In the course of the analysis of RNAs isolated from the eri-1 null mutant, we observed that the 5.8S rRNA in an eri-1 worm is longer than wild-type 5.8S rRNA (Fig. 1a). This length difference is present in all detectable 5.8S rRNA, suggesting that eri-1 mutants have an rRNA processing defect in most, if not all, cells. The mature 5.8S, 18S, and 25-28S rRNAs in eukaryotes are generated from a 35S-47S precursor RNA via a series of processing steps mediated by multiple nucleases⁵. The activity of ERI-1 as a 3′ to 5′ exonuclease^1-4 suggests that the longer 5.8S rRNA is due to an extension of the 3′ end. RNase H cleavage and 3′ end cloning on the 5.8S rRNA of wild-type and mutant C. elegans indicated that all eri-1 5.8S rRNA is at least 1 nucleotide longer than the wild-type 5.8S rRNA specifically at the 3′ end, with a substantial fraction of eri-1 5.8S rRNA containing 2 to 4 additional nucleotides (Supplementary Fig. 1 online). The 5.8S processing defect was observed in two independently derived eri-1 alleles, including another null allele (mg388, data not shown), and is rescued by an eri-1 transgene (see below), proving that loss of eri-1 causes the 5.8S processing defect. The 5.8S processing defect was not observed in other enhanced RNAi mutants with pleiotropies in common with eri-1^1,6 (data not shown).

To characterize whether ERI-1 function in rRNA trimming is an ancient feature of this orthology group, we examined the 5.8S rRNA in the S. pombe erilΔ mutant and observed a length defect similar to the C. elegans eri-1 mutant (Fig. 1b). RNase H and sequencing analysis revealed a 3′ extended 5.8S rRNA species in erilΔ containing from 2 to 8 additional 3′ nucleotides (Supplementary Fig. 2 online). This indicates that ERI-1 has a conserved, dual function in rRNA biogenesis and negative regulation of RNAi that has been inherited from the common ancestors of animals and fungi.

In C. elegans, ERI-1 forms a protein complex with Dicer (DCR-1), and eri-1 gene activity is required for the production of a variety of C. elegans endogenous siRNAs. The release of DCR-1 or other RNAi factors from the production of these endogenous siRNAs in the eri-1 mutant may be the basis of the enhanced response to exogenous dsRNAs^6,7. Co-immunoprecipitation experiments have indicated that DCR-1 interacts specifically with one of two ERI-1 protein isoforms, ERI-1b, but does not associate with the other, ERI-1a⁶. Both ERI-1a and ERI-1b contain a conserved SAP nucleic acid binding domain⁸ and 3′ to 5′ exonuclease domain¹, but ERI-1b carries an extended nematode-specific C-terminal sequence with no identifiable functional domains (Fig. 1c). To examine the role of each isoform in RNAi and rRNA biogenesis we generated transgenic C. elegans expressing each eri-1 spliced-isoform in an eri-1 null mutant background (Supplementary Fig. 3 online). Rescue of the eri-1 enhanced RNAi phenotype was assessed using unc-73 RNAi. Wild-type animals are unaffected by dsRNA to this neural development gene, while 99% (± 0%) of eri-1 mutant animals display a movement defect similar to a null mutation in unc-73 (Supplementary Fig. 3). Expression of ERI-1b in eri-1 dramatically reduced the enhanced response to unc-73 dsRNA to 6% (± 4%) affected, indicating that ERI-1b rescues animals to the wild-type lack of response to unc-73 dsRNA. ERI-1a failed to rescue the enhanced RNAi response with 85% (± 10%) of transgenic animals displaying the unc-73 phenotype (Supplementary Fig. 3). Similar results were observed for each line when tested for the enhanced RNAi-dependent phenotype of cel-1 RNAi (data not shown). Northern blot analysis indicated that ERI-1b rescued K02E2.6 endogenous siRNA biogenesis, while ERI-1a expression failed to re-establish these siRNAs (Fig. 1d). Similar results were obtained for another eri-1-dependent endogenous siRNA (Supplementary Fig. 3). Expression of either ERI-1a or ERI-1b rescued 5.8S rRNA length (Fig. 1d), indicating that both isoforms can mediate rRNA processing. This rescue was robust but incomplete, as expected with expression of ERI-1 in somatic tissues from an extrachromosomal transgene array. These data show that ERI-1a mediates 5.8S rRNA processing but may not act in RNAi pathways, while ERI-1b participates in both rRNA processing and RNAi. This supports a model in which the interaction of ERI-1b with DCR-1, possibly through its extended C-terminal domain, is important for its activity in endogenous siRNA biogenesis and inhibition of exogenous RNAi.

We examined the subcellular localization of both ERI-1 isoforms by expressing C-terminal green fluorescent protein (GFP) fusions to ERI-1a and ERI-1b in the eri-1 mutant. Rescuing GFP-tagged versions of both isoforms (Fig. 1d, and Supplementary Fig. 3) were exclusively cytoplasmic in a large variety of cell types that express the transgene (Fig. 1e). Similar results were observed for N-terminal GFP fusions to ERI-1 isoforms (data not shown). ERI-1 localization to the cytoplasm is surprising, given that rRNA processing occurs primarily in the nucleolus. However, terminal rRNA processing of the 18S is cytoplasmic ⁹. Furthermore, results from Xenopus have suggested that processing of the 5.8S rRNA from a slightly longer pre-5.8S RNA may occur in the cytoplasm¹⁰, and S. pombe Eri1 is localized to the cytoplasm². Mammalian ERI-1 localizes to the nucleolus as well as the cytoplasm⁴ (Ansel et al., co-submitted). The activity of mouse Eri-1 to trim the 5.8S rRNA on intact ribosomes, similarly to our results (see below), and its association with polysomes (Ansel et al., co-submitted) support a role for ERI-1 5.8S rRNA processing after nuclear export of ribosomes.

Most rRNA processing and ribosome assembly steps have occurred by the time the ribosome reaches the cytoplasm. In the mature ribosome, the 3′ end of the 5.8S rRNA is paired with the 5′ end of the 25-28S rRNA¹¹ (Supplementary Fig. 4 online). This 5.8S/25-28S helix is reminiscent of the 3′ hairpin of the histone mRNA, and siRNA structures that are in vitro substrates for ERI-1^1-4. We tested if each protein isoform could act on this structure in vitro using ribosomes purified from eri-1 mutants. When recombinant ERI-1a was incubated with intact eri-1 ribosomes a major product was generated with identical length to that of the wild-type 5.8S rRNA (Fig. 2a). In contrast, ERI-1b showed no activity in vitro, despite expression and purification at levels equal to ERI-1a (Fig. 2a and Supplementary Fig. 4). This is discordant with in vivo results indicating that ERI-1b expressed from a transgene can mediate rRNA trimming in the absence of ERI-1a expression (Fig. 1d), and may be due to the in vitro expressed ERI-1b enzyme lacking co-factors, such as DCR-1, which may be required for its activity. Alternatively, in vitro expression or assay conditions may fail to recapitulate critical aspects of in vivo chemistry. Similar results for enzymatic activity of ERI-1 were obtained for an oligonucleotide mimic of the 5.8S/26S structure and a synthetic siRNA (Supplementary Fig. 4 and data not shown)

Introduction of H264A and D268A mutations into the nuclease domain of S. pombe Eri1 disrupts the protein's exonuclease activity in vitro, and its negative regulation of RNAi in vivo². To verify that ERI-1 catalyzes the 3′ end processing of the 5.8S rRNA we inserted homologous mutations into C. elegans ERI-1a (H317A and D321A, “ERI-1aAA”, Supplementary Fig. 4). These mutations caused a complete loss of in vitro nuclease activity on both purified eri-1 ribosomes and an oligonucleotide 5.8S/26S substrate (Fig. 2b and Supplementary Fig. 4). Transgenic lines expressing either ERI-1aAA::GFP or ERI-1bAA::GFP at levels equal to the wild-type transgenes failed to rescue rRNA trimming defects of the eri-1 mutant in vivo (Fig. 2c and data not shown), ERI-1bAA::GFP also failed to rescue the enhanced RNAi phenotype, or re-establish expression of endogenous siRNAs in the eri-1 mutant (Supplementary Fig. 3).

The extensive conservation of 5.8S rRNA processing by ERI-1 in C. elegans, S. pombe and mouse (Ansel et al., co-submitted) suggests that this trimming is important to ribosome function. Despite this conservation, eri-1 mutant C. elegans or S. pombe do not display obvious phenotypes attributable to defects in the ribosome. In contrast in mouse, loss of Eri-1 results in cell and organismal growth defects (Ansel et al., co-submitted). This distinction could reflect subtle defects in ribosome function in all three clades that are only manifest at the gross level of organismal health in the more complex biology of the mouse. Alternatively mouse ERI-1 exonuclease may process RNAs in addition to the 5.8S rRNA, just as ERI-1 in C. elegans and S. pombe regulates endogenous siRNA produced by RNA-dependent RNA polymerases, a pathway that appears to be missing in mammals. In the budding yeast Saccharomyces cerevisiae eri-1 has been lost, along with known components of the RNAi pathway such as Dicer and Argonaute, but a CCR4-like nuclease, Ngl2p, performs analogous 3′ end trimming of the 5.8S rRNA¹². This further supports the importance of this processing in cellular fitness.

Previously, the microRNA processor Drosha and two DEAD-box helicases were shown to act in both rRNA biogenesis and small-RNA-dependent silencing¹³. Our data exposes more extensive integration between these two cellular pathways, and suggests that future analyses may uncover additional common components and origins of these pathways.

Supplementary Material