Argonaute protein identity and pairing geometry determine cooperativity in mammalian RNA silencing

Abstract

Small RNAs loaded into Argonaute proteins direct silencing of complementary target mRNAs. It has been proposed that multiple, imperfectly complementary small interfering RNAs or microRNAs, when bound to the 3' untranslated region of a target mRNA, function cooperatively to silence target expression. We report that, in cultured human HeLa cells and mouse embryonic fibroblasts, Argonaute1 (Ago1), Ago3, and Ago4 act cooperatively to silence both perfectly and partially complementary target RNAs bearing multiple small RNA-binding sites. Our data suggest that for Ago1, Ago3, and Ago4, multiple, adjacent small RNA-binding sites facilitate cooperative interactions that stabilize Argonaute binding. In contrast, small RNAs bound to Ago2 and pairing perfectly to an mRNA target act independently to silence expression. Noncooperative silencing by Ago2 does not require the endoribonuclease activity of the protein: A mutant Ago2 that cannot cleave its mRNA target also silences noncooperatively. We propose that Ago2 binds its targets by a mechanism fundamentally distinct from that used by the three other mammalian Argonaute proteins.

FIGURE 1.

Potential sources of cooperativity in the repression of a target mRNA by the small RNA-directed Argonaute complex, RISC. (A) Cooperative binding. RISC binding at multiple target sites increases site occupancy by mutually stabilizing subsequent binding of RISCs. (B) Cooperative function. RISC binding at multiple sites may increase the likelihood that repressive factors, such as nucleases, are recruited to the mRNA. (C) Multiple independent sites. Each RISC functions independently, so the multiple sites increase the probability of repression but do not influence each other.

FIGURE 2.

Extent of pairing and target site number determine both efficacy and cooperativity in small RNA-directed silencing in HeLa cells. Silencing of a Renilla luciferase reporter mRNA bearing 1–6 target sites in its 3′ UTR, relative to a firefly luciferase internal control, was determined at different siRNA concentrations. Pairing between the siRNA guide (red) to the 3′ UTR sites (black) is shown at top. IC₅₀ and Hill coefficient (n^H) were calculated for each dose-response curve. Throughout this study, values are reported as mean ± standard deviation for IC₅₀ values and n^H; error bars indicate standard error for ≥12 biological replicates. The curves correspond to the concentration-dependence of silencing expected for the mean IC₅₀ and n^H values.

FIGURE 3.

siRNA validation in HeLa cells. Each siRNA was functional in silencing a reporter containing a single perfect target site. (A) Perfect siRNA. (B) Bulged siRNA. (C) siRNA with seed plus supplementary 3′ pairing (nt 13–16). (D) siRNA with only seed pairing. The curves correspond to the concentration-dependence of silencing expected for the mean IC₅₀ and n^H values (± standard deviation) calculated from three independent trials.

FIGURE 4.

Cooperative binding of RISC requires adjacent target sites in HeLa cells. Three sites spaced 19 nt apart (A) require more siRNA to achieve half-maximal silencing, compared to three adjacent sites (B), and act noncooperatively. In contrast, a perfectly matched siRNA silences a three-site reporter with sites separated by 19 nt (C) or a reporter with three adjacent sites (D) with equal efficacy and without detectable cooperativity. The three adjacent-site experiments in this figure were performed independently from those in Figure 2. A one sample, two-tailed Student's t-test was used to calculate the P-values at 95% confidence for the Hill coefficients to determine if n^H was significantly different from the null hypothesis: n^H = 1 (i.e., noncooperative).

FIGURE 5.

Silencing in Ago2^−/− MEFs or Ago2^−/− MEFs reconstituted with mouse Ago2 or catalytically inactive, mutant Ago2^D669A or Ago1^−/− MEFs. (A) In the absence of Ago2, silencing by a perfect site (n^H = 1.6 ± 0.4; P = 0.03) is equally cooperative as a bulged site (n^H = 1.8 ± 0.3; P = 0.006). (B) Mouse Ago2 expression restored noncooperative silencing by the perfect siRNA (black; n^H = 1.0 ± 0.1); silencing directed by a bulged siRNA became less cooperative (red; n^H = 1.5 ± 0.2; P = 0.02) than in the absence of Ago2 (red in A; n^H = 1.8 ± 0.3). (C) Catalytically inactive mouse Ago2^D669A likewise restored noncooperative silencing by a perfect siRNA (black; n^H = 1.1 ± 0.1), but silencing by the bulged siRNA (red; n^H = 1.5 ± 0.3; P = 0.04), was cooperative. (D) In the absence of Ago1, silencing by the perfect siRNA was not cooperative (black; n^H = 1.1 ± 0.1), but silencing by the bulged siRNA was cooperative (red; n^H = 1.7 ± 0.2; P = 0.003). A one sample, two-tailed Student's t-test was used to calculate the P-values at 95% confidence for the Hill coefficients to determine if n^H was significantly different from the null hypothesis: n^H = 1 (i.e., noncooperative).

FIGURE 6.

Ago1 and Ago2 protein levels in MEF cells. Ago2 was detected by Western blotting using a rabbit anti-Ago1 antibody that recognizes both mouse and human Ago1 and a rabbit anti-Ago2 antibody that recognizes both mouse and human Ago2. Ago protein levels were normalized to actin, and the level of Ago protein in wild-type MEFs was set to 1. Data are mean ± standard deviation for three trials. Inset shows representative data from a single experiment.

FIGURE 7.

In the absence of Ago2, effective silencing requires adjacent sites. (A,B) Both perfect (n^H = 2.1 ± 0.3; P = 0.007) and bulged (n^H = 1.5 ± 0.3; P = 0.04) adjacent sites were silenced cooperatively in the absence of Ago2. (C,D) In Ago2^−/− MEFs, three target sites spaced 19 nt apart did not silence the reporter. (E,F) Expressing mouse Ago2 in the Ago2^−/− MEFs allowed three distant sites to silence the reporter. (G,H) Expressing catalytically inactive, mutant Ago2^D669A also allowed three distant sites to silence the reporter. (I,J) In the Ago1^−/− MEFs the three distant sites silenced the reporter. A one sample, two-tailed Student's t-test was used to calculate the P-values at 95% confidence for the Hill coefficients to determine if n^H was significantly different from the null hypothesis: n^H = 1 (i.e., noncooperative).