RNA binding proteins Smaug and Cup induce CCR4–NOT-dependent deadenylation of the nanos mRNA in a reconstituted system

Abstract Posttranscriptional regulation of the maternal nanos mRNA is essential for the development of the anterior – posterior axis of the Drosophila embryo. The nanos RNA is regulated by the protein Smaug, which binds to Smaug recognition elements (SREs) in the nanos 3’-UTR and nucleates the assembly of a larger repressor complex including the eIF4E-T paralog Cup and five additional proteins. The Smaug-dependent complex represses translation of nanos and induces its deadenylation by the CCR4–NOT deadenylase. Here we report an in vitro reconstitution of the Drosophila CCR4–NOT complex and Smaug-dependent deadenylation. We find that Smaug by itself is sufficient to cause deadenylation by the Drosophila or human CCR4–NOT complexes in an SRE-dependent manner. CCR4–NOT subunits NOT10 and NOT11 are dispensable, but the NOT module, consisting of NOT2, NOT3 and the C-terminal part of NOT1, is required. Smaug interacts with the C-terminal domain of NOT3. Both catalytic subunits of CCR4–NOT contribute to Smaug-dependent deadenylation. Whereas the CCR4–NOT complex itself acts distributively, Smaug induces a processive behavior. The cytoplasmic poly(A) binding protein (PABPC) has a minor inhibitory effect on Smaug-dependent deadenylation. Among the additional constituents of the Smaug-dependent repressor complex, Cup also facilitates CCR4–NOT-dependent deadenylation, both independently and in cooperation with Smaug.


INTRODUCTION
Poly(A) tails decorate the 3' ends of almost all eukaryotic mRNAs. Long poly(A) tails are appended to newly made mRNAs in the cell nucleus (1)(2)(3)(4)(5)(6) and then gradually shortened by 3' exonucleases in the cytoplasm (3,(7)(8)(9). Poly(A) tail shortening (deadenylation) can serve two regulatory purposes: First, deadenylation initiates the decay of most mRNAs and is a prerequisite for their complete degradation, which occurs mostly by hydrolysis of the 5' cap followed by 5'-3' degradation (3,10,11). The rates of deadenylation are specific for different mRNAs and are the major determinants of mRNA half-life (3,9,12). Second, deadenylation can contribute to the regulation of translation. Poly(A) tails stimulate the initiation of translation via a protein-mediated interaction with the mRNA 5' end (13,14). During oocyte maturation and early embryonic development of animals, it is not just the presence but the length of the poly(A) tail that affects the rate of translation, as initially shown by studies of individual mRNAs in several species (15). Transcriptome-wide analyses confirmed a strong correlation between long tails and high translation efficiency at early embryonic stages, but not in nonembryonic cells (16)(17)(18)(19)(20)(21)(22). In oocyte maturation and early development, regulated extension or shortening of poly(A) tails is used as a means for translational activation or inactivation, respectively (15,23).
Post-transcriptional regulation, by changes in poly(A) tail length or other means, is essential during early embryonic development of animals: Because the newly formed zygotic genome remains inactive during the first few cell cycles, translation depends on so-called maternal mRNAs. These are synthesized and shelved during oocyte growth and thus contributed to the developing embryo from the maternal genome via the egg (52)(53)(54). For example, regulation of the maternal nanos (nos) mRNA is important for the establishment of the anterior-posterior body axis in Drosophila. Nanos protein, the determinant for the formation of posterior structures, is made exclusively by translation of a small fraction of nos mRNA that is localized at the posterior pole. In contrast, the larger, non-localized fraction of nos mRNA is repressed (55)(56)(57). One repressor of nos is the protein Smaug, which binds to two Smaug Response Elements (SREs) in the nos 3' UTR (58)(59)(60)(61). Smaug also induces nos mRNA degradation during the first 2 1 2 h of embryo development (59,62).
Smaug and the related yeast protein Vts1p effect deadenylation of their RNA targets by the CCR4-NOT complex (21,(63)(64)(65). Smaug-dependent deadenylation contributes to the large-scale destruction of maternal mRNAs as part of the maternal-to-zygotic transition, which transfers the control over the embryo from the maternal to the zygotic genome (66). Physical and genetic interactions between Smaug and the CCR4-NOT complex have been observed repeatedly (64,(67)(68)(69), but since the interaction has not been examined with purified proteins, it is unclear whether it is direct. Among the many mRNAs deadenylated under the influence of Smaug is nos, which has a short poly(A) tail at steady-state (21,67,(70)(71)(72).
Stimulation of CCR4-NOT-catalyzed deadenylation by specificity factors has so far only been reconstituted with S. pombe CCR4-NOT and three different effectors (37,44). The data were consistent with a simple 'tethering' model in which an individual effector protein associates with a specific RNA sequence and recruits CCR4-NOT by direct interaction. The question how Smaug induces mRNA deadenylation is more complicated, as Smaug does not act on SRE-containing RNAs by itself. Rather, it initiates the assembly of a repressor complex containing six other proteins (68,73), all of which are conserved: Cup is an oocyteand embryo-specific paralog of 4E-T and repressor of translation (74)(75)(76)(77). In tethering experiments in cultured cells, Cup induces deadenylation of the bound mRNA and coprecipitates with the CCR4-NOT complex (69,78) but a direct interaction has not been demonstrated. Whether 4E-T also induces deadenylation has been controversial (79,80). A third constituent of the Smaug-dependent repressor complex is Me31B (DDX6 in mammals) (68,73,81). Me31B associates with the MIF4G domain of NOT1 on a surface opposite the CAF1 binding site (33,34,82), but Me31Bdependent deadenylation has, to our knowledge, not been reported. Me31B also binds Cup, suggesting the possibility that Cup-dependent deadenylation might be mediated by Me31B (83)(84)(85)(86). Four additional proteins are part of the Smaug-dependent repressor complex (68,73): Trailer hitch (Tral; Lsm14 in mammals), one of several proteins associating with Me31B in a mutually exclusive manner (87)(88)(89); the translation initiation factor eIF4E, which is bound by Cup (75)(76)(77)90,91); the cytoplasmic poly(A) binding protein, PABPC; and the DEAD-box protein Belle (DDX3 in mammals). Among these proteins, PABPC can facilitate deadenylation by CCR4-NOT; specifically, PABPC has been reported to promote the activity of CCR4 but inhibit CAF1 (26,92). However, since PABPC is thought to be bound to poly(A) tails in general, it is unlikely to contribute specifically to SRE-dependent deadenylation. Deadenylation of nos is disturbed in belle mutants (68), but a direct involvement of Belle in deadenylation has not been examined. Tral and eIF4E have not been tested for a role in deadenylation.
In this paper we have used biochemical reconstitution assays to address the question whether Smaug can accelerate deadenylation as an individual protein, by direct recruitment of the CCR4-NOT complex, and/or whether additional components of the Smaug-dependent translation repressor complex are also employed for the purpose of deadenylation. We find that both Smaug and Cup, individually denotes the interaction surface. NOT1 PE is a naturally occurring isoform. (B) Purification of the Dm CCR4-NOT MINI complex by gel filtration. The complex was first purified by FLAG affinity chromatography and then applied to a Superose 6 column as described in Materials and Methods. Left panel, UV profile of the column; right panel, analysis of relevant fractions by SDS polyacrylamide gel electrophoresis and Coomassie staining. Subunits of the CCR4-NOT complex are labeled with black dots and a contaminating band with an asterisk in the gel. (C) Purified proteins used in this study. Left panel, preparations of different variants of the Dm CCR4-NOT complex. Prominent contaminants are indicated with asterisks. Right panel, purified components of the SRE-dependent repressor complex. Desired polypeptides are marked. Purified proteins were separated on SDS-polyacrylamide gels and stained with Coomassie. (D) Basal activity of different variants of the Dm CCR4-NOT complex. Purified variants of the Dm CCR4-NOT complex (6.25 nM each), were incubated with excess FAM-7mer-A 20 (50 nM), and aliquots were withdrawn as indicated. Numbers at the bottom represent average deadenylation rates in nt/min plus/minus standard deviation based on n = 3. and cooperatively, directly promote deadenylation by the CCR4-NOT complex.

Expression clones and viruses
Co-expression of two or more proteins from one baculovirus vector was performed by means of the Multi-Bac expression system (93,94). For expression of the Dm CCR4-NOT variants, all ORFs were amplified from D. melanogaster cDNA. For NOT1 MINI (amino acid residues 1147-2505), NOT2, NOT3 and CAF1, N-terminal His 8 -tags were added by PCR. C-terminal FLAG-tags were appended to CCR4 and CAF40 by PCR. His 8 -NOT2 was cloned into the pFBDM vector between the SmaI and KpnI restriction sites. His 8 -NOT1 MINI was first introduced between the NheI and HindIII restriction sites of the pET28a-MBP plasmid, then cut out with PspOMI and HindIII and inserted into the pFBDM-His 8 -NOT2 plasmid between the NotI and HindIII restriction sites. This pFBDM-His 8-NOT2 His 8-NOT1 MINI plasmid was the basis for the generation of longer NOT1 variants lacking the His-tag: Additional parts of NOT1 were inserted into the NOT1 MINI plasmid between the XbaI and AgeI restriction sites until the full length NOT1 was obtained: AvrII NOT1 MINI EXT AgeI (residues 878-1570); XbaI NOT1 pE AgeI (residues 288-1570); XbaI NOT1 pC AgeI (residues 1-1570). CAF40-FLAG was cloned into the pFBDM vector between the EcoRI and HindIII restriction sites. His 8 -NOT3 was then inserted between the SmaI and KpnI restriction sites of pFBDM-CAF40-FLAG, resulting in the pFBDM-CAF40-FLAG His 8 -NOT3 plasmid. For the nuclease module, CCR4-FLAG was inserted in the pFBDM vector between the XhoI and KpnI restriction sites. His 8 -CAF1 was then cloned between the BamHI and XbaI restriction sites of pFBDM-CCR4-FLAG, resulting in the pFBDM-CCR4-FLAG His 8 -CAF1 plasmid. Point mutations in CCR4 (412D/414N to alanine) and CAF1 (53D/55E to alanine) (69) were introduced by overlap extension PCR. NOT10 and NOT11 cDNAs, cloned in pnYC-NpM and pnEA-CvH, respectively (28) were obtained from Eugene Valkov. The NOT10 sequence was removed from its vector with KpnI and NheI and inserted between the KpnI and AvrII sites of pnEK-His 8 -MBP-Cup (described below), replacing the Cup ORF. Thereby, an N-terminal His 8 -MBP-tag followed by a HRV 3C protease site was fused to NOT10. A BglII-XbaI fragment containing His 8 -MBP-NOT10 was then cut out of the pnEK vector and cloned between the BglII and AvrII restriction sites behind the second T7 promoter of the pETDuet-1 vector (Merck, Darmstadt, Germany). The resulting plasmid was cut with SalI and BglII, and NOT11, cut as an XhoI-BamHI fragment from its vector, was introduced behind the first T7 promoter. This created a fusion with an N-terminal His 6 -tag in the pETDuet-1-His 8 -MBP-NOT10 His 6 -NOT11 plasmid.
DNA sequences coding for the putative constituents of the SRE-dependent repressor complex were also amplified from D. melanogaster cDNA, except Cup, which was a kind gift of Fulvia Bono (in the pnEK vector (95)). Cterminal FLAG-tags were added to Smaug and Cup by PCR. Smaug-FLAG was inserted between the NheI and NotI restriction sites of the pFBDM vector. Cup-FLAG was cloned between the SmaI and XhoI restriction sites of the pFBDM vector. For the generation of MBP-tagged Cup fragments, MBP was cut out of pnYC-NpM with SpeI and XhoI and inserted between the equivalent sites of pnEK-Cup. An N-terminal His 8 tag was cloned, as a synthetic oligonucleotide, into the NcoI site of pnEK-MBP-Cup. The resulting plasmid, pnEK-His 8 -MBP-Cup, was the basis for the generation of truncated Cup variants (N: aa 1-417; M: aa 418-770; C: aa 771-1117; and combinations thereof), which were amplified from pFBDM-Cup-FLAG and inserted between the XhoI and AvrII restriction sites of pnEK-His 8 -MBP-Cup for E. coli expression. MBP-Cup fragments were cut out of the pnEK-His 8 -MBP vector with NcoI and AvrII and inserted into the pF-BDM vector between the NcoI and NheI restriction sites. An N-terminal His 8 -N-tag was introduced as a restriction fragment into the pFBDM-MBP-Cup fragments using the NcoI and SmaI sites, resulting in the pFBDM-His 8 -N-MBP-Cup fragment plasmids. The GST-Me31B clone in pFastBac1 (Thermo Fisher) has been described (68). A Cterminal His 8 -tag was appended to Tral by PCR, and the resulting fragment was cloned into pFastBac1 with EcoRI and NcoI. For expression of His-T7-Tral-His in E. coli, the coding sequence was cut out from the pFastBac 1 construct and transferred to pET28a (Merck, Darmstadt, Germany) by means of HindIII and Eco53kI. An N-terminal His 8 -tag was added to Belle by PCR, and His 8 -Belle was inserted between the XhoI and HindIII restriction sites of the pFast-Bac1 vector. PABPC was cloned into pET-28a using BamHI and XhoI.
All expression constructs were confirmed by Sanger sequencing. Accession numbers for the sequences amplified are listed in Supplementary Table 1, and all primers used to generate expression constructs are listed in Supplementary  Table 2. pFastBac1 and pFBDM DNA constructs were transformed into DH10MultiBac cells (Geneva Biotech) and incubated for 5 h to allow transposition into the mini-attTn7 sites of the bacmid DNA. Colonies were selected for correct bacmid DNA by antibiotic resistance and blue-white screening. Bacmid DNA was isolated via alkaline lysis (Qiagen buffers P1, P2 and P3) and isopropanol precipitation. Sf21 cells (1.5 × 10 6 cells seeded into a 6-well plate) were transfected with 10 g bacmid DNA, which had been preincubated in 200 l Ex-Cell 420 medium plus 5 l FuGENE HD transfection reagent (Promega) for 20 min at 27 • C. After 96 h, the supernatant was collected and used to infect 0.8 × 10 6 Sf21 cells/ml at a 1:500 volume ratio for virus propagation (V1 generation). The V1 generation was used to infect Sf21 cells for protein expression.

Protein expression and purification
Anti-FLAG M2 agarose and FLAG peptide were from Sigma, Ni-NTA agarose was from Qiagen, amylose resin from NEB, and Glutathione Sepharose 4B and the Superose 6 column from Cytiva. PES concentrators were from Thermo Fisher and Amicon concentrators from Merck, Darmstadt, Germany. HRV 3C protease was purified in-house by Bodo Moritz. Sf21 cells were grown as suspension cultures in Ex-Cell 420 serum-free medium (Sigma-Aldrich) at 27 • C.
For the reconstitution of hexameric Dm CCR4-NOT complexes, Sf21 cells were co-infected with a baculovirus expressing His 8 -NOT3 + CAF40-FLAG, a second virus expressing CCR4-FLAG + His 8 -CAF1, and a third virus expressing His 8 -NOT2 with one of the different NOT1 variants. The baculoviruses were used at a 2:1:2 ratio. Cells were harvested 72 h after infection, resuspended in ice-cold lysis buffer (500 mM NaCl, 50 mM HEPES-NaOH and 10 mM potassium phosphate, pH 7.6, 10% sucrose, 1 mM PMSF, 1 M pepstatin A). Sucrose stabilized the CCR4-NOT complex during freezing and thawing. Cells were lysed by sonication on ice, the lysate was cleared by centrifugation for 20 min at 20.000 x g and the supernatant applied to anti-FLAG M2 agarose matrix for 2 h under constant rotation at 6-8 • C. The matrix was washed four times with wash buffer (lysis buffer minus PMSF and pepstatin A) in a batch format, then protein was eluted by addition of wash buffer with 200 g/ml FLAG peptide. The eluate was collected after 30 min, concentrated with a PES concentrator (10 kDa MWCO) and applied to a Superose 6 column equilibrated in wash buffer. Removal of the filter from the Superose column improved protein recovery. Fractions containing the hexameric complex were pooled, concentrated as before and frozen in liquid nitrogen before final storage at -80 • C.
For reconstitution of the octameric Dm CCR4-NOT FULL complex, His 8 -MBP-NOT10 and His 6 -NOT11 were coexpressed in E. coli BL21 Star (DE3) cells with 24 h lactose autoinduction (96). Cells were harvested, resuspended in ice-cold lysis buffer containing 1 mg/ml lysozyme and 10 mM imidazole and incubated for 1 hour under constant rotation at 6-8 • C. Cells were then sonicated, the lysate was centrifuged for 20 min at 15 000 × g and the supernatant applied to Ni-NTA agarose for 2 h under constant rotation at 6-8 • C. The matrix was washed four times with wash buffer plus 20 mM imidazole, and proteins were eluted with wash buffer plus 160 mM imidazole. After 30 min, the eluate was applied to amylose resin equilibrated in wash buffer and incubated for 2 h. NOT10 and His 6 -NOT11 were eluted by a 2 h incubation with an approximately equimolar amount of HRV 3C protease and concentrated with a PES concentrator (10 kDa MWCO). A twofold excess of the heterodimer was incubated for 2 h at 8 • C with the hexameric CCR4-NOT (isoform NOT1 PC ) to reconstitute Dm CCR4-NOT FULL , which was purified on a Superose 6 column, concentrated and stored like the hexameric complex.
For production of Smaug-FLAG, Cup-FLAG, Tral-His 8 , GST-Me31B or His 8 -Belle, Sf21 cells were infected with the relevant baculoviruses. Infected cells were processed as above up to the addition of the affinity matrix. Binding to the respective matrix (anti-FLAG M2 agarose, Ni-NTA agarose, or Glutathione Sepharose 4B) was carried out for 2 h, then the matrix was washed four times with wash buffer (with 20 mM imidazole for His 8 -tagged constructs), and proteins were eluted by wash buffer containing 200 g/ml FLAG peptide or 160 mM imidazole or by on-column cleavage with HRV 3C protease for Me31B. After elution, the proteins were, if necessary, concentrated with an Amicon Ultra concentrator (10 kDa MWCO), flash frozen and stored at -80 • C. MBP-and His 8 -N-MBP-Cup fragments were also expressed in Sf21 cells by baculoviral infection. Infected cells were processed as above up to the addition of the affinity matrix. His-MBP-tagged proteins were first purified by Ni-NTA affinity chromatography as above and then applied for 2 h to an amylose agarose matrix. Bound proteins were eluted with wash buffer containing 20 mM maltose. MBP-Cup fragments were purified by direct addition of cell lysates to the amylose matrix.
E. coli BL21 Star (DE3) cells were used for the expression of His 6 -T7-PABPC and His 6 -T7-Tral-His 8 . Expression and Ni-NTA chromatography were carried out as described for the NOT10-NOT11 purification, except that the buffer contained no phosphate. At this point, Tral and PABPC were flash frozen and stored at -80 • C.
Identities of purified proteins were confirmed by western blotting. Protein concentrations were determined by SDSpolyacrylamide gel electrophoresis and Coomassie staining in comparison to a set of BSA standards. Intensities of protein bands were evaluated with Fiji. For deadenylation experiments, proteins were pre-diluted in 200 mM potassium acetate, 50 mM HEPES-KOH, pH 7.4.

Western blotting
Proteins were transferred onto nitrocellulose membrane (GE Healthcare) by wet blotting overnight at 8 • C and 27 V. Membranes were blocked with 1.5% gelatin (from cold water fish skin; Sigma-Aldrich) in 1x TBS -0.05 % Tween and incubated with the primary antibody for 2 h. The antibody against Smaug (97) was diluted 1:500 in blocking buffer. The membrane was washed three times with 1× TBS-Tween and incubated with fluorescently labelled secondary antibody (IR-Dye, LI-COR; 1:15000). The membrane was washed again three times with 1× TBS-Tween and scanned with a LI-COR Odyssey imager.

Deadenylation substrates
The SRE-only RNA and the TCE RNA as well as their mutant (SRE MUT ) versions have been described (72). The TCE RNA was called nos RNA in (72). The plasmid vector for these RNAs also encoded a poly(A) tail of about 70 nt. Compared to the version described (72), it was modified by introduction of a BsaI site into the HindIII site at the end of the poly(A) tail, so that, upon BsaI digestion, a run-off transcript was produced that ended in a straight poly(A) tail without additional nucleotides from the restriction site. The BoxB tethering construct was generated by assembly of a synthetic BamHI-XbaI restriction fragment containing two BoxB elements (98) (Supplementary Table 2) and insertion into the BglII-XbaI restriction sites of the pBSK-nLucnos plasmid (99). The control construct (nLuc-BRE MUT -A 70 RNA) contained the mutated AB Bruno recognition element (BRE) of the oskar 3' UTR (100). The pBSK-nLuc-BRE WT plasmid was first generated by amplifying the BRE from Drosophila cDNA and introducing it between the BglII-XbaI restriction sites of the pBSK-nLuc-nos plasmid.
Nucleic Acids Research, 2023, Vol. 51, No. 8 3955 The mutant BRE was then assembled from four synthetic DNA oligonucleotides (Supplementary Table 2) and introduced downstream of the nLuc ORF between the BamHI and EcoRI sites of the pBSK-nLuc-BRE WT plasmid, replacing the WT BRE.
RNAs were synthesized with T3 RNA polymerase (Promega) in the presence of [␣-32 P]UTP and 7 mM 'antireverse' cap analog (m 7,3'-O GpppG; Jena Bioscience). All radiolabeled RNAs were gel-purified. Alternatively, the RNA was labeled by incorporation of a 6-carboxyfluorescein-(6-FAM-) labelled ApG cap analog. The cap analog was added to the transcription reaction at 1 mM in the absence of GTP. After a 5 min preincubation, GTP was added to 0.2 mM, and the incubation was continued for an additional 20 min before the GTP concentration was raised to 1 mM. After an additional hour, the reaction was incubated with 1 U DNase I (Roche), the RNA was phenol-extracted and ethanol precipitated and used without further purification.

Deadenylation assays
Reactions were carried out in 20 l or multiples thereof for time-courses. The reaction buffer consisted of 50 mM potassium acetate, 30 mM HEPES-KOH, pH 7.4, 2 mM magnesium acetate, 0.15 mg/ml nuclease-free BSA (Merck-Millipore), 2 mM DTT, 800 U/ml murine RNase Inhibitor (NEB) and 3% (w/v) PEG 20000. The presence of BSA in the buffer improved the stability of the CCR4-NOT complex during extended incubations at low concentration. Yeast tRNA (0.25 mg/ml) was added as a carrier. Reactions with the FAM-7mer-A 20 RNA did not contain tRNA, unless noted otherwise. Substrate RNA was either directly incubated with CCR4-NOT complexes or pre-incubated for 20-30 min with potential activators of deadenylation at 25 • C, and deadenylation was started by addition of CCR4-NOT. Reaction temperature was 25 • C both for the human and the Drosophila CCR4-NOT complex. The reaction was stopped by addition of a two-to threefold excess of ice-cold formamide loading buffer (95% deionized formamide, 17.5 mM EDTA, pH 8.0, 0.01% bromophenol blue, 0.01% xylene cyanol). Xylene cyanol was omitted for FAM-7mer-A 20 substrates. Samples were heated to 95 • C for 3 min, cooled on ice and separated on denaturing polyacrylamide gels (19:1 acrylamide-bis acrylamide, 1× TBE, 7 M urea). Fluorescent reaction products were directly visualized by scanning with a Typhoon 9200, whereas gels with radioactive RNA were placed on storage phosphor screens overnight at -20 • C and the screens scanned the next day. Images were analysed using Fiji (101). Deadenylation rates were determined as described (102). Briefly, the modal poly(A) tail length at each time point was determined by densitometric analysis of the deadenylation pattern in each lane using Fiji. Then the modal poly(A) tail length was plotted over time and the deadenylation rate (in nucleotides/min) was derived from the slope of the fit line.
The number of experiments (n) reported in the figure legends refers to independently performed assays. In most cases, this included the use of at least two independent protein preparations.

RNA binding assays
His 8 -MBP-or MBP-tagged Cup or Cup fragments (0.6 pmol) were incubated with a 5'-32 P-labeled RNA oligonucleotide (GGGTTTAGTGCGCACGTG, 18nt; 0.5 pmol) in 10 l 16 mM HEPES, pH 7.6, 50 mM potassium acetate, 1 mM magnesium acetate, 0.8 mM ATP, 0.24 mg/ml yeast tRNA, 1 mM DTT, 10% glycerol for 15 min at room temperature and UV crosslinked (Stratalinker 1800, 254 nm, 200 mJ/cm 2 ) on ice. Reaction products were separated on SDS-polyacrylamide gels, which were dried and analyzed by phosphoimaging. For analysis of the cross-link product by affinity purification, a 10 x binding reaction was set up, and a part of each reaction was stored as input. The remaining portions were diluted in 50 mM HEPES, pH 7.6, 500 mM KCl, 10% sucrose, 6 M urea, 0.05% NP40 and incubated for 15 min at room temperature. Magnetic Ni-NTA beads (Promega) were added to the reaction, and the mix was incubated at room temperature for 1 h on a rotating wheel. The matrix was collected and washed twice with buffer as above. Before elution in SDS sample buffer, the matrix was transferred to a fresh tube. Input and elution fractions were analyzed by gel electrophoresis as above. For technical reasons, positions of protein standards indicated next to the gels are approximate.

Protein-protein interaction assays
ReLo assays were performed as described (103). In short, Drosophila S2R+ cells were seeded onto four-well chambered coverslips (Ibidi), co-transfected with the desired combination of pAc5.1 plasmids expressing the two proteins of interest, and protein localization was analyzed after two days by live confocal fluorescence microscopy. Splitubiquitin yeast two-hybrid experiments were performed as described (104). The desired combinations of bait and prey plasmids were co-transformed into NMY51 cells and plated onto SC agar lacking Leu and Trp. For the spotting assay, several colonies were pooled to prepare an overnight liquid culture. Three 10-fold dilutions were spotted onto SC agar plates either lacking Trp and Leu (control) or Trp, Leu, His and adenine (selection), and, after three days of growth, images were taken. Detailed information on all plasmids used for protein-protein interaction assays is provided in Supplementary Table 3.
To test for an interaction between Smaug and the CCR4-NOT complex by gel filtration, Smaug (15 g in 50 l) was incubated either with Dm CCR4-NOT MINI (15 g in 20 l) or protein buffer for 1 h at 8 • C. 50 l of the mixture was applied to a Superose 6 gel filtration column (bed volume ∼ 2.4 mL; Cytiva) equilibrated in lysis buffer lacking PMSF and pepstatin A. The column was run with the same buffer at 0.017 column volumes per min. Column fractions were analyzed by SDS polyacrylamide gel electrophoresis and Coomassie staining and/or by western blotting with an antibody against Smaug.

Reconstitution of the CCR4-NOT complex from Drosophila melanogaster
In order to study the mechanism of Smaug-induced deadenylation, we reconstituted and purified five different versions of the Drosophila CCR4-NOT complex (Figure 1A). Experiments were guided by earlier work on the human complex (27). Dm CCR4-NOT FULL was composed of full-length versions of all eight subunits (CAF1, CCR4, NOT1-3, CAF40, NOT10 and NOT11). Dm CCR4-NOT 10:11 lacked NOT10 and 11. Dm CCR4-NOT PE was similar, but contained the naturally occurring NOT1 splice variant PE, which lacks the 5' portion of the open reading frame (https://flybase.org). In Dm CCR4-NOT MINI EXT , NOT1 was further shortened and started at amino acid 877. Dm CCR4-NOT MINI contained the shortest NOT1 version, starting at amino acid 1147 and thus missing the tristetraprolin binding site mapped in the mammalian ortholog (43). For the production of these complexes, three Multi-Bac clones were generated, containing NOT1 and NOT2; NOT3 and CAF40-FLAG; or FLAG-CCR4 and CAF1, respectively (93,94). Complexes composed of all six subunits were produced by co-infection of insect cells with the three viruses and purified by Flag affinity-purification followed by gel filtration. For the production of Dm CCR4-NOT FULL , His 6 -MBP-NOT10 and His 6 -NOT11 were co-expressed in E. coli. After co-purification, they were mixed with the Flagpurified six subunit complex, and the assembly was purified by gel filtration. The resulting preparations were pure and generally contained approximately stoichiometric amounts of the subunits ( Figure 1B; C, left panel). Excess CCR4-CAF1 heterodimer was obtained from the same gel filtration columns ( Figure 1B; C, left panel). The basal deadenylation activities of these complexes were measured in reactions with a small synthetic RNA substrate ('FAM 7mer-A 20 ': seven nt 'body' plus 20 3'-terminal A residues) (27), carrying a fluorescent label at its 5' end. Activities of the five larger complexes were similar ( Figure 1D). We conclude that NOT10, NOT 11, and the N-terminal half of NOT1 do not increase the basal activity of CCR4-NOT. In contrast, CAF40 and/or the NOT module promote deadenylation, as the activity of the CCR4-CAF1 heterodimer was about 2% of that of the MINI complex (250 nM of the heterodimer had an activity similar to 5 nM of the MINI complex; Supplementary Figure 1A). This is in qualitative agreement with reports for the S. pombe and human CCR4-NOT complexes (27,37,38). The activity of CCR4-NOT MINI was highest at ∼1 mM Mg 2+ and ∼50 mM potassium acetate (Supplementary Figure 1B).

Smaug is sufficient to induce SRE-dependent deadenylation by the CCR4-NOT complex
In order to examine Smaug-dependent deadenylation in a fully reconstituted system, we overproduced and purified Smaug and its associates: Smaug, Cup, Me31B and Belle were produced in insect cells by means of baculovirus vectors, PABPC was made in E. coli, and Tral was made in either system ( Figure 1C, right panel). Cup was the least pure among all proteins. Attempts at further purification were not successful as the protein, after elution from the initial Flag affinity column, did not elute in a defined peak from any other column tested.
Short substrate RNAs were used to assay for Smaugdependent deadenylation: Most experiments employed the 'SRE-only' RNA, which contained two synthetic SREs, either wild-type (SRE WT ) or with a single inactivating point mutation in each (SRE MUT ) (72). Both RNAs carried a plasmid-encoded poly(A) tail of some 70 nucleotides. Linearisation of the plasmid DNA for run-off transcription was such that no non-A nucleotides were encoded at the end of the poly(A) tail (see Materials and Methods).
SRE-only RNAs were mixed with Smaug and preincubated to allow complex formation. Addition of Dm CCR4-NOT MINI resulted in rapid deadenylation of the SRE WT RNA during an incubation at 25 • C; in the SRE MUT control, the fully deadenylated product appeared only after a lag phase of ∼16 min and then accumulated at a ∼ 50fold lower rate compared to SRE WT (Figure 2A). The Dm CCR4-NOT FULL complex behaved similar to MINI (Supplementary Figure 1C). Three observations indicate that the RNA shortening visible in the gel was due to deadenylation: (1) The reaction product had the size expected for deadenylation. (2) Results similar to those seen with internally labeled RNA in Figure 2A were also obtained with 5'-endlabeled RNA (see below, Figure 6B). (3) The reaction was affected by point mutations in the active sites of the deadenylases CCR4 and CAF1 (see below, Figure 4). As negative controls, Smaug by itself did not catalyze deadenylation, and the CCR4-NOT complex by itself had a barely detectable basal deadenylation activity (Figure 2A). As a further negative control, human PTB, an RNA binding protein and regulator of splicing (105), did not accelerate deadenylation (data not shown). Several constituents of the SRE-dependent repressor complex were also inactive (see below, Supplementary Figure 6). Thus, stimulation of CCR4-NOT-dependent deadenylation by Smaug is a specific effect. Smaug-dependent deadenylation in embryo extract has previously been found to be sensitive to ATP depletion (72). However, the reconstituted deadenylation reaction proceeded in the absence of ATP and was not affected by its addition (data not shown). Smaug-dependent deadenylation was stimulated by the presence of polyethylene glycol (PEG 20000), presumably due to a macromolecular crowding effect (Supplementary Figure S2A). PEG was therefore routinely included in the deadenylation buffer. The activity of the CCR4-NOT complex itself was not affected by PEG (data not shown). miRISC-dependent deadenylation has been reported to be associated with liquidliquid phase separation (LLPS) (106), and LLPS is promoted by crowding reagents like PEG (107). In miRISCdependent deadenylation, LLPS causes rapid sedimentation of protein-RNA complexes in a microcentrifuge (106). By this criterium, no LLPS was detectable in our deadenylation assays.

Smaug interacts with NOT3
The data so far demonstrate that Smaug induces SREand CCR4-NOT-dependent deadenylation independently of its associated repressor proteins. This would be most   easily explained by SRE-bound Smaug recruiting the deadenylase through a direct interaction. Such an interaction was in fact demonstrated by analytical gel filtration ( Figure 2B): Smaug alone was recovered in low yields; only small amounts were detectable by western blotting. These eluted in fractions C8-C10, somewhat ahead of the position expected based on the molecular weight of monomeric Smaug. When Smaug was analyzed together with Dm CCR4-NOT MINI , recovery was improved, Smaug was detectable by Coomassie staining, and a large fraction of the protein co-eluted with CCR4-NOT in fractions C1-C6.
To define components of CCR4-NOT required for the interaction with Smaug, we made use of different variants of the human deadenylase complex (27). Smaug induced SRE-dependent deadenylation by the Hs CCR4-NOT FULL complex ( Figure 2C). This eight subunit complex is essentially complete except for an N-terminal deletion of NOT11 (27). Hs CCR4-NOT MINI , corresponding to Dm CCR4-NOT MINI except for N-terminal deletions of NOT2 and NOT3, catalyzed Smaug-and SRE-dependent deadenylation with an efficiency comparable to the FULL complex ( Figure 2D). CCR4-NOT CORE is a complex further simplified by omission of the NOT module (NOT2, NOT3 and a C-terminal part of NOT1) and thus consists only of CAF1, CCR4 and CAF40 bound to a central NOT1 fragment. CCR4-NOT CORE responded to Smaug very weakly. A CCR4-CAF1 heterodimer behaved similarly ( Figure 2D). We conclude that the Smaug interaction surface of the CCR4-NOT complex is conserved between D.
melanogaster and H. sapiens and is largely if not entirely contained in the NOT module.
Interactions between Smaug and individual CCR4-NOT subunits present in the MINI complex were then examined by a recently described in vivo interaction assay based on intracellular protein relocalization (ReLo assay) (103). For this assay, CCR4-NOT subunits were fused with mCherry and with the pleckstrin homology (PH) domain of rat phospholipase C␦1 and expressed as bait in Drosophila S2R+ cells; the PH domain caused their localization on the plasma membrane. Smaug, as potential prey protein, was labeled with EGFP. Whereas EGFP-Smaug by itself is expected to be cytoplasmic, an interaction with one of the bait proteins is expected to become visible as the co-localization of the two fluorescence markers on the plasma membrane. With five out of the six bait proteins tested, Smaug remained widely distributed in the cytoplasm. Membranebound NOT3, however, caused a clear re-localization of Smaug to the plasma membrane, indicating an interaction ( Figure 3B, Supplementary Figure 3A). The Smaug interaction surface of NOT3 was mapped by additional ReLo experiments with swapped tags: NOT3 fragments ( Figure  3A), fused to the PH domain and mEGFP, were combined with mCherry-tagged Smaug. The results indicated that Smaug binding is mediated by the C-terminal region of NOT3 (amino acids 687-844), which contains the NOT1interacting region and the NOT box ( Figure 3A, C, Supplementary Figure 3).
A split-ubiquitin yeast two-hybrid assay (108), in which Smaug was used as the bait, confirmed the interaction with NOT3, detectable by cell growth under selective conditions upon co-expression of bait and prey ( Figure 3D, left panel). (Note that CAF1 and CAF40 allowed growth under selective conditions in the absence of Smaug, thus no conclusions can be drawn regarding an interaction.) In the twohybrid experiment, a NOT1 fragment comprising amino acids 752-910 also showed an interaction with Smaug (Figure 3D, right panel). However, this region of NOT1 is not present in the CCR4-NOT MINI complex, which is sufficient for Smaug-dependent deadenylation, whereas it is present in the Hs CCR4-NOT CORE complex, which does not respond to Smaug. Therefore, this region of NOT1 is neither required nor sufficient for a functional interaction with Smaug, and we did not pursue the interaction. Interactions between Smaug and fragments of NOT3 were also examined by means of the two-hybrid assay. The results confirmed the ability of the C-terminal region of NOT3 to associate with Smaug ( Figure 3E). The Smaug-NOT3 interaction visible in the ReLo and two-hybrid assays is consistent with the requirement for the NOT module in deadenylation assays ( Figure 2D). Mapping of the interaction surface to the C-terminal region of NOT3 is also consistent with the activity of Hs CCR4-NOT MINI in Smaug-dependent deadenylation: In this complex, truncation of NOT3, just upstream of the NOT1-interacting region, leaves the Smaug-interacting region intact (27). In summary, in vivo interaction assays and deadenylation assays demonstrate that Smaug interacts with the NOT module of CCR4-NOT, in particular with the C-terminal region of NOT3.

Both CCR4 and CAF1 contribute to Smaug-dependent deadenylation
Earlier experiments employing RNAi or overexpression of catalytically inactive mutants suggested that CAF1 makes the main contribution to the catalytic activity of the CCR4-NOT complex in Drosophila cells (69,109). We used the reconstituted complex to examine the relative importance of CAF1 and CCR4 in vitro. Dm CCR4-NOT MINI was prepared with point mutations in the active sites of either nuclease or the combination of both. In assays employing the FAM 7mer-A 20 RNA as a substrate, the inactivation of either nuclease subunit had an unexpectedly large effect, reducing the activity of the complex by about 80% (Figure 4A). This behavior was seen with several independent preparations of the enzyme complexes and is most likely explained by either inactive subunit exerting a dominantnegative effect on the active subunit, perhaps by transiently blocking access to the 3' end. An even stronger effect of the same type has been reported for human CCR4-NOT (110). The combination of mutations in both CCR4 and CAF1 abolished the catalytic activity of the complex (Figure 4A). The Smaug-dependent reaction was also impaired by point mutations in either catalytic subunit, and again either single mutant reduced the activity by more than 50%. Combined mutation of both subunits prevented Smaugdependent deadenylation ( Figure 4B). As the inactivation of either catalytic subunit reduced deadenylation activity to a similar extent, both in basal and Smaug-stimulated deadenylation, we conclude that, under our experimental conditions, both subunits make approximately equal contributions to the activity of the CCR4-NOT complex.
In vivo, the substrate for deadenylation is not naked poly(A), but a complex of poly(A) and the cytoplasmic poly(A) binding protein, PABPC. Moreover, PABPC is part of the SRE-dependent repressor complex (68). The S. pombe and human PABPC orthologues stimulate the basal activities of the cognate CCR4-NOT complexes (26,92) but S. pombe PABPC was somewhat inhibitory to deadenylation accelerated by the Puf3 protein (44). Thus, we sought to examine the effect of PABPC on deadenylation by Drosophila CCR4-NOT. With the FAM 7mer-A 20 RNA, relatively high concentrations of Drosophila PABPC were necessary for complete binding ( Supplementary Figure 4A), perhaps because the excess tRNA used as carrier competed for binding, although we cannot exclude partial inactivity of the PABPC preparation. As expected, the protein stimulated deadenylation by Dm CCR4-NOT MINI both at saturating and sub-saturating concentrations. PABPC itself was devoid of deadenylation activity (Supplementary Figure 4B). In a more detailed deadenylation time course, PABPC was initially inhibitory, and the stimulatory effect became apparent only at later times. Size distributions of the RNA were also more heterogeneous in the presence of PABPC ( Figure 5A). Presumably, PABPC initially sequesters the 3' end of the RNA from attack by the nuclease; once the enzyme has gained a foothold on the substrate, PABPC facilitates further deadenylation. Versions of CCR4-NOT carrying inactivating point mutations in either CCR4 or CAF1 responded similarly to PABPC: Both CCR4 and CAF1 were able to degrade poly(A) covered by PABPC, although CCR4 appeared to be slightly more efficient ( Figure 5B). The deadenylation of RNAs carrying long poly(A) tails was assayed in the absence of tRNA. Under these conditions, PABPC had again a weak stimulatory effect at low concentrations (5-10 nM; Figure 5C). Native gel electrophoresis suggested that, at 10 nM, about one copy of PABPC was bound per RNA molecule (Figure 5C; Supplementary Figure 4C). Higher concentrations of PABPC (20 -80 nM) were inhibitory ( Figure 5C). Native gel electrophoresis suggested that these concentrations resulted in the binding of up to four copies of PABC (Supplementary Figure 4C). Thus, PABPC binding is likely to extend into the RNA body and may interfere with binding of CCR4-NOT to upstream RNA sequences. In contrast to basal deadenylation, the Smaug-dependent degradation of an A 70 tail by the CCR4-NOT complex was inhibited by PABPC at any concentration. Whereas inhibition was modest upon binding of about one copy of PABPC, additional PABC inhibited strongly ( Figure 5D, Supplementary Figure 4C).

Smaug makes deadenylation processive
Accessory factors often boost the activity of nucleic acidpolymerizing or -degrading enzymes by increasing their processivity (111,112). A deadenylase can be judged to be processive by two criteria: First, under conditions of excess substrate over enzyme, largely or completely deadenylated products will co-exist with untouched substrate because a processive enzyme will act repeatedly, without dissociation, on the small fraction of substrate to which it is first bound, leaving the rest for later rounds. A completely distributive enzyme, in contrast, will remove single nucleotides from random RNA molecules and thus shorten an excess of substrates in a synchronous manner through multiple rounds of association and dissociation. Second, if a poly(A) tail is completely degraded without intermittent dissociation of the enzyme, the rate at which an individual poly(A) tail is shortened will be independent of the concentration of the processive nuclease or its ratio to substrate. Thus, the time at which the deadenylation end product is first seen will be independent of the nuclease concentration; only the amount of end product present at this time will increase with increasing nuclease concentration. In contrast, for a distributive enzyme the time required for the deadenylation end product to appear will be shorter with higher enzyme concentrations. This is because a high enzyme concentration will drive the association with substrate preceding every catalytic event.
The processivity of the CCR4-NOT complex by itself was first assessed in reactions with the FAM 7mer-A 20 RNA. As shown in Figure 6A, fully deadenylated products did not co-exist with untouched substrate and became visible only at later time points when all of the substrate had already been shortened to a significant extent. Also, fully deadenylated product first became visible at earlier time points in proportion with increasing enzyme concentration (e. g. ∼32 min at a 1:8 ratio of CCR4-NOT:RNA versus 4 min at a 1:2 ratio); in other words, the rate of shortening of an individual poly(A) tail was dependent on the nuclease concentration. By both criteria, the activity of CCR4-NOT with this substrate was distributive.
To determine whether Smaug-dependent deadenylation is processive, we pre-incubated SRE-containing TCE WT RNA (see Materials and Methods) with a threefold excess of Smaug over SREs. Limiting amounts of Dm CCR4-NOT MINI were then added, and time-dependent deadenylation was measured. Biphasic kinetics were observed: In an initial burst phase, fully deadenylated end product was already present at the first time point, only five seconds after the start of the reaction. At this early time point, most of the substrate RNA had not been attacked, and partially shortened intermediates of deadenylation were not visible. Thus, the reaction was processive. In the second phase of the reaction, additional deadenylation product accumulated at a decreasing and overall much lower rate ( Figure 6B). Since the initial reaction was so fast, we could not test the prediction that the time at which the end product first appeared should be independent of enzyme concentration. However, the amount of RNA that was completely deadenylated within the burst phase was approximately stoichiometric with the CCR4-NOT complex: 5 nM CCR4-NOT complex deadenylated approximately 12% of 50 nM substrate RNA within 5 s. Thus, the initial burst phase represented the complete deadenylation of approximately one substrate RNA per enzyme complex, whereas the subsequent slow increase in deadenylated RNA presumably reflected the rate-limiting transition of the enzyme to new substrate molecules. In the control reaction lacking Smaug, no deadenylation was detectable under these conditions ( Figure 6B), but weak distributive activity was visible with longer incubation times (Supplementary Figure 5).

Cup also induces deadenylation
Additional constituents of the Smg-dependent repressor complex (Cup, Me31B, Tral, Belle) were also titrated into deadenylation assays containing the CCR4-NOT complex. Me31B was inactive in the presence or absence of ATP (Supplementary Figure 6). Tral and Belle were also inactive, and so was a combination of Tral and Me31B (data not shown). However, Cup consistently stimulated deadenylation ( Figure 7A). Thus, the Cup-dependent stimulation of deadenylation reported by Igreja and Izaurralde (78) is a direct effect. MINI and FULL complexes were similar in their response to Cup (Supplementary Figure 1C). Like Smaug-dependent deadenylation, the Cup-dependent reaction was also strongly stimulated by PEG (Supplementary Figure 2B). As expected, Cup-dependent deadenylation was not SRE-dependent (Supplementary Figure 6A). We separated Cup into three non-overlapping fragments, the N-terminal, middle and C-terminal domains (N, M and C) (78). The three fragments as well as the NM and MC combinations were initially produced as His-N-MBP fusion proteins ( Figure 7B, Supplementary Figure 7B). The phage N peptide (98) allowed binding to a deadenylation substrate containing two box B elements 16 nucleotides upstream of an A 70 tail. All Cup fragments prepared in this manner were able to stimulate deadenylation of the box Bcontaining substrate ( Figure 7C, left panel). No deadenylation was visible when Cup fragments were replaced by His-N-MBP as a control, and the Cup fragments had no activity on their own. The ability of the M and C fragments to stimulate deadenylation is in agreement with the results of Igreja and Izaurralde obtained in cells (78). Surprisingly, all Cup fragments also stimulated deadenylation of a substrate RNA in which the box B elements were replaced by a control sequence ( Figure 7C, right panel); only the M fragment had weaker activity with this substrate.
The activity of wild-type Cup and the box B-independent activity of the Cup fragments suggested that the protein and its fragments can bind RNA and recruit the CCR4-NOT complex. In fact, Cup has been identified as an RNA binding protein by proteome-wide UV cross-linking screens (113,114), and so has the related protein 4E-T in mammals (115). Unfortunately, purified Cup and its fragments did not yield interpretable results in electrophoretic mobility shift experiments and were not clean enough for nitrocellulose filter-binding experiments. Thus, UV cross-linking to a radiolabeled synthetic RNA oligonucleotide was used to assess the ability of Cup to bind RNA. For these experiments, use of the His-N-MBP tag was avoided because of the RNA binding activity of the N peptide. Instead, all five Cup fragments were purified as MBP fusion proteins from baculovirus-infected Sf21 cells ( Figure 7D). All these Cup fragments promoted deadenylation, albeit weakly (Supplementary Figure 7C), and all could be cross-linked to RNA ( Figure 7D). As controls, the MBP tag alone and BSA were not cross-linked under the same conditions (Supplementary Figure 7D). Migration of the major cross-link products in the SDS gel varied as expected from the molecular weights of the Cup variants, providing evidence that the signals were in fact derived from Cup. In order to further examine the identities of the cross-link products, we generated His-MBP fusions of full-length Cup (in Sf21 cells) and of the M, MC and C fragments (in E. coli). We were unable to generate the N and NM fragments with this type of fusion. Fusion proteins were cross-linked to RNA, and their identities were ascertained by their binding, under denaturing conditions, to a Ni-NTA matrix (Supplementary Figure 7E). Some proteolytic products common to the M and MC fragments were co-purified on the Ni-NTA column, but a control protein, GST-Tral, was not, confirming the specificity of the pulldown. Thus, full-length Cup as well as all fragments tested are able to bind RNA.
We wished to test whether Smg and Cup can simultaneously stimulate CCR4-NOT-dependent deadenylation. However, under standard reaction conditions, Smaugdependent deadenylation was so efficient ( Figure 6) that no further stimulation by Cup would have been detectable. Thus, the reaction was weakened by omission of PEG from the buffer. Under these sensitized conditions, the simultaneous presence of Smaug and Cup indeed led to significantly improved deadenylation of the SRE WT RNA compared to either protein alone; stimulatory effects were reproducibly more than additive ( Figure 7E). Each protein was used at a concentration that individually was saturating for deadenylation and close to saturation in RNA binding (Supplementary Figure 7F, G). Thus, Smg and Cup bound to the same RNA can cooperate in the stimulation of the CCR4-NOT complex.

DISCUSSION
During early embryonic development of Drosophila, Smaug is responsible for the degradation of hundreds of maternal mRNAs, thus making a major contribution to the maternalto-zygotic transition (65,116). Its best-studied target is the nanos mRNA. Smaug represses nos both by preventing its translation and by inducing its deadenylation by the CCR4-NOT complex. As Smaug binds the nos 3'-UTR in the company of six other proteins, the mechanism by which it induces deadenylation is potentially complex. In order to examine the mechanism of the deadenylation reaction, we have reconstituted the eight subunit Drosophila CCR4-NOT complex and also overproduced and purified Smaug as well as the other constituents of the Smaugdependent repressor complex. Activity assays with these proteins revealed that Smaug on its own is able to induce an efficient, processive deadenylation by the CCR4-NOT complex. As a second component of the repressor complex, Cup was also able to stimulate CCR4-NOT-catalyzed deadenylation. A third component, PABPC, modestly promoted basal deadenylation, but was moderately inhibitory to Smaug-dependent deadenylation. The other components of the repressor complex did not affect deadenylation in our in vitro assays. The inactive components included Me31B, even though the protein is able to bind NOT1 (33,34,82).
Our data support a simple tethering model for the ability of Smaug to accelerate deadenylation: Smaug binds SREs with high affinity (63,117). As shown here, the protein also interacts directly with the CCR4-NOT complex via the C-terminal domain of NOT3 and induces a processive activity of CCR4-NOT. Smaug-induced deadenylation proceeded at a rate too fast to be measured by manual pipetting. The amount of RNA deadenylated in the initial burst appeared to be stoichiometric with respect to the deadenylase, although this conclusion is limited by the accuracy with which the concentration of CCR4-NOT could be determined. These data indicate that Smaug promotes deadenylation by preventing the dissociation of the deadenylase from its substrate. This corresponds to the mechanism by which deadenylation effectors Mmi1, Puf3 and Zfs1 from S. pombe stimulate their cognate CCR4-NOT complex (37,44). Smaug-dependent deadenylation was initially speculated to be more complex, as the reaction was sensitive to ATP depletion in Drosophila embryo extracts (72). Deadenylation induced by miRNAs behaved similarly (118). However, the apparent ATP requirement for miRNAdependent deadenylation was later explained by the accumulation of AMP, presumably derived from ADP by the reversible adenylate kinase reaction; AMP inhibited deadenylation (119). We have independently found that the apparent ATP-dependence of deadenylation in embryo extract was only observed with substrate RNAs produced by runoff transcription that ended in a few non-A residues down-stream of the poly(A) tail due to the restriction site used at the time to linearize the template; deadenylation was not inhibited by ATP depletion when the poly(A) tails were added by poly(A) polymerase (C. Temme and EW, unpublished data). Perhaps degradation of the non-A residues is more sensitive to AMP inhibition. Regardless, our reconstitution experiments confirm that Smaug-dependent deadenylation does not require ATP.
Deadenylation effectors typically interact with CCR4-NOT in a complex manner, employing multiple short interaction motifs embedded in intrinsically disordered regions (37,44,120). For example, contacts to CCR4-NOT involve four domains of the Drosophila Pumilio protein (109) or multiple tryptophan-containing motifs of the miRNAassociated GW182 proteins (33,34,39). Many effectors also interact with more than one subunit of CCR4-NOT. For example, the Drosophila Nanos protein is itself an effector of deadenylation and uses redundant binding sites to recruit CCR4-NOT: The dominant site forms one short ␣helix contacting NOT1 in its C-terminal domain and a second ␣-helix binding the NOT box of NOT3 (121). Likewise, Drosophila Roquin has a binding site for CAF40 and redundant binding sites for the NOT module (51), and mammalian TTP interacts both with NOT1 (43) and CAF40 (46). In the case of Smaug, the C-terminal region of NOT3 provides the major interaction surface, but our data do not exclude additional interactions. Although the ReLo and two-hybrid assays were performed in the presence of all cellular proteins, both the Smaug -CCR4-NOT interaction in gel filtration and the stimulation of deadenylation by Smaug were observed with highly purified preparations. Thus, the interaction between Smaug and NOT3 is almost certainly direct. As in the case of other deadenylation effectors, multiple regions of Smaug appear to be involved as our attempts to map an individual NOT3 binding domain in the protein were not successful.
In addition to Smaug, Cup was also able, on its own, to stimulate CCR4-NOT-dependent deadenylation. The ability to induce deadenylation seems to be distributed over much of the protein, as each of the three non-overlapping fragments was active. Our results are consistent with earlier results (78) that the M and C fragments of Cup associate with CCR4-NOT and, when tethered to a reporter RNA, induce deadenylation in cells. Our results in the reconstituted in vitro system indicate that this effect of Cup is direct. The ability of Cup and its fragments to induce deadenylation even in the absence of tethering suggests they are able to bind RNA. Unfortunately, we have been unable to purify Cup or its fragments to homogeneity, as the proteins did not elute in clean peaks from any column tested. Thus, RNA binding activity in nitrocellulose filter-binding assays could not be attributed to Cup as opposed to contaminations. Upon electrophoresis in native gels, all complexes formed between RNA and the Cup preparation remained stuck in the wells so that, again, the protein responsible for binding could not be identified with certainty. However, UV cross-linking assays were consistent with RNA binding of Cup and its fragments. We have also tested whether Cup interacts with CCR4-NOT. Whereas CCR4-NOT formed a well-defined peak in the included volume of a gel filtration column ( Figure 1B), Cup eluted in the void volume, indicat-ing that it is aggregated and/or assumes a large volume due to its intrinsically disordered nature. When Cup and CCR4-NOT were analyzed in combination, both eluted in the void volume (data not shown). While this is evidence of an interaction, the specificity is questionable in view of the disordered and presumably aggregated state of Cup.
Although Cup is able to bind RNA and induce deadenylation on its own, its in vivo activity probably depends on the protein being recruited to specific mRNAs by Smaug (75) and other RNA binding proteins, for example Bruno (74,76). The presence of two deadenylation effectors, Smaug and Cup, in a single complex is not without precedent: Nanos and Pumilio and, in some cases, Brat, cooperate in the regulation of Drosophila hunchback and other mRNAs (122,123). Both Nanos and Pumilio can individually elicit deadenylation by CCR4-NOT (see above) (109,121,124). When both Smaug and Cup were used at near-saturating concentrations, their effects on deadenylation were not merely additive, but a modest degree of cooperativity was reproducibly observed. Cooperativity is not altogether surprising, as Smaug and Cup are thought to interact directly (75). The cooperative effect was seen in the absence of a crowding reagent; in the presence of PEG, the Smaugdependent reaction is so efficient that one cannot expect an additional stimulation by Cup.
Previous knock-down experiments and overexpression of catalytically dead polypeptides in cultured Drosophila cells suggested that CAF1 may be the dominant catalytic subunit of CCR4-NOT in vivo (69,109). The interpretation of these experiments was limited by incomplete protein depletion and an uncertain degree to which WT protein was replaced by the inactive version in the CCR4-NOT complex. CCR4 is important for the deadenylation of nos (67) (see below), and genetic evidence for the importance of CCR4 catalytic activity has been presented (123). Our reconstitution experiments now provide clear evidence that both CCR4 and CAF1 are active nucleases. Individual inactivation of either subunit reduced the activity of the complex by more than 50%. This apparent interdependence of the two activities may be most easily explained by the inactive subunit transiently blocking access to the 3' end. Nevertheless, inactivation of either CCR4 or CAF1 reduced the activity of the complex to a similar extent, both in unassisted, basal deadenylation and in the Smaug-dependent reaction, suggesting comparable contributions of the two nucleases to the activity of CCR4-NOT. However, since the activity of CAF1 is highly sensitive to pH and the concentrations of Mg 2+ and Zn 2+ (31), our results do not exclude the possibility that one catalytic subunit plays a dominant role in vivo.
Using a fully reconstituted CCR4-NOT complex from S. pombe, Webster et al. (92) found that the enzyme's basal activity is facilitated by PABPC (Pab1 in S. pombe). Similar conclusions have been reached for the human deadenylase (26), although reconstitution was limited to a CCR4-CAF1 heterodimer. In accordance with these data, we also observed that the basal deadenylation activity of the Drosophila CCR4-NOT complex was moderately enhanced by PABPC at appropriate concentrations. In contrast, when added to a Smaug-containing deadenylation reaction, PABPC caused a modest inhibition; excess protein was strongly inhibitory. Webster et al. (44) also reported a weak inhibitory effect of PABPC on the activity of S. pombe CCR4-NOT activated by an RNA binding protein. In summary, PABPC can be described as being largely permissive for specific deadenylation, i.e. the strong effects of the deadenylation effectors that have been tested are dominant compared to the modest effects of PABPC.
Webster et al. (92) and Yi et al. (26) found that the catalytic activity of S. pombe or human CCR4 was stimulated by PABPC, whereas CAF1 was unable to degrade PABPCbound poly(A). In contrast, we observed that both catalytic subunits of Drosophila CCR4-NOT are able to cope with a PABPC-poly(A) complex. This discrepancy may reflect species-specific differences. However, Stupfler et al. (125) reported that mouse CAF1 (CNOT7) digested PABPCbound poly(A) and the naked polymer with similar efficiencies; thus, it seems more likely that the effect of PABPC depends on the reaction conditions, to which CAF1 is highly sensitive. Mice lacking both CCR4 orthologues (CNOT6 and CNOT6L) were viable, had no overt phenotype, and fibroblasts derived from them had a normal poly(A) tail length distribution. In contrast, the deletion of NOT1 or combined deletion of both CAF1 orthologues (CNOT7 and CNOT8) were lethal, and their reduced expression led to an accumulation of long poly(A) tails (126). These data show that CCR4-NOT is essential, but the catalytic activities of CCR4-type subunits are not. Thus, CAF1-type subunits must be able to degrade PABPC-covered poly(A) in vivo; any impediment imposed by PABPC cannot be absolute. The ability of CAF1 to degrade poly(A) in the presence of PABPC is also enhanced in the presence of BTG2/TOB (38,125,127).
Our data reveal that Cup plays a direct role in the deadenylation of nos and probably many other RNAs. Thus Smaug, acting on CCR4-NOT by itself and promoting the recruitment of Cup (75), accelerates deadenylation directly and presumably also indirectly. However, the experiments capture only select aspects of the complicated in vivo situation: Smaug and its partner Cup are not the only effectors of nos deadenylation; piRNAs are also involved (128), and accordingly a deletion of Smaug only partially stabilizes nos RNA (129). Whereas nos loses its entire poly(A) tail in the Smaug-and CCR4-NOT-dependent in vitro reaction, the RNA persists with an oligo(A) tail in vivo (see Introduction). Possibly, the complete and extremely rapid deadenylation reaction observed in vitro is tempered by conditions prevailing in the embryo. However, a competing poly(A) tail extension reaction catalyzed by the non-canonical poly(A) polymerase GLD2 (encoded by wispy) presumably plays a more important role (130).
In the early Drosophila embryo, translational efficiency is tied to poly(A) tail length (21). Therefore, Smaug-induced deadenylation, even though it does not go to completion in the embryo, contributes to the translational repression of the non-localized fraction of nos RNA (67). However, nos translation is also repressed by mechanisms independent of deadenylation (68,73,75). Although CCR4-NOT can repress translation independently of its deadenylation activity (39,40,86,131), the deadenylase is an unstable and/or substoichiometric constituent of the Smaug-dependent repressor complex (68) and therefore unlikely to play a major role in deadenylation-independent repression.
Deadenylation not only contributes to translational repression, but is also the first step in the decay of nos, as shown by the stabilization of nos in smg and twin mutants (67). Cup stabilizes deadenylated RNA by inhibiting decapping (78). Thus, the degradation of Cup during the first three hours of embryonic development (132) might be thought to be essential for the further degradation of deadenylated nos. However, nos is cleared rapidly and completely in wispy mutants (130); thus, even in the presence of Cup, deadenylation appears to be sufficient to induce the decay of the nos mRNA.

DATA AVAILABILITY
All data are incorporated into the article and its online supplementary material.