Activating silent Argonautes

Multiple Argonaute proteins are implicated in gene silencing by RNA interference (RNAi), but only one is known to be an endonuclease that can cleave target mRNAs. Chimeric Argonaute proteins now reveal an unexpected mechanism by which mutations distal to the catalytic center can unmask intrinsic catalytic activity, results hinting at structurally mediated regulation.

During RNAi, RNAs of 20–30 nucleotides post-transcriptionally downregulate target mRNAs containing complementary sequences. The guide RNA molecules in this pathway serve to recruit key protein components including the enzyme Argonaute (Ago), which functions as the catalytic engine of the RNA-induced silencing complex (RISC). Within the RISC, Argonaute accepts small RNA duplexes and selects one strand as the guide to mediate mRNA silencing by Ago-catalyzed target-mRNA cleavage or translational silencing. Humans have four Argonaute proteins (Ago1–Ago4) associated with RNAi; however, only one is known to cleave target RNAs through its 'slicer' activity, whereas others silence mRNAs through translational repression followed by decay¹. But what is the basis for this marked difference in Ago function, given the remarkable sequence and architectural similarity of these proteins? Two papers by Hauptmann et al.² and Schürmann et al.³ in this issue of Nature Structural & Molecular Biology now reveal the roles of noncatalytic structural features within Argonaute that can influence enzymatic activity. This work elucidates the mechanism of the functional divergence of Ago family members and provides a basis for further analysis of their biological and biochemical behavior.

The Argonaute clade of proteins includes four structural domains: N-terminal (N), PAZ, MID and PIWI (Fig. 1a). These domains form a bilobed architecture, a structure that helps explain how a guide RNA is recognized through two conserved sets of interactions^4,5,6. The 5′-phosphate and the first nucleotide of the guide strand are anchored by the MID domain, whereas the 3′-hydroxyl end is bound by the PAZ domain (Fig. 1b). The PIWI domain contains a catalytic tetrad of acidic residues that trigger the endonucleolytic cut in a target RNA⁵ (Fig. 1a,b). The human Ago1 and Ago4 proteins lack the intact catalytic tetrad, and neither has RNA-cleaving activity. Surprisingly, however, both the catalytically active Ago2 and the catalytically inactive Ago3 possess these residues necessary for RNA cleavage, although only Ago2 is an active slicer enzyme. This leads to the following two questions: what renders Ago3 to be catalytically inactive, and what does this imply about the evolution of multiple Ago variants?

Figure 1: Model of the elements in the N domain that affect cleavage activity.

(a) Domain architecture of Ago2 with the chimeric residues in the N domain indicated in red and the acidic catalytic residues with asterisks; aa, amino acids. (b) Model of the Ago2 crystal structure with the N-terminal regions affecting PIWI cleavage activity highlighted in red and the acidic catalytic residues with asterisks⁴. Nucleotides 1–8 of a guide RNA are drawn in, and the ends of the RNA are noted in red. The linker regions (L1 and L2) are drawn in gray. (c) Counterclockwise rotation of the Ago2 representation from b to match the orientation in Hauptmann et al.².

Hauptmann et al.² tackled these questions with chimeric proteins, first combining domains from human Ago2 and Ago3. The PIWI domain from Ago3, in the context of the remaining domains from Ago2, was found to be catalytically competent, thus confirming that the Ago3 PIWI domain does not lack intrinsic functionality. This led to the hypothesis that Ago3 prevents RNA cleavage with additional inhibitory elements. By analyzing the sequence conservation between the two Argonaute proteins, the authors designed and tested a series of chimeras, ultimately revealing two sequences in the N domain that inhibit the activity of the PIWI domain. One comprises the N terminus of the protein and forms little secondary structure, whereas the other is located toward the C-terminal end of the N domain and contains two α-helices (Fig. 1c). The authors speculated that the amino acids in the loops of these elements help Ago2 align the catalytic center to cleave a target RNA.

A complementary approach, called DNA family shuffling, was used by Schürmann et al.³ to create a library of chimeric Argonaute proteins. This technique can generate up to tens of thousands of chimeric cDNAs at once and has been used for vaccine development and protein engineering⁷. In this modified version of PCR, highly similar gene sequences are fragmented with enzymes, heated and reannealed to allow the single-stranded DNA to bind homologous sequences. PCR is then used to extend the overlapping regions and create chimeric full-length genes. With DNA family shuffling, Schürmann et al.³ isolated comparable parts of the N domain that overlap with the regions outlined above. One part (motif I) contains just five amino acids at the N-terminal end of the protein, and the other (motif II) contains a segment at the C-terminal end of the N domain (Fig. 2). Protein modeling based on the human Ago2 crystal structure showed that these two motifs reside adjacent to the PIWI domain. Similarly to Hauptmann et al.², Schürmann et al.³ speculated that motif I in Ago2 helps position the catalytic residues to activate Argonaute, specifically a glutamate whose contributions to catalysis were not recognized until the recent crystal structures of Argonaute were published⁵. They predicted that the unique region in Ago3 prevents this reorientation, inhibiting the activity of the protein. They further proposed that motif II inhibits guide-target pairing with additional amino acids in the loop between the α-helices; this bulky loop blocks a channel in Argonaute needed to propagate base-pairing with target RNAs.

Figure 2: Minimal changes required to impart activity to Argonaute.

Ago2 is the only Argonaute known to cleave target RNAs through its 'slicer' activity and needs no mutations (denoted by black check). To become an active slicer enzyme (denoted by green check) on perfect small-RNA duplexes, Ago1, similar to Ago2, required mutation of both the N and PIWI domains, whereas Ago3 required only mutation of the N domain. No chimeric protein containing Ago4 domains generated a competent enzyme (denoted by red x). Motifs I and II identified in Schürmann et al.³ are indicated for Ago3.

The same trend of inhibitory regions in the N domain is observed with human Ago1. Ago1 has a PIWI domain lacking the acidic residues, but once it is made catalytically active by the appropriate mutation, the N domain of Ago1 still prevents cleavage (Fig. 2). By replacing just the residues at the N terminus of the protein, Hauptmann et al.² created a fully competent slicer enzyme. These results, in combination with data from other species, indicate that the regulatory role of the N domain is well conserved among Argonautes. The crystal structure of a bacterial Argonaute from Thermus thermophilus revealed that the N domain prevents proper base-pairing with target RNA through steric clashes⁸. However, once this part of the N domain (corresponding to the first 106 amino acids of the protein) was deleted, the T. thermophilus Ago was rendered catalytically inactive, a result suggesting that this domain could stabilize the active complex. In Drosophila melanogaster, mutations to the N-terminal lobe also affected the activity of the PIWI domain⁹. Finally, another study using alanine scanning identified amino acids in the same region of Ago2 that could affect RISC assembly¹⁰.

Although it is now clear that mutations distal to the catalytic center can dramatically change the activity of Argonaute, it remains unclear why there are four distinct Argonaute proteins in humans. In considering the evolution of the Ago proteins, it is notable that Ago4 is the most divergent of the four proteins and is not expressed in most human cell lines¹¹. Using their library of chimeric proteins, Schürmann et al.³ noted that no chimera containing any part of Ago4 was competent for slicing, reinforcing the idea that Ago4 is evolutionarily and functionally distinct from the other human Ago proteins (Fig. 2). The library created by Schürmann et al.³, containing >35,000 chimeras, could help unravel the mystery of gene duplications for Argonaute. This library could probe Argonaute structural features' complex effects not only on RNA cleavage but also on gene silencing. It may be that the different natural Argonautes can fine-tune the rate and extent of mRNA suppression, to provide exquisite control over biological outcomes.

Beyond Argonaute's intrinsic behavior, the interactions between Argonaute and other proteins also contribute to gene-silencing efficiency. Various proteins assist Argonaute in translational repression of mRNAs, including glycine-tryptophan proteins that directly bind the PIWI domain¹². The studies from Hauptmann et al.² and Schürmann et al.³ raise the interesting possibility that such interactions could activate Ago3 or inactivate Ago2 by rearranging Argonaute's N-terminal extensions. Dramatic conformational rearrangements mediated by the chaperone machinery are already known to help create a mature RISC^13,14. Corroborating the results discussed here, a related recently published crystal structure of Ago1 confirms the regulatory role of the N domain¹⁵. Future studies should reveal the properties of different Argonaute complexes and how they contribute to small-RNA biology.