eIF1A augments Ago2-mediated Dicer-independent miRNA biogenesis and RNA interference

MicroRNA (miRNA) biogenesis and miRNA-guided RNA interference (RNAi) are essential for gene expression in eukaryotes. Here we report that translation initiation factor eIF1A directly interacts with Ago2 and promotes Ago2 activities in RNAi and miR-451 biogenesis. Biochemical and NMR analyses demonstrate that eIF1A binds to the MID-domain of Ago2 and this interaction does not impair translation initiation. Alanine mutation of the Ago2-facing Lys56 in eIF1A impairs RNAi activities in human cells and zebrafish. The eIF1A-Ago2 assembly facilitates Dicer-independent biogenesis of miR-451, which mediates erythrocyte maturation. Human eIF1A (heIF1A), but not heIF1A(K56A), rescues the erythrocyte maturation delay in eif1axb knockdown zebrafish. Consistently, miR-451 partly compensates erythrocyte maturation defects in zebrafish with eif1axb knockdown and eIF1A(K56A) expression, supporting a role of eIF1A in miRNA-451 biogenesis in this model. Our results suggest that eIF1A is a novel component of the Ago2-centered RNA induced silencing complexes (RISCs) and augments Ago2-dependent RNAi and miRNA biogenesis.

microRNAs (miRNAs) are ~22 nucleotide (nt) endogenous noncoding RNAs involved in gene expression regulation. Their genes are usually transcribed by RNA polymerase II or III 1,2 . The resulting primary miRNAs (priRNAs) contain characteristic hairpins, which are excised in the nucleus by Drosha/DGCR8 to yield the pre-miRNAs 3,4 . Subsequently, the Exportin-5/Ran-GTP complex translocates the pre-miRNAs to the cytoplasm 5,6 , where they are engaged by DICER to form a RISC Loading Complex (RLC). RLC includes TRBP and Ago2 [7][8][9] , while ADAR1 facilitates pre-miRNA loading 10 . In the canonical miRNA biogenesis pathway, DICER removes the terminal loop region to yield the mature miRNA [11][12][13] . Recent studies revealed that miR-451 is produced by an alternative DICERindependent pathway, where pre-miR-451 is loaded directly onto Ago2 and sliced on the 3′ hairpin arm, as guided by the 5′ end of the hairpin, yielding a 30nt cleaved species [14][15][16] . A 3′ resection activity by PARN trims ~ 7-nt to produce the 23-nt miR-451 17 . However, the mechanisms of DICER-independent miRNA biogenesis have remained elusive. miRNAs are incorporated into effector ribonucleoprotein complexes called RNA induced silencing complexes (RISCs) to exert RNA interference. RISCs can suppress translation without mRNA degradation, destabilize mRNAs by deadenylation or directly degrade mRNAs by Ago2 slicer activity 12,[18][19][20] , and inhibit translation at the stage of initiation or elongation 21,22 . The core of RISCs is a member of the Ago protein family 23 . In mammals, this family consists of four members (Ago1-4). Although these four Ago proteins can suppress translation of their target mRNAs, only Ago2 (also known as eIF2C2) is endonucleolytically active 19,20 . Ago2 is composed of four globular domains (N, PAZ, MID and PIWI) and two structured linker domains (L1 and L2) 24 . MID and PAZ domains harbor the 5′-phosphate and 3′-hydroxyl termini of miRNA, respectively 25,26 , while the PIWI domain is responsible for the cleavage-slicer-activity 27,28 . Crystal structures of human Ago2 revealed that the 5′-end of miRNA is tethered to Argonaute through interactions with a binding pocket that is mostly formed by the MID domain 29,30 .
Eukaryotic translation initiation factor eIF1A is composed of a globular domain (GD) and two unstructured tails (N-and C-terminal tail), which are absent in prokaryotic initiation factor-1 (IF1). The eIF1A globular domain consists of an IF homologous oligonucleotide/ oligosaccharide binding (OB) fold, and an additional subdomain C-terminal to the OB fold that contains two alpha helices and two extended strands, which pack to the helix on opposite sites. The globular domain contains a large RNA binding face 31 , which is directed towards the ribosome 32,33 Here we find that the GD of eIF1A, which is conserved from IF1 to eIF1A, directly binds to Ago2, and that the MID-domain of Ago2 interacts with eIF1A but does not affect eIF1A's function in translation initiation. eIF1A forms a complex with Ago2 and promotes Ago2 cleavage activity in RNAi, and enhances Ago2 activity in miR-451 production in human cells and the developing zebrafish. The results support the notion that eIF1A is a component of the Ago2-centered RISCs.

eIF1A interacts with Ago2
To identify new components of the protein networks that participate in DICER-independent miRNA biogenesis and Ago2-mediated RNAi, we generated a human embryonic kidney 293 (HEK293) cell line that stably expresses Flag-HA (haemagglutinin)-Ago2 for immunoprecipitation analyses. Eukaryotic translation initiation factor eIF1A is a previously unrecognized Ago2-interaction protein that is immunoprecipitated by Ago2 in cell lysates and colocalizes with Ago2 in mammalian cells ( Fig. 1a-b). To test the specificity of the eIF1A-Ago interaction, we tested for possible interactions with other members of the Ago family using HEK293 cells stably expressing Ago 1-4 34 . Transiently overexpressed eIF1A showed interaction with Ago2, but not Ago1, Ago3 or Ago4 (Fig. 1c).

eIF1A binds to Ago2 in an RNA-independent manner
To identify which of the individual domain(s) of Ago2 interacts with eIF1A, we examined the interaction of recombinant-RNase A and Benzonase-treated-L1, PAZ, MID and PIWI domains of Ago2 with GST-tagged-eIF1A in pull down assays. RNA free eIF1A specifically interacted with the RNA free MID-domain ( Fig. 2a and Supplementary Fig. 1a). To further characterize the interaction between human eIF1A fragments and Ago2, we screened recombinant GST-tagged eIF1A and eIF1A mutants (N-tail deletion, C-tail deletion, and globular domain, Supplementary Fig. 1b-c) for interaction with Ago2. GST-tagged full length, N-tail deletion (ND), C-tail deletion (CD) and globular domain (GD) of eIF1A interacted with Ago2 but not the GST-tag alone, indicating that the eIF1A-GD is required and sufficient for the interaction with Ago2 (Fig. 2b). We further investigated the interaction between RNase A and Benzonase-treated eIF1A and the MID-domain by Nuclear Magnetic Resonance (NMR) spectroscopy. Unlabeled, Sumo-tagged MID domain titrated into a 15 Nlabeled sample of eIF1A at molar ratio of 1:1 (eIF1A:SMT3-MID) (Fig. 2c) resulted in broadening of eIF1A resonances characteristic of the complex being in an intermediate exchange regime. The resonances of the eIF1A-GD experience more broadening when compared to the unstructured tails. For instance, resonances of Val55, Lys56 and Lys67 residues of 15 N-eIF1A-GD show broadening upon MID domain titration (Fig. 2c). These results suggest that the interaction of eIF1A and Ago2 primarily involves the GD of eIF1A. Furthermore, when unlabeled RNA free eIF1A was titrated to the RNA free 15 N-MIDdomain at molar ratio of 1:1, we observed broadening of the MID resonances indicative of direct binding between the two proteins (Fig. 2d). We verified that eIF1A does not bind to the SMT3-tag (strong signals in Fig. 2d). These NMR data demonstrate that eIF1A directly interacts with Ago2 in an RNA independent manner.

Ago2 interaction does not impair eIF1A translation functions
Based on the observed spectral changes (Fig. 2c) we generated three eIF1A mutants: V55A, K56A and K67A. To test whether the mutants are folded we recorded a TROSY HSQC spectrum of eIF1A K56A . The spectrum clearly shows that eIF1A K56A is properly folded ( Supplementary Fig. 2). To examine the roles of Val55, Lys56 and Lys67 of eIF1A in full length eIF1A-Ago2 interaction, we generated HEK293 cell lines stably expressing single eIF1A mutants (V55A, K56A or K67A) involved in Ago2-eIF1A interaction. As determined by immunoprecipitation, all three mutations reduced eIF1A-Ago2 interaction (Fig. 3a), suggesting that these amino acids in eIF1A are important for the interaction between Ago2 and full-length eIF1A. In an in vitro dual luciferase translation assays, elevated eIF1A K56A increases cap-dependent and IRES-dependent translation, to a similar degree as wild type eIF1A (Fig. 3b), suggesting that the K56A mutation does not compromise the activity of eIF1A in translation initiation. In addition, eIF1A K56A does not repress polysome complex formation activities in polysome profiling analyses (Fig. 3c), indicating that the K56A mutation does not generally impair translation.. Taken together, the data suggests that eIF1A directly interacts with Ago2 in an RNA-binding-independent manner, the MID-domain of Ago2 binds to the GD of eIF1A, and that the interaction between eIF1A and Ago2 does not impair eIF1A functions in translation initiation.

eIF1A promotes Ago2-mediated miRNA-guided RNAi in vitro
To examine whether eIF1A plays a role in Ago2-mediated RNAi, we employed a human HMGA2 3′-UTR mutant which harbors a characterized native let-7 target site 35 (Fig. 4a). This assay evaluates the dependence of eIF1A and its mutants on let-7a/Ago2 mediated cleavage of the target mRNA. In the RNAi in vitro assay reactions with added eIF1A or eIF1A mutants without Ago2, no cleavage product was observed, suggesting that the cleavage is Ago2 dependent (Fig. 4b). Interestingly, we found that high concentrations of eIF1A increased HMGA2 3′-UTR mutant cleavage while eIF1A mutants (V55A, K56A and K67A) significantly reduced the level of cleavage products (Fig. 4b-c), indicating that eIF1A augments Ago2-mediated RNAi activities in vitro. To investigate the effects of eIF1A in Ago2-mediated miRNA-guided RNAi activities in cells, we used green fluorescent protein (GFP) reporters containing let-7 targeted sequences 36 . Because transient knockdown of eIF1A in HEK293 cells led to significant decrease of cell viability ( Supplementary Fig.  3), we employed specific eIF1A-Ago2 interaction interference with eIF1A K56A . Increased eIF1A levels elevated let-7 RISC activity, while eIF1A (K56A) mutant compromised let-7 RISC activity ( Fig. 4d-e). Elevated eIF1A and Ago2 decrease cellular GFP mRNA levels as determined by qPCR (Fig. 4f). In contrast, eIF1A K56A caused increased cellular GFP mRNA levels (Fig. 4f). These results suggest that eIF1A forms a complex with Ago2 and promotes Ago2-mediated RNAi in human cells in vitro.

eIF1A promotes Ago2-mediated miRNA-guided RNAi in vivo
To further evaluate the roles of eIF1A on Ago2-mediated RNAi in an animal model, we employed zebrafish. Both zebrafish paralog eIF1AXa and eIF1AXb interact with Ago2 ( Supplementary Fig. 4a-b). eif1axb mRNA level is much higher (> 25-fold) than that of eif1axa in the developing zebrafish ( Supplementary Fig. 4c-d) therefore we focused on eif1axb here. We employed a lower dosage of eif1axb Morpholino (MO, 62.5μM, 1nl), which does not significantly affect zebrafish phenotypes at 18 h.p.f and 48 h.p.f ( Supplementary Fig. 4e-f). Previous studies reported that miR-126 targets c-Myb and suppresses c-Myb expression in zebrafish 37 . We next examined if eIF1A plays a role in Ago2-mediated miR-126-guided c-Myb suppression in zebrafish at 28 h.p.f. Zebrafish treated with eif1axb MO (62.5μM, 1nl) showed increased c-Myb in comparison to the control group (Fig. 4g). Zebrafish with treatment of eif1axb MO (62.5μM, 1nl) plus human eIF1A (heIF1A, 1μg/μl mRNA, 1nl) presented decreased c-Myb levels compared to the group treated with only eif1axb MO (62.5μM, 1nl). In contrast, zebrafish treated with human eIF1A K56A (1μg/μl mRNA, 1nl) and eif1axb MO (62.5μM, 1nl) show increased c-Myb (Fig.  4g). In addition, heIF1A K56A increased c-Myb mRNA accumulation in zebrafish (Fig. 4h), indicating that interfering with eIF1A-Ago2 interaction impairs the miR-126-mediated suppression of c-Myb expression in zebrafish, even though the level of miR-126 remain is not significantly decreased under these conditions (Fig. 4i). These results demonstrated that eIF1A-Ago2 complex promotes miR-126 RISC activities in vivo. Overall the above results demonstrate that eIF1A augments Ago2-mediated miRNA-guided RNAi both in vitro and in vivo.

Discussion
Ago2-mediated miRNA biogenesis and processing are essential for the development in eukaryotes. We found that eIF1A directly binds to Ago2 in an RNA-independent manner, and plays a role in both Ago2-mediated miRNA-guided RNAi and miRNA biogenesis (Fig.  6). Using biochemistry and NMR analyses, we demonstrated that MID-domain of Ago2 binds to the globular domain of eIF1A. eIF1A promotes Ago2-mediated RNAi and miR-451 biogenesis in vitro and in vivo. eIF1A augments erythrocyte maturation in zebrafish through mediating miR-451 surveillance. The previously unrecognized roles of eIF1A in Ago2mediated miRNA processes identified here provide insights into the understanding of the mechanisms of miRNA-guided RNAi and miRNA biogenesis.
Interestingly, point mutation of K56A in eIF1A leads to dissociation of eIF1A-Ago2 interaction. The eIF1A (K56A) mutant does not impair translation initiation. In contrast, eIF1A(K56A) impairs Ago2-mediated miRNA guided RNAi and Ago2-dependent miR-451 production in vitro and in zebrafish in vivo. Consistently, miR-451 partly reverts reduced hemoglobinization and delayed maturation of erythrocytes in eif1axb knockdown zebrafish with overexpressed heIF1A (K56A). These data demonstrate that eIF1A-Ago2 interaction promotes Ago2-mediated RNAi and miRNA production but does not compromise eIF1A functions in translation initiation.
Recently, groups of Ameres et al. 40 , Cheloufi et al. 14 , and Cifuentes et al. 15 have reported that recombinant Ago2 is sufficient for slicer activity with large quantities of RNA fragments in RNAi and miR-451 processing in vitro. These data elegantly evidenced that Ago2 is a central effector of miR-451 generation and RNAi processes. On the other hand, many studies demonstrated that the Ago2-centered RISC complex is necessary for efficient RNAi activity with long and/or structured mRNAs and miR-451 processing in vivo 23,41 . It is possible that other components in RISC are needed to resolve mRNA secondary structure, to scan the long and/or structured mRNAs for recognition of miRNA seed regions in low abundance, and to generate mature miR-451. Currently, the component list of RISC is increasing. Here we show that eIF1A directly binds to Ago2 and augments miR-451 biogenesis and RNAi processes, suggesting that eIF1A is an important but previously unrecognized factor of RISC.
MID-domain of Ago2 is essential for miRNA docking in RISC mediated RNAi and pre-miR-451 loading in Dicer-independent miR-451 biogenesis. Here, we found that eIF1A directly binds to MID-domain of Ago2 and eIF1A promotes miRNA-guided RNAi and miR-451 generation.
Our findings reveal that eIF1A directly interacts with Ago2 and augments Ago2-mediated miRNA-guided RNAi and DICER-independent miRNA biogenesis. The newly identified eIF1A-Ago2 complex together with its functions in miRNA processes provides insights in understanding how Ago2 mediates miRNA processes in translation, development and diseases.

Immunoprecipitation, MS and Western blot assays
The stable HEK293 cell line expressing Flag-HA (as a control) or Flag-HA-Ago2 were lysed with lysis buffer containing 20mM Tris HCl pH7.4, 137mM NaCl, 10% glycerol, 1% NP-40, EDTA-free protease inhibitor (Roche) followed by 10μg/ml Ribonuclease A treatment (25°C, 1hr). The Ago2 interacted proteins were immunoprecipitated by anti-HA monoclonal antibodies (Cell Signaling, 2367S, 20μl for immunoprecipitation assay) immobilized to Dyna beads. The purified proteins were analyzed by silver staining and Mass Spectrometry via LC-MS/MS on an LTQ Orbitrap Velos mass spectrometer (Thermo Scientific, German) equipped with a Thermo Fisher Scientific nanospray source, an Agilent 1100 Series binary HPLC pump, and a Famos autosampler. The spectral data were searched with SEQUEST against a database containing the human International Protein Index (IPI) protein sequence database: (http://www.ebi.ac.uk/IPI/) together with the reversed complement. eIF1A was transiently overexpressed in stable HEK293 cell lines expressing Agos (Ago1~ 4). The cell lysates were treated with 10μg/ml Ribonuclease A (25°C, 1hr).

Endogenous miRNA-guided RISC activity assays
To read miRNA-guided RISC activity, pcDNA-GFP-let 7 (complementary sequence of miRNA let-7, 5′-AACTATACA ACCTACTACCTCA-3′) were generated in HEK293 cells as previously described 36 . The pcDNA-GFP-let 7 was transfected into HEK293 cells followed by 1.6mg/ml G418 selection for two weeks. The resistant colonies were propagated into stable cell lines. For let-7 RISC activity assay, 2μg of expression constructs were transfected into 5×10 5 cells of the GFP-reporter stable cell lines in six-well plates, the cells were transferred to 10cm plates after 24h. After 36hr, total proteins were extracted with standard RIPA buffer together with protease inhibitor (Roche) and total RNAs were extracted using Trizol reagents (Invitrogen). Control Morpholino (MO) or 62.5μM eif1axb MO plus 1μg/μl human eIF1A or mutant mRNAs were injected into one-cell stage zebrafish embryos. Total RNAs and proteins were isolated at 48 h.p.f. for the analyses of c-Myb proteins and mRNAs.

In Vitro Translation Assays
The dicistronic reporter construct, containing the Renilla luciferase sequence after the 5′ UTR followed by the CrPV IRES and the firefly luciferase sequence, has been previously described 42 . The reporter construct plasmid was linearized with BamHI (New England Biolabs) and transcribed in vitro with an ARCA cap and poly (A)-tailed using the mMessage Machine T7 Ultra Kit (Ambion). The mRNAs were purified with RNeasy Plus Mini Kit (Qiagen) with RNase-free H 2 O. Human eIF1A and mutants (V55A, K56A and K67A, with anti-Flag antibodies, 0.15mg/ml, 2μl per reaction per reaction) were purified with Dyna beads (Invitrogen) from HEK293 cells with transient transfection of plasmids encoding Flag-tagged eIF1A and eIF1A mutants. In vitro translation reactions (50μl per reaction) were carried out using Rabbit Reticulocyte Lysate (Promega, 35μl per reaction) with 2mM magnesium acetate and 60mM potassium acetate, which was incubated at 30°C for 100 min. Translation of the reporter genes was measured using the Dual-Glo luciferase assay (Promega). Three independent experiments were performed.

O-Dianisidine staining
O-dianisidine (Sigma) solution at 0.7 mg/ml was prepared in fresh ethanol and protected from exposure to light. The staining solution was prepared by mixing 2 ml of water, 2 ml of 0.7 mg/ml O-dianisidine solution, 0.5 ml of 100 mM sodium acetate, and 100 μl 3% hydrogen peroxide. PTU treated 48 h.p.f embryos were transferred to 12-well plates and 1 ml of the staining solution was added. Stained embryos were kept in the dark for 15 minutes, washed three times with 1 × PBS, and fixed with 4% paraformaldehyde.

May-Grünwald/Giemsa staining
Erythrocytes collection of 20 zebrafish (56 h.p.f) of each group was performed by cutting the zebrafish tails. Circulating erythrocytes were collected with 1 × PBS with 10% FBS with cytospin. The erythrocytes on slides were stained with May-Grünwald/Giemsa solutions (Polysciences), imaged and photographed using a Nikon 80i Upright Microscope. The pixel area of the nucleus and cytoplasm was quantified for 100-145 cells per sample in three independent experiments using ImageJ software and the nuclear: cytoplasmic ratio calculated for each cell.

Image acquisition
In O-Dianisidine staining assays, stained embryos were imaged and photographed using a Leica M80 Microscope with a Nikon D200 digital camera using an adjustable flash system. Erythrocytes with May-Grünwald/Giemsa staining were imaged by Nikon 80i Upright Microscope in Nikon Image Center at Harvard Medical School. Fluorescence and/or DAPI stained images were taken with a Nikon Ti Inverted Fluorescence Microscope with Perfect Focus System in Nikon Image Center at Harvard Medical School. Images of zebrafish embryos were taken by a Leica M80 Microscope with a Nikon D200 digital camera.

Supplementary Material
Refer to Web version on PubMed Central for supplementary material.     The primary miRNA (priRNA) is cleaved by Drosha-DGCR8 pathways to generate pre-miRNA within the nucleus. Exportin-5 in complex with Ran-GTP exports the pre-miRNA to the cytoplasm, where the pre-miRNA is bound by DICER to form a RISC Loading Complex that includes Ago2. In the canonical miRNA biogenesis pathway, DICER removes the terminal loop region to yield the mature miRNA. The pre-miR-451 is loaded directly onto Ago2 and sliced on the 3′ hairpin arm. The miR-144 biogenesis is DICER-dependent. The globular domain (GD) of eIF1A directly binds to MID-domain of Ago2 and forms an eIF1A-Ago2 complex promoting Ago2-mediated RNAi and miR-451 biogenesis. MID-domain of Ago2 binds to the GD of eIF1A and does not impair eIF1A functions in translation initiation. The 5′-term of guide strand of miRNA-miRNA* duplex is docked onto a pocket with residues mainly from MID-domain. The long and structured mRNAs are scanned by RISCs to recognize seed regions in miRNAs. After perfect or imperfect complementary guide with miRNAs, mRNAs are nicked by PIWI domain resulting in RNAi. eIF1A augments Ago2-mediated RNAi, miR-451 production and erythrocyte maturation. Black arrows are from previous reports of other groups 14,15,43 ; blue arrows are from this study.