Unfortunately, there is no basis for predicting just how tended, imperfect dsRNA would also be predicted by the many RNAs are edited in any given species, nor is it fact that all RNAs can be software-folded into imperfect known how physiologically significant the editing events dsRNA structures and yet, only a negligibly small fracare. It is perhaps revealing that nearly all genes found to tion of these is A-to-I edited in vivo. give rise to edited transcripts are expressed in nervous Theoretical considerations as well as new experimentissue and encode ion channels and neurotransmitter tal evidence (Reenan et al., 2000) link editing, along receptors. In these molecular machines, single amino with its site selectivity and efficiency, to splicing. First, acid substitutions can induce subtle property changes splicing must follow editing because the intronic ECS to adjust response characteristics to altered functional sequence acquires a cis-acting role in editing. Thus, requirements, for instance during development. It thus splicing of the ECS-containing intron may be artificially appears that site-selective RNA editing by A-to-I substislowed down relative to that of other introns to allow tution has evolved to permit fine tuning of certain neurofor site-selective editing before intron removal. Indeed, physiological processes, and hence failure to edit partic- the predicted dsRNA structures required for editing apular sites may not engender a discernible phenotype. pear unfavorable for intron removal, given that a 5 splice However, in the best-studied example, that of Q/R site site is engaged in the predicted dsRNA structures of editing of transcripts encoding the AMPA receptor sub- most edited pre-mRNAs and, hence, cannot interact unit GluR-B, premature death is caused in mice when with small nuclear ribonucleoprotein particles (Reyes the Q/R site, a molecular determinant for single-channel et al., 1996) to initiate intron removal by spliceosome conductance and Ca2 permeability (Seeburg et al., assembly (Lamond, 1993). 1998), remains unedited (Brusa et al., 1995). Substituting Whether the site selectivity of A-to-I RNA editing rests in the GluR-B gene the unedited codon CAG with the in part on the interaction with the splicing machinery edited codon CGG has no apparent phenotypic conse- remains to be determined. However, that the dsRNA quence on the gene-manipulated mouse (Kask et al., structure for RNA editing needs to be resolved for effi-1998). Thus, the Q/R site–unedited form of GluR-B ap- cient splicing has now been revealed from genetic evipears not to be needed, and one wonders from an evolu- dence (Reenan et al., 2000), which indicates that the tionary point of view why the GluR-B gene does not dsRNA structure in a newly identified edited pre-mRNA specify the edited version on the exon sequence. needs unwinding for correct intron removal. Puzzling A-to-I editing is a nuclear process (see Figure 1). It genetic findings in Drosophila have for some time linked a particular mutation in an ATP-dependent RNA helicase gene to the reduced expression of a voltage-gated Na