DCP1 forms asymmetric trimers to assemble into active mRNA decapping complexes in metazoa

Abstract

DCP1 stimulates the decapping enzyme DCP2, which removes the mRNA 5′ cap structure committing mRNAs to degradation. In multicellular eukaryotes, DCP1-DCP2 interaction is stabilized by additional proteins, including EDC4. However, most information on DCP2 activation stems from studies in S. cerevisiae, which lacks EDC4. Furthermore, DCP1 orthologs from multicellular eukaryotes have a C-terminal extension, absent in fungi. Here, we show that in metazoa, a conserved DCP1 C-terminal domain drives DCP1 trimerization. Crystal structures of the DCP1-trimerization domain reveal an antiparallel assembly comprised of three kinked α-helices. Trimerization is required for DCP1 to be incorporated into active decapping complexes and for efficient mRNA decapping in vivo. Our results reveal an unexpected connectivity and complexity of the mRNA decapping network in multicellular eukaryotes, which likely enhances opportunities for regulating mRNA degradation.

In eukaryotes, removal of the mRNA 5′ cap structure is catalyzed by the decapping enzyme DCP2 (1, 2); to be fully active and/or stable, DCP2 requires additional proteins (1, 2). Yeast DCP2 interacts directly with DCP1 and this interaction is required for decapping in vivo and in vitro (3–7). In humans, the DCP2-DCP1 interaction requires additional proteins, which together assemble into multimeric decapping complexes that also include the enhancers of decapping 3 and 4 (EDC3 and ECD4), and the DEAD-box protein DDX6/RCK (8, 9).

DCP2 is highly conserved and most information on DCP2 activation stems mainly from studies in S. cerevisiae and S. pombe (3–7). Fungi, however, lack EDC4 as well as many extensions and additional domains present in decapping activators of metazoan orthologs (8–10). For example, all eukaryotic DCP1 proteins contain an N-terminal EVH1 domain (3, 5, 6); however, DCP1 orthologs from metazoa and plants also have a proline-rich C-terminal extension (9, 10). The sequence of this extension is not conserved except for a 14-residue short motif (motif I, MI) conserved in metazoa with the exception of C. elegans (Fig. 1A and Fig. S1) and a C-terminal domain conserved in plants and metazoa (Fig. 1A, referred to as TD).

Fig. 1.

Structure of the DCP1-trimerization domain. (A) Domain organization of human DCP1a (HsDCP1a), indicating the EVH1 domain, motif I (MI), and the trimerization domain (TD). (B and C) Crystal structures of human DCP1a-TD and of the D. melanogaster orthologue, respectively. Chains A, B, and C are shown in green, blue, and red, respectively, with the N-termini of chains A and C in the front and the N terminus of chain B in the back. Residues from the hydrophobic core are shown as sticks. A total of 4,700 Ų is buried by the interfaces of the three chains in HsDCP1a-TD. (D and E) Asymmetric arrangement of the chains. Views down the two pseudodyads of HsDCP1a-TD, relating chains A and B or chains C and B, respectively. (F) Superposition and comparison of the three chains of HsDCP1a-TD with a close-up of the helix elbow. Interhelical angles are 86°, 95°, 86°, for chains A, B, and C, respectively. Strictly conserved side chains are shown as sticks.

The DCP1 C-terminal domain is predicted to adopt an α-helical conformation. In this work, we show that this domain trimerizes in an asymmetric fashion. We solved the crystal structure of the trimerized domain for both human and D. melanogaster DCP1 and show that the trimer adopts an unprecedented fold, with no current similarities in the protein database. We further show that DCP1 trimerization is required for the assembly of active decapping complexes and for mRNA decapping in vivo. The conservation of structurally critical residues indicates that this domain adopts a similar fold in DCP1 orthologs of other multicellular eukaryotes. Consequently, within mRNA decapping complexes in these organisms, the stoichiometry of the protein components is likely more complex than previously thought.

Results and Discussion

Crystal Structure of the Human DCP1a Trimerization Domain.

To investigate the role of the conserved C-terminal domain of human DCP1a in mRNA decapping we expressed the domain in E. coli (DCP1a residues S539 to L582). Using static light scattering measurements coupled with size exclusion chromatography, we found unexpectedly that the purified domain forms stable trimers in solution (Table S1). We have termed this domain the DCP1-trimerization domain (DCP1-TD). Furthermore, although the recombinant polypeptide contains 51 residues per monomer (i.e., 44 from DCP1a-TD and seven from the expression vector), NMR spectroscopy yields >115 peaks in the ¹⁵N-HSQC spectrum (Fig. S2), suggesting that in solution the trimers are asymmetric (assuming a single trimeric assembly).

We determined the crystal structure of this unusual assembly to a resolution of 2.3 Å (R_work = 20.8, R_free = 25.2, Table S2). The structure reveals an unprecedented, antiparallel bundle of three kinked α-helices (two up, one down), in which the structural environment for a given side-chain differs for each of the three molecules (Fig. 1 B–F). The central sequence (K544-L571) of each molecule is α-helical, with a strong kink at D558 that separates helix α1 from helix α2 with an elbow angle of ≈90° (Fig. 1F). At the kink, residue L554 from helix α1 makes van der Waals contacts with F561 from helix α2, while D558 caps the N terminus of helix α-2 by hydrogen-bonding with the peptide NH-groups of residues S560 and F561 (Fig. 1F). Alignment of DCP1 sequences from various species shows only these three residues are invariant (Fig. 2A, asterisks), indicating the elbow is a conserved structural feature of DCP1 trimerization domains from multicellular eukaryotes.

Fig. 2.

DCP1 trimerization occurs in vivo and is required for the assembly of active decapping complexes. (A) Structure-based alignment of the trimerization domain (TD) of DCP1 orthologs from Homo sapiens (Hs), Xenopus laevis (Xl), Danio rerio (Dr); Drosophila melanogaster (Dm), Aedes aegypti (Aa), Caenorhabditis elegans (Ce), Arabidopsis thaliana (At), and Oryza sativa (Os). Residues from the hydrophobic core of the trimerization domain are shaded blue, residues mediating conserved intrachain hydrogen bonds are shaded green. The portion of the sequence that is α-helical in all three chains of HsDCP1a-TD is indicated. Triangles mark mutated residues in mutant-1 (green), mutant-2 (red) and mutant-3 (blue). Gray letters indicate residues with variable conformation, likely to be affected by crystal packing. (B and C) DCP1 trimerization in vivo. GFP- and HA-tagged proteins were coexpressed in human cells as indicated. Cell lysates were immunoprecipitated using anti-GFP antibodies. GFP-tagged maltose binding protein (GFP-MBP) served as a negative control. Protein samples were analyzed by Western blotting using anti-HA and anti-GFP antibodies. (D) Lysates from D. melanogaster (Dm) S2 cells expressing HA-GST or HA-GST-DmTD together with GFP-tagged DmTD, were immunoprecipitated using anti-HA antibodies. Inputs and immunoprecipitates were analyzed as described above. The asterisk indicates cross-reactivity of the anti-HA antibody with an endogenous protein. (E) GFP-tagged proteins were expressed in human cells, as indicated. Cell lysates were immunoprecipitated using anti-GFP antibodies. The immunoprecipitates were tested for decapping activity using in vitro synthesized ³²P-labeled capped mRNA. (F) Samples corresponding to panel (E) were analyzed by Western blot to ensure that equivalent amounts of DCP1a wild-type and mutants were used for the decapping assay.

The three polypeptide chains (termed A, B, and C in Fig. 1 B–F) superimpose over the central sequence, with a maximal C_α r.m.s.d. of 1.15 Å (chain A and B, residues 544–571). To assemble the trimer, chain B interacts with chains A and C in an antiparallel fashion, causing helices α1 of chain A and α2 of chain C to interact in parallel. Chains A and B are thus related by a pseudotwofold axis close to F561 (Fig. 1D), while chains C and B are related by a pseudotwofold axis close to I552 (Fig. 1E). This arrangement places most hydrophobic side chains into a densely packed core (Figs. 1B and 2A, residues shaded in blue), explaining the stability of the trimer. The resulting DCP1a trimerization domain could be characterized as a novel fold if the chains were connected in cis.

Using the structural information, we constructed three mutants designed to either disrupt the interface with chain A (Mutant-1, L554S, F561R, L565S; Fig. 2A, green triangles), or chain C (Mutant-2, I552S, I555R, L562E; Fig. 2A, red triangles) or to completely prevent oligomerization (Mutant-3, L551R, I555S, F561R, L565S; Fig. 2A, blue triangles). Static light scattering shows all mutants remain monomeric (Table S1). Notably, we did not detect dimeric assemblies suggesting that the DCP1 C-terminal domain oligomerizes as trimer exclusively.

Crystal Structure of the Drosophila melanogaster DCP1 Trimerization Domain.

Despite the apparent stability of this unusual asymmetric homotrimeric assembly, we could not formally rule out the possibility that this particular protein sequence causes a unique artifact. Therefore, we crystallized the trimerization domain of D. melanogaster DCP1 (residues L328 to D366), which is only 36% identical to the human DCP1a trimerization domain (identity calculated over the central sequence of the TDs: HsDCP1a 544–571, DmDCP1 331–358; Fig. 2A). We found it also trimerizes with the same topology but in a different crystal packing environment and with two independent copies in the asymmetric unit (Fig. 1C, Fig. S3, and Table S2). Moreover, when D. melanogaster DCP1 is mutated at structural positions equivalent to those in human DCP1a mutants (Mut-1, Mut-2, and Mut-3), trimerization is again disrupted (Table S1). We therefore conclude that the C terminus of DCP1 contains a trimerization domain, which is conserved in metazoa but absent in fungi.

Functional Analysis of the Human and D. melanogaster DCP1 Trimerization Domain.

Further analysis showed DCP1 also oligomerizes in vivo. In human HEK293 cells, we coexpressed hemagglutinin (HA)-tagged DCP1a with green fluorescent protein (GFP)-tagged wild-type or mutant DCP1a, and then performed coimmunoprecipitations with anti-GFP antibodies. HA-DCP1a coimmunoprecipitated with GFP-DCP1a, but not with the negative control protein GFP-MBP (Fig. 2B, lane 9 vs. 8). As expected, the trimerization domain (GFP-DCP1a-TD) is sufficient for oligomerization (Fig. 2C, lane 6); while for DCP1a mutants 1, 2 and 3 oligomerization is strongly impaired (Fig. 2B, lanes 10–12). Accordingly, deleting the trimerization domain prevents DCP1a oligomerization (Fig. 2B, lane 13). Similarly, we confirmed that the trimerization domain of D. melanogaster DCP1 is also sufficient for oligomerization in vivo (Fig. 2D).

Because we were able to disrupt oligomerization in vivo using information from the crystal structure, we infer DCP1 in human and D. melanogaster cells exist as asymmetric trimers. Accordingly, the possibility that the observed DCP1a self-association is indirect and results from the incorporation of several monomeric DCP1a copies into larger decapping complexes can be ruled out, as it is inconsistent with the crystal structures and with the findings described below.

We next showed human DCP1a trimerization is required for assembly of active decapping complexes. We immunopurified GFP-DCP1a (wild-type and mutants) from HEK293 cells and, using an m⁷G-capped RNA substrate, tested for decapping activity in vitro. Decapping activity coimmunopurified with GFP-DCP1a as reported before (8, 10) (Fig. 2E, lane 3). This activity likely comes from the associated DCP2 (as a positive control, compare an immunoprecipitation of GFP-DCP2, Fig. 2E, lane 9) since adding nucleotide diphosphate kinase converts the m⁷GDP product to m⁷GTP (8, 10) (Fig. 2E, lanes 10 and 11). Strikingly, trimerization-defective mutants of DCP1a or DCP1a-ΔTD did not copurify with decapping activity (Fig. 2E, lanes 4–7), although the amounts of these proteins in the decapping assay were comparable to those of wild-type DCP1a (Fig. 2F).

Trimerization Is Required for DCP1a to Interact with DCP2 and EDC4.

Two scenarios can explain why DCP1a trimerization-defective mutants fail to copurify with decapping activity: either these mutants are not incorporated into DCP2-containing complexes, or they fail to stimulate DCP2 decapping activity. To discriminate between these possibilities, we examined the association of DCP1a with additional components of the decapping complex (i.e., EDC4, DCP2, EDC3, and DDX6/RCK). These all coimmunoprecipitated with DCP1a, as reported before (8, 11–14) (Fig. 3 A–D, lanes 9). DCP1-DCP2 association is likely stabilized by endogeneous EDC4 or other components (8, 9, 12, 14). We found DCP1a trimerization-defective mutants 1, 2, and 3 and DCP1a-ΔTD were strongly impaired in the interaction with EDC4 and DCP2 (Fig. 3 A and B, lanes 10–13). Therefore, the fact that DCP1a mutants that cannot trimerize also lack decapping activity can be explained by the failure to associate with DCP2 and EDC4. Notably, the trimerization domain itself is not sufficient to coimmunoprecipitate EDC4 or DCP2 (Fig. 3 E and F, lane 6). These results show that the oligomerization of DCP1-TD in vivo does not require an association with EDC4 or DCP2 and that additional sequences in DCP1 are necessary for binding EDC4 and DCP2.

Fig. 3.

Role of trimerization in the interactions of DCP1 with partner proteins. (A–F) GFP- and HA-tagged proteins were expressed in human cells as indicated above the panels. Cell lysates were immunoprecipitated using anti-GFP antibodies. GFP-MBP served as a negative control. Protein samples were analyzed by Western blotting using anti-GFP and anti-HA antibodies. DCP1a interaction with EDC4 (A). DCP1a interaction with DCP2 (B). DCP1a interaction with EDC3 (C). DCP1a interaction with DDX6/RCK (D). Interaction of DCP1a and DCP1a-TD with EDC4 (E) or DCP2 (F).

EDC3 and DDX6/RCK assemble with DCP1a independently of DCP2 and EDC4. Indeed, loss of trimerization does not affect the interaction of DCP1a with EDC3 and DDX6/RCK (Fig. 3 C and D, lanes 10–13). However, EDC3 and DDX6 do not interact with a DCP1a mutant that lacks the conserved motif MI (Fig. 3 C and D, lane 14), while DCP2 and EDC4 still do (Fig. 3 A and B, lane 14). Despite the lack of interaction with EDC3 and DDX6, DCP1a-ΔMI still oligomerizes (Fig. 2B, lane 14), demonstrating once again that the observed self-association of DCP1 is independent of other decapping factors and occurs by an asymmetric trimerization. Furthermore, immunopurified DCP1a-ΔMI complexes retain decapping activity in vitro, while immunopurified DCP1a-ΔTD complexes do not (Fig. 2E, lanes 8 vs. 7). Consequently, a minimal decapping complex consisting of DCP1, DCP2 and EDC4 may be sufficient for decapping activity.

Specificity of DCP1 Trimerization In Vivo.

Human cells express two DCP1 orthologs, DCP1a and DCP1b, whose trimerization domains exhibit 50% identity over the central sequence (HsDCP1a 544–571, HsDCP1b 580–607; Fig. 2A). It was therefore of interest to investigate whether human DCP1a and DCP1b heteromerize. We observed that HA-DCP1b coimmunoprecipitated with GFP-DCP1a, indicating that DCP1a and DCP1b can indeed heteromerize (Fig. 4A, lane 6). Consistently, we found heteromerization required the trimerization domain (Fig. 4A, lane 7).

Fig. 4.

Specificity of DCP1 trimerization. (A–C) GFP- and HA-tagged proteins were expressed in human cells as indicated above the panels. Cell lysates were immunoprecipitated using anti-GFP antibodies. GFP-MBP served as a negative control. Protein samples were analyzed by Western blotting using anti-GFP and anti-HA antibodies.

To investigate the specificity of this interaction further, we substituted the trimerization domain of human DCP1a with the equivalent domain from D. melanogaster DCP1 and examined whether the chimeric protein (DCP1a-DmTD) oligomerized and interacted with wild-type DCP1a and DCP1b. As expected, the chimeric protein homomerized, indicating that D. melanogaster DCP1-TD trimerizes independently of the additional flanking sequences (Fig. 4B, lane 8). In contrast, we observed no interaction between the chimeric DCP1a-DmTD and DCP1a or DCP1b (Fig. 4 A and C, lane 8, and B, lane 6). Thus, heteromerization is observed only between highly related sequences. Furthermore, BLAST searches using the trimerization domains of DCP1a and DCP1b failed to identify any significantly similar sequence in the human genome. Thus, we conclude that DCP1a and DCP1b are unlikely to heteromerize with alternative, unrelated proteins.

Trimerization Enhances DCP1a Accumulation in P-Bodies.

DCP1a and partners localize to P-bodies (8, 9, 15–17). We therefore tested the localization of DCP1a mutants, to define what interactions are critical for P-body accumulation. As reported (8, 9), GFP-DCP1a accumulates in cytoplasmic foci corresponding to endogenous P-bodies (Fig. 5A). Deleting the conserved motif MI did not affect this pattern (Fig. 5B). In contrast, the trimerization-deficient mutants and DCP1a-ΔTD dispersed throughout the cytoplasm, although some signal was still detectable in P-bodies (Fig. 5 C and D and Fig. S4). The isolated trimerization domain did not localize to P-bodies (Fig. 5E). Importantly, the chimeric DCP1a protein containing the D. melanogaster trimerization domain did localize to P-bodies in human cells (Fig. 5F). We conclude that trimerization per se, independently of the sequence of the trimerization domain, is required for DCP1a to efficiently accumulate in P-bodies.

Fig. 5.

Trimerization enhances accumulation of DCP1a in P-bodies. (A–F) Representative confocal fluorescent micrographs of fixed human HeLa cells expressing wild-type GFP-DCP1a or the mutants indicated on the left. Cells were stained with antibodies cross-reacting with EDC4 and a nuclear human antigen (23). The merged images show the GFP signal in green and the EDC4 signal in red. The fraction of cells exhibiting a staining identical to that shown in the representative panel was determined by scoring at least 100 cells in three independent transfections performed per protein. (Scale bar, 10 μm.)

DCP1 Trimerization Is Important for Efficient Decapping in Vivo.

To test the biological significance of the identified DCP1 trimerization domain and its functional relevance in mRNA decapping we established a complementation assay in D. melanogaster cells, which express a single DCP1 paralog. In the complementation assay, endogenous DCP1 was depleted using a dsRNA targeting the DCP1 ORF. A dsRNA targeting GFP was used as a negative control. Wild-type DCP1 and a DCP1 mutant lacking the trimerization domain were then tested for the ability to restore decapping in DCP1-depleted cells. Transcripts encoding the proteins were made resistant to the dsRNA by introducing mutations that disrupt base pair interactions with the dsRNA without altering the protein sequence.

To assay mRNA decapping we monitored miRNA-mediated mRNA degradation. We previously showed that miRNA targets are degraded by exonucleolytic digestion of the mRNA body only after they are first deadenylated and then decapped (18, 19). Therefore, we could prevent miRNA targets from being degraded by inhibiting decapping (18, 19). In our assay, we used the F-Luc-CG3548 reporter consisting of the firefly luciferase (F-Luc) ORF fused to the 3′ UTR of the D. melanogaster gene CG3548, which is silenced by miR-12 (19). Each transfection mixture included either a plasmid encoding the primary miR-12 transcript or the corresponding control vector without an insert, plus a transfection control plasmid expressing Renilla luciferase (R-Luc).

Our results showed the F-Luc-CG3548 mRNA is degraded in a miR-12-dependent manner (Fig. 6A, lane 2 vs. 1). Depleting DCP1 did not restore reporter mRNA levels. This result is expected since we previously showed that, in S2 cells, at least two decapping activators must be codepleted to restore the levels of miRNA reporters (19). Accordingly, codepletion of endogenous DCP1 plus EDC4 from S2 cells strongly inhibited decapping of the F-Luc-CG3548 reporter, leading to the accumulation of F-Luc-CG3548 mRNA as a fast migrating deadenylated form (Fig. 6A, lane 4). We therefore performed the complementation assay in this sensitized background.

Fig. 6.

Trimerization is required for efficient decapping in vivo (A–C) D. melanogaster S2 cells were treated with dsRNA targeting the ORF of DCP1 and EDC4 mRNAs. Control cells were treated with GFP dsRNA. These cells were subsequently transfected with a mixture of three plasmids: one expressing the F-Luc-CG3548 reporter; another expressing miR-12 primary transcripts (+miR-12) or the corresponding empty vector (-); and a third expressing Renilla luciferase (R-Luc). Plasmids (5 ng) encoding wild-type HA-DCP1, HA-DCP1ΔTD or HA-MBP were included in the transfection mixtures, as indicated. RNA samples were analyzed by Northern blot. Firefly luciferase mRNA levels were normalized to those of the Renilla luciferase. For each condition, the normalized values of F-Luc mRNA were set to 100 in the absence of miR-12. (B) Shows that HA-DCP1, HA-DCP1ΔTD and HA-MBP are expressed at comparable levels. (C) The decay of F-Luc-CG3548 mRNA was monitored in the presence (black lines) or absence (red lines) of miR-12 in depleted cells following inhibition of transcription by actinomycin D. The levels of the F-Luc-CG3548 mRNA were normalized to rp49 mRNA and plotted against time. mRNA half-lives (t_1/2) calculated from the decay curves are indicated in minutes.

We thus tested whether in cells codepleted of DCP1 and EDC4, we could restore mRNA degradation by expressing wild-type DCP1 or the DCP1 mutant lacking the trimerization domain. We restored target mRNA degradation in cells depleted of the endogenous DCP1 by expressing the dsRNA-resistant version of wild-type DCP1 (Fig. 6A, lane 6). In contrast, the DCP1-ΔTD mutant was impaired in restoring mRNA degradation indicating that decapping is still inhibited in these cells (Fig. 6A, lane 8). DCP1 and DCP1-ΔTD were expressed at comparable levels (Fig. 6B). Notably, in control cells, neither DCP1 nor DCP1-ΔTD inhibited mRNA degradation in a dominant negative manner (Fig. S4). Finally, the changes in steady state levels of F-Luc-CG3548 mRNA in cells depleted of DCP1 and EDC4 were accompanied by corresponding changes in the half-life of the mRNA (Fig. 6C). We conclude that DCP1 trimerization is required for efficient mRNA decapping in vivo.

Concluding Remarks.

In this study, we identified a protein trimerization domain in the DCP1 protein. In metazoa, this domain is physiologically important for DCP1 to be incorporated into active mRNA decapping complexes and for mRNA decapping in vivo. This is remarkable, not only because trimers are rare among homomeric protein assemblies, but particularly because such an entirely asymmetric arrangement has not been described before in a physiological context (20).

Trimerization of DCP1 reveals an unexpected connectivity and complexity of the decapping network in multicellular eukaryotes, as both EDC3 and EDC4 are known or presumed to form homodimers (11, 13, 15, 21). We find that DCP1a can interact with DCP2 and EDC4 independently of the interaction with EDC3 and DDX6/RCK. Therefore, we can now conceive of multimeric assemblies consisting of (i) DCP1, EDC4, and DCP2, (ii) DCP1, EDC3, and DDX6 or (iii) combinations of both. Presently, we do not know precisely how DCP1 trimers associate with EDC3 and EDC4 dimers and why DCP1 oligomerizes in a nonsymmetric fashion. It shall be challenging to determine the structural details of such assemblies and their possible significance for regulating mRNA decapping or for promoting P-body formation.

Methods

Detailed experimental procedures are given in the SI Text. Briefly, the trimerization domains of human DCP1a (residues 539–582) and D. melanogaster DCP1 (residues 328–366), N-terminally tagged with GST, were expressed in the E. coli strain BL21 Gold (DE3) (Stratagene) at 25 °C overnight. The proteins were purified by glutathione affinity step (GSTrap HP column; GE Healthcare). After cleaving the GST with PreScission protease, the complexes were further purified by gel filtration (HiLoad 26/60 Superdex 75 columns; GE Healthcare). The structure of the human protein was solved using single-wavelength anomalous dispersion data, collected from a crystal with selenomethionine-substituted protein. The structure of the D. melanogaster protein was solved by molecular replacement using the structure of the human protein as search model. Coimmunoprecipitation assays and immunofluorescence were performed as described in ref. 22. Decapping assays were performed as described in ref. 10, with the modifications indicated in the SI Text.