ADARs convert adenosines to inosines by hydrolytic deamination. The stereochemistry of the proposed tetrahedral intermediate has not been investigated. However, given the similarities of ADARs and ADATs (adenosine deaminases that act on tRNAs) to cytidine deaminases (CDAs), the intermediate is drawn as if water attacks from the same side of the base as observed with the CDA enzymes.

Human, fly, and worm ADAR open reading frames (ORFs) are shown with Z-DNA binding motifs (oval with a Z), dsRBMs (gray ovals), and catalytic domains (blue). Splice forms are designated by a letter of the alphabet following the ADAR name (6), and ORF lengths correlate with the relative number of amino acids for the particular splice-form. Insertions (triangle, apex up) and deletions (triangle, apex down) relative to other splice forms are indicated. An ADAR3 has been detected in mammals, but so far deaminase activity cannot be detected with this protein (26, 27). For comparison an ADAT is also shown.

(a) The cartoon illustrates that ADARs can target double-stranded regions in 5′ and 3′ untranslated regions (UTRs), exons, and introns. An exon complementary sequence (ECS) is shown in green. (b) The hairpin targeted by ADARs at the R/G site of gluR-B, -C, and -D is shown as an example of a structure required for deamination in a codon. The identities of sequences in blue are strictly conserved, and those indicated with a red dot covary so as to maintain an unpaired state (31). (c) The structure targeted by ADARs in the 3′ UTR of syntaxin (unc-64) is shown as an example of a structure that mediates deamination in a UTR (D. P. Morse et al., unpublished data).

ADARs are capable of a wide range of specificities depending on the structure of the RNA substrate. (a) RNAs that are predicted to form rod-shaped molecules of similar lengths. While an 800– base-pair dsRNA is deaminated promiscuously by an ADAR, the ~1700 nucleotide HDV antigenome is specifically targeted at the amber/W site (red). (b) ADARs (green) reacting with four dsRNAs of differing stabilities. Because ADARs change AU base-pairs to IU mismatches, ADAR substrates become increasingly single-stranded as the reaction proceeds. The model proposes that an ADAR reaction stops when the RNA is too single-stranded to be bound by an ADAR. Substrates that are shorter, or contain mismatches, are more selectively deaminated because it takes fewer deaminations to reach the critical “thermodynamic” threshold. In the far left and far right panels blue lines represent a specific sequence. The sequence is modified more selectively when placed between internal loops, as in the barbell molecule.

ADARs and ADATs share sequence similarities with the cytidine deaminase (CDA) family. At the top an alignment of six sequences includes representative members of the ADARs, and the ADATs that act at A37 (Tad1p). Highly conserved sequences are shown in capital letters (identities four of six shown), and regions proposed to function in binding zinc are shaded in gray. The S. cerevisiae enzyme Tad2p represents the ADATs that target A34 of tRNA. This group of enzymes shows similarities to the ADARs and Tad1p ADAT, as well as the cytidine deaminase sequence shown below, here represented by the E. coli enzyme. To the right, amino acids that coordinate zinc (red) and the glutamate involved in proton shuttling (blue) are shown interacting with cytidine, according to the crystal structure of E. coli cytidine deaminase (ECCDA) (112). The same color scheme is used to highlight residues proposed to serve a similar function in the ADARs and ADATs. Nomenclature and accession numbers for the sequences are as follows: hADAR1a, human ADAR1a, NM-001111; hADAR2a, human ADAR2a, NM-001112; dADARa, D. melanogaster ADARa, AF208535; ceADR2, C. elegans ADR2, AF051275; dTad1p, D. melanogaster Tad1p, AF192530; scTad1p, S. cerevisiae Tad1p, AJ007297; scTad2p, S. cerevisiae Tad2p, AJ242667; ECCDA, E. coli cytidine deaminase, M60916.

A model of how ADARs evolved.