Biogenesis of miRNAs in mammals

3.1.2. Dicing of the pre-miRNA into miRNA/miRNA* duplex

Once into the cytoplasm, the pre-miRNA is recognised by Dicer. It was first described that this enzyme binds via its Piwi Argonaute Zwille (PAZ) domain the 3' overhang of the pre- miRNA prior to its cleavage near the terminal loop by the two RNAse III catalytic domains into an RNA duplex of approximately 22 nt with 2 nt 3' overhangs (Figure 4) (Bernstein et al., 2001; Hutvágner et al., 2001; MacRae et al., 2006; Zhang et al., 2004). More recently it was shown that in metazoans Dicer also anchors the 5' end of the pre-miRNA through a basic motif (5' pocket), which recognises its 5'-terminal phosphate group with the cleavage site being principally determined by the distance from this end (Park et al., 2011). Dicer has been described to interact during this step with the HIV-1 TAR RNA-binding protein (TRBP) and with the PACT protein. These two cofactors are not required for the processing activity itself, but it seems that they have a role in facilitating RISC assembly and their depletion strongly affects mature miRNAs accumulation in the cells. They are considered to constitute the RISC- loading complex in association with Dicer and Ago proteins (Chendrimada et al., 2005; Haase et al., 2005; Lee et al., 2006). Therefore, some argue that RISC assembly is an event coupled with dicing -independent of ATP-, but others argue that these processes uncoupled and ATP- dependent are (Yoda et al., 2010).

In rare cases, pre-miRNA processing can be Dicer-independent such as for miR-451. Its maturation has been shown to rely on Ago2, via endogenous cleavage of the pre-miRNA to generate an intermediate 3' end, which is further trimmed giving rise to its mature form (Cheloufi et al., 2010; Cifuentes et al., 2010).

3.1.3. Mature strand selection and loading into RISC

After processing by Dicer, the mature or "guide" strand, which possesses a biological function, is incorporated into the Ago-containing RISC - thus referred to as "miRISC". In some cases the miRNA*, which is usually degraded, can also be incorporated (Figure 4) (Carmell et al., 2002). Initially, regarding the mechanism of strand selection, it was shown that the extremity of the duplex with the weakest thermodynamical pairing defines the 5' of the guide strand (Khvorova et al., 2003; Schwarz et al., 2003). Consistently, recent studies have shown that Dicer, through its helicase domain, could sense the thermodynamic stability of the duplex’s ends (Noland et al., 2011; Sakurai et al., 2011; Welker et al., 2011). Noland et al., thus proposed a model in which the RNA duplex, following dicing, is repositioned within Dicer complexes in an orientation to allow correct guide strand incorporation into RISC

(Noland et al., 2011). Finally, it seems that in animals a 5'U tends to improve the loading of the guide strand. This strand selection mechanism is distinct from its contribution to weakening base pairing at the 5!-end (Seitz et al., 2011).

Whereas in Drosophila mismatched miRNA duplexes and perfectly matched siRNAs duplexes are actively sorted respectively into AGO1- and AGO2-RISC complexes (Förstemann et al., 2007; Tomari et al., 2007), in human, all four Ago proteins (Ago1-4) incorporate miRNAs indiscriminately of their sequence, showing very similar preferences for the structures of small RNA duplexes (Liu et al., 2004a; Meister et al., 2004; Yoda et al., 2010). Central mismatches promote RISC loading, and thus their features are reminiscent of Drosophila AGO1 (Yoda et al., 2010). Among the four human Ago, only Ago2 has been shown to be capable of miRNA- and siRNA-mediated target cleavage through an RNAse H- like fold (Liu et al., 2004a; Meister et al., 2004). Ago proteins are multidomain proteins that contain an N-terminal, a PAZ, a middle (MID), and a PIWI domain. They adopt a bilobate architecture, with one lobe formed by the N-terminal and PAZ domains, and the other by the MID and PIWI domains (Jinek and Doudna, 2009). The RNAse H-like fold is found in the PIWI domain (Song et al., 2004). MID and PIWI domains junction forms a binding pocket for a deep anchoring of the 5' monophosphate group of the miRNA first nucleotide (Ma et al., 2004; Parker et al., 2005), the 3' end being reversibly bound to the PAZ domain in a preformed hydrophobic pocket (Lingel et al., 2003, 2004; Ma et al., 2004; Song et al., 2003;

Yan et al., 2003). It is interesting to note that slicing has been shown to occur between the bases of the target paired with nt 10-11 of the small RNA (Elbashir et al., 2001a, 2001c) and that modelling of a full length siRNA-target duplex gave consistent results for this specificity, placing the target RNA in the slicer catalytic site of Ago, at a fixed distance measured from the 5' end of the guide strand. Perfect complementarity around these base pairs would thus probably ensure correct orientation of the scissile phosphate group of the target in the active cleavage site (Jinek and Doudna, 2009).

The process during which Dicer substrates are loaded into RISC is not yet fully understood, but the incorporation of a single strand of the duplex, either siRNA or miRNA, seems to be a stepwise process. Recent studies and inferred models have proposed that it involves two steps:

a first step with the physical association of the duplex with Ago, and a second step of activation during which the passenger strand is eliminated by Ago, or in the presence of accessory proteins (Gu et al., 2011a; Kawamata and Tomari, 2010; Ye et al., 2011). Indeed, after association with the diced substrate, an active process of elimination by cleavage of the

Ago proteins (Matranga et al., 2005; Miyoshi et al., 2005; Rand et al., 2005; Wang et al., 2009a). In addition, it has also been suggested that before their binding by Dicer, some pre- miRNAs could interact with Ago2, which would result in the cleavage of the pre-miRNA hairpin 12 nt from its 3' end. Thus, this would facilitates the nicked miRNA* removal after pre-miRNA dicing (Diederichs and Haber, 2007). On the other hand, many studies have shown that RISC-loading requirements differ for miRNAs and siRNAs. Indeed, most miRNA duplexes contain mismatches, thus preventing Ago-mediated slicing, and Ago proteins that lack slicer activity are unable to cleave the passenger strand. Even for Ago2-loading of siRNAs, it was demonstrated that slicing enhances strand separation, but this is not an absolute requirement, as a "bypass mechanism" exists when cleavage by the small RNA is not feasible (Matranga et al., 2005). Therefore, an RNA helicase activity is thought to mediate the unwinding and removal of the passenger strand of the miRNA duplex, and this step could be performed by Dicer or by a specific helicase, such as for example P68 or RNA Helicase A (Robb and Rana, 2007; Salzman et al., 2007).