Biochemical and structural basis of Dicer helicase function unveiled by resurrecting ancient proteins

Significance

Among animals, the contribution of Dicer’s helicase to recognition and elimination of viral double-stranded RNA (dsRNA) varies from phylum to phylum. Vertebrate Dicers show no helicase activity, while an arthropod ortholog uses helicase translocation to efficiently move dsRNA into Dicer’s cleavage site. The biochemical and structural basis of Dicer’s helicase function, and the evolutionary events that contribute to divergence in function, have remained unknown. This study shows how ancient Dicer helicase tightly binds dsRNA and couples adenosine triphosphate (ATP) hydrolysis to movement along dsRNA. The data reveal how components of this intricate system declined along different clades of animal evolution.

Abstract

A fully functional Dicer helicase, present in the modern arthropod, uses energy from ATP hydrolysis to power translocation on bound dsRNA, enabling the processive dsRNA cleavage required for efficient antiviral defense. However, modern Dicer orthologs exhibit divergent helicase functions that affect their ability to contribute to antiviral defense. Moreover, mechanisms that couple ATP hydrolysis to Dicer helicase movement on dsRNA remain enigmatic. We used biochemical and structural analyses of ancestrally reconstructed Dicer helicases to map evolution of dsRNA binding affinity, ATP hydrolysis and translocation. Loss of affinity for dsRNA occurred early in Dicer evolution, coinciding with a decline in translocation activity, despite preservation of ATP hydrolysis activity. Ancestral nematode Dicer also exhibited significant decline in ATP hydrolysis and translocation, but studies of antiviral activities in the modern nematode Caenorhabditis elegans indicate Dicer retained a role in antiviral defense by recruiting a second helicase. Cryogenic electron microscopy (cryo-EM) analyses of an ancient metazoan Dicer allowed capture of multiple helicase states revealing the mechanism that connects each step of ATP hydrolysis to unidirectional movement along dsRNA. Our study rationalizes the diversity in modern Dicer helicases by connecting ancestral functions to observations in extant enzymes.

Antiviral defense in vertebrates involves RIG-I-like receptors (RLRs; Retinoic acid-Inducible Gene-like receptors) that recognize viral dsRNA to trigger an interferon response, while invertebrates of the arthropod and nematode phyla use Dicer to cleave viral dsRNA and trigger antiviral RNA interference (RNAi) (1). RLRs and Dicer have related Superfamily 2 (SF2) helicase domains, suggesting this helicase domain was involved in antiviral defense in a common ancestor (1, 2). Helicase domains of extant Dicer enzymes have widely different activities that likely reflect an evolutionary arms race with viruses, and the need to acquire new functions, such as the microRNA (miRNA) processing required for gene regulation in extant animals (3, 4).

Drosophila melanogaster Dicer-1 (dmDcr1) has a defunct helicase domain incapable of ATP hydrolysis and instead uses its Platform/PAZ (Piwi-Argonaute-Zwille) domains to recognize miRNA precursors (pre-miRNAs), which are cleaved by Dicer’s RNase III domains to produce mature miRNAs (5). By contrast, Dicer-2 (dmDcr2), encoded by a duplicated Dicer gene specific to arthropods, has a specialized helicase domain that couples ATP-driven dsRNA translocation to processive cleavage of long viral dsRNAs (Fig. 1A) (6–10). Humans have a single Dicer (hsDcr) that is specialized for miRNA production, but its helicase domain lacks the ability to hydrolyze ATP (Fig. 1B). Phylogenetic analyses indicate this loss occurred at the onset of vertebrate evolution (11–13). Accordingly, most evidence indicates mammalian Dicers only play a role in antiviral defense in special conditions, such as in mammalian germ and stem cells, where processing is ATP-independent (14, 15).

Fig. 1.

Phylogenetic analysis of diversity in Dicer helicase function. (A–C) Cartoon depictions of (A) dmDcr-2 binding long dsRNA with helicase domain, (B) hsDcr binding pre-miRNA with Platform/PAZ domains, and (C) ceDCR-1 antiviral complex binding pre-miRNA and long dsRNA using Platform/PAZ domains and DRH-1 helicase, respectively (10, 12, 16, 17). RDE-4 is shown with dsRNA-binding domains in maroon. (D) Domain organization of AncD1D2, with colored rectangles showing conserved domain boundaries indicated by amino acid number. Domain boundaries were defined as in previous studies (13). (E) Summarized phylogenetic tree constructed by constraining maximum likelihood tree to known species relationships in Dicer-1 clade. Font colors at nodes correlate with data for ancestral protein.

The lack of helicase function in hsDcr may have arisen due to competition between helicase domains of Dicer and RLRs, as both recognize viral dsRNAs in the cytoplasm. Indeed, RNAi and RLR-stimulated interferon pathways antagonize one another in mammals (18–20). Also, arthropods like Drosophila, where extant dmDcr2’s helicase domain retains a role in antiviral defense, do not have RLRs. Alternatively, loss of Dicer helicase function may have begun early in animal evolution, driven by competition for RNA binding between Dicer’s helicase domain and its own Platform/PAZ domains. Here, dominance of RLRs in viral dsRNA recognition, rather than causing loss of Dicer helicase function, would be an evolutionary consequence of its decline. Caenorhabditis elegans Dicer (ceDCR-1) functions in antiviral defense with an RLR called DRH-1, which is crucial for dsRNA cleavage in vitro and for antiviral defense in vivo (Fig. 1C) (16, 21, 22). This suggests that ceDCR-1’s helicase domain was subject to the same evolutionary pressures that led to loss of vertebrate Dicer helicase function, and that Dicer began to lose helicase function early in animal evolution, before divergence of vertebrates and invertebrates.

Ancestral protein reconstruction (APR) provides a powerful tool for dissecting Dicer helicase evolution (13, 23). Previously, we used APR to track evolution of Dicer’s helicase domain in animals, mapping differences in ATP hydrolysis between arthropod and vertebrate Dicers (13). Here, we extend biochemical and phylogenetic analyses to map evolution of dsRNA translocation by Dicer’s helicase domain (Fig. 1D). This analysis, together with cryo-EM analyses of an ancient Dicer helicase domain, produced a more comprehensive understanding of Dicer evolution. We find that early metazoan Dicer helicase bound dsRNA with high affinity and coupled dsRNA binding with efficient ATP hydrolysis to drive robust translocation along dsRNA. Helicase affinity for dsRNA began to decline in the Dicer-1 clade before onset of bilaterians, declining further in deuterostome and protostome lineages. While ATP hydrolysis activity is progressively lost along both clades, our work suggests that loss of dsRNA binding is the primary mechanism underlying loss of translocation in the Dicer-1 clade. Structural analyses of ancient Dicer helicase translocating on dsRNA provide previously unrecognized details in SF2 helicase function, likely operative in modern RLRs.

Results

Ancestral Protein Reconstruction Using Alternative Dicer Helicase Phylogeny.

The maximum likelihood (ML) phylogenetic tree of our prior studies fit best with a model where an early metazoan Dicer (AncD1D2) underwent a gene duplication to produce Dicer-1 and Dicer-2, and one copy was lost in most modern animals (13). While this model is supported by multiple lines of phylogenetic evidence (23, 24), finer details of the ML tree did not fit the species tree of life, either due to incomplete lineage sorting or long branch attraction. This is not unexpected since the ML tree, or “gene” tree, was based on phylogenetic analyses of only two domains (helicase and DUF283, Fig. 1D) of the Dicer gene, while species trees are typically built from multigene or whole genome datasets. Dicer gene duplications and subsequent gene loss could also explain the gene tree-species tree incongruence.

To identify the evolutionary windows where Dicer helicase functions were lost, we repeated phylogenetic tree reconstruction using the helicase domain and DUF283 (henceforth referenced as helicase) and constrained the Dicer-1 clade of the tree to match evolutionary relationships from the consensus tree of life (Fig. 1E). While the species tree recapitulated some clade relationships of the gene tree, some nodes were replaced by new nodes in the species tree, providing the opportunity to perform additional analyses (SI Appendix, Fig. S1 A and B). Addition of the hypothetical ancestor of bilaterian Dicer-1 (AncD1_BILAT) was particularly advantageous, offering the opportunity to determine whether loss of ATP-dependent functions is unique to deuterostomes and their vertebrate descendants, or if the trend extends into invertebrate protostomes. Interestingly, ancestral vertebrate Dicer-1 (AncD1_VERT) primary sequence remained identical when predicted from either tree. APR is inherently inexact as phylogenies and reconstructed enzymes are statistical hypotheses with a degree of uncertainty. Here, the similarity in biochemical trends between ancestral constructs from gene and species trees, even when equivalent nodes produce different sequences, lends additional confidence to the conclusions we draw from these trends (13).

Ancestral Animal Dicer Helicase Translocates on dsRNA.

To determine whether ancient animal Dicers translocate on dsRNA, we adopted a gel-based streptavidin-displacement assay that monitors ability of translocating helicases to remove dsRNA-bound streptavidin (Fig. 2A) (25, 26). dsRNA was ³²P-labeled at the 5’ end of the sense (top) strand to allow visualization of radioactive species on a native gel and conjugated with biotin at the 5’ end of the antisense (bottom) strand to allow streptavidin binding (Fig. 2B and SI Appendix, Fig. S2).

Fig. 2.

Ancient metazoan Dicer translocates along dsRNA. (A) Schema depicting streptavidin-displacement assay for translocation. Green oval, biotin covalently linked to RNA; Maroon circle, tetravalent streptavidin. Color-coded Dicer domains as labeled. (B) Streptavidin-displacement assays for AncD1D2 on BLT (Left) and 3’ovr (Right) dsRNA in the absence and presence of 5 mM ATP. Yellow highlighted P, 5’ ³²P-label. (C) Graph showing streptavidin displacement for ancestral constructs from species-constrained tree for BLT (bold) and 3’ovr (dashed) 42-bp dsRNA with 5 mM ATP. Data were fit to a pseudo-first-order rate equation. Data points, mean ± SD (n ≥ 3). (D) Graph showing quantification of TLC analyses monitoring hydrolysis of ATP (100 µM) by select ancestral helicases with BLT (bold) and 3’ovr (dashed) 42-bp dsRNA (400 nM). Data are fit as in (C) (n ≥ 3). (E) Michaelis–Menten curves of ATP hydrolysis by AncD1D2, AncD1_BILAT, and AncD1_DEUT with BLT (bold) and 3’ovr (dashed) dsRNA. (F) Binding isotherms for ancestral helicase-DUF283 proteins (HELDUF; Fig. 1D) from gel-shift assays where fraction bound was determined from radioactivity of dsRNA_free and dsRNA_bound. Data were fit to calculate dissociation constant, K_d, using the Hill formalism, where fraction bound = 1/(1 + [K_dⁿ/[P]ⁿ]). Data points, mean ± SD (n ≥ 3).

AncD1D2 displayed robust translocation on dsRNAs with blunt (BLT) and 3’ overhang (3’ovr) ends (Fig. 2 B and C and SI Appendix, Table S1). With BLT dsRNA, some ATP-independent streptavidin displacement was observed (Fig. 2 B, Left). Possibly AncD1D2 binds to ends as well as internal regions of BLT dsRNA, leading to stacking of multiple helicases, even without ATP, which in turn displaces streptavidin. This phenomenon was not observed with 3’ovr (Fig. 2 B, Right), or any other ancestral proteins. Quantification of multiple assays revealed that the bilaterian ancestor, AncD1_BILAT, also exhibited translocation activity, albeit reduced from AncD1D2, for both BLT and 3’ovr dsRNA (Fig. 2C and SI Appendix, Fig. S3A and Table S1). AncD1_DEUT, the deuterostome descendant of bilaterian Dicer-1, showed minimal displacement activity, with or without ATP (Fig. 2C and SI Appendix, Fig. S3B). In these assays, we used dsRNA with a 5’ monophosphate, based on observations from invertebrate Dicers, which do not require a specific 5’ phosphorylation state (7, 16). Further, while a 5’ triphosphorylated BLT dsRNA terminus is optimal for RIG-I (1), the domain responsible for recognition of triphosphates in RIG-I, the CTD, is not found in Dicers. We used BLT dsRNA to mimic viral dsRNA and 3’ovr to mimic cellular pre-miRNA for in vitro Dicer helicase assays. A third class of Dicer substrates, the endogenous (endo) siRNA precursors, typically has structured, frayed or closed loop ends that require cofactors like Loqs-PD to promote helicase engagement (27, 28).

AncD1_VERT did not translocate on dsRNA, consistent with the absence of ATPase activity shown in prior work (Fig. 2C and SI Appendix, Fig. S3C) (13). A similar loss of translocation was observed with AncD1_NEM, the common ancestor of protostome nematode Dicer-1 (Fig. 2C and SI Appendix, Fig. S3D). This independent loss along two descendant lineages of AncD1_BILAT suggests the decline in helicase function between AncD1D2 and AncD1_BILAT progressed until dsRNA translocation activity was lost in modern Dicer-1. While we did not test the ancestor of Arthropod Dicer-1, we can confidently predict loss of helicase function based on the degeneration of Dicer-1 ATPase sequences in modern arthropods (5, 29).

We performed ATP hydrolysis and dsRNA binding assays to determine how these reactions underpin the trends in dsRNA translocation. Using a thin-layer chromatography (TLC) assay, we performed multiple-turnover ATPase assays with 100 µM ATP and the ancestral protein, with and without excess dsRNA. Like the gene tree constructs (13), ancestral Dicer helicases from the species tree required dsRNA to catalyze robust ATP hydrolysis (Fig. 2D and SI Appendix, Fig. S4A). However, subtle differences in hydrolysis efficiency did not mirror the larger differences observed in dsRNA translocation between AncD1D2, AncD1_BILAT, and AncD1_DEUT (Fig. 2 C and D and SI Appendix, Fig. S4 and Table S1). We also performed Michaelis–Menten analyses on select ancestral helicases. We observed ~twofold to threefold decrease in the efficiency of hydrolysis (k_cat/K_M) from AncD1D2 to AncD1_BILAT. (Fig. 2E and SI Appendix, Table S1). For AncD1_DEUT, the net k_cat/K_M value was similar to AncD1_BILAT for BLT dsRNA but reduced by ~threefold for 3’ovr (Fig. 2E and SI Appendix, Table S1). The preservation of some ATP hydrolysis in AncD1_DEUT suggested its lack of translocation could not be entirely explained by loss of hydrolysis.

Both ATP hydrolysis assays were performed with excess dsRNA compared to protein (multiple turnover conditions) to minimize contributions of dsRNA affinity to observed differences. Due to practical limitations, the translocation assay was performed with excess protein (single turnover; see Materials and Methods). To assess effects of dsRNA binding on translocation trend, we performed gel shift assays incubating AncD1D2, AncD1_BILAT, or AncD1_DEUT with BLT dsRNA in the presence of ATP, using conditions similar to the translocation assay (Materials and Methods). AncD1D2 had a dissociation constant (K_D) of 79 nM compared to 140 nM for AncD1_BILAT and 942 nM for AncD1_DEUT (Fig. 2F and SI Appendix, Fig. S5). The gel shift results indicate that progressive decline in translocation from AncD1D2 to AncD1_BILAT to AncD1_DEUT is related to loss of dsRNA affinity as the Dicer-1 helicase evolved. Thus, while ATP hydrolysis also declines along the Dicer-1 clade, our inability to detect translocation in AncD1_DEUT may also be attributed to reduction of dsRNA affinity. Possibly, the streptavidin-displacement assay is not sensitive enough to capture translocation below a certain limit of detection. However, as discussed subsequently, cryo-EM analyses suggest AncD1_DEUT is truly incapable of translocation even when bound to dsRNA. Importantly, ancestral Dicer helicases from the gene tree show a similar trend in streptavidin displacement activity. Ancient AncD1D2 and AncD1_{ARTH/LOPH/DEUT} helicases translocate along dsRNA (SI Appendix, Fig. S6), but this activity is progressively lost along the Dicer-1 clade, mirroring decline in dsRNA affinity and ATP hydrolysis and accounting for some of the inherent uncertainty in APR (13).

C. elegans Dicer Recruits an RLR, DRH-1, for Translocation in the Antiviral Complex.

ceDCR-1 functions in multiple pathways in vivo, and when targeting viral dsRNA, functions in the antiviral complex with an RLR helicase, DRH-1, and a dsRNA binding protein, RDE-4 (Fig. 1C) (21). Our recent work shows ATP hydrolysis by both helicases is required for cleavage, but ceDCR-1 hydrolysis is far less efficient than DRH-1 hydrolysis (16), consistent with the decline in hydrolysis and translocation observed in AncD1_NEM. Structural data in the recent study show DRH-1 localized to internal regions of dsRNA, suggesting that it fuels translocation of the antiviral complex (16). To understand how modern C. elegans antiviral complex translocates, we used the streptavidin-displacement assay. We first quantified translocation by the C. elegans antiviral complex using extant dmDcr2 as control. dmDcr2 acts as a single protein to cleave viral dsRNA in fruit flies, and in vitro studies show it couples ATP hydrolysis to translocation (7, 30). To focus on translocation, for both organisms, we made mutations in Dicer to eliminate RNase III cleavage activity (Fig. 3 A and B). Robust ATP-dependent translocation was observed for the ceDCR-1^RIII/DRH-1/RDE-4 complex, as well as dmDcr2^RIII, although the C. elegans antiviral complex was more efficient, showing similar levels of translocation at lower protein concentrations (Fig. 3 A and B and SI Appendix, Fig. S7 A and B). Consistent with prior studies of cleavage and ATP hydrolysis catalyzed by the ceDCR-1 antiviral complex and dmDcr2, translocation was more efficient with BLT than 3’ovr dsRNA (Fig. 3B) (9, 16).

Fig. 3.

Extant invertebrate antiviral Dicers translocate along dsRNA. (A) Streptavidin-displacement assay for ceDCR-1^RIII/DRH-1/RDE-4 with BLT 42-bp dsRNA, with and without 5 mM ATP. (B) Graph quantifying streptavidin displacement data for 50 nM ceDCR-1^RIII/DRH-1/RDE-4 and 200 nM dmDcr2^RIII, for BLT and 3’ovr dsRNA with 5 mM ATP. Data were fit to a pseudo-first-order rate equation. Data points, mean ± SD (n = 3). (C) Streptavidin-displacement assay for ceDCR-1/DRH-1/RDE-4 with BLT 42-bp dsRNA with deoxynucleotide patch, with and without 5 mM ATP. Cartoons indicate deoxynucleotides (red) and gel migration of different species. (D) Streptavidin-displacement assay for ceDCR-1/DRH-1^K320A/RDE-4 and ceDCR-1^G36R/DRH-1/RDE-4 with BLT 42-bp dsRNA with deoxynucleotide patch with 5 mM ATP. Cartoons as in (C). (E) Streptavidin-displacement assay for ceDCR-1/DRH-1 with BLT 42-bp dsRNA containing deoxynucleotides with 5 mM ATP. Cartoons as in (C). (F) Graph quantifying streptavidin displacement data for ceDCR-1/DRH-1/RDE-4 and indicated mutants, using BLT 42-bp dsRNA with deoxynucleotides and 5 mM ATP. Data were fit as in (B).

To delineate contributions of each protein of the antiviral complex to translocation, we designed a version of the assay that did not require RNase III mutation, using 42-bp dsRNAs with deoxynucleotides at predicted ceDCR-1 cleavage sites (SI Appendix, Figs. S2 and S8). Wildtype C. elegans antiviral complex showed ATP-dependent streptavidin displacement on this substrate, with minimal dsRNA cleavage (Fig. 3C). A point mutation in the Walker A motif of DRH-1’s helicase domain (DRH-1^K320A) abolished all translocation activity, while mutating the Walker A motif in ceDCR-1 (ceDCR-1^G36R), or omitting RDE-4 from the complex, did not affect translocation (Fig. 3 D–F). Thus, DRH-1, not ceDCR-1, is responsible for translocation by the antiviral complex. While ceDCR-1 helicase mutant retained streptavidin displacement activity, it did not cleave dsRNA (compare Fig. 3 C and D), suggesting that ceDCR-1’s helicase retains an ATP-dependent activity important for cleavage (16).

AncD1D2 Undergoes ATP-Dependent Conformational Changes That Are Coupled to Translocation.

To understand how active ancient Dicer helicase used ATP hydrolysis to move on dsRNA, we determined cryo-EM structures of AncD1D2 bound to 27-bp BLT dsRNA without nucleotide, with ADP-aluminum fluoride (ADP-AlF_x), and with ATP, at resolutions 3.0 to 3.5 Å (Fig. 4 and SI Appendix, Figs. S9–S13 and Table S2). Image processing recovered five structural snapshots: state A^0ground (no nucleotide, ground state), state B^end‡ (endbound transition state), state C^post,closed (post hydrolysis closed state), state D^post,ground (post hydrolysis ground state) and state E^internal‡ (internal transition state) (Fig. 4). Each state represented a unique conformation of the helicase as it cycles through ATP hydrolysis and couples this to motor function. Consistent with prior models (31), we observed the tandem RecA domains, Hel1 and Hel2, wrapped around dsRNA in a C-shape with the C-terminal dsRBM fold, DUF283, which is connected to Hel2 by a pincer domain (Fig. 4 A–E, far left column) (31). Hel2i, an SF2 helicase-specific insertion between Hel1 and Hel2, was not visible in states A^0ground and B^end‡, likely due to flexibility, but was observed in states C-E. As AncD1D2 progresses through the ATPase cycle, the Hel1 α-helix containing motif Ia, termed the spring helix (32), and the Hel2 helix containing motif IVa, termed the loop helix for its connection to the Hel2 loop (33), moved relative to each other to switch between semiclosed and closed conformations (Fig. 4 A–E, second column) (34, 35).

Fig. 4.

Cryo-EM structures of AncD1D2 in complex with BLT 27-dsRNA in different stages of ATPase cycle. Colored atomic models of AncD1D2 in complex with BLT 27-dsRNA in (A) state A^0ground, ground state without nucleotide, (B) state B^end‡, transition state following ADP-AlF_x binding to ground state, (C) state C^post,closed, post hydrolysis closed state following state B^end‡, (D) state D^post,ground, post hydrolysis semiclosed state in internal dsRNA segment following state C^post,closed, (E) state E^internal‡, transition state after initial ATPase cycle at dsRNA terminus. Hel2i is invisible in states A^0ground and B^end‡ likely due to subdomain flexibility. Colors: Hel1 (purple), Hel2 (sea green), Hel2i (yellow), Pincer (red), DUF283 (dark gray), dsRNA (cornflower blue). Distance across ATPase cleft calculated by measuring distance between central point of Hel1 spring helix and Hel2 loop helix. The dsRNA position column shows zoom-in of contacts between spring and loop helices and dsRNA 3’ tracking strand. Polar contact between V70 from spring helix, and labeled PO₄ group in state A, was used to benchmark helicase position on dsRNA for all states. The nucleotide state column shows ATPase pocket and occupying nucleotide.

In state A^0ground, AncD1D2 was bound near the dsRNA terminus with backbone amines of G93 of motif Ib and V70 of motif Ia contacting phosphate oxygens of the 3’ terminal P27 and P26 respectively (Fig. 4A, dsRNA position column). In B^end‡, C^post,closed and E^internal‡, V70 was shifted, but still interacting near the dsRNA terminus, while D^post,ground, was positioned at an internal dsRNA segment, with V70 contacting P19 (Fig. 4 A–E, dsRNA position column). Combining information from the position of AncD1D2 on dsRNA, the relative positions of spring and loop helices, and the orientation of nucleotide in the ATPase pocket, allowed correlation of each snapshot with stages of the ATPase cycle (Fig. 4 A–E, nucleotide state column).

ATP-Stimulated Closure of the Helicase Is the First Step in AncD1D2 Translocation.

In state A^0ground, Hel1 (motifs Ia, Ib) and Hel2 (motifs IVa, IVb, V) contact the end of 27-bp dsRNA with ~8-bp footprint (Fig. 5A and SI Appendix, Fig. S9A). To interpret helicase movement along dsRNA as AncD1D2 cycles through ATP binding, hydrolysis, and product release, we assigned initial Hel1-bound 3’ strand phosphates as P1 and P2. In A^0ground, motif Ib (92-VGDMD-96) binds the first 3’ phosphate (P1, equivalent to P27 in Fig. 4) and motif Ia (68-NTV-70) binds the adjacent phosphate (P2) while motif IVa (406-IVGH-409) binds P3 (Fig. 5A). Motif V contacts the next phosphate and ribose (SI Appendix, Fig. S9A). This contiguous interaction between the helicase domain and the first four phosphates on the 3’ strand is the ground state configuration prior to addition of ATP. In this conformation, the ATPase cleft between Hel1 and Hel2 is semiclosed, as determined by the distance separating the spring helix and loop helix (Fig. 5A).

Fig. 5.

Chronology of AncD1D2 ATPase cycle coupled to translocation on dsRNA tracking strand. (A–E) Zoom-in of RecA–dsRNA interface, showing select Hel1 and Hel2 motifs in contact with dsRNA 3’ tracking strand for states A-E. Phosphates (red and yellow) connect RNA bases (tubes). Key residues are shown in spring and loop helices. Motifs are labeled and highlighted with lines depicting position in helicase. Phosphate position is benchmarked by assigning PO₄ group contacting V70 in state A^0ground as P2, and correlating helicase movement on dsRNA to stages of ATPase cycle, as determined by nucleotide state in ATPase pocket and state of ATPase cleft. Interactions between protein and RNA, dashed lines. Distance (in Å) between center of spring helix and center of loop helix was used to depict semiclosed or closed state. Arrows represent chronological transition from state to state.

We added ATP, NaF, and Al(NO₃)₃ to generate ADP-AlF_x and mimic the transition state of a helicase-bound ATP. This led to state B^end‡ and showed closure of the ATPase cleft around the ATP analog, switching the helicase from the ground state to the transition state (compare Fig. 5A with Fig. 5B). The spring and loop helices moved closer together to bridge ATP, causing Hel1 to move 3’ to 5’ to form the high energy transition state (Fig. 5 A and B and SI Appendix, Fig. S9 A and B). V70 and H409 are moved into proximity in the transition state where they both contact P2 (Movie S1).

ATP Hydrolysis Is Coupled to Helicase Movement Along 3’ Tracking Strand.

For Dicer’s helicase to translocate, it needs to progress beyond the transition state by coupling ATP hydrolysis to motion that results in net 3’ to 5’ movement. To visualize active translocation, we incubated AncD1D2 with 27-bp BLT dsRNA and ATP at 37 °C for 5 to 10 min before freezing grids. We solved multiple cryo-EM structures that showed AncD1D2 in internal dsRNA segments with ADP in the ATPase pocket indicating a post-ATP hydrolysis state. Focused 3D classification yielded a unique class density depicting a helicase conformation close to the dsRNA terminus, C^post,closed, where Hel1 and Hel2 remained in the closed conformation. V70 and H409 contact P3 instead of P2, indicating movement of the helicase by one nucleotide in the 3’ to 5’ direction compared to B^end‡ (Fig. 5C and SI Appendix, Figs. S9C and S12A). The presence of ADP indicates that C^post,closed represents the helicase conformation following removal of the scissile γ-PO₄ from ATP. C^post,closed is a structural snapshot not previously reported, and establishes that the helicase moves in the 3’ to 5’ direction while the ATPase cleft is still closed (Movie S2). This observation offers insight relevant to other dsRNA-stimulated SF2 helicases (Discussion).

Multiple attempts to capture AncD1_DEUT in the translocation state yielded 2D classes with density for the helicase exclusively at the dsRNA terminus. Orientation bias of the frozen sample prevented reconstruction of high-resolution 3D density maps, but 2D classes support biochemical assays and show a lack of translocation despite conditions promoting dsRNA terminus binding (SI Appendix, Fig. S14 A and B). Possibly AncD1_DEUT is missing an essential helicase component that enables coupling of hydrolysis to translocation or is incapable of sustaining binding to internal dsRNA regions and dissociates rapidly (SI Appendix, Fig. S14C).

AncD1D2’s Transition from Closed to Semiclosed Completes Translocation Along One Base-Pair.

Remaining 3D classes from the post-ATP hydrolysis dataset show AncD1D2 in internal dsRNA regions (SI Appendix, Figs. S9D and S12A). The consensus model, with all subdomains visible, is assigned as state D^post,ground (Fig. 5D and SI Appendix, Fig. S12A). In D^post,ground, spring and loop helices are in the semiclosed conformation of the ground state, despite presence of ADP in the ATPase pocket (Fig. 5D and SI Appendix, Fig. S9D). This observation suggests progression from C to D is the final stage of the ATPase cycle: relaxation to ground state. The return of Hel1 and Hel2 to ground state changes contacts between spring and loop helices, and the dsRNA tracking strand (Fig. 5 C and D). In C^post,closed, H409 sidechain (IVa) contacts the same P3 backbone phosphate as V70, but in D^post,ground, H409 has drifted away from P3 in the 3’ to 5’ direction (Fig. 5D and Movie S3). Hel1, now disconnected from Hel2, appears to pull on P3 as it relaxes in the 5’ to 3’ direction, while Hel2 drifts back to its ground state in the opposite direction.

Comparison of D^post,ground subclasses (D.1 to D.4) showed the Hel2 loop adopting different positions, while the remainder of the helicase remained roughly identical (SI Appendix, Figs. S12A and S15A). These Hel2 loop movements changed H409 contacts in different state D subclasses (SI Appendix, Fig. S15 A and B and Movie S4). The Hel2 loop also contacts the Hel2i bundle via salt bridges in proximity to conserved residues (SI Appendix, Fig. S15 C and D) and widens the dsRNA major groove as it transitions from the dsRNA terminus to internal segments (SI Appendix, Fig. S16 A and B). dsRNA bending caused by major groove expansion enables extensive contacts between internal dsRNA and DUF283 that may contribute to AncD1D2 translocation by facilitating binding to internal dsRNA (SI Appendix, Figs. S9D and S16 C and D). AncD1D2 DUF283, as in dmDcr2, contacts the dsRNA minor groove and major groove with dsRBM region 1 and region 3 residues respectively (SI Appendix, Fig. S16 C and D) (28). Previous analysis of dmDcr2 structures attributed dsRNA bending to an extended Hel2i loop contacting dsRNA at the expanded major groove (10). However, AncD1D2 lacks the extended Hel2i loop, suggesting that dmDcr2’s Hel2i loop is a modern adaptation for improving dsRNA contact. Indeed, Hel2i in structures of dmDcr2 bound to Loqs-PD, a Dicer cofactor, is rotated ~9.4 Å toward dsRNA at an angle of ~12.5° compared to Hel2i from AncD1D2 (SI Appendix, Fig. S17). Loqs-PD binds dmDcr2 at the Hel2–Hel2i–pincer junction where it modulates helicase function, possibly by coordinating Hel2i movement (10). Hel2i’s proximity to dsRNA in the dmDcr2 structure permits contact with the dsRNA terminal 5’ phosphate, and the absence of anchoring interactions mediated by either the extended Hel2i loop or the presence of Loqs-PD may explain why AncD1D2 Hel2i is too flexible for cryo-EM observation in endbound states A and B. Conversely, Hel1, Hel2, and DUF283 adopt similar positions for dmDcr2 and AncD1D2 (SI Appendix, Fig. S17).

To further visualize the impact of ATPase pocket closure on Hel1 and Hel2, we solved a 3.1 Å cryo-EM structure of AncD1D2 bound to ADP-AlF_x in E^internal‡, a state that depicts the start of the second ATPase cycle (Fig. 5E and SI Appendix, Fig. S9E). In E^internal‡, the helicase returns to the closed conformation of the transition state as a new ATP mimic is bound to initiate a second round of hydrolysis and continue translocation along dsRNA. V70 and H409 are again in position to bind the same phosphate, this time the adjacent P4 phosphate. The use of one ATP to move one base pair on dsRNA is consistent with prior studies showing dmDcr2 hydrolyses ~23 ATP molecules to produce one 21-bp siRNA (36).

The higher quality of the E^internal‡ electron density map allowed closer analysis of AncD1D2 RecA movement during hydrolysis. Comparing B^end‡ and C^post,closed suggests that the two RecA domains move as one rigid body during or immediately after ATP γ-PO₄ bond cleavage (Fig. 5 B and C). Examination of the ATPase pocket in E^internal‡ confirmed the AlF_x mimic of the γ-PO₄ in proximity to motif VI, while E143 (DECH) is poised to coordinate water for attack on the ß-γ PO₄ bond (SI Appendix, Fig. S18 A, Left). Closure of the ATPase cleft creates a network of polar interactions between Hel1 and Hel2, similar to contacts observed in RIG-I (SI Appendix, Fig. S18 A, Right and SI Appendix, Fig. S19). In C^post,closed, hydrolysis of the γ-PO₄ disrupts connection between the nucleotide and Hel2 (SI Appendix, Fig. S18 B, Left). However, the network of interactions between Hel1 and Hel2 was largely maintained, causing the helicase to remain closed despite the inability of ADP to bridge the RecA domains (SI Appendix, Fig. S18 B, Right). In the transition between C^post,closed and D^post,ground, the network breaks down as Hel2 drifts away from Hel1 (SI Appendix, Fig. S18C). This order of events suggests that formation of these contacts during closure of the RecA domains enables the helicase to move as a single body during the hydrolysis step of the ATPase cycle. Subsequent disruption of these contacts coincides with Hel1 and Hel2 returning to the semiclosed state, adopting a conformation primed for the next round of ATP binding and hydrolysis.

Discussion

Biochemical Properties of Ancient Dicer-1 Helicases Explain Diversity in Modern Helicase Function.

Subfunctionalization of Dicer-1’s helicase domain in bilaterians was likely stochastic, beginning with progressive loss of dsRNA binding and translocation, while ATP hydrolysis remained relatively intact (Fig. 2). Decline in dsRNA binding and translocation may have been sufficient to render Dicer’s helicase defunct as a competitor with RLRs for viral dsRNAs or with Platform/PAZ for endogenous dsRNAs. This would remove selection pressure for helicase nonfunctionalization and allow retention of vestigial ATP hydrolysis up to AncD1_DEUT, before being lost entirely in AncD1_VERT and hsDcr. Thus, components of Dicer helicase distant from the ATPase pocket were initially lost and this loss was sufficient to disable helicase function, inadvertently resulting in vestigial retention of ATPase motifs in nonfunctional helicases like hsDcr. In the protostome clade, minimal ATP hydrolysis is retained by AncD1_NEM helicase (Fig. 2D). Possibly this ATP hydrolysis is also vestigial, but it clearly serves a function in modern ceDCR-1 where it is required for proper dsRNA cleavage (16), but not translocation (Fig. 3). Conversely, the ATPase pocket of modern dmDcr1 is degenerate and incapable of hydrolysis, possibly because retention of the dmDcr2 duplicate hastened the nonfunctionalization of dmDcr1. Thus, the stochastic nature of Dicer-1’s evolution is reflected in diverse modern adaptations to competition for cellular or viral dsRNA and loss of helicase function.

Mechanistic Basis of AncD1D2 Translocation Along dsRNA.

Variation in helicase translocation is governed by the capacity to bind dsRNA, hydrolyze ATP, and couple these activities to movement. The streptavidin-displacement assay correlates with translocation for related RLRs (25, 26), but this correlation is not assured for our quasi-artificial Dicer helicase constructs. However, cryo-EM analysis of select ancestral helicases, in combination with observations for modern dmDcr2 and ceDCR-1, supports the notion that streptavidin displacement is caused by Dicer helicase translocation. Structural analyses of AncD1D2 helicase in multiple states (Figs. 4 and 5) provide insights into helicase function that can be extended to modern Dicers and RLRs. The ground state of AncD1D2, A^0ground, captured without nucleotide, exists in a similar semiclosed state to structures of modern dmDcr2 bound to dsRNA termini (Fig. 4A) (10). ATP binding caused Hel1 and Hel2 RecA domains to close further to form a high-energy transition state observed in B^end‡ and E^internal‡ AncD1D2 (Fig. 5 A and E and Movie S1). Closed transition states have been observed in RLR helicases containing transition state mimics (34, 37) and in AAA+ helicases (38), but the mechanism that governs coupling of ATP binding and hydrolysis in Dicer to successive transitions between semiclosed and closed states, and the resulting unidirectional translocation, had remained enigmatic. By isolating C^post,closed, the post hydrolytic closed state, and D^post,ground, the post hydrolytic semiclosed state, we find that movement of the helicase on dsRNA occurs in the closed state, concurrent with ATP hydrolysis. Conformational changes in Hel2 as the γ-PO₄ is cleaved and dislodged from motif VI drive movement of Hel2 in the 5’ direction, while interactions between Hel2 and Hel1 ensure that Hel1 is pulled in the same direction (SI Appendix, Fig. S18). Our cryo-EM snapshots depict 3’ to 5’ movement as the transition from B^end‡ to C^post,closed, which is followed by relaxation from C^post,closed to D^post.ground (Fig. 5 B–D and Movies S2 and S3). Structural analysis of AncD1D2 helicase thus provides insight into the second half of the ATPase cycle, revealing movement of the closed helicase on dsRNA, followed by relaxation of the closed post hydrolysis state to the semiclosed ground state.

Combining Cryo-EM with Evolutionary Biochemistry Enables Comprehensive Structure–Function Analysis of Dicer Helicase.

Cryo-EM analyses of AncD1D2 revealed mechanistic insights into translocation, but components of this mechanism were progressively lost in different Dicer-1 clades, eventually resulting in the decline of helicase function in hsDcr and ceDCR-1 (Fig. 6). AncD1_BILAT and AncD1_DEUT contain conserved motifs for hydrolyzing ATP (Q, I, II, III, and VI) and contacting dsRNA (Ia, Ib, Ic, IV, IVa, IVb, V, Va, and DUF283) despite the decline in helicase function. Integrating biochemical and structural insights with sequence conservation data enables identification of amino acid substitution events that may explain this decline (SI Appendix, Fig. S20).

Fig. 6.

Model of metazoan Dicer helicase evolution. AncD1D2 possessed helicase function: efficient ATP hydrolysis coupled to dsRNA binding and translocation. AncD2_ARTH retains helicase function, and core functions are passed down to extant dmDcr2 for antiviral defense and endo-siRNA processing. AncD1_BILAT retains ATP hydrolysis function but is diminished in dsRNA binding and translocation, presumably due to mutations at Hel2–Hel2i–pincer hinge (Inset), (SI Appendix, Fig. S20). Further degeneration at hinges and at dsRNA contact sites (yellow) lead to loss of helicase function in nematode clade. Deuterostome Dicer helicase accumulates hinge mutations, losing translocation function. AncD1_VERT helicase accumulates additional hinge mutations as well as dsRNA contact mutations leading to inability to form semiclosed conformation and hydrolyze ATP (SI Appendix, Fig. S20) (13). All models apart from AncD1D2 are AlphaFold 3 predictions (39).

Between AncD1D2 and AncD1_BILAT, substitutions at the Hel2–Hel2i–pincer hinge (227-SYN-229 and 495-KNK-497) appear to have driven early loss of helicase function. The KNK motif in AncD1D2’s pincer domain is substituted to MEK in AncD1_BILAT and further degenerated to LEA in AncD1_DEUT (Fig. 6 and SI Appendix, Fig. S20). Mutations in the equivalent KEK motif in mammalian RIG-I disrupt interferon signaling, underlining its importance to helicase function (31). Moreover, mutation of F225 at the dmDcr2 Hel2–Hel2i–pincer hinge compromises dsRNA binding but spares ATP hydrolysis (7, 30). Hel2i is too flexible to be visible in the endbound states A and B but is observed once AncD1D2 progresses into the dsRNA stem in states C-E, suggesting that the Hel2–Hel2i–pincer hinge enables helicase attachment to internal dsRNA segments (Fig. 4). The conserved contacts between pincer, Hel2, and Hel2i possibly mediate helicase function by enabling conformational changes important for helicase movement, and loss of these contacts in AncD1_BILAT and AncD1_DEUT may explain decline in dsRNA affinity and translocation. The inability of AncD1_DEUT to progress into internal dsRNA segments can be explained by an intact Hel1–Hel2 interface which is capable of ATP hydrolysis but is rendered unproductive by a nonfunctional Hel2–Hel2i–pincer hinge (Fig. 6 and SI Appendix, Fig. S14). Our cryo-EM structures indicate that ATP hydrolysis occurs before the completion of the translocation cycle during the relaxation step, supporting a model where AncD1_DEUT undergoes futile hydrolysis, uncoupled from unidirectional helicase movement (Fig. 5).

Alternatively, the loss of conservation at this Hel2–Hel2i–pincer hinge may reflect changes in the requirement for binding helicase cofactors. In dmDcr2, Loqs-PD, a cofactor that is important for dmDcr2 activity on endo-siRNA precursors, binds at the Hel2–Hel2i–pincer hinge (10, 28). The changes observed in AncD1_BILAT and AncD1_DEUT may reflect the evolution of unique binding interfaces for ancestral cofactors whose absence in our analysis would explain decline in helicase function. This caveat extends to all conclusions gleaned from our biochemical and structural assays, which may not mimic optimal in vivo conditions where these ancestral enzymes existed. AncD1_DEUT as well as its descendants, AncD1_VERT and hsDcr, also contain solvent-exposed loops inserted into Hel2 that may play a role in binding cofactors or inhibiting helicase function as Dicer-1 transitions into its role in regulating miRNA processing (Fig. 6) (3).

Further loss of helicase function along the nematode clade is explained by the degeneration of dsRNA contact motifs combined with loss of Hel2–Hel2i–pincer contacts, a phenomenon also observed in AncD1_VERT and hsDcr (SI Appendix, Fig. S20). However, unlike vertebrates which lost all Dicer helicase function and conceded their role in innate immunity to RLRs, nematodes developed a unique adaptation by recruiting a second helicase, DRH-1, to specifically aid in viral dsRNA translocation while maintaining miRNA processing (Fig. 3). Our analysis reveals multiple transitions in helicase evolution explaining how an intricate helicase mechanism that couples ATP hydrolysis and dsRNA binding to movement along dsRNA has progressively degraded during animal evolution to yield a diverse set of modern Dicers.

Materials and Methods

Detailed descriptions of procedures are in SI Appendix, Materials and Methods. Phylogenetic analysis was performed as described (13). Ancestral Dicer helicases and modern full-length Dicers were expressed in Sf9 insect cells. RNAs were procured from IDT. Proteins were purified by affinity, ion exchange or heparin chromatography, and gel filtration (13). Gel shift and ATP hydrolysis assays were as described (13). Streptavidin displacement assays are described in Materials and Methods and cryo-EM data collection, refinement, and validation protocols are in SI Appendix, Table S2.

Supplementary Material

Appendix 01 (PDF)

Movie S1.

Closure of Dicer helicase RecA domains is caused by ATP binding. Transition from State A to State B is depicted. Labels as in Fig. 5.

Movie S2.

ATP hydrolysis triggers helicase movement in the closed state. Transition from State B to State C is depicted. Interactions between V70 and H409 sidechains and phosphate backbone are depicted with dashed lines. Labels as in Fig. 5.

Movie S3.

Helicase transitions from post hydrolysis closed state to post hydrolysis semi-closed state. Transition from State C to State D is depicted. Labels as in Fig. 5.

Movie S4.

Flexibility of Hel2 loop as revealed from state D subclassification. Movement of Hel2 loop affects position of flanking motifs: H409 (IVa) and Q427 (IVb).

Movie S5.

Model of Dicer helicase function depicting complete coupling of ATPase cycle to translocation on dsRNA.

Data, Materials, and Software Availability

Density maps and coordinate data for states A, B, C, D, and E are deposited at the Electron Microscopy Data Bank (EMDB) and Protein Data Bank (RSCB PDB) under EMD accession codes EMD-48678 (40), EMD-48691 (41), EMD-48697 (42), EMD-48708 (43), and EMD-48710 (44), and PDB accession codes 9MW6 (45), 9MW7 (46), 9MW8 (47), 9MX3 (48), and 9MX5 (49), respectively. Ancestral sequences are in SI Appendix, Table S3. All other data are included in the manuscript and/or supporting information.