Independent screens by the Vaucheret and Baulcombe laboratories for co-suppression-defective mutants in Arabidopsis have recovered mutations in a RdRP-related gene, named sgs2 (suppressor of gene silencing) [17] or sde1 (silencing defective) [18]. For the purposes of this review, the gene is referred to as sgs2/sde1. Mutants were isolated by screening for individuals defective in transgene-induced PTGS. Several other RdRP-related genes exist in Arabidopsis, but no mutations were recovered in those genes.

Because co-suppression in plants can be triggered by viral infection and, in fact, is hypothesized to be a viral defense mechanism, both groups investigated the effects of viral infection on sgs2/sde1 mutants. Was virally induced PTGS defective? Were the mutants super-susceptible to viral infection? Dalmay and colleagues [18] found that PTGS triggered by tobacco mosaic virus and tobacco rattle virus is not defective in sde1 mutants. They interpret their data as meaning that SDE1 is not required for virally induced PTGS, but only for transgene-induced PTGS. They suggest that virally encoded RdRP activity can substitute for the putative RdRP activity of SDE1. In contrast, Mourrain and colleagues [17] found that sgs2 mutants had increased susceptibility to infection by certain viruses, such as a cucumovirus, but not others, such as a tobamovirus or a potyvirus. They interpret these results as indicating that PTGS indeed acts as a defense against at least some viruses. Interpretation of infection data is complicated by the fact that many viruses can inhibit PTGS: tobamovirus and potyvirus strongly inhibit PTGS, whereas cucumovirus inhibits it only weakly [17,23]. Logically, the loss of a plant component of PTGS, such as SGS2/SDE1, will have no effect on a virus that fully blocks PTGS by itself, but would allow a higher level of infection by a virus that disables PTGS only partially or not at all. At any rate, these results suggest that the Dalmay et al. model is unlikely to be correct for all viruses; in at least some cases, viral RdRP cannot replace SGS2/SDE1.

RdRP activity has been known in plants for decades, although its physiological function has been unclear [24]. Many models for PTGS in plants postulate an RNA amplification step that could be accomplished by RdRP activity (for example in [25]), but it has yet to be shown definitively that RdRP is in fact involved in PTGS. In light of these models, it is intriguing that proteins related to the purified tomato RdRP have now been linked to PTGS in diverse organisms. It should be noted, however, that these proteins have not yet been shown to have RdRP activity. They contain large regions of sequence conservation with tomato RdRP, which may encode RdRP activity, but they also contain extensive regions of divergent sequences (see Figure 1) [15,17,18].

Analysis of genome sequence data has shown that C. elegans and Arabidopsis contain several RdRP-related genes. In C. elegans, for example, there are three rrf (RdRP family) genes in addition to ego-1. Thus far, six other RdRP-related genes have been found in the Arabidopsis genome in addition to sgs2/sde1. Genome sequencing has uncovered RdRP-related genes in many other organisms, including several plant species, such as wheat and petunia, and the fission yeast Schizosaccharomyces pombe. It is not yet clear though how many of these RdRP-related proteins actually function in PTGS. Some of them may have other cellular functions, perhaps in RNA metabolism. Indeed, ego-1 was identified originally as being important in development of the C. elegans germ line, and was only later shown to function in RNAi; the connection between development and RNAi is not yet clear, but one possibility is that EGO-1 protein has different functions in the two processes. Since alleles of only one RdRP-related gene, sgs2/sde1, were recovered as suppressors of PTGS, perhaps the related genes in Arabidopsis do not function in PTGS. Alternatively, some or all of these genes may be redundant, so that simultaneous mutations in two or more of them would have to be induced in order to see a phenotype. In addition, if any RdRP-related gene is essential, it would not have been identified because the screens for silencing-defective mutants did not allow for recovery of lethal mutations.

RdRP-related genes are apparently absent from at least one organism that is susceptible to RNAi: Drosophila melanogaster. Genome sequence data are nearly complete for Drosophila, yet no gene with homology to either tomato or viral RdRP has been found. Since viral and cellular (for example, tomato) RdRP have little amino acid homology to each other, perhaps RdRP function in Drosophila is accomplished by an enzyme with yet a different amino acid sequence. Alternatively, RNAi in Drosophila may not require RdRP activity at all.

Molecular characterization of another Arabidopsis sgs gene, sgs3, by Mourrain et al. [17] revealed that it encodes a novel protein. In general, sgs3 mutants behave like sgs2 mutants, including having an increased susceptibility to cucomovirus infection. Thus, it appears that SGS3 is as critical to PTGS as is SGS2/SDE1. No SGS3 relative has been found yet in any other organism, including those with fully sequenced genomes such as C. elegans and Drosophila and this result is consistent with SGS3 functioning in a plant-specific aspect of PTGS.

The extensive and growing genome sequence database is an important resource for studying the mechanics of PTGS and the role of RdRP-related genes in this and other cellular processes. Three protein families are now known to be important to PTGS. Genome sequence information has allowed researchers to identify paralogs and orthologs very quickly, and provides a focus for future experiments. For example, reverse genetic approaches can be used to generate deletion mutations in these related genes for use in studying their possible role in PTGS. We can expect a more complete molecular portrait of PTGS to be revealed in the near future.