The ERI-6/7 helicase is a negative regulator of exogenous RNAi and as we have shown here, is required for a particular suite of endogenous siRNAs in what is now emerging as a multidimensional set of endogenous RNAi pathways. ERI-6/7 is, like the Argonaute ERGO-1, required for the generation and/or stability of two classes of siRNAs, oocyte- and embryo-specific 26G siRNAs and the later generated somatic secondary 22G siRNAs corresponding to the same loci. These 22G siRNAs reduce target mRNA levels, similar to secondary siRNAs in the exogenous RNAi pathway. Our analysis of the ERGO-1/ERI-6/7 pathway has two major surprises: first, that this pathway targets a relatively small number of loci in the genome, a set of duplicated genes with extensive nucleic acid homology. This points to a dedicated surveillance pathway for such gene duplications. Because so many of the components of the ERGO-1/ERI-6/7 pathway are conserved across phylogeny, this duplicated gene silencing pathway is likely to be general. Secondly, the detailed deep sequencing analysis of eri-6/7-dependent small RNA revealed the presence of a passenger strand and (imperfect) phasing between 26G siRNAs, adding to our understanding of the mechanism of endogenous siRNA biogenesis and the role of the Argonaute ERGO-1.
The set of eri-6/7
target genes revealed by our deep sequencing analysis consists of pairs or larger groups of genes that share extensive DNA sequence homology, have a small number of introns and are poorly conserved, even in other Caenorhabditae
. The poor conservation and few introns support the model that these genes have recently been acquired by C. elegans
, perhaps via horizontal gene transfer, for example, from viruses. RNAi pathways have been implicated in antiviral defense, and the ERGO-1/ERI-6/7 pathway may constitute elements of such a viral surveillance pathway. Our data suggest that viral surveillance extends beyond the initial infection. Newly acquired viral genomes may tend to integrate at multiple loci so that extensive nucleotide sequence homology between disparate loci may be a signature of such genes and continue to be silenced by the eri-6/7
small RNA pathway. Alternatively, these duplicated genes may be novel DNA transposons. We did not find evidence of target site duplication at the boundaries of the homologous sequences, nor did we find terminal inverted repeats. If these are transposons, they may no longer are be active, even in eri
mutants; although the level of sequence identity is high, it is not as high as of active DNA transposons in C. elegans 
. Also, we have not found evidence of mutator activity in eri
The targeting of these genes with extensive nucleotide homology suggests that the ERI-6/7 helicase, the ERGO-1 Argonaute protein and other ERI proteins must specifically generate or load siRNAs from the duplicated segments of such gene pairs. To achieve this specificity for duplicated genes, transcripts of every gene may be compared to every other gene and extensive but not perfect nucleotide homology over a distance of hundreds of nucleotides may be assessed by this system. Such a system would need to detect the distinct genome location or the small level of nucleotide sequence divergence that would distinguish surveillance of another transcript from the same gene from a transcript emerging from a duplicated distinct gene.
The most obvious phenotype of the eri-6/7 or ergo-1 null mutations is enhanced response to exogenous RNAi. Our data shows that eri-6/7 mutants have a reduced brood size. The reduced fitness could be attributed to over-expression of specific target genes that act in specific pathways or by more systemic effects induced by a general accumulation of unwanted RNAs. The lack of functional annotation for the majority of eri-6/7 target genes suggests some of these genes may not be functional, though about half of the target genes are weakly conserved in C. briggsae. Thus, the surveillance of these duplicated genes does not appear to subserve any key function for development or survival in the lab. However, this surveillance program uses sufficient C. elegans small RNA machinery that when the eri-6/7 system is disabled, the ability of the animal to respond to exogenous double stranded RNA is enhanced. The small RNA demands of this pathway point to its importance to the organism.
Our data provides insight into the structure of the double-stranded intermediate in 26G siRNA generation by the identification of the passenger strand and the first large scale sequence analysis of the passenger strand in a slicer-defective Argonaute mutant. An obvious candidate for duplex siRNA generation is Dicer/DCR-1, which was shown to be required for 26G siRNA biogenesis 
; A helicase domain mutation (dcr-1/eri-4(mg375))
in DCR-1 specifically abolishes endogenous siRNAs but not microRNAs 
, indicating that the requirement for Dicer is unlikely an indirect effect via a microRNA target gene that acts in 26G siRNA biogenesis. However, the 26G siRNA duplexes are not canonical Dicer products in terms of the lengths of the antisense and passenger strands, the 5′ overhang, and the strong bias for a G as the 5′ nucleotide. Our observation of variable phasing between the 5′ ends of the 26G siRNAs within genes, suggest there is a processive activity that generates 26G siRNAs. This is most likely the RdRP RRF-3, preferentially using a guanylate as an initiation nucleotide, that in conjuction with an endonuclease, possibly Dicer, generates a 26G siRNA duplex and continues doing so along the mRNA starting at a neighboring cytosine in the mRNA template 
. The structure of the duplex suggests that it is modified by other enzymes, such as the ERI-1 3′ exonuclease, to produce the 19 nucleotide passenger strand (Figure S6
). Our eri-1
mutant deep sequencing data did not provide evidence for the existence of longer passenger strand precursors; Possibly such precursors are not stable.
acts in the same pathway as the Argonaute ERGO-1 but unlike eri-6/7
is not required for the passenger strand opposite of the 26G siRNA. This suggests that eri-6/7
acts, with eri-1
and other eri
genes, in the production of a 26G siRNA duplex precursor, while ergo-1
acts on the duplex after biogenesis, removing the passenger strand possibly by slicing, similar to the function of another slicing-capable Argonaute RDE-1 in exogenous RNAi in C. elegans
, and similar to the roles of Argonautes in flies and mammals 
. Site-directed mutagenesis experiments of the catalytic amino acids DDH are necessary to provide more direct evidence of slicing versus other ways of passenger strand destabilization by ERGO-1. The role of ERGO-1 in passenger strand removal versus siRNA biogenesis could also explain the weaker reduction in siRNAs seen in ergo-1(tm1860)
mutants versus eri-1
mutants. Alternative explanations are that the ergo-1(tm1860)
allele is a partial loss-of-function, although the deletion removes more than one third of the PAZ domain, or that other Argonautes are partially redundant with ergo-1
The molecular function of the ERI-6/7 helicase is unclear. The homologous protein Mov10 in humans associates with Argonaute 
and the fly homolog Armitage is required for RNA induced silencing complex (RISC) formation 
. Thus it is possible that ERI-6/7 interacts with ERGO-1 and functions in the assembly of an active effector complex. eri-6/7
does not act in the sperm-specific 26G siRNA pathway that involves the Argonautes ALG-3/4 in place of ERGO-1; it will be of interest to determine if another helicase functions in 26G endogenous siRNA generation in this pathway.
Vasale et al.
and Gent et al. 
have proposed a two-step model for siRNA generation in the ERGO-1 pathway. Downstream of 26G siRNAs, 22G siRNAs are generated by RNA-dependent RNA polymerases RRF-1 and EGO-1. Our data suggests that these events are actually spaced in time, with 26G siRNA generation first in the developing embryo and subsequent RdRP-mediated 22G siRNA generation occurring post-embryonically.
mRNA levels of eri-6/7
target genes are down-regulated in wild type worms compared to eri
mutants. This is explained in part by routing of the endogenous siRNAs into a nuclear co-transcriptional silencing pathway that involves the Argonaute NRDE-3. eri-6/7
-dependent siRNAs are also likely to associate with other Argonautes, such as SAGO-1 and SAGO-2, since at least two eri-6/7
-dependent endogenous siRNAs, assayed by Northern blotting, were shown to associate with SAGO-1 and -2 
. How these Argonautes affect target gene expression is unknown, but the lack the catalytic residues required for slicing, suggests that they direct mRNA degradation by some means other than slicing or that they inhibit translation.
mutants, that are defective in some endogenous RNAi pathways, the exogenous RNAi pathway is more active. The opposing functions of eri-6/7
(and other eri
genes) in exogenous RNAi and endogenous RNAi have been explained by a competition model in which the exogenous RNAi pathway competes with the endogenous RNAi pathway for limiting factors 
. The alg-3/-4
double mutant does not show an enhanced RNAi phenotype 
, suggesting that the limiting factors that the exogenous RNAi pathway competes for are only part of the embryo- and soma-specific ERGO-1/ERI-6/7 endogenous siRNA pathway. The observation that overexpression of the SAGO-1 and -2 proteins that interact with secondary siRNAs causes an Eri phenotype and an enhanced accumulation of endogenous siRNAs 
, shows that these SAGOs could be the limiting factors in the exogenous RNAi pathway. It remains possible that one or more target genes regulated by the eri
genes also act in RNAi pathways. Several target genes encode proteins with potential RNA modifying capability, such as a few helicases, dsRNA binding proteins and a PAZ domain protein.
Whereas in mouse oocytes endogenous siRNAs are formed from antisense pseudogene transcripts, C. elegans has RNA-dependent RNA polymerases, in this case possibly RRF-3, that can produce antisense transcripts. How RRF-3 is recruited to target mRNAs to generate antisense siRNAs is unknown. Possibly, short double stranded RNAs of mRNAs base pairing with antisense transcripts generated from homologous genes recruit the RNA-dependent RNA polymerase. Only a few eri-6/7 target genes have been annotated as pseudogenes, but it remains possible that among the annotated coding genes targeted in eri-6/7 mutants are also pseudogenes. Thus, similar to the function of some pseudogenes in mouse oocytes, pseudogenes in C. elegans may play an important role in endogenous RNAi.