message diversity on such minute scales. This has precluded analysis of ADAR activity in individual neurons, particularly in relatively small organisms such as Drosophila. An alternative strategy relies on ADAR’s ability to edit the amber nonsense codon (UAG) to the tryptophan codon (UIG), which has been used to generate molecular reporters for editing activity by coupling a modified ADAR substrate containing an editable Visualizing adenosine-to-inosine RNA editing in the Drosophila nervous system James E C Jepson 1,2 , Yiannis A Savva 1 , Kyle A Jay 1,3 & Robert A Reenan 1 amber codon to a downstream output such as β-galactosidase 9–11 . However, to our knowledge no reporter system for editing activ- ity has been validated in vivo. Here we reverse-engineered a de novo editing site in the context of an artificial UAG stop codon in Aequorea victoria GFP mRNA, which results in full-length GFP only after splicing and A-to-I editing, thus yielding a fluorescent output dependent on dADAR activity. We used this reporter sys- tem to detect biologically relevant spatial and temporal endog- enous dADAR activity in the fly nervous system, responsiveness to an autoregulatory feedback loop in which dADAR edits its own transcript to fine-tune enzyme function 12,13 and inter-individual variation in neuronal dADAR activity. RESULTS Engineering a fluorescent reporter of ADAR activity To generate an in vivo reporter for dADAR activity, we used structural information gleaned from comparative genomics 8 . The Drosophila synaptotagmin-1 (Syt1) mRNA is edited at four posi- tions (A–D), of which editing at site D is most robust 2 . The struc- ture that directs editing at sites C and D has been partially solved and consists of a complex dsRNA pseudoknot formed between the edited exon and two editing site complementary sequence elements (E1 and E2) located in the downstream intron 8 . The duplex structure that directs editing at site D consists largely of intronic sequence (Fig. 1a). Using this fact, we generated a chimeric gene construct that would express full-length GFP only when acted upon both by RNA splicing and dADAR modification. First, we inserted the entire intron 9 of Syt1, containing all splicing signals and the E2 cis element that directs editing at site D, directly into the GFP coding sequence, artificially generating two GFP ‘exons’ (Fig. 1b). We arranged the GFP sequences such that the second position guanosine of a conserved GFP tryptophan codon (UGG) would be positioned at the normally edited third position adenosine of the isoleucine codon (AUA) in the RNA structure of Syt1. We mutated this tryptophan to an amber stop codon (UGG to UAG), recreat- ing a potentially editable adenosine at the analogous position to Syt1 site D. Finally, we engineered element E2 with structurally compensatory changes (as well as a single synonymous change in 1 Department of Molecular Biology, Cell Biology and Biochemistry, Brown University, Providence, Rhode Island, USA. 2 Department of Neuroscience, Thomas Jefferson University, Philadelphia, Pennsylvania, USA. 3 Department of Biochemistry and Biophysics, University of California, San Francisco, California, USA. Correspondence should be addressed to R.A.R. (robert_reenan@brown.edu). RECEIVED 8 JUNE; ACCEPTED 17 NOVEMBER; PUBLISHED ONLINE 25 DECEMBER 2011; DOI:10.1038/NMETH.1827 npg © 2012 Nature America, Inc. All rights reserved.