Loquacious-PD facilitates Drosophila Dicer-2 cleavage through interactions with the helicase domain and dsRNA

Significance

Drosophila melanogaster use RNA interference to respond to a viral infection. Dicer-2 cleaves viral double-stranded RNA (dsRNA), producing siRNAs that silence viral gene expression. Dicer-2 recognizes the ends of dsRNA, and this property likely evolved to distinguish between viral and cellular dsRNA. Loquacious-PD (Loqs-PD), a dsRNA binding protein, is not required for Dicer-2’s antiviral activity. However, by allowing Dicer-2 to cleave in a termini-independent manner, Loqs-PD facilitates cleavage of endogenous substrates with more complex termini. Our studies are significant because they provide a mechanistic basis for how Loqs-PD modulates Dicer-2 activity. For example, they reveal a previously unrecognized protein–protein interaction interface on the helicase domain of Dicer-2.

Abstract

Loquacious-PD (Loqs-PD) is required for biogenesis of many endogenous siRNAs in Drosophila. In vitro, Loqs-PD enhances the rate of dsRNA cleavage by Dicer-2 and also enables processing of substrates normally refractory to cleavage. Using purified components, and Loqs-PD truncations, we provide a mechanistic basis for Loqs-PD functions. Our studies indicate that the 22 amino acids at the C terminus of Loqs-PD, including an FDF-like motif, directly interact with the Hel2 subdomain of Dicer-2’s helicase domain. This interaction is RNA-independent, but we find that modulation of Dicer-2 cleavage also requires dsRNA binding by Loqs-PD. Furthermore, while the first dsRNA-binding motif of Loqs-PD is dispensable for enhancing cleavage of optimal substrates, it is essential for enhancing cleavage of suboptimal substrates. Finally, our studies define a previously unrecognized Dicer interaction interface and suggest that Loqs-PD is well positioned to recruit substrates into the helicase domain of Dicer-2.

There are two Dicer genes in Drosophila melanogaster, Dcr-1 and Dcr-2, that produce micro-RNAs (miRNAs) and short interfering RNAs (siRNAs), respectively (1, 2). Dcr-2 is required to initiate antiviral RNA interference (RNAi), in which viral double-stranded RNA (dsRNA) is cleaved to produce siRNAs capable of silencing viral gene expression (3, 4). In vitro studies indicate Dcr-2 recognizes dsRNA termini and exhibits termini-dependent cleavage (5, 6). For example, dsRNA with blunt (BLT) termini are cleaved processively in an ATP-dependent manner, while dsRNA with 2-nt 3′overhanging (3′ovr) termini elicit distributive cleavage that occurs in the absence of ATP. Dcr-2’s helicase domain plays an important role in termini discrimination (5, 6) and is required to mount an antiviral response (7, 8), suggesting that the termini preferences of Dcr-2 likely arose to distinguish between viral and cellular dsRNA.

Loquacious-PD (Loqs-PD), a dsRNA-binding protein (dsRBP), is required for the biogenesis of a subset of endogenous-siRNAs (endo-siRNAs) (9, 10) but is not required for antiviral RNAi (8). Early studies found that endo-siRNAs map to dsRNA originating from convergent transcription, inverted repeats, and transposons (11–16). Given the sensitive termini dependence of Dcr-2, many endo-siRNA precursors are predicted to be poor substrates. We recently showed that, in vitro, Loqs-PD minimizes the termini dependence of Dcr-2 and facilitates cleavage of suboptimal substrates, including predicted endo-siRNA precursors (6). This suggests Loqs-PD evolved to expand the range of Dcr-2 endogenous substrates; however, the mechanism by which Loqs-PD modulates Dcr-2 substrate recognition and processing is unknown.

Loqs-PD is one of four protein isoforms encoded by the gene loqs. Loqs-PA and Loqs-PB, homologs of TRBP, interact with Dcr-1 during miRNA biogenesis (9, 17–20). Loqs-PC is rarely expressed and has no known function (9). Loqs-PD is the only Loqs isoform capable of facilitating Dcr-2–dependent endo-siRNA biogenesis (19, 20). It contains two dsRNA-binding motifs (dsRBMs) separated by a short linker, with the rest of the protein predicted to be largely unstructured. Only the C-terminal 22 amino acids are unique to the PD isoform (Fig. 1 A and C), and studies performed in S2 cells indicate they are important for endo-siRNA silencing (21) and for interactions with Dcr-2 (21, 22). However, studies monitoring the interaction between Loqs-PD and Dcr-2 by immunoprecipitation have noted varying degrees of association (9, 10, 23), and so far, studies with purified proteins have not been performed. During RISC assembly, Dcr-2 interacts with another dsRBP, R2D2, and it is unclear whether R2D2 and Loqs-PD compete for the same binding site (23) or bind to unique sites (22).

Fig. 1.

Design and purification of Loquacious-PD and its truncations. (A) Schematic of Loqs-PD and N- and C-terminal truncations. dsRBMs are shown as boxes, and the isoform-specific C terminus is colored red. (B) Coomassie-stained SDS/PAGE gel of purified Loqs-PD and truncations. Molecular mass markers were run in first and last lanes with sizes indicated (kDa). (C) Primary sequence of Loqs-PD. dsRBMs are underlined, and the 22 C-terminal, isoform-specific amino acids are colored in red.

Using purified components, we performed a series of biochemical experiments to investigate the mechanism by which Loqs-PD modulates Dcr-2 activity. We show that Loqs-PD directly interacts with Dcr-2 in an RNA-independent manner, and this interaction, as well as Loqs-PD binding to dsRNA, are both required for Loqs-PD function. We discovered the first dsRBM of Loqs-PD is uniquely required to enhance cleavage of suboptimal substrates but not an optimal substrate. Finally, we report an unrecognized Dicer–dsRBP interaction interface and describe its potential implication for the function of Loqs-PD.

Results

Purification of Loqs-PD Truncations.

We previously showed that Loqs-PD modifies Dcr-2 cleavage activity, but the mechanism by which Loqs-PD accomplishes this is unknown. To identify regions of Loqs-PD required to alter Dcr-2 activity, we designed and purified a series of N- and C-terminal truncations (Fig. 1 A and B). By precedent (24), each construct was named based on domains or features it contained. For example, the smallest construct, LR₂C, contained the linker region between dsRBMs (L), the second dsRBM (R₂), and the C-terminal tail (C), while the N-terminal region (N) and the first dsRBM (R₁) were deleted. NR₁LR₂C_Δ22 lacked the C-terminal 22 amino acids, which are the only amino acids unique to the PD isoform (Fig. 1C). While many dsRBPs form homodimers in solution (25–27), Loqs-PD and all of its truncations were found to be monomers by sedimentation equilibrium experiments (Fig. S1).

Fig. S1.

Loqs-PD and its truncations are monomeric. Representative analytical ultracentrifugation equilibrium data (red symbols) and the corresponding global single ideal species fits (solid black lines) are shown for each Loqs-PD truncation construct (n = 3, NR₁LR₂C; n = 1, NR₁LR₂C_Δ22, R₁LR₂C, and LR₂C). Depicted are the high and low concentrations of each sample (square and circular symbols, respectively) at the highest speed (18,144 × g) with 50% of data points shown for clarity. The floating molecular mass fit solution is indicated. Fit residuals are indicated below and contain all data points.

Loqs-PD Requires the C-Terminal 22 Residues to Fully Enhance Dcr-2 Cleavage.

To determine which domains of Loqs-PD were required to affect Dcr-2 cleavage activity in vitro, we performed single-turnover cleavage assays using dsRNAs with BLT or 3′ovr termini (created by annealing 106-nt sense and antisense RNAs), with Dcr-2^WT alone or supplemented with Loqs-PD or its truncations (Fig. 2 A and B, Table 1, and Fig. S2). As in our prior studies (6), Dcr-2 alone (-, dotted) cleaved 106 BLT dsRNA at a faster rate (k_obs, 0.12 ± 0.02 min⁻¹) than 106 3′ovr dsRNA (k_obs, 0.01 ± 0.02 min⁻¹), emphasizing that dsRNA with BLT termini is an optimal substrate compared with dsRNA with 3′ovr, or other non-BLT termini (suboptimal substrates). Inclusion of full-length Loqs-PD (NR₁LR₂C, black lines) dramatically increased the rates of cleavage for both BLT (k_obs, 1.98 ± 0.05 min⁻¹) and 3′ovr (k_obs, 0.73 ± 0.06 min⁻¹) 106 dsRNA, and a similar rate enhancement was observed when the N-terminal 135 residues were removed (R₁LR₂C, green lines). Conversely, removal of the C-terminal 22 amino acids from Loqs-PD (NR₁LR₂C_Δ22, red lines) severely compromised the ability of Loqs-PD to stimulate Dcr-2 cleavage activity, for both BLT and 3′ovr dsRNA. Thus, the N terminus of Loqs-PD is dispensable for Loqs-PD effects on Dcr-2 cleavage, while the C-terminal 22 amino acids are essential.

Fig. 2.

Loqs-PD truncations affect Dcr-2 cleavage rates in a substrate-dependent manner. Graphs plot single-turnover cleavage over time for 106 BLT (A) and 3′ovr (B) dsRNA (1 nM) with 30 nM Dcr-2 at 25 °C, in the absence or presence of an equimolar amount of Loqs-PD or its truncations. Portions of Top graphs are enlarged below. Data were fit to the pseudofirst order rate equation, y = y_O + A × (1 − e^−kt), where y is fraction cleaved [(cleaved)/(cleaved + uncleaved)], A is amplitude of rate curve (>0.5), y₀ is baseline (=0), k is pseudofirst order rate constant, and t is time. Data points are mean ± SEM (n = 3). (C) PhosphorImage shows single-turnover cleavage of ³²P-internally labeled esi-2^hp (1 nM) with Dcr-2 (30 nM), ±Loqs-PD or its truncations (120 nM), ±5 mM ATP. Cleavage products were separated by 12% denaturing PAGE, and 10-nt RNA ladder is on the left. *, RNA trapped in well, possibly due to disordered N terminus of Loqs-PD. (D, Top) Single-turnover cleavage of esi-2^hp, with 5 mM ATP and Loqs-PD or truncations, was quantified from data as in C (cleavage products divided by total radioactivity in lane) and plotted relative to cleavage with full-length Loqs-PD. Data points are mean ± SEM (n = 3). Paired t test—ns, P > 0.05; *P < 0.02; **P < 0.005. (D, Bottom) Predicted secondary structure of esi-2^hp colored according to mFold (58).

Table 1.

Summary of k_obs, t_1/2, and K_d values

	Cleavage of 106 BLT		Cleavage of 106 3′ovr		Binding of 106 BLT
Loqs-PD	kobs, min−1	t1/2, min	kobs, min−1	t1/2, min	Kd, nM	h
—	0.12 ± 0.02	5.98	0.01 ± 0.02	66.3	n/a	n/a
NR1LR2C	1.98 ± 0.05	0.35	0.73 ± 0.06	0.95	8.97 ± 0.66	1.1 ± 0.1
R1LR2C	1.73 ± 0.18	0.40	0.60 ± 0.07	1.16	1.36 ± 0.12	1.8 ± 0.2
NR1LR2CΔ22	0.21 ± 0.02	3.25	0.06 ± 0.07	12.0	9.34 ± 0.75	1.1 ± 0.1
LR2C	1.28 ± 0.07	0.54	0.19 ± 0.04	3.57	29.01 ± 1.3	3.8 ± 0.6
His-LR2C	1.53 ± 0.06	0.45	n.d.	n.d.	29.93 ± 1.2	3.3 ± 0.4
His-LR2CK,A	0.14 ± 0.01	4.81	n.d.	n.d.	1,631 ± 43	5.0 ± 0.5
His-LR2CKKK,EAA	0.12 ± 0.01	5.62	n.d.	n.d.	1,318 ± 69	3.5 ± 0.5
His-NR1LR2CFF,AA	n.d.	n.d.	n.d.	n.d.	9.47 ± 1.1	1.0 ± 0.1

Values shown are mean ± SEM (n = 3). n/a, not applicable; n.d., not determined.

Fig. S2.

Single-turnover Dcr-2 cleavage assays with Loqs-PD truncations. Single-turnover cleavage assay of ³²P-end–labeled 106 BLT (A) or 3′ovr dsRNA (B) (1 nM) with Dcr-2 (30 nM) in the presence or absence of Loqs-PD or its truncations (30 nM). Cleavage products were separated by 17% denaturing PAGE (n = 3); a representative PhosphorImage is shown.

Loqs-PD Requires both dsRBMs to Enhance Dcr-2 Cleavage of Suboptimal Substrates.

Unexpectedly, additional truncation to remove the first dsRBM (LR₂C, blue lines) revealed substrate-dependent effects. LR₂C increased the Dcr-2 cleavage rate for 106 BLT dsRNA to levels approaching that observed after addition of full-length Loqs-PD (k_obs, 1.28 ± 0.07 min⁻¹ vs. k_obs, 1.98 ± 0.05 min⁻¹). However, while LR₂C slightly increased the Dcr-2 cleavage rate for 106 3′ovr dsRNA, the rate was ∼fourfold slower than that observed in the presence of NR₁LR₂C (k_obs, 0.19 ± 0.04 min⁻¹ vs. k_obs, 0.73 ± 0.06 min⁻¹). To extend these results to a natural, endogenous substrate, we tested a dsRNA derived from esi-2, a Drosophila endo-siRNA precursor that gives rise to abundant endo-siRNAs in vivo (13–15). esi-2 contains 20 inverted repeats capable of forming multiple stem-loop, or hairpin, structures, and we used a substrate with a single inverted repeat flanked by noncomplementary sequences. This substrate, referred to as esi-2^hairpin (esi-2^hp) [previously referred to as pre-sl by Miyoshi et al. (22)], is predicted to form a single hairpin with single-stranded overhangs at each terminus (Fig. 2D, Bottom). While the endogenous termini of esi-2 have not been defined, we previously showed that, in vitro, esi-2^hp recapitulates the Loqs-PD–dependent siRNA production (6) observed in vivo (9, 10). We performed single-turnover cleavage assays of esi-2^hp with Dcr-2^WT alone, or supplemented with Loqs-PD or its truncations, in the presence (+) or absence (−) of ATP (Fig. 2 C and D). Dcr-2 alone was unable to appreciably cleave esi-2^hp, while addition of NR₁LR₂C resulted in cleavage and siRNA-sized cleavage products. As with the 3′ovr 106 dsRNA substrate, R₁LR₂C enhanced cleavage of esi-2^hp to the same extent as NR₁LR₂C, while NR₁LR₂C_Δ22 and LR₂C showed a significantly decreased ability to promote cleavage. All cleavage events were dependent on ATP (Fig. 2C). Thus, while esi-2^hp differs from 3′ovr 106 dsRNA in that its cleavage is completely dependent on Loqs-PD, it is a suboptimal substrate and, like 3′ovr dsRNA, requires both dsRBMs for cleavage.

In addition to the siRNA-sized cleavage products of esi-2^hp, larger products of ∼33 and ∼43 nts accumulated in a Loqs-PD–dependent manner (Fig. 2C). In vivo, esi-2 is processed into two adjacent endo-siRNAs, esi-2.1 and esi-2.2, leaving the ∼42 nt loop region as a byproduct (14) (Fig. S3A). To get information about the identity of bands observed in our in vitro cleavage assays, we performed Northern blots in which we probed for a region that encompasses esi-2.1 (nucleotides 30–60, red), the predominant endo-siRNA observed from esi-2 (13–15), the loop region (nucleotides 79–109, green), or the 3′ end of esi-2^hp (nucleotides 158–190, blue) (Fig. S3B). The “red” probe primarily detected siRNA-sized products, suggesting that siRNA-sized products in Fig. 2C include esi-2.1. The loop probe primarily detected ∼43 nt-sized products, suggesting the ∼43-nt band in Fig. 2C corresponds to the loop region of esi-2^hp. Finally, the 3′-end probe detected multiple bands, including likely intermediates, and an ∼33 nt product, suggesting the ∼33-nt band in Fig. 2C corresponds to the hairpin base. These data are consistent with cleavage of esi-2^hp to produce two siRNAs and byproducts that include the hairpin loop and base and agree with prior analyses of esi-2 processing in vivo (14).

Fig. S3.

Northern blot analysis of esi-2^hp. (A) Model for esi-2^hp cleavage. Locations of esi-2.1 and esi-2.2, the two most abundant endo-siRNAs derived from esi-2 (13, 14), are shown in magenta and cyan, respectively. Filled arrowheads mark predicted cleavage sites by Dcr-2. (B, Top) Northern blot of esi-2^hp (1 nM) following cleavage by Dcr-2 (30 nM) in the presence or absence of Loqs-PD (120 nM). Cleavage products were separated by 12% denaturing PAGE and 10-nt RNA ladder is shown on the left. The same blot was probed iteratively with the probes listed below each representative PhosphorImage (n = 3). (B, Bottom) Cartoon of esi-2^hp with position of probes used above.

Loqs-PD and Its Truncations Bind dsRNA with High Affinity.

To gain insight into the differential ability of the Loqs-PD truncations to enhance Dcr-2 cleavage activity, we measured their dsRNA binding affinity. We performed gel mobility shift assays with each Loqs-PD variant and 106 BLT dsRNA (Fig. 3). NR₁LR₂C bound dsRNA with high affinity, exhibiting a K_d of ∼9 nM (Table 1). We observed two faint bands of slower mobility, but the majority of bound or shifted dsRNA appeared as a diffuse smear (Fig. 3A, Top Left), suggesting a subset of complexes dissociate during electrophoresis (28). NR₁LR₂C_Δ22 bound dsRNA with the same affinity as NR₁LR₂C (K_d of ∼9 nM), and the pattern of shifted dsRNA was also similar (Fig. 3A, Top Right), indicating deletion of the 22 C-terminal amino acids does not compromise dsRNA binding. R₁LR₂C bound dsRNA with a slightly higher affinity (K_d of ∼1.4 nM) than NR₁LR₂C, suggesting the N-terminal region is inhibitory to dsRNA binding. Further, there was a dramatic change in the pattern of shifted dsRNA. For R₁LR₂C, we observed the sequential appearance of ∼6 distinct complexes (Fig. 3A, Bottom Left), consistent with a maximal protein to dsRNA stoichiometry of 6:1. In structures with dsRNA, dsRBMs bind ∼16 bp along one face of the dsRNA helix such that another dsRBM can bind opposite the first (29, 30). To accommodate six molecules of R₁LR₂C (∼32 bp) on a 106-bp substrate, we predict binding occurs along opposite faces of the dsRNA. LR₂C bound dsRNA with slightly lower affinity (K_d of ∼29 nM), consistent with the loss of one of the two dsRBMs. We observed one prominent shift along with a faint second shift. LR₂C also exhibits a Hill coefficient >1, suggesting some form of cooperativity may be operative. As summarized in Fig. 3B, removal of the C-terminal 22 amino acids does not alter binding affinity from that of the full-length protein, while removal of the N terminus increases affinity and deletion of both the N terminus and first dsRBM decreases affinity.

Fig. 3.

Loqs-PD and its truncations bind dsRNA with high affinity. (A) Representative PhosphorImages for gel shift experiments with 106 BLT dsRNA (10 pM), ³²P-end–labeled on the sense strand and incubated with indicated concentrations of Loqs-PD or its truncations. Free dsRNA was separated from bound dsRNA by native PAGE on a 4% 19:1 polyacrylamide gel. (B) Radioactivity in gels, as in A, was quantified to generate binding isotherms. dsRNA_total and dsRNA_free were quantified to determine fraction bound (1 − (dsRNA_free/dsRNA_total)), and data were fit using the Hill formalism, fraction bound = 1/(1 + (K_dⁿ/[P]ⁿ)), where K_d is the dissociation constant, n is the Hill coefficient, and [P] is the protein concentration. Data points are mean ± SEM (n = 3 unless marked otherwise; *n = 2; ^#n = 1).

The C-Terminal 22 Residues of Loqs-PD Are Necessary for Interaction with Dcr-2.

Previous studies suggest the C-terminal 22 residues of Loqs-PD are required to interact with Dcr-2 (21, 22). However, the interaction has not been monitored with purified proteins. Whether Loqs-PD and Dcr-2 interact in the absence of RNA also is untested. To address these questions, we used purified proteins in pull-down experiments with His-tagged Loqs-PD variants and untagged Dcr-2. To facilitate formation of a stable complex, we used a Dcr-2 variant in which both RNaseIII and helicase activity were disrupted by point mutations (Dcr-2^RIII,K34A). We found that NR₁LR₂C was able to pull down Dcr-2 (Fig. 4A, lane 7), confirming a direct interaction. Additionally, our recombinant proteins were free of RNA as measured by A_260/280, suggesting this interaction is RNA-independent. By contrast, NR₁LR₂C_Δ22 was unable to pull down Dcr-2 (Fig. 4A, lane 8), confirming that the C terminus of Loqs-PD is required for a direct, RNA-independent interaction with Dcr-2. R₁LR₂C and LR₂C were both able to pull down Dcr-2 (Fig. 4A, lanes 9 and 10), although R₁LR₂C pulled down slightly less Dcr-2 that NR₁LR₂C or LR₂C (Fig. 4B). Thus, Loqs-PD directly binds Dcr-2, and its C-terminal 22 amino acids are required for this interaction.

Fig. 4.

C-terminal 22 amino acids of Loqs-PD mediate direct interaction with Dcr-2. (A) Coomassie-stained SDS/PAGE gels show input (Left, 5% of total) and pull-down (Right, 100%) of Dcr-2^RIII,K34A (D, 2 µM) in the absence (−) or presence of His-tagged Loqs-PD or truncations (L, 4 µM). Molecular mass markers (kDa) are on Left. (B) Data as in A were quantified to determine amount of Dcr-2 pulled down by each Loqs-PD variant. Values were normalized to Dcr-2 in the input and plotted relative to amount pulled down by His–Loqs-PD (His-NR₁LR₂C). Data points are mean ± SEM (n = 3); paired t test—ns, P > 0.05; **P < 0.003. (C) Pull-down performed as in A with increasing amounts of PD22 peptide (P, 10, 20, 40 µM) as a binding competitor. (D) Data were quantified as in B for pull-downs as in C. Data points are mean ± SEM (n = 3). Paired t test—ns, P > 0.05; *P < 0.04.

To determine whether the C-terminal 22 residues of Loqs-PD alone were able to bind Dcr-2, we synthesized the 22-residue peptide (PD22) and performed a competition experiment in which we pulled down Dcr-2 with LR₂C in the presence of increasing amounts of PD22 (Fig. 4 C and D). PD22 effectively competed for binding with LR₂C as seen by the dose-dependent decrease in the amount of Dcr-2 pulled down by LR₂C (Fig. 4C, lanes 9–11 compared with lane 8). To control for nonspecific effects from the high concentration of peptide used, a mutated version of PD22 (PD22^mut, described below) was tested at the highest concentration of peptide assayed. PD22^mut did not compete for the interaction between LR₂C and Dcr-2 (Fig. 4C, lane 12), confirming the specificity of the PD22 interaction. We attempted direct binding studies by fluorescence polarization using a fluorescein-labeled version of PD22 but were unable to saturate binding without using prohibitively high concentrations of Dcr-2. Without quantitative binding studies, we cannot rule out that other portions of Loqs-PD contribute to binding, but our analyses indicate the C-terminal 22 amino acids of Loqs-PD directly interact with Dcr-2.

dsRNA Binding Is Required by Loqs-PD to Affect Dcr-2 Cleavage Activity.

Our cleavage assays with NR₁LR₂C_Δ22 (Fig. 2) confirmed that the C-terminal 22 residues of Loqs-PD were essential for enhancing Dcr-2 cleavage activity. Our pull-downs (Fig. 4) provided an explanation in that those residues were required for interaction with Dcr-2. An outstanding question was whether dsRNA-binding by Loqs-PD was also required to enhance Dcr-2 cleavage activity. To test this, we disrupted dsRNA-binding activity of Loqs-PD in a construct capable of enhancing Dcr-2 activity. We selected LR₂C because it contained a single dsRBM yet was able to enhance Dcr-2 activity toward a 106 BLT dsRNA to a similar extent as full-length Loqs-PD (Fig. 2A). dsRBMs contain a highly conserved KKxxK motif, which mediates direct interaction with the phosphate backbone of dsRNA (29–31). To disrupt the dsRNA-binding activity of LR₂C, we mutated lysine 301, present in the KKxxK motif of dsRBM2 (Fig. 5 A and B). We performed gel shift assays of 106 BLT dsRNA with His-LR₂C or His-LR₂C^K,A (Fig. 5C). Indeed, mutation of lysine 301 to alanine resulted in a ∼55-fold reduction in binding affinity by His-LR₂C^K,A compared with His-LR₂C (Fig. 5D and Table 1). The presence of the 6xHis tag had no effect on dsRNA binding (K_d ∼ 29 nM vs. K_d ∼ 30 nM, respectively) (Table 1). To ensure that the decrease in dsRNA-binding affinity was due to the mutation and not a secondary affect of protein misfolding, we compared His-LR₂C and His-LR₂C^K,A by circular dichroism (CD) spectroscopy (Fig. S4A). CD spectra reflect the secondary structure composition of a protein (32), and there was no significant difference between the His-LR₂C and His-LR₂C^K,A spectra, suggesting the K301A mutation did not grossly affect protein folding. Thus, LR₂C^K,A was properly folded but had greatly reduced affinity for dsRNA.

Fig. 5.

dsRNA binding is required for LR₂C to affect Dcr-2 cleavage of an optimal substrate. (A) Schematic of LR₂C construct and location of mutation to disrupt dsRNA binding. (B) Coomassie-stained SDS/PAGE gel of purified His-LR₂C and His-LR₂C^K,A with molecular mass markers on the left. (C) PhosphorImage of gel shift assay for 106 BLT dsRNA (10 pM), ³²P-end–labeled on the sense strand and incubated with indicated concentrations of His-LR₂C or His-LR₂C^K,A. Free dsRNA was separated from bound dsRNA by native PAGE on a 4% 19:1 polyacrylamide gel. *, trace amounts of ssRNA present with high concentrations of His-LR₂C^K,A. (D) Radioactivity in gels, as in C, was quantified to generate binding isotherms as in Fig. 3B. Data points are mean ± SEM (n = 3). (E) Coomassie-stained SDS/PAGE gels show input (Left, 5% of total) and pull-down (Right, 100%) using Dcr-2^RIII,K34A (2 µM) with His-tagged Loqs-PD constructs (4 µM). Molecular mass markers are to the left. (F) Quantification as in Fig. 4B was performed for data as in E and plotted relative to LR₂C. Data points are mean ± SEM (n = 3). Paired t test—ns, P > 0.05. (G) Graphs as in Fig. 2A show single-turnover cleavage over time for 106 BLT dsRNA (1 nM) with Dcr-2 (30 nM), in the absence or presence, of an equimolar amount of wild-type or mutant His-LR₂C (30 nM). Data points are mean ± SEM (n = 3).

Fig. S4.

A second dsRBM mutant, His-LR₂C^KKK,EAA, phenocopies His-LR₂C^K,A. (A) CD spectra of His-LR₂C, His-LR₂C^K,A, and His-LR₂C^KKK,EAA. (B) Schematic of LR₂C construct and position of mutations to disrupt dsRNA binding. (C) Coomassie-stained SDS/PAGE gel of purified His-LR₂C, His-LR₂C^K,A, and His-LR₂C^KKK,EAA. (D) Gel shift assays of 106 BLT dsRNA (10 pM), ³²P-end–labeled on the sense strand and incubated with indicated concentrations of His-LR₂C, His-LR₂C^K,A, or His-LR₂C^KKK,EAA. Free dsRNA was separated from bound dsRNA by native PAGE on a 4% 19:1 polyacrylamide gel. Representative PhosphorImages are shown for each Loqs-PD construct. *, trace amounts of ssRNA present with high concentrations of His-LR₂C^K,A and His-LR₂C^KKK,EAA. (E) Radioactivity in gels, as in D, was quantified to generate binding isotherms. Binding isotherms were fit using the Hill formalism. Data points are mean ± SEM (n = 3). (F) Pull-downs of Dcr-2^RIII,K34A (2 µM) with His-tagged Loqs-PD constructs (4 µM). Representative Coomassie-stained SDS/PAGE gels of input (Left, 5% of total) and pull-down (Right, 100%) are shown (n = 3). (G) Quantification of pull-downs as in F. Data points are mean ± SEM (n = 3). Paired t test—n.s., P > 0.05. (H) Quantification of single-turnover cleavage assays for 106 BLT dsRNA (1 nM) with Dcr-2 (30 nM), in the absence, or presence, of an equimolar amount of wild-type or mutant His-LR₂C (30 nM). Data points are mean ± SEM (n = 3).

We performed pull-downs with His-LR₂C^K,A and found that the dsRBM mutation had no effect on Dcr-2 binding (Fig. 5 E and F). This result emphasized that the interaction between Dcr-2 and Loqs-PD is independent of dsRNA. After determining that His-LR₂C^K,A had greatly reduced dsRNA-binding affinity but was still capable of interacting with Dcr-2, we tested whether His-LR₂C^K,A could affect Dcr-2 cleavage activity. We performed single-turnover cleavage assays of 106 BLT dsRNA by Dcr-2^WT alone (−) or supplemented with His-LR₂C or His-LR₂C^K,A (Fig. 5G). His-LR₂C^K,A was unable to increase the rate of Dcr-2 cleavage (Table 1), indicating dsRNA binding is required under the conditions tested. In addition to the single point mutant, we made a more severe mutant in which all three lysines of the KKxxK motif were mutated to EAxxA (Fig. S4 B and C). We obtained similar results for K,A and KKK,EAA mutants in all of the above experiments (Table 1 and Fig. S4). Thus, Loqs-PD must bind dsRNA as well as Dcr-2 to enhance Dcr-2 cleavage activity.

Loqs-PD Binds the Hel2 Subdomain of Dcr-2’s Helicase.

Previous studies indicate Loqs-PD interacts with the helicase domain of Dcr-2 (21), but the exact binding interface is unknown. The helicase domain of Dcr-2 contains two RecA-like domains (Hel1 and Hel2) separated by a Hef-like insertion domain (Hel2i). To identify the region of Dcr-2 that binds Loqs-PD, we coupled protein cross-linking with mass spectrometry (XL–MS), in which hybrid peptides, resulting from intra- or interprotein cross-links, are identified and sequenced by liquid chromatography and tandem MS (LC–MS/MS) (33–35). We performed chemical cross-linking with disuccinimidyl suberate (DSS), a homo-bifunctional NHS-ester cross-linker that primarily reacts with primary amines of lysine side chains or the N terminus (36).

When treated with DSS, Dcr-2 migrated slightly slower during SDS/PAGE (D vs. D+xl) (Fig. 6A, compare lanes 1 and 6). Loqs-PD and its truncations migrated slightly faster after DSS treatment, with broader, more diffuse bands (L vs. L+xl; Fig. 6A, compare lanes 2–5 and 7–10). In both cases, the altered SDS/PAGE mobility is likely due to intraprotein cross-linking. When Dcr-2 was incubated with NR₁LR₂C and treated with DSS, the main Dcr-2 band (D+xl) shifted to a higher molecular mass species, suggesting formation of a covalent adduct between Loqs-PD and Dcr-2 (D+L+xl) (Fig. 6A, compare lanes 6 and 11). Consistent with the requirement of the C-terminal 22 amino acids for interacting with Dcr-2 in pull-down assays (Fig. 4 A and B), the D+L+xl species was greatly reduced when cross-linking was performed with NR₁LR₂C_Δ22 (Fig. 6A, compare lanes 11 and 12). Cross-linking performed with Dcr-2 and R₁LR₂C or LR₂C also resulted in the D+L+xl species (Fig. 6A, lanes 13 and 14). The agreement between our cross-linking and pull-downs suggests DSS cross-linking captures the native interaction between Loqs-PD and Dcr-2.

Fig. 6.

Loqs-PD interacts with the Hel2 domain of Dcr-2 via an FDF-like motif. (A) DSS cross-linking of OSF–Dcr-2^RIII (2 µM), in the presence or absence, of Loqs-PD or its truncations (4 µM). Representative Coomassie-stained SDS/PAGE gel of individual proteins without (lanes 1–5, D and L) or with DSS (lanes 6–10, D+xl and L+xl) or together with DSS (lanes 11–14, D+L+xl) (n > 3). Brackets mark bands excised for subsequent XL–MS analysis. (B) Schematic [xVis (59)] of Dcr-2 and Loqs-PD with color-coded domains depicting intra- and interprotein cross-links identified by MS/MS analysis using ProteinProspector2 (57) in D+L+xl species (n = 2). (C) Homology model of Dcr-2 helicase domain with Hel1 (cyan), Hel2i (orange), and Hel2 (magenta). The Dcr-2–derived peptide that cross-linked to Loqs-PD is shown (blue), with the site of DSS cross-linking, lysine 501, in stick representation. (D, Top Left) Complex of DDX6-C (gray) and EDC3-FDF (red). FDF motif is shown in stick representation (PDB ID code 2WAX). (D, Top Middle) Hel2 from Dcr-2 homology model, colored as in C. (D, Top Right) Structural superposition of DDX6-C:EDC3-FDF complex and homology model of Dcr-2 Hel2 domain. (D, Bottom) Sequences of Homo sapiens EDC3 and C-terminal 22 residues from Loqs-PD. The FDF-motif in EDC3 and the putative FDF-motif in Loqs-PD are shaded in red. (E) Coomassie-stained SDS/PAGE gels of input (Left, 5% of total) and pull-down (Right, 100%) for Dcr-2^RIII,K34A (2 µM) with His-tagged Loqs-PD constructs (4 µM). Molecular mass markers are to the left. (F) Quantification as in Fig. 4B for data as in E. Data points are mean ± SEM (n = 3). Paired t test—ns, P > 0.05; ***P < 0.001.

To identify the sites of cross-linking between Dcr-2 and Loqs-PD, we analyzed the in-gel tryptic digest of the D+L+xl species by LC–MS/MS. We identified peptides mapping to both Dcr-2 and Loqs-PD (Table S1), suggesting the second shift we observed by SDS/PAGE was indeed due to Loqs-PD cross-linking to Dcr-2. We identified 18 Dcr-2–Dcr-2 cross-links and one Dcr-2–Loqs-PD cross-link from two replicates (Fig. 6B and Table S2). The sole Loqs-PD–Dcr-2 cross-link and 11/18 Dcr-2–Dcr-2 cross-links were identified in both replicates. We predict all identified Dcr-2–Dcr-2 cross-linked peptides reflect intraprotein rather than interprotein cross-linking because the difference in SDS/PAGE mobility between untreated (D) and treated (D+xl and D+L+xl) samples was very slight. In samples treated with DSS, some protein remained trapped in the wells and may correspond to Dcr-2–Dcr-2 interprotein cross-links that were too large to enter the gel. The sole interprotein cross-link was between the penultimate residue of Loqs-PD, K358, and K501 in Dcr-2, which is located in the Hel2 subdomain of the helicase domain. As a control, we analyzed the in-gel tryptic digest of the D+xl species by LC–MS/MS and identified 16 Dcr-2–Dcr-2 cross-links, and no cross-linked peptides corresponding to Loqs-PD, in two replicates. Nine of 16 Dcr-2–Dcr-2 cross-links were identified in both replicates. Given that the Loqs-PD–Dcr-2 cross-link occurs in the C-terminal 22 residues of Loqs-PD, which are required for interaction with Dcr-2, we predict the reciprocal site of cross-linking in Dcr-2 correctly identifies the interaction surface, Hel2.

Table S1.

Peptide coverage of Loqs-PD and Dcr-2

	Replicate 1		Replicate 2
Sample:	D+xl	D+L+xl	D+xl	D+L+xl
Dcr-2	574 (89.7)	536 (91.2)	505 (87.6)	678 (90)
Loqs-PD	12 (49.1)*	77 (75.1)	0 (0)	106 (76.9)

Shown is the number of unique peptides (% coverage) identified from Dcr-2 and Loqs-PD in the D+xl and D+L+xl gel bands from two independent experiments.

^*Possible contamination from excising gel bands with the same razor blade.

Table S2.

Loqs-PD–Dcr-2 cross-linked peptides

m/z	z	Error, ppm	Peptide 1, Loqs-PD	Peptide 2, Dicer-2	Xlink Res, Loqs-PD	Xlink Res, Dcr-2
840.4472	+3	−2.0	DFEFIK*I	TTEAK*FVLFTADK	358	501
935.4938	+3	−3.2	DFEFIK*I	TTEAK*FVLFTADKER	358	501
701.8723	+4	−3.0	DFEFIK*I	TTEAK*FVLFTADKER	358	501

Shown are cross-linked peptides corresponding to the sole interprotein cross-link identified in Fig. 6 (n = 2). Representative m/z and error values from a single experiment are shown. K*, site of DSS cross-linking.

Loqs-PD Interacts with Dcr-2 Through an FDF-Like Motif.

DSS contains an eight-carbon linker (11.4 Å) between reactive NHS-ester moieties. When cross-linking occurs between lysine side chains, the alpha carbons of each lysine should be within ∼24 Å [Lys1(6.4 Å)–DSS(11.4 Å)–Lys2(6.4 Å)] (37). To verify the specificity of cross-linking, we determined how many of the identified cross-links met this distance constraint. Since the only identified interprotein cross-link to Loqs-PD was in the helicase domain of Dcr-2, we focused analyses on the helicase domain. There are no high-resolution structures available for a Dicer helicase domain, but there are structures available for related helicases from the RIG-I–like and DEAD-box families. We generated a homology model of Dcr-2’s helicase domain using Robetta (38) (robetta.bakerlab.org) (Fig. 6C). From our combined XL–MS data, we identified seven intraprotein cross-links within the helicase domain. Using our homology model, we measured the distance between alpha carbons of cross-linked residues and found that 7/7 were within 24 Å (Fig. S5A); this suggests our homology model is accurate and that we are detecting structurally plausible cross-links.

Fig. S5.

Intra- and interprotein cross-links between Dcr-2 and Loqs-PD. (A) Intraprotein cross-links within the helicase domain of Dcr-2 mapped on the homology model of Dcr-2’s helicase domain. Distances are measured from the alpha carbon of each reactive lysine and measure in angstroms (Å). (B) Structural superposition of the DDX6-C:EDC3-FDF complex (PDB ID code 2WAX) and a homology model of Hel2 from the Dcr-2 helicase, with DDX6-C omitted for clarity. The Dcr-2–derived peptide that cross-linked to Loqs-PD is shown in blue, with the site of DSS cross-linking, lysine 501, shown in stick representation. The phenylalanines of the FDF motif in EDC3-FDF are shown in stick representation. The reactive lysine in Loqs-PD, K358, is two residues away from the FDF-like motif. Since the orientation of Loqs-PD is unknown, the distance from the alpha carbon of Dcr-2-K501 to the alpha carbon of the amino acid either two residues before or after the FDF motif of EDC3 is shown (Å). The two measurements account for two possible orientations of Loqs-PD.

We were unable to find documented examples of protein–protein interactions mediated by Hel2 for other RIG-I–like helicases, so we expanded our search to the closely related DEAD-box helicase family (39). DDX6, a DEAD-box helicase involved in mRNA decapping and degradation, interacts with several proteins via its second RecA motif (DDX6-C) (40), which is analogous to Hel2 in Dcr-2. DDX6-C–interacting proteins, such as EDC3, typically bind DDX6 through short peptide interactions involving a Phe–Asp–Phe sequence, known as an FDF motif (40–42) (Fig. 6D, Bottom). In a crystal structure of DDX6-C and a peptide derived from EDC3 (EDC3-FDF), the two phenylalanines of the FDF motif in EDC3 pack into a hydrophobic pocket on the surface of DDX6-C, distal from the RNA- and ATP-binding sites of DDX6 (43) (Fig. 6D, Left). Interestingly, Loqs-PD contains a putative FDF-like motif in its C-terminal 22 amino acids (Fig. 6D, Bottom), raising the possibility that Loqs-PD binds Dcr-2 in a manner analogous to EDC3-FDF and DDX6-C.

We structurally aligned Hel2 of our Dcr-2 helicase model with DDX6-C and noticed that the peptide we identified as cross-linking to Loqs-PD (shown in blue), while not superimposable, was in close proximity to the EDC3-FDF binding site in DDX6-C (Fig. 6D, Right). To determine if the FDF-like motif of Loqs-PD could bind Dcr-2 similarly to the DDX6-C–EDC3-FDF interaction, we measured the distance between K501 in Dcr-2 and residues in EDC3-FDF that would correspond to K358 of Loqs-PD in either orientation. Both distances were <24 Å, consistent with the distance constraints of DSS cross-linking (Fig. S5B). Beyond the FDF-like motif, Loqs-PD and EDC3 have low sequence similarity, and thus, we cannot confidently predict other interactions, or the orientation of Loqs-PD, from the DDX6-C–EDC3-FDF crystal structure.

While additional studies are required to elucidate the detailed interface, our XL–MS data and comparative modeling suggest Loqs-PD interacts with Hel2 of Dcr-2 using a putative FDF-like motif. To directly test this hypothesis, we mutated both phenylalanines in the FDF-like motif of full-length Loqs-PD to alanines (Fig. S6 A and B). We performed pull-downs of Dcr-2 with His-tagged Loqs-PD constructs, including the FDF-like motif mutant NR₁LR₂C^FF,AA (Fig. 6 E and F). Mutation of the FDF-like motif reduced the interaction between Dcr-2 and NR₁LR₂C^FF,AA (compare lanes 6 and 7) to the same extent as deleting the C-terminal 22 residues (compare lanes 7 and 8). This suggests the FDF-like motif in the C-terminal tail of Loqs-PD is required for interaction with Dcr-2. We performed gel shift assays with NR₁LR₂C^FF,AA and saw no difference in dsRNA-binding affinity compared with NR₁LR₂C, suggesting that protein folding has not been grossly perturbed (Table 1 and Fig. S6 C and D). Finally, we compared the ability of the FDF-like motif mutant (NR₁LR₂C^FF,AA) to promote cleavage of esi-2^hp to that of either wild-type Loqs-PD (NR₁LR₂C) or the Loqs-PD construct lacking the C-terminal 22 amino acids (NR₁LR₂C_Δ22) (Fig. S6 E and F). Mutation of the FDF-like motif completely abolished the ability of Loqs-PD to enhance cleavage of esi-2^hp and is comparable to entirely removing the C-terminal 22 amino acids. Thus, the isoform-specific C terminus of Loqs-PD contains an FDF-like motif that is required for direct interaction with Dcr-2 and to promote cleavage of a suboptimal substrate.

Fig. S6.

Loqs-PD interacts with Dcr-2 through an FDF-like motif. (A) Schematic of NR₁LR₂C^FF,AA construct and positions of mutations in the FDF-like motif. (B) Coomassie-stained SDS/PAGE gel of purified His-NR₁LR₂C and His- NR₁LR₂C^FF,AA. (C) Gel shift assays of 106 BLT dsRNA (10 pM), ³²P-end–labeled on the sense strand and incubated with indicated concentrations of NR₁LR₂C or His-NR₁LR₂C^FF,AA. Free dsRNA was separated from bound dsRNA by native PAGE on a 4% 19:1 polyacrylamide gel. Representative PhosphorImages are shown for each Loqs-PD construct. (D) Radioactivity in gels, as in C, was quantified to generate binding isotherms. Binding isotherms were fit using the Hill formalism. Data points are mean ± SEM (n = 3). (E) PhosphorImage shows single-turnover cleavage of ³²P-internally labeled esi-2^hp (1 nM) with Dcr-2 (30 nM), ±His-Loqs-PD variants (120 nM) in the presence of 5 mM ATP. Cleavage products were separated by 12% denaturing PAGE, and 10-nt RNA ladder is on the left. (F) Single-turnover cleavage of esi-2^hp, with 5 mM ATP and His-Loqs-PD variants, was quantified from data as in E (cleavage products divided by total radioactivity in lane) and plotted relative to cleavage with wild-type, full-length Loqs-PD. Data points are mean ± SEM (n = 3). Paired t test— ns, P > 0.05; **P < 0.002; ***P < 0.001.

Discussion

Since its discovery, Loqs-PD has been implicated in endo-siRNA biogenesis, but a detailed mechanistic understanding of its function is lacking. Based on our prior study (6) and the biochemical experiments presented here, we propose an integrated model for Loqs-PD function in Dcr-2–dependent siRNA biogenesis (Fig. 7).

Fig. 7.

Model of substrate-dependent requirements for Loqs-PD to affect Dcr-2 cleavage. (A) Loqs-PD interacts with Dcr-2 in the absence of nucleotide or dsRNA via the FDF-like motif at the C terminus of Loqs-PD and Hel2 of the Dcr-2 helicase domain. Subdomains of each protein are individually colored and labeled accordingly. (B) Interaction with Hel2 positions Loqs-PD at the base of Dcr-2’s C-shaped helicase domain. (C) Model for termini-dependent Loqs-PD variants dictated by intrinsic termini preference of Dcr-2 in the presence of ATP.

In our model, Loqs-PD and Dcr-2 directly interact in the absence of dsRNA (Fig. 7A), consistent with our pull-down and cross-linking data (Figs. 4 A and B, 5 E and F, and 6A). Based on our XL–MS data and subsequent mutational analysis (Fig. 6), the interaction is mediated by an FDF-like motif in the isoform-specific C terminus of Loqs-PD and Hel2 of the Dcr-2 helicase domain. Results from our cleavage assays with NR₁LR₂C_Δ22 (Fig. 2) confirm the necessity of the C-terminal 22 residues for Loqs-PD to fully enhance Dcr-2 cleavage toward both optimal and suboptimal substrates. We predict the defect results solely from an inability of NR₁LR₂C_Δ22 to interact with Dcr-2 (Fig. 4 A and B) as NR₁LR₂C_Δ22 had unaltered dsRNA binding compared with NR₁LR₂C (Fig. 3). Based on these data, in Fig. 7 we depict the C-terminal 22 amino acids of Loqs-PD, including the FDF-like motif, interacting with Hel2 of Dcr-2 in all conditions and propose it is necessary for Loqs-PD function. Based on our homology model and structures of RNA-bound RIG-I–like helicases (44–48), we predict Loqs-PD is positioned such that it would pull a dsRNA into the C-shaped helicase domain of Dcr-2 (Fig. 7B) and that this positioning is critical to its function (see below).

In Fig. 7C, we depict the dsRNA–Dcr-2 complex in different conformations for optimal vs. suboptimal substrates, reflecting the intrinsic termini preference of Dcr-2 (6). In the presence of ATP and an optimal substrate (e.g., BLT dsRNA), Dcr-2 is proposed to undergo a conformational change in which the helicase domain clamps onto the dsRNA (closed conformation) to hold it along the body of the enzyme, which in turn promotes processive cleavage (Fig. 7C, Top Left). For a suboptimal substrate (e.g., 3′ovr or esi-2^hp) in the presence of ATP, Dcr-2 is proposed to exist predominantly in an open conformation with the dsRNA positioned outside of the helicase domain, thus favoring distributive cleavage (Fig. 7C, Top Right). Based on the experiments reported here, we predict the substrate-dependent conformations of the dsRNA–Dcr-2 complex dictate which domains of Loqs-PD are required to enhance cleavage.

For an optimal substrate, either R₁LR₂C or LR₂C is sufficient to enhance Dcr-2 cleavage (Figs. 2A and 7C, Left). In Fig. 7C, Bottom, R₁LR₂C/LR₂C is shown interacting with Dcr-2 and the dsRNA substrate, consistent with our findings that dsRNA binding (Fig. 5) and Dcr-2 binding (Figs. 2 and 4 A and B) are both necessary for Loqs-PD to affect Dcr-2 activity. For an optimal substrate, we predict the dsRNA–Dcr-2 complex is in the closed conformation, which based on crystal structures of RNA-bound RIG-I–like helicases (44–48) would position the dsRNA and Hel2 in close proximity such that LR₂C is sufficient to simultaneously bind both. Thus, we depict LR₂C (highlighted as solid lines in Fig. 7C, Bottom) holding the dsRNA in the correct orientation relative to the helicase domain to stabilize the closed conformation of the dsRNA–Dcr-2 complex.

For suboptimal substrates, in contrast, R₁LR₂C is the only variant sufficient to fully enhance Dcr-2 cleavage (Figs. 2 B–D and 7C, Right). For a suboptimal substrate, we predict the dsRNA–Dcr-2 complex is predominantly in an open conformation, which based on low-resolution cryo-EM reconstructions of human Dicer bound to an siRNA (49) may position the dsRNA and Hel2 farther apart such that LR₂C is no longer sufficient to simultaneously bind both (Fig. 7C, Top Right). This model is consistent with our findings that LR₂C interacted with Dcr-2 comparable to full-length Loqs-PD (Fig. 4 A and B) and bound dsRNA (Fig. 3) yet was not sufficient to fully enhance cleavage of suboptimal substrates (Fig. 2 B–D). Given that LR₂C leads to a partial increase in cleavage of suboptimal substrates compared with NR₁LR₂C, we hypothesize that Dcr-2 occasionally transitions into the closed conformation (Fig. 7C, dashed arrow) such that LR₂C can stabilize it to promote cleavage. In our model, inclusion of the first dsRBM extends the reach of R₁LR₂C, allowing it to now simultaneously bind both Dcr-2 and the dsRNA substrate. We predict this allows R₁LR₂C to reposition the substrate within the helicase domain of Dcr-2 such that it can now adopt the closed conformation and be stabilized by LR₂C (Fig. 7C, Bottom). Additional studies are needed to fully elucidate the structures and dynamics of the different conformations discussed.

As a general mechanism, we propose Loqs-PD coordinates Dcr-2 binding with dsRNA binding to promote or stabilize a conformational change in the helicase domain of Dcr-2, which correlates with increased cleavage. Consistent with this model, Loqs-PD has no effect in the absence of ATP (6) (Fig. 2C), which we predict is required for the conformational change in Dcr-2. In vivo, under ATP-replete conditions, we expect Loqs-PD directly facilitates endo-siRNA biogenesis by this mechanism, although we cannot rule out the possibility that other factors may further enhance the efficiency.

Many of the annotated Dicer–dsRBP interactions require the helicase domain of Dicer (50, 51), but the exact interface is unknown, with the exception of human TRBP and Dicer. Biochemical and structural studies indicate the third dsRBM of TRBP interacts with Hel2i of Dicer (52, 53). A recent study suggests this interaction is conserved in the fly homologs Dcr-1 and Loqs-PB (20). In contrast, our data indicate that Loqs-PD primarily interacts with Hel2 of Dcr-2’s helicase (Fig. 6), identifying an additional Dicer–dsRBP interaction interface. There are conflicting reports as to whether Loqs-PD and R2D2 simultaneously interact with Dcr-2 (22) or whether their binding is mutually exclusive (21, 23). Additional studies are required to determine whether R2D2 binds the same Hel2 interface we have described for Loqs-PD or interacts with Hel2i of Dcr-2 in a manner analogous to TRBP and Dicer. It remains to be seen if other ATP-dependent Dicers such as Schizosaccharomyces pombe Dcr1 and Caenorhabditis elegans DCR-1 interact with dsRBPs similarly to Dcr-2 and Loqs-PD. Protein–protein interactions mediated by small motifs located in disordered regions have become a dominant theme among RNP assemblies (40, 54).

Materials and Methods

Protein Expression and Purification.

Loqs-PD and Dcr-2 were purified from Escherichia coli and Sf9 cells, respectively, as described (6) (SI Materials and Methods).

Synthesis of PD22 Peptide.

PD22 and PD22^mut peptides were chemically synthesized as described (55) (SI Materials and Methods). Peptide sequences were as follows: PD22: VSIIQDIDRYEQVSKDFEFIKI; PD22^mut: VSIIQDIDRYEQVSKDAEAIKI.

In Vitro Transcription of RNA Substrates.

We prepared 106 dsRNA as described (6). In the plasmid, each RNA strand was flanked by a hammerhead (5′ side) and HDV (3′ side) ribozyme to ensure accurate termini. ³²P-end–labeled 106 sense RNA was annealed with 106 BLT or 3′ovr antisense RNA to generate 106 BLT and 106 3′ovr dsRNA, respectively. esi-2^hp was cloned into the same ribozyme plasmid and prepared as described for 106 dsRNA with minor changes (SI Materials and Methods for details). Sequences of 106 dsRNAs and esi-2^hp are in SI Materials and Methods.

Gel Shift and Cleavage Assays.

Gel shift and single-turnover cleavage assays were performed as described (6) with minor changes (SI Materials and Methods for details).

Pull-Down Assays.

Dcr-2^RIII,K34A (2 µM) and His-Loqs-PD (4 µM) were incubated together in pull-down buffer (25 mM Tris, pH 8, 175 mM KCl, 10 mM MgCl₂, 10 mM imidazole, 1 mM TCEP, 5% glycerol, 0.1% nonidet P-40) for 1 h at 4 °C and added to prewashed His-Select Resin (Sigma-Aldrich) for 2 h at 4 °C. Resin and bound proteins were pelleted by centrifugation, and unbound protein (supernatant) was removed. Resin was washed with chilled pull-down buffer, and bound protein was eluted in pull-down buffer containing 300 mM imidazole. Proteins were resolved on a 4–15% gradient gel and stained with Coomassie Brilliant Blue. Competition pull-down assays were performed as described above with addition of PD22 (10, 20, and 40 µM) or PD22^mut (40 µM). Bound proteins were resolved on a 4–20% gradient gel by SDS/PAGE.

Chemical Cross-Linking.

Dcr-2^RIII,K34A or OSF-Dcr-2^RIII, and Loqs-PD and its truncations, were dialyzed into cross-linking buffer (20 mM Hepes, pH 7.8, 100 mM KCl, 10 mM MgCl₂, 1 mM TCEP, 5% glycerol). Cross-linking reactions were assembled with Dcr-2 (2 µM) and/or Loqs-PD (4 µM) and incubated (25 °C, 30 min). DSS (5 mM in DMSO; Sigma-Aldrich) was added to make 100–400 µM final, and cross-linking was quenched after an additional 30 min at 25 °C with 30 mM Tris, pH 8. Cross-linked proteins were resolved by SDS/PAGE (4–15%) and detected with Coomassie Brilliant Blue.

MS and Identification of Cross-Linked Peptides.

Bands corresponding to D+L+xl and D+xl were excised and subjected to in-gel digestion by trypsin and Lys-C. Peptides were extracted, reduced, treated with iodoacetamide, and analyzed using a nano-LC–MS/MS system equipped with a nano-HPLC pump (2D-ultra; Eksigent) and a maXis II ETD mass spectrometer (Bruker Daltonics). The maXis II ETD mass spectrometer was equipped with a captive spray ion source.

Cross-linked peptides were identified using the webserver version of ProteinProspector2 (v5.18.0/1) (prospector.ucsf.edu/prospector/mshome.htm). A custom database was made containing amino acid sequences of Dcr-2^RIII,K34A or OSF-Dcr-2^RIII, His-Loqs-PD, and 20 decoy proteins. Loqs-PD and Dcr-2 sequences were each randomized 10 times using Decoy Database Builder (56) to generate decoy targets. Up to three missed cleavages were allowed. The MS1 and MS2 mass tolerances were both set to 11 ppm. DSS was specified as the cross-linker. Spectra were annotated as potential cross-linked products if the ProteinProspector total cross-linked product score was >20, and the score difference was >0. To identify high-confidence cross-link products (57), a score difference >8.5 was used, and spectra were manually verified.

Homology Modeling.

A homology model of the Dcr-2 helicase domain (residues 1–539) was generated using the Robetta webserver (robetta.bakerlab.org); reference parent structure was DDX3X [Protein Data Bank (PDB) ID code 4PXA].

SI Materials and Methods

Chemicals and Reagents.

ATP and DSS were purchased from Sigma-Aldrich. α-[³²P]ATP (3,000 Ci/mmol) and γ-[³²P]ATP (6,000 Ci/mmol) were purchased from Perkin-Elmer.

Purification of Dcr-2 and Dcr-2 Point Mutants.

Dcr-2^WT, OSF–Dcr-2^RIII, and Dcr-2^RIII,K34A were purified as described previously (6).

Design and Cloning of Loqs-PD Truncations and Point Mutants.

Loqs-PD truncations were PCR-amplified from the parental expression vector (Loqs-PD–pET28a-p) using forward and reverse primers that introduce or included NdeI and NotI restriction sites, respectively. The PCR products and pET28a-p were digested with restriction enzymes, gel-purified, and ligated together. Each construct was verified by DNA sequencing. The primers (5′ to 3′) used to generate the Loqs-PD truncations are listed below:

• NdeI_1_Forward: GGC CCG CAT ATG GAC CAG G

• NdeI_136_Forward: GG CAT ATG CCC GTC AGC ATT CTG C

• NdeI_206_Forward: GG CAT ATG GGC GCG CAG CTG CCG GAA TCG

• NotI_337_Reverse: CT GGC GGC CGC TTA TTC GCC CTC CAA CTC GCC

• NotI_359_Reverse: GTG GTG CTC GAG TGC

Point mutations were introduced by site-directed mutagenesis and confirmed by DNA sequencing. The following primers (5′ to 3′) were used to introduce point mutations into a given Loqs-PD expression plasmid:

• 206_K,A_Forward: GCA AGG GCA AAA GCG CGA AGA TAG CCA AGC G

• 206_K,A_Reverse: CGC TTG GCT ATC TTC GCG CTT TTG CCC TTG C

• 206_KKK,EAA_Forward: CAA GGG CAA AAG CGA GGC GAT AGC CGC GCG CTT GGC CGC

• 206_KKK,EAA_Reverse: GCG GCC AAG CGC GCG GCT ATC GCC TCG CTT TTG CCC TTG

• FF,AA_Forward: GAG CAA GTC TCT AAA GAT GCT GAG GCC ATC AAG ATC TAA GCG GCC

• FF,AA_Reverse: GGC CGC TTA GAT CTT GAT GGC CTC AGC ATC TTT AGA GAC TTG CTC

Purification of Loqs-PD and Loqs Truncations/Point Mutants.

Loqs-PD was expressed from the pET28a-p expression plasmid in BL21 DE3 Gold E. coli. Cells were grown to midlog phase (OD₆₀₀ ∼ 0.8), and protein expression was induced with 0.5 mM IPTG at 30 °C for 4–6 h. Cells were harvested by centrifugation and stored at −80 °C until purification. Cells were resuspended in lysis buffer (50 mM Tris, pH 8, 300 mM NaCl, 5 mM imidazole, 1 mM TCEP, 5% glycerol) plus protease inhibitor (Roche Complete EDTA free). Lysozyme was added to ∼0.5 mg/mL and incubated on ice for 20–30 min, and cells were lysed by homogenization (three to four passes). Insoluble cellular debris was pelleted by centrifugation (25,000 × g for 30 min). The clarified lysate was incubated with 4 mL of Ni-NTA resin (Qiagen) pre-equilibrated in lysis buffer for 1 h at 4 °C to bind His-tagged protein. The resin and bound protein were pelleted; washed sequentially with lysis buffer, high-salt buffer (lysis buffer + 1 M NaCl), and stringent wash buffer (lysis buffer + 40 mM imidazole); and eluted with a stepwise imidazole gradient of elution buffer (100, 200, and 400 mM imidazole + lysis buffer). Eluted protein was dialyzed overnight into buffer A (30 mM Hepes, pH 7, 50 mM NaCl, 1 mM TCEP, 5% glycerol) + 100 mM NaCl at 4 °C in the presence or absence of PreScission protease, depending on whether the 6xHis tag was to be removed or retained, respectively. The following morning, protein was diluted back into buffer A (with 50 mM NaCl) and further purified by ion-exchange chromatography (anion-Q-HP). Protein was eluted from the IEX column by a 0–100% gradient of buffer B (buffer A + 1,000 mM NaCl). Peak fractions were combined and concentrated. Glycerol was added to 20% final, and aliquots were flash-frozen in liquid nitrogen and stored at −80 °C.

Loqs-PD truncations were purified as described for full-length Loqs-PD with the following exceptions. Loqs-PD truncation R₁LR₂C was induced at 20 °C for 8 h, and Loqs-PD mutant NR₁LR₂C^FF,AA was induced at 25 °C for 5 h. Loqs-PD truncations and mutants lacking the N-terminal 135 amino acids (e.g., R₁LR₂C and LR₂C) were dialyzed directly into buffer A (with 50 mM NaCl) before IEX.

All purified proteins were confirmed by intact mass analysis (ESI-MS) (His-tagged proteins were found to have lost the N-terminal methionine) and were free of RNA as measured by OD_260/280 < 0.7.

Synthesis of PD22 Peptide.

PD22 and PD22^mut peptides were chemically synthesized using solid-phase peptide synthesis (SPPS) with Fmoc-protected amino acids [Protein Technologies, Inc. (PTI)] and TentaGel R RAM resin to generate a C-terminal amide (0.19 mmol/g; Rapp Polymere). Peptides were synthesized on a Prelude X peptide synthesizer (PTI) with a standard synthesis scale of 30 μmol per peptide. Both peptides were synthesized with the pseudoproline VS (200 mmol; PTI) in place of the individual amino acids Val and Ser at positions 13 and 14, respectively, to improve peptide synthesis quality. To further improve peptide quality by reducing aspartimide formation, Oxyma Pure (0.1 M; Novabiochem) was added to 20% piperdine during Fmoc deprotection.

To generate fluorescently labeled PD22, an N-terminal fluorescein tag was added offline to PD22 on resin using the standard standard peptide synthesis protocol [200 mM 5 (6)-carboxyfluoroscein; Acros Organics].

Peptides were cleaved from resin for 3 h using 3.7 mL TFA, 100 μL water, 100 μL TIS, and 100 μL EDT. After cleavage, peptides were precipitated in cold diethyl ether (35 mL), stored in −20 °C overnight, and then washed with diethyl ether (3 × 30 mL), and finally peptides were pelleted by centrifugation and dried overnight in a desiccator before HPLC purification.

Crude peptides were purified by reverse phase HPLC on a Phenomenex Jupiter 4-μm Proteo C12 90 Å (250 × 21.20 mm) column with a water/ACN gradient in 0.1% TFA. Purified peptides were lyophilized and purity validated by analytical HPLC on a Phenomenex Jupiter 4-μm Proteo C12 90 Å (150 × 4.6 mm) column with a linear gradient of 5–90% ACN over 30 min. Masses of the peptides were confirmed by LC/MS on an Agilent Technologies single quadrupole LC/MS system (model: G6120B; serial: SG16094108) using an Agilent 2.7-μm Poroshell C18 120 Å (4.6 × 50 mm) column.

Peptide sequences were as follows:

• PD22: VSIIQDIDRYEQVSKDFEFIKI

• PD22^mut: VSIIQDIDRYEQVSKDAEAIKI

In Vitro Transcription of RNA Substrates.

We prepared 106 dsRNA as described in ref. 6. DNA templates for in vitro transcription were generated by PCR from plasmids containing 106 sense, 106 BLT antisense, or 106 3′ovr antisense sequences. Each RNA sequence was flanked by a hammerhead ribozyme (5′ side) and HDV ribozyme (3′ side) to ensure accurate transcript lengths. Following in vitro transcription with T7 RNA polymerase at 37 °C, Turbo DNase (Ambion) was added to digest the DNA template (37 °C for 15 min). RNA was ethanol precipitated and resuspended in ribozyme cleavage buffer (20 mM Tris, pH 8, 10 mM MgCl₂) plus T4 PNK (NEB) and incubated at 37 °C for 2–3 h. In the absence of ATP, T4 PNK catalyzes the removal of the 2′3′-cyclic phosphate left over from the self-cleaving ribozyme (60). The solution was treated with Proteinase K (37 °C for 15 min; Invitrogen), phenol:chloroform extracted, and ethanol precipitated. Isolated RNA was resuspended in formamide loading buffer (95% formamide, 25 mM EDTA, 0.025% xylene cyanol, 0.025% bromophenol blue) and separated on a denaturing gel (8% 19:1 polyacrylamide/bisacrylamide, 1× TBE, 8 M urea) (constant 30 W for ∼70 min). RNA was visualized by UV-shadowing, and bands corresponding to 106 RNAs were excised. Gel slices were crushed and RNA was extracted in “crush and soak” buffer (500 mM NH₄CH₃CO₂, 0.1 mM EDTA, pH 8, 0.1% SDS) for 12–18 h and subsequently ethanol-precipitated. The 106 sense strand was ³²P-end–labeled by treatment with T4-PNK and γ-[³²P]ATP and repurified on a 8% 19:1 denaturing gel. To generate 106 BLT and 106 3′ovr dsRNA, radiolabeled 106 sense RNA was annealed with 1.1× molar excess of unlabeled (nonradioactive) 106 BLT antisense or 106 3′ovr antisense in annealing buffer (50 mM Tris, pH 8, 20 mM KCl), incubated at 95 °C for ∼3 min, and allowed to slowly cool to room temperature (∼1.5–2 h). dsRNA was separated from excess ssRNA on an 8% 19:1 native gel (8% 19:1 polyacrylamide/bisacrylamide, 1× TBE), gel-extracted, and ethanol-precipitated. Radiolabeled dsRNA was resuspended in annealing buffer and quantified by liquid scintillation counting.

We PCR-amplified esi-2^hp from D. melanogaster genomic DNA as described (6). To clone esi-2^hp into the ribozyme-construct plasmid, a second PCR introduced XbaI and PstI restriction sites. The PCR product and ribozyme-construct plasmid were digested with restriction enzyme, gel-purified, and ligated together. The construct was verified by DNA sequencing. The primers (5′ to 3′) used to generate the ribozyme-esi-2^hp construct are listed below:

• Ribo_esi2_forward: GGT TTC CCT CTA GAT TCA ATC TGA TGA GGC CGT TAG GCC GAA ACA CTT CGG TGT CAT TGA ATG TAG CGC CCT GG

• Ribo_esi2_reverse: GGC TGG GTC TGC AGC CGA CCC TTC TGT TCT TAA GGG CCC CCC AGT AT

A DNA template for in vitro transcription was generated by PCR from the plasmid containing esi-2^hp flanked by a hammerhead ribozyme (5′ side) and HDV ribozyme (3′ side). esi-2^hp was transcribed using T7 RNA polymerase at 37 °C, and α-[³²P]ATP was included to uniformly label the transcript. Following transcription, Turbo DNase was added to digest the DNA template (37 °C for 15 min). RNA was ethanol-precipitated, resuspended in formamide loading buffer, and separated on a denaturing gel (8% 19:1 polyacrylamide/bisacrylamide, 1× TBE, 8 M urea) (constant 30 W for ∼70 min). RNA was detected by autoradiography on film, and the band corresponding to esi-2^hp was excised. Gel slices were crushed and RNA was extracted in crush and soak buffer for 12–18 h and subsequently ethanol-precipitated. Radiolabeled dsRNA was resuspended in annealing buffer and was quantified by liquid scintillation. Cold esi-2^hp was generated as described above with the exception that α-[³²P]ATP was omitted.

Sequences of dsRNA Substrates.

106 dsRNAs.

The following are the 106 dsRNA sequences:

• 106-nt sense RNA: 5′ggcaaugaaagacggugagcuggugauaugggauaguguucacccuuguuacaccguuuuccaugagcaaacugaa acguuuucaucgcucuggagugaauaccaa3′

• 106-nt BLT antisense RNA: 5′uugguauucacuccagagcgaugaaaacguuucaguuugcucauggaaaacgguguaacaagggugaacacuaucc cauaucaccagcucaccgucuuucauugcc3′

• 106-nt 3′ovr antisense RNA: 5′gguauucacuccagagcgaugaaaacguuucaguuugcucauggaaaacgguguaacaagggugaacacuauccca uaucaccagcucaccgucuuucauugccaa3′

• 106-nt sense RNA annealed to 106-nt BLT antisense RNA = 106 BLT-BLT dsRNA (106 bps) 106-nt sense RNA annealed to 106-nt 3′ovr antisense RNA = 106 3′ovr-3′ovr dsRNA (104 bps, 2 nt 3′overhangs)

esi-2^hp.

The following is the esi-2^hp sequence:

• 5′auugaauguagcgcccugguagccuguaguuugacuccaacaaguucgcucccggcgcuucacaggcgcuggaaaaucuuaaccgccggaagucacuuccgcuggcuuugauuuuccagcgucugucgagcggaaggagggacuuguuugaguccaacuacaggauacuggggggcccuuaagaacagaa3′

Gel Shift (EMSA) Assays.

Loqs-PD was serially diluted into binding buffer (25 mM Tris, pH 8, 100 mM KCl, 10 mM MgCl₂, 1 mM TCEP) plus 20% glycerol at 4 °C. To start the binding reaction, 10 µL of protein was added to an equal volume of the 2× dsRNA mix [binding buffer plus 20 pM dsRNA and 0.5 U/µL RNaseOUT (Invitrogen)] and incubated on ice for 30 min to allow the reaction to come to equilibrium. Bound dsRNA was resolved from unbound dsRNA by gel electrophoresis on a 4% native gel (19:1 acrylamide/bis-acrylamide 0.5× TBE). The gel was run at a constant voltage of 150 V for 3 h at 4 °C. The gel was then placed on filter paper and dried for 1 h at 65 °C under vacuum. Radiolabeled RNA was detected by autoradiography using a PhosphorImager screen and a Typhoon Trio Scanner. dsRNA_total and dsRNA_free were quantified using ImageQuant5 to determine the fraction bound as fraction bound = 1 − (dsRNA_free/dsRNA_total). Binding isotherms were fit using the Hill formalism, fraction bound = 1/(1 + (K_dⁿ/[P]ⁿ)), K_d is the dissociation constant, n is the Hill coefficient, and [P] is the protein concentration. Prism was used for curve-fitting analysis.

Single-Turnover Cleavage Assays.

Single-turnover cleavage assays were performed at 25 °C with 1 nM ³²P-end–labeled 106 BLT or 3′ovr dsRNA and 30 nM protein in binding buffer plus 1 U/µL RNaseOUT and 5 mM ATP. Reactions were started by adding protein to the reaction mixture (final total volume of 25 µL). A 5-µL aliquot was removed at each time point and added directly to 1.4 volumes of 2× formamide loading buffer (95% formamide, 25 mM EDTA, 0.025% xylene cyanol, 0.025% bromophenol blue). Samples were then heated for 3 min at 95 °C immediately before gel electrophoresis. Cleavage products were separated from the precursor on a 17% denaturing gel (19:1 polyacrylamide/bisacrylamide, 1× TBE, 8 M urea). The gel was run at a constant wattage of 30 W for 1.33 h at 25 °C. The gel was then frozen and radiolabeled RNA detected by autoradiography using a PhosphorImager screen and a Typhoon Trio Scanner. Cleaved and uncleaved dsRNA were quantified using ImageQuant5, and the fraction cleaved was determined as fraction cleaved = (cleaved)/(cleaved + uncleaved). Fraction dsRNA cleaved versus time was fit to the pseudo-first order rate equation: y = y₀ + A × (1 − e^−kt), A is amplitude of rate curve (>0.5), y₀ is baseline (=0), k is pseudo-first order rate constant, t is time. Prism was used for curve-fitting analysis.

Northern Blot Analysis.

Single-turnover cleavage reactions were performed using cold esi-2^hp. Cleavage products were resolved by 12% denaturing PAGE and transferred to a nylon membrane. Carbodiimide-mediated cross-linking was performed to increase the sensitivity for small RNAs (61). DNA probes were chemically synthesized and end-labeled using γ-[³²P]ATP and T4 PNK. The blot was stripped between each probing and confirmed to be devoid of residual signal.

Probe used to detect the duplex region of esi-2^hp (nucleotides 30–60) is as follows:

• 5′-AAGCGCCGGGAGCGAACTTGTTGGAGTCAAA-3′

Probe used to detect the loop region of esi-2^hp (nucleotides 79–109) is as follows:

• 5′-AAAGCCAGCGGAAGTGACTTCCGGCGGTTAA-3′

Probe used to detect the 3′-end region of esi-2^hp (nucleotides 158–190) is as follows:

• 5′-TTCTGTTCTTAAGGGCCCCCCAGTATCCTGTAG-3′

CD.

CD spectra were collected on a AVIV CD spectrometer model 410 at 25 °C. His-LR₂C, His-LR₂C^K,A, and His-LR₂C^KKK,EAA were dialyzed into CD buffer [10 mM K₂HPO₄/KH₂PO₄, pH 7.4, 100 mM (NH₄)₂SO₄] + 0.5 mM TCEP. Each protein was diluted with CD buffer to yield 2.5 mL of 1 µM final. A 10-mm path length quartz cuvette was used, and CD spectra were collected from 203 nm to 260 nm at 1-nm wavelength increments (Dynode voltage < 500 V). Three scans were averaged for buffer alone and each protein. The data were smoothed, and the background resulting from buffer was subtracted from each of the protein samples. The units were then converted from millidegrees to mean residue elipticity, [θ], degrees⋅cm²/dmol (62).

Equilibrium Sedimentation.

Equilibrium sedimentation assays on Loqs-PD and its truncations were performed on a Beckman Coulter Optima XL-A Analytical Ultracentrifuge. Loqs-PD samples were dialyzed into AUC buffer (30 mM Tris, pH 8, 150 mM NaCl, 1 mM TCEP), and equilibrium radial distributions of three concentrations each (NR₁LR₂C and NR₁LR₂C_Δ22: 10, 20, and 40 µM; R₁LR₂C: 7.5, 15, and 30 µM; LR₂C: 12.5, 25, and 50 µM) at four speeds (6,532–18,144 × g) were recorded via UV absorbance at 280 nm. Using the AUC analysis software package HETEROANALYSIS, data were globally fit to a single ideal species with floating molecular mass using nonlinear least squares (63). Buffer density and protein partial specific volume calculations were performed with SEDNTERP (64).