Dicer family proteins are involved in the production of small RNA molecules and have several conserved motifs (Figure 1c) 767: two amino-terminal DExH-Box helicase domains, a PAZ (Piwi/Argonaute/Zwille) domain, tandem RNase III domains and a carboxy-terminal dsRNA binding domain. A single Dicer protein is involved in both the siRNA and miRNA pathways in C. elegans 676869. In contrast, different Dicer proteins are assigned to the siRNA and miRNA pathways in Drosophila 17. Dcr-1, which retains a PAZ domain but lacks an amino-terminal helicase domain (Figure 1c), is involved in the miRNA pathway 17. On the other hand, Dcr-2 seems to lack a full-length PAZ domain but has the helicase domain (Figure 1c), and is involved in the RNAi pathway 17. In addition, a distantly related RNase III emzyme, Drosha, is involved in the maturation of miRNA precursors 7071.

We identified one drosha and two Dicer genes in the Tribolium genome. One gene (Tc-Dcr-1) clearly codes for the ortholog of Dm-Dcr-1 and Ce-Dcr-1. The sequence of the second Tribolium Dicer does not clearly cluster with Dm-Dcr-2 (Figure 1a, b). However, as it shares some similarities in domain architecture with Dm-Dcr-2 (Figure 1c, and see below), we tentatively call it Tc-Dcr-2.

A ScanProsite search 72 has revealed that, in contrast to Dm-Dcr-1, which lacks a helicase domain, Tc-Dcr-1 retains both the helicase and PAZ domains (Figure 1c). This domain architecture makes Tc-Dcr-1 more similar to Ce-Dcr-1. Tc-Dcr-2 also has both domains, but the PAZ domain is more diverged (Figure 1c). ScanProsite shows high scores for the PAZ domains of Ce-Dicer-1, Tc-Dcr-1, and Dm-Dcr-1 (scores of 24, 23 and 30, respectively), while the PAZ domain in Tc-Dcr-2 shows a lower score (score 17) (see Materials and methods for a brief explanation of these scores). Dm-Dcr-2, which lacks a full-length PAZ domain, shows a much lower score for the PAZ domain region (score 8). Tc-Dcr-2 also lacks the carboxy-terminal dsRNA binding domain. The diverged PAZ domain and the lack of the dsRNA binding domain make Tc-Dcr-2 more similar to Dm-Dcr-2 (Figure 1c).

A group of dsRBM-containing proteins act with Dicer to load small RNA molecules into a silencing complex. In Drosophila, each Dicer protein acts with a particular dsRBM protein: Loquacious (Loqs) for Dcr-1, R2D2 for Dcr-2, and Pasha for Drosha 101112131473. Interestingly, these proteins seem to determine the specificity of Dicer proteins, since Drosophila Dcr-1, which normally processes miRNAs, can instead produce siRNA in a loqs mutant 1114. This suggests that differences in these dsRBM-containing proteins might affect the efficiency of RNAi in different organisms.

We found clear orthologs of Drosophila loqs and pasha in Tribolium (Figure 2). In contrast, the Tribolium genome contains two R2D2-like genes (we named one of them Tc-R2D2 and the other Tc-C3PO), but orthology with Drosophila R2D2 is not as clear as for the other dsRBM proteins (Figure 2).

In conclusion, Drosophila and Tribolium have the same number of Dicer proteins. However, similarity of domain architecture of Tc-Dcr-1 to Ce-Dcr-1 (rather than to Dm-Dcr-1) suggests that, in addition to Tc-Dcr-2, Tc-Dcr-1 could also be involved in both the miRNA and RNAi pathways, perhaps contributing to the robust RNAi response in Tribolium. The presence of an additional R2D2-like protein might also help make Tribolium hypersensitive to dsRNA molecules taken up by cells.

Argonaute proteins are core components of RISC and miRNP, and are involved in siRNA-based as well as miRNA-based silencing 216. Some Argonaute proteins are also involved in transcriptional silencing as a component of RITS 7475. Different Argonaute proteins are used for each process 16. For instance, in Drosophila, Ago-1 and Ago-2 are predominantly used for miRNA and siRNA pathways, respectively 18, while Piwi, Aubergine (Aub), and Ago-3 are used for transcriptional silencing 76777879. Argonaute proteins contain two distinctive domains: a PAZ domain and a PIWI domain 16. The PAZ domain seems to be involved in dsRNA binding, while the PIWI domain possesses RNase activity.

There is a striking expansion of Argonaute proteins in C. elegans (27 Argonaute proteins have been identified) 80. As in Drosophila, these Argonaute proteins function in different processes. Rde-1 and Ergo-1 have been identified to act in the RNAi pathway 980, while Alg-1 and Alg-2 are important for the miRNA pathway 81. Yigit et al. 80 identified yet another class of Argonaute proteins, the secondary Argonautes (Sago), that interact specifically with the siRNAs produced via RdRP amplification but not with the initial siRNAs. These results led the authors to propose a two-step model: first, the primary siRNAs, which are produced from the initial dsRNA, bind specifically to the initial Argonautes (Rde-1 or Ergo-1), and second, subsequent amplification by RdRP leads to the production of secondary siRNAs, which exclusively bind to secondary Argonaute proteins. This two-step recognition is proposed to be required for amplification of the RNAi effect, and at the same time possibly reducing off-target effects. As the secondary Argonaute proteins lack critical metal binding residues in the catalytic RNAse H-related PIWI domain, they are predicted to recruit other nucleases for degradation of target mRNAs 80.

Both Tribolium and Drosophila have five Argonaute genes. To investigate the orthology relationships of these genes we calculated a tree based on an alignment of the PIWI domains of all Tribolium and Drosophila Argonaute proteins, a representative selection of C. elegans paralogs and the single Schizosaccharomyces pombe Argonaute protein (Figure 3; see Additional data file 1 for the alignment).

A single miRNA class Argonaute (Ago-1 in Drosophila and Alg-1/Alg-2 in C. elegans) is present in Tribolium (Tc-Ago-1). For the siRNA class Argonautes, we found two Ago-2 paralogs in Tribolium (Tc-Ago-2a and Tc-Ago-2b) that probably stem from a duplication in the lineage leading to beetles. These two proteins are clearly orthologous to Drosophila Ago-2; however, the relationships to C. elegans Rde-1 and Ergo-1 are not resolved in our analysis. The duplication of Ago-2 in Tribolium might lead to higher amounts of Tc-Ago2 protein and, hence, an enhanced RNAi response.

For the Piwi/Aub class Argonautes, which are involved in transcriptional silencing, we find one Tribolium ortholog (Tc-Piwi) of the Drosophila Piwi and Aub. One additional protein of this family (Tc-Ago3) is orthologous to a recently described Drosophila protein, Dm-Ago3 7782. All these insect PIWI-type proteins are orthologous to the C. elegans Prg-1 and Prg-2.

Importantly, we do not find any homologs of secondary Argonaute proteins (represented by Ce-Ppw-1 and Ppw-2 in our tree) in either Tribolium or Drosophila (Figure 3). Furthermore, we confirmed that all Tribolium and Drosophila Argonaute proteins do have the metal binding residues of the PIWI domain, unlike the C. elegans secondary Argonaute proteins, which lack them 80. The only exception is Drosophila Piwi, which has a lysine instead of a histidine in the third position. These data, along with the fact that the Tribolium genome lacks an ortholog of RdRP (see below), suggest that the two-step RNAi mechanism of RdRP-mediated amplification followed by secondary Argonaute function is not conserved in either Tribolium or Drosophila. The different abilities of Drosophila and Tribolium to perform systemic RNAi might, therefore, depend on factors other than the Argonaute repertoire in these insects.

Systemic RNAi relies on the distribution of the trigger dsRNA, its uptake and subsequent efficient gene knockdown in cells. The distribution of the dsRNA trigger leads to its dilution 83. Hence, a mechanism for enhancing the signal may be required for efficient silencing. RdRP is a key for the amplification of the RNAi effect in C. elegans as well as in several plants 19208485. It is possible that Tribolium has a similar amplification mechanism. However, we do not find a gene encoding an RdRP-related protein in the Tribolium genome by BLAST searches. Moreover, a BLAST search of all metazoan genes in the NCBI database identified RdRP genes only in several Caenorhabditis species and a cephalochordate Branchiostoma floridae 86. Even some nematode species outside Caenorhabditis do not seem to carry RdRP genes. All other eukaryotic RdRPs belong to plants, fungi or protists, suggesting that RdRP is not conserved in animals (Figure 4). The lack of an RdRP gene in Tribolium suggests that the strong RNAi response in Tribolium does not rely on amplification of the trigger dsRNA by RdRP.

In C. elegans, several tissues, such as the nervous system, are refractory to RNAi, apparently due to the expression of eri-1 87. Abundant siRNA accumulates in eri-1 mutants, suggesting that Eri-1 is involved in siRNA degradation 87. The eri-1 gene encodes an evolutionarily conserved protein that contains a SAP/SAF-box domain and DEDDh family exonuclease domain 87. The expression level and/or tissue specificity of eri-1 homologs might cause differences in sensitivity to dsRNA among organisms.

We have identified an eri-1-like gene in Tribolium. 5' and 3' rapid amplification of cDNA ends (RACE) analysis has revealed that this gene encodes a 232 amino acid protein (see Materials and methods for details). We also found a close homolog of this gene in Drosophila (CG6393, Dm-snipper). Interestingly, these genes are lacking the amino-terminal SAP/SAF-box domain. Also, phylogenetic analysis using the nuclease domain (Additional data file 1) reveals that the insect homologs cluster together, while Ce-Eri-1 and its human ortholog (3'hExo; three prime histone mRNA exonuclease 88) compose another subclass. We subsequently noticed that there are at least three subclasses of nucleases closely related to Eri-1 in metazoans: the Eri-1/3'hExo subclass, the Pint1 (Prion Interactor 1 89, also named Prion protein interacting protein (PrPIP) in 90) subclass, and the Snipper subclass (Figure 5). Humans as well as sea urchins have all three subclasses of nucleases. C. elegans has at least two types of these nucleases, which belong to the Eri-1/3'hExo and Pint1 subclassses, respectively. In addition, it contains another nuclease (Cell-death-related nuclease 4 (Crn-4) 91), whose position relative to the three subclasses of nucleases is unclear. Crn-4 clusters with C. elegans Eri-1 (Additional data file 2), but this affinity is questionable since Crn-4 does not share the amino-terminal region that is conserved in other members of the Eri-1/3'Exo subclass. The Tribolium and Drosophila nucleases, with their vertebrate and sea urchin orthologs, compose a distinct subclass (Snipper subclasss). This suggests that Drosophila and Tribolium lack nucleases belonging to the Eri-1 subclass, and that the insect nucleases might have a function other than siRNA digestion.

Recently, the Drosophila nuclease has been characterized as Snipper (Snp) 90; therefore, we have named the Tribolium ortholog Tc-Snp. Although Snp can cleave RNA as well as DNA molecules in vitro, Snp seems to have no role in RNAi in Drosophila 90. This supports our idea that the Snp subclass nucleases might not have an important role in the RNAi pathway. In conclusion, it is unlikely that nucleases related to Eri-1 are causing the differential sensitivity to dsRNA in Tribolium and Drosophila.

Sid-1 is the best characterized protein involved in systemic RNAi in C. elegans 5359. The Sid-1 protein contains a long amino-terminal extracellular domain followed by an array of transmembrane domains, which are inferred to form a channel for dsRNA molecules 5359. Mosaic analysis in C. elegans using a sid-1 overexpression construct showed that Sid-1 is cell-autonomously required for receiving the systemic RNAi signal (it is still possible that Sid-1 is also involved in the RNAi spreading step) 53. Overexpression of sid-1 in Drosophila culture cells also enhances the ability of the cells to uptake dsRNA from the culture media, further suggesting an important role for Sid-1 in dsRNA uptake 59. C. elegans carries two additional sid-1 like genes, tag-130 (also known as ZK721.1) and Y37H2C1, although their functions are unclear. Many vertebrate species also have sid-1 homologs 5392. However, Drosophila, which does not show a robust systemic RNAi response, lacks sid-1-like genes, leading to the hypothesis that the presence or absence of a sid-1-like gene is the primary determinant of whether or not systemic RNAi occurs in an organism 2853929394.

We have identified three sid-1-like genes in the Tribolium genome. We have decided to call these genes sil (sid1-like; Tc-silA-C) instead of Tc-sid-1, because of uncertainty about the orthology of insect sid1-like genes to C. elegans sid-1 (see below). RT-PCR and RACE analyses have revealed the full-length sequences (Tc-SilA, 764 amino acids; Tc-SilB, 732 amino acids; Tc-SilC, 768 amino acids, see Materials and methods for details). Like C. elegans Sid-1, all three proteins contain a long amino-terminal extracellular domain followed by 11 transmembrane domains predicted by TMHMM server version 2.0. InterProScan identified no additional motifs or domains.

To determine whether the presence of sil genes correlates with the presence of systemic RNAi in insects, we have searched the genome of several insects using the Tc-SilA protein sequence as a query (Table 1). The honeybee (Apis mellifera; Hymenoptera) and a parasitic wasp (Nasonia vitripennis; Hymenoptera) each contain a single sid-1-like gene. The silkworm moth (B. mori; Lepidoptera) has three sid-1-like genes. We have determined the full-length sequences of these genes in Bombyx (see details in Materials and methods). As previously mentioned, D. melanogaster does not have any sid-1-like genes. We have confirmed that none of the 11 additional Drosophila species whose genomes have been sequenced carry sid-1 family genes. In addition, two mosquito species (Anopheles gambiae and Aedes aegypti) also lack sid-1-like genes, suggesting the early loss of sid-1-like genes in the dipteran lineage.

The presence of three sil genes in Tribolium is consistent with their hypothesized importance to a robust systemic RNAi response. It has also been shown that parental RNAi is possible in Nasonia 42, which is consistent with the presence of a sil gene in this insect. On the surface, the lack of sid-1-like genes in dipterans seems to correlate with the apparent lack of systemic RNAi response in these insects. However, reports that some tissues in Drosophila as well as in mosquitos are capable of taking up dsRNA 333435363745 (MJ Gorman, personal communication) suggest that such correlations might be misleading. Moreover, Bombyx carries three sil genes, yet does not show a robust systemic RNAi response (S Tomita, unpublished data; R Futahashi and T Kusakabe, personal communications). This apparent breakdown in the correlation between systemic RNAi and sil genes (Table 1) raises the question of whether sid-1-like genes are the determinant of presence/absence of systemic RNAi in insects.

We have analyzed the expression of sil genes to provide a clue about the function of these genes in Tribolium. in situ hybridization analysis shows that all three sil genes are expressed uniformly in embryos; however, silA and silB seem to be expressed at lower levels than silC (data not shown). Semi-quantitative RT-PCR reveals that all sil genes are expressed throughout all developmental stages (Additional data file 3). silA and silB expression level is uniform through the larval to adult stages, while silC has peak expression at the pupal stage.

We have performed phylogenetic analyses using the carboxy-terminal conserved region (the region corresponding to the second to tenth transmembrane domains; Additional data file 4) to solve the orthology of Sid1-like proteins. Both neighbour-joining and maximum-likelihood analyses produce the same tree with slightly different bootstrap values (see Figure 6a for the neighbour-joining tree). In these trees, all three C. elegans proteins comprise a distinct cluster. Two of the Tribolium Sil proteins (Tc-SilA and Tc-SilB) also comprise a separate cluster, while Tc-SilC clusters with honeybee as well as vertebrate Sid-1-like proteins. Bombyx Sil proteins belong to this cluster; however, they comprise a distinct sub-cluster in this branch. This result is somewhat puzzling since it appears to suggest multiple occurrences of lineage-specific duplication. Alternatively, the expansion of sil genes might be ancient, but the paralogs might have been subjected to lineage specific parallel constraints (perhaps to target a species specific ligand), leading to convergent sequence similarity. The clustering of the three C. elegans homologs might be due to a long branch attraction caused by their highly diverged sequences. The clustering of vertebrate Sid-like proteins with Tc-SilC and the honeybee proteins might suggest a conserved function in this cluster.

Although the carboxy-terminal transmembrane region shows a high degree of identity between all Sid-1-like proteins, the amino-terminal extracellular region is less conserved (Additional data files 4 and 5). We noticed, however, that there are several segments in the extracellular region that are shared by insect and vertebrate Sid-1-like proteins (Figure 6b; see also Additional data file 5 for dot-matcher alignments). Interestingly, C. elegans Tag-130, but not Sid-1, also shares these amino-terminal motifs (Figure 6a, Additional data file 5), raising questions about the orthology of insect/vertebrate Sid-like proteins and C. elegans Sid-1. Sil proteins in insects and vertebrates might instead be orthologous to C. elegans Tag-130.

Although our phylogenetic analysis is inconclusive on the orthology of insect Sil proteins, the sequence similarity of the amino-terminal extracellular region between Sil proteins and C. elegans Tag-130 suggests that these proteins may share similar functions. To gain further insight into the function of sil genes, we have analyzed whether tag-130 has any function in systemic RNAi in C. elegans. We obtained two deletion alleles of tag-130 from the Caenorhabditis Genetics Center. One allele, tag-130^gk245, has been described to have a 711 bp deletion that removes the promoter region as well as the first 221 bp of the coding region (73 amino acids) (Additional data file 6). We have confirmed this deletion by PCR. We have also determined that the other allele, tag-130^ok1073, has a 689 bp deletion spanning several exons that encode transmembrane domains (exons 14 to 17; see Additional data file 6 for the detailed deleted region). RT-PCR analysis has revealed that tag-130^gk245 lacks tag-130 gene transcription, suggesting that this is a null allele. We have detected two different forms of mRNA transcribed in tag-130^ok1073, both of which encode truncated proteins (Additional data file 6). These proteins lack several transmembrane domains, suggesting that tag-130^ok1073 is also a null allele. To determine whether these mutants are susceptible to systemic RNAi, we fed them unc-22 dsRNA expressing E. coli. The N2 wild-type strain was used as a positive control, and sid-1^sq2, a null allele for sid-1 5359, was used as a negative control. If tag-130 is involved in systemic RNAi, mutations in the tag-130 gene should prevent the unc-22 RNAi twitching effect 95. However, almost 100% of individuals carrying either tag-130 deletion allele show a twitching phenotype upon administration of unc-22 feeding RNAi, while none of the sid-1 individuals showed twitching (Table 2). These data indicate that tag-130 is not necessary for the systemic RNAi response in C. elegans. By extension, the greater sequence similarity of insect Sil proteins to Tag-130 than to Sid-1 suggests that Sil proteins might not be involved in systemic RNAi in Tribolium.

Another screen for C. elegans mutants lacking systemic RNAi led to the discovery of several additional genes involved in the systemic RNAi response, including rsd-2, rsd-3, and rsd-6 60. Mutants for these genes still retain the systemic RNAi response in somatic cells, but germ-line cells lack the ability to respond to dsRNA 60. The Rsd-2 protein contains no particular motifs, while Rsd-6 has a Tudor domain, which is found in some RNA binding proteins 60. A yeast two-hybrid analysis found that Rsd-2 interacts directly with Rsd-6, suggesting that these proteins act together 60. We do not find Tribolium homologs for rsd-2 or rsd-6 in the genomic sequence of Tribolium (Table 3) or in several other insects whose genomes have been sequenced, which suggests that the Rsd-2/Rsd-6 system is either not conserved in insects, or is evolving too rapidly to be detected across long evolutionary distances.

The third gene, rsd-3, encodes a protein that contains an epsin amino-terminal homology (ENTH) domain 60. ENTH domains are often found in proteins involved in vesicle trafficking, suggesting the possible involvement of endocytosis in systemic RNAi 60. We found a homolog for Rsd-3 in Tribolium (Tc-Rsd3). Drosophila also carries a protein similar to Rsd-3 (Epsin-like).

In addition, the Rsd-3 protein has a close relative in C. elegans, Epn-1, whose Drosophila counterpart (Liquid Facets; Lqf) has been reported to be involved in Notch signaling 969798. We found a Tribolium ortholog for Epn-1/Lqf, which we named Tc-Lqf. Although there is no report implying the involvement of Epn-1/Lqf family proteins in systemic RNAi, their high degree of identity with Rsd-3 proteins suggest that such a role is possible.

Since Drosophila (which seems to lack a systemic RNAi response) also carries Rsd-3-like proteins (Table 3), it does not seem likely that these proteins determine the presence or absence of systemic RNAi in insects. However, it might be still possible that the expression level and/or tissue specificity of rsd-3-like genes affect the degree of RNAi efficiency.

Another piece of evidence that suggests the involvement of endocytosis in dsRNA uptake comes from a study using Drosophila S2 culture cells 6162. Among the factors identified in this study as necessary for dsRNA uptake are a number of proteins whose functions are implicated in endocytosis 6162 (Table 4). Also, several scavenger receptors, such as Eater and Sr-CI, were found to be important for dsRNA uptake 6162 (Table 4). Scavenger receptors are known to act as receptors for large molecules and/or microbes (for a review of scavenger receptors, see 99). Since these receptors act in phagocytosis, a type of endocytosis, they could potentially act as receptors for dsRNA molecules in an endocytic process. Although the factors identified in S2 cells might reflect mechanisms of dsRNA uptake specific to hemocyte-like cells 100, it is possible that some of them function in other tissues as well.

We have identified the Tribolium orthologs of these genes (Table 4). In many cases, they show clear one-to-one orthology. In the case of Sr-CI, however, we found only one Tribolium homolog (Tc-Sr-C), in contrast to four closely related paralogs in Drosophila 100. The case of eater is even more complicated. eater encodes a Nimrod family protein that contains multiple NIM-type EGF domains 101102. A BLAST search using Drosophila Eater (CG6124) identifies several predicted Tribolium proteins that contain NIM repeats. Among them, an NCBI predicted protein (XP_969372) and Eater are reciprocal best hits, suggesting that they might be orthologous. However, the similarities among Nimrod family proteins in both insects make it difficult to assign orthology. Detailed sequence analysis will be required to definitively determine the orthology of these genes.

As in the case of rsd-3, the fact that both Drosophila and Tribolium carry these genes might suggest that these factors do not determine the presence or absence of systemic RNAi in insects. Yet, it is still possible that a difference in tissue specificity and/or expression level might affect the efficiency of dsRNA uptake from the outside environment. Expression analysis of these genes might help determine whether these factors are broadly involved in dsRNA uptake in insects.