Mechanistic analysis of the enhanced RNAi activity by 6-mCEPh-purine at the 5' end of the siRNA guide strand.

A key approach for improving siRNA efficacy is chemical modifications. Through an in silico screening of modifications at the 5'-end nucleobase of the guide strand, an adenine-derived compound called 6-(3-(2-carboxyethyl)phenyl)-purine (6-mCEPh-purine) was identified to improve the RNAi activity in cultured human cells and in vivo mouse models. Nevertheless, it remains unclear how this chemical modification enhances the siRNA potency. Here, we used a series of biochemical approaches to quantitatively evaluate the effect of the 6-mCEPh-purine modification at each step in the assembly of the RNAi effector complex called RISC. We found that the modification improves the formation of mature RISC at least in two different ways, by fixing the loading orientation of siRNA duplexes and increasing the stability of mature RISC after passenger strand ejection. Our data will provide a molecular platform for further development of chemically modified siRNA drugs.

INTRODUCTION

RNA interference (RNAi) is an evolutionarily conserved gene silencing mechanism that is triggered by double-stranded RNAs (dsRNAs). Long dsRNAs are first processed by the RNase III enzyme Dicer into ∼21 nt small interfering RNAs (siRNAs), which are then incorporated into the effector complex called RNA-induced silencing complex (RISC) for direct cleavage of complementary target RNAs (Kobayashi and Tomari 2016). In mammals, dsRNAs longer than ∼30 nt activate innate immune pathways, which ultimately lead to cell death (Alexopoulou et al. 2001; Barber 2005; Gantier and Williams 2007; Estornes et al. 2012). However, prediced or synthetic ∼21 nt siRNA duplexes can mediate efficient RNAi without adverse immune responses (Elbashir et al. 2001; Marques and Williams 2005), enabling their utilization even in humans. Recently, the first siRNA drug Onpattro, which targets transthyretin (TTR) mRNA and reduces misfolded TTR amyloids, was approved for treating patients suffering hereditary ATTR amyloidosis (Adams et al. 2018).

Chemical modifications of siRNAs are key to improving the efficacy of RNAi, especially in vivo. Indeed, both small RNA strands of Onpattro are heavily, but not completely, modified with 2′-O-methylation in order to improve the chemical stability and silencing activity (Adams et al. 2018). Rational design of further chemical modifications has promising potentials for future siRNA drug development (Khvorova and Watts 2017; Egli and Manoharan 2019; Setten et al. 2019). To this end, it is critical to understand the precise effect of chemical modifications on the assembly and function of RISC.

RISC contains a member of the Ago subfamily of Argonaute protein family. siRNA duplexes are first loaded into Ago by the aid of the Hsc70/Hsp90 chaperone machinery (Iki et al. 2010; Iwasaki et al. 2010; Johnston et al. 2010; Miyoshi et al. 2010; Tsuboyama et al. 2018). The resultant complex containing Ago and a siRNA duplex is called “pre-RISC.” Subsequently, one of the two siRNA strands, called the passenger strand, is ejected from Ago, forming “mature RISC” that contains Ago and the single-stranded guide strand (Matranga et al. 2005; Miyoshi et al. 2005; Rand et al. 2005; Leuschner et al. 2006). The selection of the guide strand is often asymmetric; in general, the strand whose 5′ end is less stably paired is more favored as the guide strand (Schwarz et al. 2003). Importantly, the direction of duplex loading (i.e., the orientation of the siRNA duplex in pre-RISC) predetermines which of the two strands remains in mature RISC as the guide strand. Mature RISC then recognizes target RNAs complementary to the guide strand and mediates endonucleolytic cleavage, translational repression, and/or deadenylation and decay (Huntzinger and Izaurralde 2011; Iwakawa and Tomari 2015). In humans, there are four Ago paralogs (Ago1–Ago4), among which Ago2 is a major endonuclease or “slicer” for target cleavage (Liu et al. 2004; Meister et al. 2004; Park et al. 2017b).

Ago proteins are characterized by having a bilobed architecture, with one lobe consisting of the N and PAZ domains and the other lobe containing the MID and PIWI domains (Song et al. 2004; Yuan et al. 2005; Wang et al. 2008). The pocket called “5′ nucleotide-binding pocket” is located at the interface between the MID and PIWI domains, which recognizes the monophosphate and the nucleobase at the 5′ end of the guide strand (Wang et al. 2008; Elkayam et al. 2012; Nakanishi et al. 2012; Schirle and MacRae 2012). A previous structural and biophysical analysis of the isolated MID domain of human Ago2 suggested that the 5′ nucleotide-binding pocket has a nucleobase preference for uracil (U) and adenine (A), compared to cytosine (C), and guanin (G) (Frank et al. 2010). The recognition of the 5′-end nucleobase was also confirmed in the structure of full-length human Ago2 (Elkayam et al. 2012; Schirle and MacRae 2012).

It is thought that the binding of the 5′ end of the guide strand in the 5′ nucleotide-binding pocket plays a critical role in anchoring siRNAs within Ago. Accordingly, Shinohara et al. hypothesized that chemically modifying the 5′-end nucleobase of the guide strand could be an effective strategy to enhance the siRNA knockdown potency by increasing the affinity of the siRNA duplex in Ago2 (Shinohara et al. 2021). After an in silico screening of modifications, they designed the compound 6-(3-(2-carboxyethyl)phenyl)purine (6-mCEPh-purine) (Fig. 1A), an adenine-derived nucleotide analog bearing a hydrophobic moiety and an acidic functional group at the position 6, which occupies the empty space around the 6th position of the adenine nucleobase and creates additional interactions in the pocket. As anticipated, Shinohara et al. found that 6-mCEPh-purine markedly enhances RNAi activity in cultured cells and in an in vivo mouse model (IC50 = 0.073 mg/kg for 6-mCEPh-purine and >0.3 mg/kg for A at 168 h after tail vein injection; Shinohara et al. 2021). Nevertheless, the molecular mechanisms underlying the observed improvement by 6-mCEPh-purine remains unclear. Here, using a series of cell-free biochemical assays, we carefully evaluated the effect of the 6-mCEPh-purine modification at each fundamental step in RISC assembly, and found that 6-mCEPh-purine enhances the formation of mature RISC at least in two different ways. Our analysis provides a molecular platform for further development and optimization of siRNA chemical modifications.

FIGURE 1.

The 6-mCEPh-purine modification improves the target cleavage. (A) Chemical structure of the 6-mCEPh-purine modification. (B) Four siRNA duplexes used in this set of experiments. They bore different nucleotides (A, U, 6-mCEPh-purine or G) at the 5′ end of the guide strand and 5-nitroindole at position 19 of the passenger strand. Both the guide and passenger strands had a 5′ monophosphate. (C) Scheme of RISC assembly and target cleavage. The target RNA was 5′ cap-radiolabeled. (D) A representative result of the target cleavage assay. The upper and lower bands correspond to the full-length and cleaved target, respectively. (E) Quantification of the target cleavage assay. The graph shows the average ± SD from three independent experiments using the same set of reagents. The ANOVA P-values are given in an inset under the graphs. The results of pairwise comparison on the effect of 6-mCEPh-purine compared to natural nucleotides are indicated by asterisks (*) P < 0.05, (**) P < 0.005, (***) P < 0.0005. The detailed results of statistical analyses are summarized in Supplemental Table 1.

RESULTS

The 6-mCEPh-purine modification enhances target cleavage

As a starting point for understanding how the 6-mCEPh-purine modification enhances RNAi in vitro, we decided to perform a well-established target cleavage assay using a series of siRNA duplexes bearing A, U, 6-mCEPh-purine or G at the 5′ end of the guide strand (Fig. 1B). Changing the guide 5′-end nucleobase inevitably alters the base-pairing state and the thermodynamic asymmetry of the siRNA duplexes, which may affect which strand is chosen as the guide. To equalize this effect, we introduced the universal base 5-nitroindole (Supplemental Fig. 1) to position 19 of the passenger strand, which lies across from the 5′ end of the guide strand (Loakes and Brown 1994; Kawamata et al. 2011). We also prepared a 5′ cap-radiolabeled target RNA bearing a complementary sequence to the guide strand, but with an adenine at the position opposite to the guide 5′-end nucleobase (Fig. 1C); this position is called “t1A,” which does not form a base pair with the guide 5′ nucleobase but is instead directly recognized by Ago2 (Schirle et al. 2015). To assemble RISC, we incubated the four siRNA duplexes with different guide 5′-end nucleobases in the lysate of HEK293T cells overexpressing Ago2. We then added the target RNA and examined the time courses of the target cleavage reaction. Predictably, 6-mCEPh-purine showed significantly stronger target cleavage than natural nucleotides (Fig. 1D,E; see Materials and Methods and Supplemental Table 1 for statistical analyses; the same hereinafter), confirming the in cells and in vivo results (Shinohara et al. 2021).

The 6-mCEPh-purine modification improves mature RISC formation

To more directly examine the effect of the 6-mCEPh-purine modification on RISC assembly, we radiolabeled the guide strand 5′ end in the four siRNA duplexes (Fig. 2A) and monitored the formation of pre-Ago2-RISC and mature Ago2-RISC in Ago2-overexpressing HEK293T cell lysate, using a previously established assay that utilizes native agarose gel electrophoresis (Yoda et al. 2010; Kawamata and Tomari 2011; Kawamata et al. 2011; Kwak and Tomari 2012). In this assay, a nonradiolabeled, uncleavable 30-nt 2′-O-methyl target oligonucleotide complementary to the guide strand was included in the reaction mix to trap mature Ago2-RISC (Fig. 2B, also see below). As shown in Figure 2C,D, the amounts of pre-Ago2-RISC were comparable among the four different duplexes. In contrast, siRNAs bearing 5′ A or U produced much more mature Ago2-RISC than that with 5′ G (Fig. 2C,E), consistent with the result of the target cleavage assay (Fig. 1E). Importantly, the 6-mCEPh-purine modification significantly promoted the formation of mature Ago2-RISC compared to nonmodified A (Fig. 2C,E), which also agrees with the observed enhancement in target cleavage (Fig. 1E).

FIGURE 2.

The 6-mCEPh-purine modification improves mature RISC formation. (A) Four siRNA duplexes used in this set of experiments. Each of them bore different nucleotides (A, U, 6-mCEPh-purine or G) at the 5′ end of the guide strand and 5-nitroindole at position 19 of the passenger strand. The guide strand was radiolabeled at the 5′ monophosphate, whereas the passenger strand had a nonradiolabeled 5′ monophosphate. (B) Scheme of RISC assembly. A nonradiolabeled, uncleavable target oligonucleotide complementary to the guide strand was added to trap mature Ago2-RISC. (C) A representative result of the native agarose gel assay. (D,E) Quantification of pre-Ago2-RISC (D) and mature Ago2-RISC (E) formation. The quantified signals were normalized to the mature Ago2-RISC value of 6-mCEPh-purine at 120 min. The graphs show the average ± SD from three independent experiments using the same set of reagents. The ANOVA P-values are given in insets under the graphs. Note that the ANOVA P-value for pre-Ago2-RISC in D is imprecisely determined, because the distribution of residuals was slightly skewed toward higher values even after log-transformation. The results of pairwise comparison on the effect of 6-mCEPh-purine compared to natural nucleotides are indicated by asterisks (*) P < 0.05, (**) P < 0.005, (***) P < 0.0005. The detailed results of statistical analyses are summarized in Supplemental Table 1.

The guide strand selection is slightly enhanced by the 6-mCEPh-purine modification

How does the 6-mCEPh-purine modification promote mature RISC formation without apparent changes in the amount of pre-RISC formed? Given that the guide and passenger strands cannot be distinguished in the pre-RISC signal on the native agarose gel, one possibility is that 6-mCEPh-purine helps the asymmetric selection of the guide strand by fixing the orientation of duplex loading. To investigate this hypothesis, we radiolabeled the passenger strand in the four siRNA duplexes (Fig. 3A), incubated the duplexes in Ago2-overexpressing HEK293T cell lysate together with a 2′-O-methyl oligonucleotide complementary to the passenger strand, and observed the formation of pre-Ago2-RISC and “passenger-derived” mature Ago2-RISC on native agarose gel (Fig. 3B). The amounts of “passenger-derived” mature RISC (Fig. 3C,E) were generally much lower than those of “guide-derived” mature RISC (Fig. 2C,E), indicating that the four duplexes have intrinsically strong asymmetry in guide-strand selection presumably due to the relatively weak thermodynamic stability between 5-nitroindole and natural nucleotides (Loakes and Brown 1994). Nevertheless, the 6-mCEPh-purine modification reduced the formation of passenger-derived mature RISC, compared to A, U and G (Fig. 3C,E), suggesting that 6-mCEPh-purine can further enhance the guide strand selection upon duplex loading. However, the observed difference was small, leaving a possibility of another reason(s) for the observed improvement of guide-derived mature RISC formation by 6-mCEPh-purine in Figure 2C.

FIGURE 3.

The guide strand selection is slightly enhanced by the 6-mCEPh-purine modification. (A) Four siRNA duplexes used in this set of experiments. Each of them bore different nucleotides (A, U, 6-mCEPh-purine or G) at the 5′ end of the guide strand and 5-nitroindole at position 19 of the passenger strand. The passenger strand was radiolabeled at the 5′ monophosphate, whereas the guide strand had a nonradiolabeled 5′ monophosphate. (B) Scheme of RISC assembly. A nonradiolabeled, uncleavable target oligonucleotide complementary to the passenger strand was added. (C) A representative result of the native agarose gel assay. (D,E) Quantification of pre-Ago2-RISC (D) and mature Ago2-RISC (E) formation. The quantified signals were normalized to the mature Ago2-RISC value of 6-mCEPh-purine at 120 min. The graphs show the average ± SD from three independent experiments using the same set of reagents. The ANOVA P-values are given in insets under the graphs. The results of pairwise comparison on the effect of 6-mCEPh-purine compared to natural nucleotides are indicated by asterisks (*) P < 0.05, (**) P < 0.005, (***) P < 0.0005. The detailed results of statistical analyses are summarized in Supplemental Table 1.

6-mCEPh-purine enhances mature RISC formation independently of the guide strand selection

Another possibility for the improvement of mature RISC formation is that the 6-mCEPh-purine modification enhances the anchoring of the guide strand in Ago2 after the ejection of the passenger strand. To test this hypothesis, we replaced the 5′ monophosphate of the passenger strand with an amino linker (Fig. 4A), which blocks loading of this strand into the nucleotide-binding pocket of Ago2 (Chiu and Rana 2002; Czauderna et al. 2003). Accordingly, the orientation of duplex loading is fixed in such a way that the strand with 5′ 6-mCEPh-purine, A, U or G is always chosen as the guide (Fig. 4B). Still, we observed enhanced formation of mature RISC with 6-mCEPh-purine (Fig. 4C,E), whereas the amounts of pre-RISC formed were similar among the four duplexes (or slightly smaller for 6-mCEPh-purine) (Fig. 4C,D). This observation indicates that the 6-mCEPh-purine modification can enhance mature RISC formation independently of the guide strand selection. Similar enhancement of mature RISC-like signals on the native agarose gel was observed with a different siRNA sequence derived from firefly luciferase (Supplemental Fig. 2), although the identities of the bands remain to be validated.

FIGURE 4.

6-mCEPh-purine enhances mature RISC formation independently of the guide strand selection. (A) Four siRNA duplexes used in this set of experiments. Each of them bore different nucleotides (A, U, 6-mCEPh-purine or G) at the 5′ end of the guide strand and 5-nitroindole at position 19 of the passenger strand. The guide strand was radiolabeled at the 5′ monophosphate, whereas the passenger strand held a 5′ amino linker that fixes the loading orientation by blocking this strand from being anchored in the 5′ nucleotide-binding pocket of Ago2. (B) Scheme of RISC assembly. A nonradiolabeled, uncleavable target oligonucleotide was added. (C) A representative result of the native agarose gel assay. (D,E) Quantification of pre-Ago2-RISC (D) and mature Ago2-RISC (E) formation. The quantified signals were normalized to the mature Ago2-RISC value of 6-mCEPh-purine at 120 min. The graphs show the average ± SD from three independent experiments using the same set of reagents. The ANOVA P-values are given in insets under the graphs. The results of pairwise comparison on the effect of 6-mCEPh-purine compared to natural nucleotides are indicated by asterisks (*) P < 0.05, (**) P < 0.005, (***) P < 0.0005. The detailed results of statistical analyses are summarized in Supplemental Table 1.

Different 5′-end nucleotides do not affect the duplex loading efficiency

In all the native agarose gel experiments described above, the formation of pre-RISC was consistently similar among the duplexes bearing the different nucleotides (A, U, G, and 6-mCEPh-purine). However, because pre-RISC is continuously converted into mature RISC during RISC assembly, it still remained unclear if the 6-mCEPh-purine modification has any direct effect on the efficiency of duplex loading per se. To precisely address this point, a 2′-O-methyl modification was introduced at the position 9 of the passenger strand, which blocks slicing and ejection of the passenger strand (Fig. 5A,B; Miyoshi et al. 2005; Leuschner et al. 2006) and allows us to purely monitor the efficiency of duplex loading. Our data showed that pre-Ago2-RISC formation was virtually equal among the four different siRNA duplexes (Fig. 5C,D). Taken all together, our data suggest that the 6-mCEPh-purine modification does not affect the efficiency of siRNA duplex loading into Ago2 to form pre-RISC, but increases mature RISC stability after the passenger strand is ejected.

FIGURE 5.

The efficiency of duplex loading is not affected by different 5′-end nucleobases. (A) Four siRNA duplexes used in this set of experiments. Each of them bore different nucleotides (A, U, 6-mCEPh-purine or G) at the 5′ end of the guide strand and 5-nitroindole at position 19 of the passenger strand. The guide strand was radiolabeled at the 5′ monophosphate, whereas the passenger strand held a 5′ amino linker that fixes the loading orientation and a 2′-O-methyl modification that blocks passenger ejection. (B) Scheme of RISC assembly. A nonradiolabeled, uncleavable target oligonucleotide was added. (C) A representative result of the native agarose gel assay. (D) Quantification of pre-Ago2-RISC formation. The quantified signals were normalized to the pre-Ago2-RISC value of 6-mCEPh-purine at 120 min. The graph shows the average ± SD from three independent experiments using the same set of reagents. The ANOVA P-values are given in an inset under the graphs. The results of pairwise comparison on the effect of 6-mCEPh-purine compared to natural nucleotides are indicated by asterisks (*) P < 0.05, (**) P < 0.005, (***) P < 0.0005. The detailed results of statistical analyses are summarized in Supplemental Table 1.

The target does not influence enhanced mature RISC formation by 6-mCEPh-purine

As described above, we routinely include a complementary nonradiolabeled 30-nt 2′-O-methyl oligonucleotide in our native agarose gel assay, which enables detection of trapped mature RISC as a sharp signal on the native agarose gel by avoiding its binding to heterogenous mRNAs present in the lysate; in the absence of the target, the signal of mature Ago2-RISC is detected as a smeared signal on the gel (Supplemental Fig. 3). However, it has been reported that interactions between mature RISC and its targets can promote destabilization and/or “unloading” of the guide strand from human Ago2-RISC (Ameres et al. 2010; Cazalla et al. 2010; De et al. 2013; Park et al. 2017a). To evaluate the effect of the 6-mCEPh-purine modification on RISC assembly in the absence of targets, we treated the Ago2-overexpressing HEK293T lysate with micrococcal nuclease (MNase) to digest and deplete endogenous mRNAs (Svitkin and Sonenberg 2004). After quenching the MNase activity by EGTA, we then performed the native agarose gel assay without adding a 2′-O-methyl target oligonucleotide (Fig. 6A,B). In this condition, mature Ago2-RISC can be detected as a discrete signal from pre-Ago2-RISC even without the target RNAs (Supplemental Fig. 3). Still, we observed significantly enhanced formation of mature Ago2-RISC with 6-mCEPh-purine, as well as comparable formation of pre-Ago2-RISC among the different duplexes (Fig. 6C–E). We concluded that the observed enhancement of mature RISC formation by 6-mCEPh-purine is independent of target-mediated unloading or destabilization of the guide strand.

FIGURE 6.

The target does not influence enhanced mature RISC formation by 6-mCEPh-purine. (A) Four siRNA duplexes used in this set of experiments. Each of them bore different nucleotides (A, U, 6-mCEPh-purine or G) at the 5′ end of the guide strand and 5-nitroindole at position 19 of the passenger strand. The guide strand was radiolabeled at the 5′ monophosphate, whereas the passenger strand held a 5′ amino linker that fixes the loading orientation. (B) Scheme of RISC assembly. The lysate was treated with MNase for the depletion of endogenous mRNAs, and no target was added. (C) A representative result of the native agarose gel assay. (D,E) Quantification of pre-Ago2-RISC (D) and mature Ago2-RISC (E) formation. The quantified signals were normalized to the mature Ago2-RISC value of 6-mCEPh-purine at 120 min. The graphs show the average ± SD from three independent experiments using the same set of reagents. The ANOVA P-values are given in insets under the graphs. The results of pairwise comparison on the effect of 6-mCEPh-purine compared to natural nucleotides are indicated by asterisks (*) P < 0.05, (**) P < 0.005, (***) P < 0.0005. The detailed results of statistical analyses are summarized in Supplemental Table 1.

Passenger strand cleavage is not affected by the 6-mCEPh-purine modification

It is known that human Ago2 can cleave the passenger strand within pre-RISC, which facilitates the passenger strand ejection and mature RISC formation (Matranga et al. 2005; Miyoshi et al. 2005; Rand et al. 2005; Leuschner et al. 2006; Park and Shin 2015). Therefore, the efficiency of passenger strand ejection can be estimated by monitoring the production of the cleavage fragment of the passenger strand, which is rapidly degraded in normal conditions but can be trapped for detection by a 2′-O-methyl oligonucleotide complementary to the passenger strand (Matranga et al. 2005). To directly test if the 6-mCEPh-purine modification affects the passenger strand cleavage, we used the same setting as in Figure 4 but radiolabeled the 3′ end of the passenger strand instead of the 5′ end of the guide strand (Fig. 7A). The trapped 13-nt passenger cleavage fragment was accumulated at similar levels among the different nucleotides at the guide 5′ end (Fig. 7B,C), indicating that 6-mCEPh-purine does not affect the efficiency of the passenger strand cleavage (Fig. 7D). This result corroborates the idea that the 6-mCEPh-purine modification stabilizes mature RISC after passenger strand ejection.

FIGURE 7.

The 6-mCEPh-purine modification does not impact the passenger cleavage. (A) Four siRNA duplexes used in this set of experiments. They bore different nucleotides (A, U, 6-mCEPh-purine or G) at the 5′ end of the guide strand and 5-nitroindole at position 19 of the passenger strand. (B) Scheme of RISC assembly and passenger cleavage. The passenger was 3′ radiolabeled. (C) A representative result of the passenger cleavage assay. The upper and lower bands correspond to the full-length and cleaved passenger strand, respectively. (D) Quantification of the target cleavage assay. The graph shows the average ± SD from three independent experiments using the same set of reagents. The ANOVA P-values are given in an inset under the graph. The results of pairwise comparison on the effect of 6-mCEPh-purine compared to natural nucleotides are indicated by asterisks (*) P < 0.05, (**) P < 0.005, (***) P < 0.0005. The detailed results of statistical analyses are summarized in Supplemental Table 1.

DISCUSSION

The effects of siRNA chemical modifications on the RNAi activity have been analyzed in many studies (Khvorova and Watts 2017; Egli and Manoharan 2019; Setten et al. 2019). However, most of the previous studies have relied on luciferase reporter assays in cultured cells, which can only evaluate the final outcome of target silencing. In this study, we used a series of biochemical approaches including native agarose gel electrophoresis and analyzed the mechanistic effect of the 6-mCEPh-purine modification at the 5′ end of the guide strand at each step during Ago2-RISC assembly. We found that 6-mCEPh-purine improves the formation of mature RISC at least in two different ways (Fig. 8); (1) 6-mCEPh-purine enhances the guide strand selection by fixing the loading orientation of siRNA duplexes (Fig. 3), and more importantly, (2) 6-mCEPh-purine increases the stability of mature RISC after passenger strand ejection (Figs. 4–6). On the other hand, no apparent difference was observed for the amount of pre-Ago2-RISC formed (Fig. 5). Thus, 6-mCEPh-purine does not quantitatively affect the efficiency of duplex loading, even though it can qualitatively change the orientation of the duplex in pre-Ago2-RISC (Fig. 3).

FIGURE 8.

A model for the molecular mechanism by which 6-mCEPh-purine enhances the RNAi activity. 6-mCEPh-purine modification improves the formation of mature RISC in two different ways: (1) 6-mCEPh-purine helps the guide strand selection. (2) 6-mCEPh-purine increases the stability of mature RISC after passenger strand ejection.

It has been suggested that anchoring of the guide 5′ monophosphate in the 5′ nucleotide-binding pocket is the first critical event in the loading of siRNA duplexes into Ago (Elkayam et al. 2012; Nakanishi et al. 2012; Schirle and MacRae 2012; Iwasaki et al. 2015). It is therefore reasonable to speculate that the guide 5′ nucleobase, adjacent to the 5′ monophosphate, is also anchored in the pocket at the step of duplex loading. However, enhancement by 6-mCEPh-purine was undetectable in the efficiency of pre-Ago2-RISC formation and became apparent only at the step of mature Ago2-RISC formation. There are two possible explanations for this counterintuitive observation. First, the “double-stranded state” of siRNA duplexes in pre-Ago2-RISC may act as an additional anchor point for Ago2, masking the effect of the 5′ nucleobase anchoring; this internal anchor point is lost after passenger ejection, making the 5′ nucleobase anchoring more important (Supplemental Fig. 4, Model 1). This is reminiscent of what we have previously proposed for Drosophila Ago1, where the enhancement effect of preferred nucleobases (A and U) was more prominent in mature Ago1-RISC than in pre-Ago1-RISC (Kawamata et al. 2011). Alternatively, but not mutually exclusively, the 5′ nucleobase may not be fully anchored in the 5′ nucleotide-binding pocket immediately upon duplex loading due to some structural constraints, and may become fixed in there only after ejection of the passenger strand (Supplemental Fig. 4, Model 2). In either case, the 6-mCEPh-purine modification at the guide 5′ nucleobase leads to the formation of more mature Ago2-RISC (Fig. 4), which explains higher RNAi efficiency observed in vitro (Fig. 1), in cells and in vivo (Shinohara et al. 2021).

Although siRNA therapeutics have a great potential, there is still room for improvement in their efficacy while reducing their dosage and cost (Khvorova and Watts 2017; Egli and Manoharan 2019; Setten et al. 2019). Our current study will help to provide a molecular platform for mechanistic understanding of chemically modified siRNAs, which will contribute to further development of RNAi drugs. Notably, 5′-(E)-vinylphosphonate (5′-E-VP), which mimics the 5′ phosphate of the siRNA guide strand, is known to enhance not only the metabolic stability and also the binding affinity to Ago2, leading to better silencing in vitro and in vivo (Lima et al. 2012; Prakash et al. 2016; Elkayam et al. 2017; Haraszti et al. 2017). Given that 6-mCEPh-purine is a 5′ nucleobase modification that is theoretically compatible with the 5′-E-VP modification, it would be interesting to analyze in the future if and how these two adjacent chemical modifications synergize to enhance the siRNA efficiency.

MATERIALS AND METHODS

Cell culture

HEK 293T cells were cultured in Dulbecco's modified Eagle's medium (DMEM) (Sigma) supplemented with 10% (v/v) Fetal Bovine Serum (FBS) (Sigma) at 37°C in 5% CO2.

Overexpression of FLAG-tagged hAgo2 protein in HEK 293T cells

HEK293T cells at ∼80% confluence were transfected with 10 µg/10-cm dish of pIRESneo-FLAG-HA-Ago2 (Meister et al. 2004) by using Lipofectamine 3000 (Thermo Fisher). The cells were harvested after 48 h.

Cell lysate preparation

HEK293T cells were washed three times with cold PBS (pH 7.4) and collected by centrifugation at 1000g for 3 min at 4°C. The cell pellet was resuspended in two packed-cell volume of lysis buffer (30 mM HEPES-KOH [pH 7.4], 100 mM potassium acetate 2 mM magnesium acetate) (Haley et al. 2003) containing 5 mM DTT and 1× EDTA-free Complete Protease Inhibitor tablets (Roche) and subjected to Dounce homogenization. Subsequently, the lysate was clarified by centrifugation at 17,000g for 30 min at 4°C. The supernatant was flash frozen in liquid nitrogen and immediately stored at −80°C in single-use aliquots.

Preparation of siRNA duplexes

The sequences, chemical modifications, and radiolabeling of the siRNA duplexes are shown in respective figures. The guide and passenger strands were heat-annealed in lysis buffer as previously described (Haley et al. 2003).

Preparation of the target mRNA for cleavage assay

The 182-nt target mRNA for the cleavage assay was in vitro transcribed using T7-Scribe Standard RNA IVT Kit (Cellscript) from the PCR product amplified from pGL3-basic vector (Promega) as previously described (Haley et al. 2003; Naruse et al. 2018). The target mRNA was gel purified and radiolabeled at the 5′ cap using the ScriptCap m7G Capping System (Cellscript) and [α-32P] GTP (PerkinElmer) according to the manufacturer's instructions.

Target cleavage assay

Target mRNA cleavage assays were performed in 20 µL reactions mixtures as described (Haley et al. 2003) with the following modifications: 2 µL of 200 nM 5′-phosphorylated siRNA duplexes were incubated at 25°C with 10 µL of lysate from HEK293T cells overexpressing FLAG-tagged Ago2 and 6 µL of 40× reaction mix (containing ATP, the ATP regeneration system and the RNase inhibitor; described in detail in Haley et al. 2003). Then, 2 µL of ∼10 nM cap-radiolabeled target mRNA was added. An amount of 2 µL of the reaction mixture was taken at each time point, mixed with 8 µL of low-salt PK solution (0.125% SDS, 12.5 mM EDTA, 12.5 mM HEPES-KOH (pH7.4), 12.5% Proteinase K), and incubated at 55°C for 10 min. An equal volume of 2× formamide dye (10 mM EDTA pH 8.0, 98% (w/v) deionized formamide, 0.025% (w/v) xylene cyanol, 0.025% bromophenol blue) was then added and incubated at 68°C for 5 min. The cleavage products were analyzed on an 8% denaturing polyacrylamide gel. Gels were dried and imaged by Typhoon FLA 7000 (GE Healthcare Life Sciences) and quantified using MultiGauge software (Fujifilm Life Sciences). Graphs were prepared using IgorPro (WaveMetrics).

Native agarose gel analysis

Native agarose gel analysis was performed essentially as previously described (Kawamata et al. 2009). Briefly, the 5′ end of the guide strands was radiolabeled by [γ-32P] ATP (PerkinElmer) and T4 Polynucleotide Kinase (Takara). The radiolabeling efficiency was comparable among A, U, 6-mCEPh-purine and G (∼1.05–1.27-fold, depending on the experiments) and the specific radioactivity was further normalized before being annealed to a 1.5-fold excess amount of the nonradiolabeled passenger strand. 1.4% (w/v) agarose gels (Low Range Ultra Agarose, Bio-Rad Laboratories) were cast vertically between glass plates with 1.5–2 mm-thick side spacers and a 0.5 mm-thick bottom spacer (16 cm × 16 cm). The reaction mixture containing 10 µL Ago2-expressing HEK293T cell lysate, 6 µL 40 × reaction mix, 2 µL 15% (w/v) Ficoll 400 in lysis buffer, 2 µL of ∼50 nM siRNA duplexes with the 5′ monophosphate of the guide strand radiolabeled, and 2 µL of 50 nM 2′-O-methyl target oligonucleotide (5′-mUmCmUmUmCmAmCmUmAmUmAmCmAmAmCmCmUmAmCmUmAmCmCmUmC/5-nitroindole/mAmCmCmUmU-3′) was incubated at 25°C. This target oligonucleotide was complementary to the siRNA guide strand and contained a 5-nitroindole at the position opposite the 5′-end nucleotide of the guide strand. In Figure 3, the passenger strand instead of the guide strand was radiolabeled and the corresponding target oligonucleotide (5′-mUmCmUmUmCmAmGmUmGmAmGmGmUmAmGmUmAmGmGmUmUmGmUmAmUmAmAmCmCmUmU-3′) was used. After incubation, the complexes were resolved by native agarose gel electrophoresis with 0.5× TBE buffer at 300 V at 4°C for 1.5 h. Complexes were detected by PhosphorImager (FLA-7000 image analyzer, Fujifilm) and quantified using MultiGauge software (Fujifilm). The background signals were appropriately subtracted by using an empty space for each lane on the gel as the “0” value and the graphs were prepared using IgorPro (WaveMetrics).

MNase treatment for digestion of endogenous mRNAs

First, naïve HEK293T cell lysate was fivefold diluted in lysis buffer. Then, MNase (2,000,000 gel unit/mL; NEB) was 125-fold diluted in the diluted native cell lysate. Finally, 100 µL of Ago2-expressing HEK293T cell lysate was incubated with 4 µL of the diluted MNase solution in the presence of 1 mM Ca(OAc)2 at 25°C for 20 min. 4 mM EGTA (final concentration) was added to stop the reaction. The native gel assay using MNase-treated lysate was performed in the absence of target oligonucleotide (Fig. 6; Supplemental Fig. 3).

Detection of passenger strand cleavage

Detection of passenger strand cleavage was performed as described (Matranga et al. 2005) with the following modifications: The 3′ end of the passenger strand was radiolabeled by [α-32P] cordycepin-5′-triphosphate (PerkinElmer) and yeast poly(A) polymerase (Thermo Fisher Scientific), while the 5′ end of the guide strand was nonradioactively monophosphorylated. Typically, in 20 µL reactions, 2 µL of siRNA duplexes were incubated at 25°C with 10 µL of lysate from HEK293T cells overexpressing FLAG-tagged Ago2, 2 µL of 1 µM 2′-O-methyl oligonucleotide complementary to the passenger strand, and 6 µL of 40× reaction mix. An amount of 2 µL of the reaction mixture was taken at each time point, mixed with 8 µL of low-salt PK solution, and incubated at 55°C for 10 min. An equal volume of 2× formamide dye was then added and incubated at 68°C for 5 min. The 3′ cleavage fragments of the passenger strand were analyzed on an 15% denaturing polyacrylamide gel. Gels were dried and imaged by Typhoon FLA 7000 (GE Healthcare Life Sciences) and quantified using MultiGauge software (Fujifilm Life Sciences). Graphs were prepared using IgorPro (WaveMetrics).

Statistical analyses

Significance of the observed differences was assessed by an analysis-of-variance (ANOVA) on a model with two explanatory variables: the 5′ nucleotide identity and a variable related to time by a bijective function (for mature RISC accumulation: variable = 1–2–t/τ; for pre-RISC decrease: variable = 2–t/τ; with t indicating time and τ indicating an estimated doubling or halving time [10 min for Figs. 1, 5, and 7; 50 min for Figs. 2–4 and 6]). These log-transforming functions were chosen because the abundance of mature RISC and pre-RISC should vary linearly to these variables in a simplified kinetics model. In every analysis throughout this study, ANOVA always found a significant effect (P < 0.05) of this time-dependent variable. Whenever a significant effect was also found for the 5′ nucleotide identity, the significance of the effect of an interaction between these two variables was also assessed. Whenever 5′ nucleotide identity, or its interaction with the time-dependent variable, was found to have a significant effect (P < 0.05), pairwise comparisons with false discovery rate correction were performed to compare the effect of 6-mCEPh-purine to natural nucleotides. Such pairwise comparisons were also performed by an ANOVA model with the same two explanatory variables. The results of statistical analyses are summarized in Supplemental Table 1. The results of pairwise comparison for the 5′ nucleotide identity (the effect of 6-mCEPh-purine compared to natural nucleotides) are indicated as asterisks in the figures (* P < 0.05; ** P < 0.005; *** P < 0.0005). Conditions for applicability of the ANOVA were verified by Levene's test (for variance homogeneity) and Shapiro Wilk's test (for residual normality); whenever conditions were not met, ANOVA was performed on log(values) instead of values themselves. Log-transformation was always sufficient to make the data suitable for ANOVA, except for pre-RISC abundance in Figure 2 (where the distribution of residuals was slightly skewed toward higher values, making the estimation of P-value by ANOVA imprecise for that analysis). Scripts and data files are available at https://github.com/HkeyHKey/Brechin_et_al_2020.

SUPPLEMENTAL MATERIAL

Supplemental material is available for this article.