A non-proteolytic role for ubiquitin in deadenylation of MHC-I mRNA by the RNA-binding E3-ligase MEX-3C

The regulation of protein and mRNA turnover is essential for many cellular processes. We recently showed that ubiquitin—traditionally linked to protein degradation—directly regulates the degradation of mRNAs through the action of a newly identified family of RNA-binding E3 ubiquitin ligases. How ubiquitin regulates mRNA decay remains unclear. Here, we identify a new role for ubiquitin in regulating deadenylation, the initial and often rate-limiting step in mRNA degradation. MEX-3C, a canonical member of this family of RNA-binding ubiquitin ligases, associates with the cytoplasmic deadenylation complexes and ubiquitinates CNOT7(Caf1), the main catalytic subunit of the CCR4-NOT deadenylation machinery. We establish a new role for ubiquitin in regulating MHC-I mRNA deadenylation as ubiquitination of CNOT7 by MEX-3C regulates its deadenylation activity and is required for MHC-I mRNA degradation. Since neither proteasome nor lysosome inhibitors rescued MEX-3C-mediated MHC-I mRNA degradation, our findings suggest a new non-proteolytic function for ubiquitin in the regulation of mRNA decay.

M essenger RNA (mRNA) turnover plays a critical role in the regulation of the majority of cellular processes. Up to 50% of the changes seen in gene expression are estimated to occur at the level of mRNA stability 1,2 , although how mammalian mRNA decay is regulated remains poorly understood. Although ubiquitin is traditionally associated with protein degradation, we recently identified a role for ubiquitin in the degradation of mRNA 3 . Of the more than 600 E3 ubiquitin ligases described, at least 15 contain an RNA-binding domain 4 in addition to the RING domain, which defines the largest family of E3 ligases and is required for the ubiquitination reaction. In a small interfering RNA (siRNA) ubiquitome screen, we identified MEX-3C, a canonical member of this novel family of RNA-binding ubiquitin E3-ligases, which regulates the cell surface expression of major histocompatibility complex (MHC) class I proteins, via the post-transcriptional regulation of MHC-I mRNA. MEX-3C therefore provides a direct link between ubiquitination and mRNA degradation 3 .
The four members of the mammalian MEX-3 gene family (MEX-3A-D) each contain two RNA-binding KH domains and a ubiquitin E3-ligase RING domain 5 . This family has arisen by gene duplication from the MEX-3 orthologue in Caenorhabditis elegans, which also has two RNA-binding KH domains but lacks the RING domain. MEX-3C binds the 3 0 untranslated region (UTR) of its target mRNA HLA-A2 (an MHC-I allotype) through its KH domains and together with its cargo mRNA shuttles from the nucleus to the cytosol 3,5 . This HLA-A2 mRNA substrate bound to MEX-3C cannot be translated into protein, a function conserved with its C. elegans MEX-3 ancestor that also binds mRNA but lacks the RING domain 6 . However, although the C. elegans MEX-3 inhibits translation of its substrate mRNA 6 , MEX-3C not only inhibits translation but also induces the degradation of its target mRNA in a ubiquitin-dependent manner 3 . In the absence of a RING domain, MEX-3C is still able to inhibit substrate translation, but there is an absolute requirement for the RING domain, and therefore E3-ubiquitin ligase activity, for HLA-A2 mRNA degradation. RINGless MEX-3C therefore behaves more like its C. elegans ancestor, in that its substrate HLA-A2 mRNA is effectively sequestered and not translated, but is no longer degraded.
Eukaryotic mRNAs are protected from decay at their 5' and 3' ends by the cap and poly(A) tail, respectively. The degradation of mRNAs starts with the removal of the poly(A) tail by a process called deadenylation. This process is mediated by the concerted action of two complexes, namely CCR4-NOT and PAN2-PAN3. Studies in several model organisms show that deadenylation is a rate-limiting step for mRNA degradation 7 , and its impaired regulation is associated with a variety of cellular conditions in mammalian cells 8 . However, how mammalian deadenylation is regulated remains poorly understood.
Our characterization of MEX-3C's E3 ligase activity in the regulation of mRNAs decay now establishes a new role for ubiquitin in the regulation of deadenylation. Here, we show that MEX-3C associates with different members of the cytoplasmic deadenylation complexes and ubiquitinates CNOT7, the main catalytic subunit of the CCR4-NOT deadenylation machinery. Ubiquitination of this subunit (CNOT7) by MEX-3C regulates its deadenylation activity and is required for HLA-A2 mRNA degradation. Moreover, since neither proteasome nor lysosome inhibitors, nor the use of ubiquitin mutants that prevent the formation of protein degradation signalling K11-and K48-linked chains, rescued MEX-3Cmediated mRNA degradation, our findings point to a new non-proteolytic function for ubiquitin in the regulation of mRNA decay.

Results
MEX-3C interacts with the major deadenylation complexes. To establish the ubiquitin-related mechanism responsible for MEX-3C-mediated degradation of mRNAs, we first sought to identify MEX-3C-binding partners involved in mRNA degradation, or potential ubiquitination substrates. We performed a pull-down with RINGless MEX-3C expressed in HEK293T cells in the presence of RNAse-I, followed by mass spectrometry analysis (Supplementary Table 1). The rationale behind this experiment was that the RINGless mutant form of MEX-3C, which is unable to ubiquitinate, should act as a substrate trap and remains bound to its ubiquitination substrates. Analysis of the RINGless MEX-3C interactome reflected MEX-3C's involvement in different stages of RNA metabolism, primarily mRNA processing, with an emphasis on mRNA degradation (Fig. 1a).
The cytosolic degradation of eukaryotic mRNAs requires the initial shortening of the 3 0 -poly(A) tail (deadenylation) and subsequent removal of the 5 0 -cap (decapping). Deadenylation is mediated by the concerted action of two complexes, CCR4-NOT and PAN2-PAN3, both of which were found in association with MEX-3C and subsequently confirmed in immunoprecipitated (IP) blots following RNAse-I treatment (Supplementary Table 1, Fig. 1b and Supplementary Fig. 1a for control IP). PolyA-binding protein interacts with MEX-3C through its RNA cargo 5 . To control for effective RNAse-I treatment, we showed that under the experimental conditions used, our mass spectrometry analysis did not identify PolyA-binding protein bound to MEX-3C (Supplementary Table 1).
The identification of MEX-3C bound to the cytosolic deadenylation complexes is especially relevant as, despite deadenylation of mRNAs being the initial and often rate-limiting step in mRNA degradation 9 , little is known about how these deadenylases are regulated in mammalian cells. To determine whether MEX-3C promotes the deadenylation of its endogenous model substrate, HLA-A2 mRNA in-vivo, we used an reverse transcription-PCR-based assay to measure the length of HLA-A2 mRNA poly(A) tail in MEX-3C-expressing cells (Fig. 1c) following fluorescence-activated cell sorting (FACS) ( Supplementary Fig. 1b). MEX-3C promotes the shortening of the poly(A) tail of HLA-A2 mRNA ( Fig. 1c right panel), but not of the ACTIN control mRNA (Fig. 1c left panel); an activity that requires its ubiquitin-ligase activity as it is not seen with its RINGless mutant form (Fig. 1c). Similar results were obtained for FF-Luc-HLA-A2 3 0 UTR reporter mRNA 3 . Taken together, these results led us to hypothesize that MEX-3C's E3 ligase activity controls mRNA decay through the regulation of deadenylation (Fig. 1d).
MEX-3C's E3-ligase activity regulates deadenylation. The above results are reminiscent of the two-step miRNA-mediated repression of mRNAs by the CCR4-NOT complex, which requires an initial translational repression step, followed by the degradation of the target mRNA by deadenylation 11 . As MEX-3C interacts with Argonaute proteins 5 , which play a key role in RNA silencing, we asked whether MEX-3C's E3-ligase activity regulates the transition between these two steps by triggering deadenylation.
We therefore set up an in-vitro deadenylation assay using a specific (fluorescein labelled) RNA substrate for CNOT7 (ref. 12). Strep-MEX-3C pull-downs from wild-type (WT) and RINGless MEX-3C-expressing cells were incubated with a 5 0 -fluorescein labelled specific RNA substrate (Flc-5 0 -UCUAAAUA 20 ) to assay the deadenylation activity of CNOT7 over time. Degradation of 5 0 -fluorescein-labelled RNA deadenylation substrate was visualized by denaturing polyacrylamide gel electrophoresis. The deadenylation substrate was readily degraded following WT MEX-3C IP, an effect not seen with RINGless MEX-3C (Fig. 3b,c). These results confirm that MEX-3C's E3-ligase activity is required for the deadenylation of its substrate.
Ubiquitination of CNOT7 regulates its deadenylation activity. Since Caf1 (CNOT7/8) is the major catalytic component of the CCR4-NOT deadenylation complex, and was the only deadenylation subunit whose depletion prevented MEX-3C-mediated degradation of FF-Luc-HLA-A2 3 0 UTR mRNA, we determined whether CNOT7 was a ubiquitination target of MEX-3C. Endogenous or overexpressed (HA-tagged) CNOT7 was IP under denaturing conditions to prevent interaction with other proteins. Ubiquitin immunoblot analysis showed CNOT7 ubiquitination with WT but not RINGless MEX-3C (Fig. 4a,b), confirming that CNOT7 is indeed ubiquitinated in a MEX-3C-dependent manner.
This ubiquitination did not promote CNOT7 protein degradation ( Fig. 1b and Supplementary Fig. 3a  . ACTIN mRNA was assayed as control (left panel) and Ao PCR controls for loading. Ao: refers to PCR products using primers to amplify the last protein half-life was unaffected following cycloheximide treatment either in the absence (shMEX-3C) or in the presence of exogenously expressed WT or RINGless MEX-3C ( Supplementary Fig. 3c).
We then wanted to determine which CNOT7 lysine residue is ubiquitinated by MEX-3C and use these findings to ask how MEX-3C-mediated ubiquitination of CNOT7 affects its deadenylation activity in-vitro, and the stability of the FF-Luc-HLA-A2 3 0 UTR reporter mRNA in-vivo. The choice of lysines to be mutated by site-directed mutagenesis was based on the structure of CNOT7 and the lysine residues (K200 and K206) reported to be ubiquitinated in mass spectrometry data sets 13,14 . A representation of CNOT7's structure highlights (in blue) the exposed lysine residues mutated ( Supplementary Fig. 4a). (The CNOT7 K203R variant was toxic to cells and therefore excluded from this and further experiments). In comparison to WT CNOT7, ubiquitination of the K206R mutant was mildly impaired, whereas ubiquitination was completely lost with the 4K (K196R,K200R,K203R,K206R) CNOT7 mutant (Fig. 4c).
To control for the structural integrity of CNOT7 4K-mutant, we performed the in-vitro CNOT7 deadenylation assay following CNOT7 pull-down in the absence of exogenous MEX-3C as previously described (Suzuki et al.) 12 . This CNOT7 4K-mutant remains functionally active in the absence of MEX-3C ( Supplementary Fig. 4b) suggesting that the folding of CNOT7 4K-mutant and its ability to form a functional deadenylation complex (Fig. 4d) remained intact. Although ubiquitination is not required for CNOT7's basal deadenylation, these results highlight that this deadenylation activity can be modulated in a novel ubiquitin-dependent manner in mammalian cells for specific mRNAs.
A new non-proteolytic function for ubiquitin in mRNA decay. In addition to its role in proteolysis, ubiquitin provides a signal for a range of non-proteolytic functions 15 , by virtue of forming chains of distinct topologies depending on whether they are linked through one of its seven Lysine (K) residues or at the N-terminus 15,16 . Substrates modified by K48-linked polyubiquitin chains are targeted to proteasomes for degradation. In contrast, K63-linked chains provide non-proteolytic signals, as characterized in DNA damage and repair pathways, kinase signalling pathways and endocytosis 16 . Since neither proteasome nor lysosome inhibitors rescued MEX-3C-mediated degradation of the FF-Luc-HLA-A2 3 0 UTR mRNA ( Fig. 5a and Supplementary  Fig. 4c for expression levels), a new non-proteolytic function for ubiquitin in the regulation of mRNA decay was suggested.
To determine the ubiquitin chain linkage required for MEX-3Cmediated degradation of HLA-A2 mRNA, we used a range of Lysine-to-Arginine Ubiquitin-Green Fluorescent Protein (UB-GFP) mutants 15 . These ubiquitin mutants are particularly useful as the co-translational cleavage of GFP from ubiquitin provides a quantitative surrogate marker for mutant ubiquitin expression 15 (Fig. 5b). None of the ubiquitin lysine mutants rescued endogenous HLA-A2 protein levels ( Fig. 5b) from MEX-3C downregulation, and this was in keeping with RINGless MEX-3C, which lacks E3-ubiquitin ligase activity 3 , and is still able to inhibit HLA-A2 translation without triggering its mRNA degradation. Similar results were obtained for firefly luciferase protein levels (measured as relative luciferase activity against renilla luciferase), when the FF-Luc-HLA-A2-3 0 UTR reporter was used 3 (Fig. 5c).
We therefore determined the effect of the ubiquitin mutants on MEX-3C-mediated mRNA degradation. Degradation of the FF-Luc-HLA-A2-3 0 UTR reporter mRNA was rescued by ubiquitin mutants that cannot form K6-and K63-linked chains (Fig. 5d) suggesting an important role for these lysine residues in deadenylation. Furthermore, CNOT7 ubiquitination by MEX-3C was significantly reduced with the K6R and K63R ubiquitin mutants ( Supplementary Fig. 4d). Conversely, neither MEX-3Cmediated ubiquitination of CNOT7 ( Supplementary Fig. 4d) nor the degradation of FF-Luc-HLA-A2-3 0 UTR target mRNA was impaired in the presence of K11-or K48-linkage ubiquitin mutants, which are traditionally associated with protein degradation signals (Fig. 5d). Taken together, these results suggest a nonproteolytic function for ubiquitin in the regulation of mRNA decay by MEX-3C.

Discussion
We have identified a new role for ubiquitin in the regulation of deadenylation, the initial and rate-limiting step in mRNA degradation. MEX-3C, a member of the recently described family of RNA-binding ubiquitin E3-ligases 4 , associates with the cytoplasmic deadenylation complexes, and ubiquitinates CNOT7, the main deadenylase subunit of the CCR4-NOT machinery. Ubiquitination of CNOT7 by MEX-3C promotes its deadenylation activity and therefore MHC-I mRNA degradation. In addition to its established role in protein regulation/ degradation, ubiquitination provides a critical signal for many other cellular regulatory functions 15   a new non-proteolytic role for ubiquitin in the regulation of mRNA decay. Ubiquitination of CNOT7 did not lead to its degradation ( Fig. 1b and Supplementary Fig. 3a-c) or did proteasomal or lysosomal inhibitors rescue MEX-3C-mediated degradation of its target mRNA (Fig. 5). Degradation of the target mRNA was, however, rescued in the presence of ubiquitin mutants that cannot form K6-and K63-linked chains, presumably due to reduced ubiquitination of CNOT7 ( Supplementary Fig. 4d). These results add a new signalling function to K6-and K63-linked ubiquitin chains. K6-linked chains have recently been associated with parkin's regulation of mitochondrial quality control 17 and to the stabilization of RING1b 18 and BRCA1/BARD 19 , both involved in histone modification and DNA repair. K63-linked chains mediate different processes including endocytosis, assembly of DNA repair complexes and the activation of the nuclear factor-kB pathway. In fact, MEX-3C was recently reported to activate the nuclear factor-kB pathway by ubiquitinating RIG-I after viral infection in a K63-linked manner 20 . Ubiquitination of CNOT7 was not a prerequisite for basal deadenylation activity. However, the loss of specific mRNA degradative activity with the CNOT7 4K-mutant, highlights a novel role for ubiquitin in regulating deadenylation of certain mRNAs in the mammalian system. This is important since the mechanisms that regulate deadenylation in mammalian cells are highly regulated but poorly understood. Deadenylation and RNA turnover play an important role in a broad range of cellular conditions including development, mRNA surveillance, DNA damage, cell differentiation and cancer 8 . Understanding how ubiquitin regulates mRNA abundance and protein production will provide a better mechanistic understanding of different disease states.
A number of potential mechanisms may account for MEX-3C's ability to regulate mRNA decay through ubiquitination. Ubiquitin may induce conformational changes in CNOT7 that activate deadenylation. Alternatively, ubiquitinated CNOT7 provides a scaffold to recruit accessory proteins for activation of the degradation machinery. A similar ubiquitin-mediated regulation has been observed in the activation and disassembly of the spliceosome at distinct steps of the splicing reaction 21 .
MEX-3C is not the only ubiquitin E3-ligase to bind and regulate RNA, but belongs to a family of at least 15 RNA-binding proteins with ubiquitin ligase activity. Another prominent member of this family is CNOT4, itself a component of the CCR4-NOT deadenylation complex, and has been best studied in yeast where its orthologue (Not4) has multiple functions. These include nuclear transcriptional regulation, mRNA maturation and quality control 22 , co-translation protein quality control 23 and proteasome assembly 24 . Not4's contribution to mRNA deadenylation by Ccr4 (yeast homologue of CNOT6) and Caf1 (yeast homologue of CNOT7/8) is unclear. By analogy to MEX-3C, CNOT4 may also play a role in activation of the deadenylases in the CCR4-NOT complex and mRNA degradation.
Previous studies 25 had suggested a link between ubiquitination and mRNA decay in the turnover of some AU-rich (ARE) mRNAs. Overexpression of deubiquitylating enzymes of the UBP family prolonged the half-life of specific ARE-mRNAs 25 . Furthermore proteasome inhibition prevented the rapid turnover of ARE-mRNAs, without altering the stability of non-ARE mRNAs. This latter result is in agreement with our data as HLA-A mRNA lacks AREs. Together these results suggest a differential role for ubiquitin in the regulation of mRNAs and highlight the diversity of this system. It will be critical to further dissect the mechanisms responsible for these different types of regulation. Further studies on MEX-3C and the ubiquitindependent regulation of its mRNA substrates will therefore provide an excellent platform to delineate how ubiquitin controls mRNA degradation. It is remarkable that the role of ubiquitin now extends beyond protein degradation to include the regulation and turnover of nucleic acids.

Methods
Cells, plasmids and transfections. HEK293T cells were grown in RPMI-1640 medium supplemented with 10% FCS. Cells were transfected using 293-Transit Reagent (Mirus Bio) and analysed by flow cytometry or immunoblotting at 48 or 72 h following transfection.
The Streptag-His-MEX-3C and myc-MEX-3C proteins and the FF-Luc-HLA-A2-3 0 UTR reporter are as previously described 3 . The UB-GFP mutants are as described by Boname J.M. et al. 15 . The pCMV5-HA-CNOT7 construct is previously described 10 . The Lysine mutant forms of CNOT7 were made by sitedirected mutagenesis as described in Mittal S. et al. 10 .
For the luciferase quantitative PCR assays, an TK Renilla luciferase reporter (pRL-TK) gene was co-transfected at a 1:20 ratio to provide an internal control. All assays were performed in triplicate, with the Renilla-luciferase control used to standardize transfection efficiency. Results are relative to control levels (set as 1), and expressed as the mean ± s.d. of at least three independent experiments.
IP and immunoblotting. For IPs, cells were lysed 72 h post transfection in 1% NP-40 in Tris-Buffered Saline (TBS) with 1 mM ZnCl 2 , 0.5 mM phenylmethyl sulphonyl fluoride (PMSF), 10 mM iodoacetamide (IAA) and Roche complete protease inhibitor for 30 min on ice. Strep-His-tagged proteins were IP with Streptactin sepharose beads (IBA GmbH) for 2 h. After three washes in lysis buffer, samples were eluted in SDS sample buffer (10 min at 98°C). (For myc-tagged MEX-3C proteins, IPs were done as previously described in Cano et al. 3 ) IP proteins were then separated by SDS-PAGE, and transferred to polyvinylidene difluoride (Millipore) for immunoblotting. The membranes were blocked for 1 h at room temperature, and incubated with primary antibodies overnight at 4°C in PBST containing 5% milk. Antibodies used were: rabbit polyclonal anti-RKHD2 (MEX-3C; Abcam) used at 1:5,000 dilution and rabbit anti-CNOT1 (Proteintech), rabbit anti-CNOT7 (abN1C1, GeneTex), rabbit anti-CNOT3 (abC2C3, GeneTex), Rabbit anti-PAN2 (kindly provided by Dr Jens Lykke-Andersen, University of California San Diego, USA)-all used at 1:1,000 dilution. Rabbit Anti-PCB2 was used as negative control. Membranes were developed in West Pico Extended Chemiluminescent substrate (Thermo FisherScientific). Full images of western blots and gels are shown in Supplementary Fig. 5.
For mass spectrometry analysis of RINGless MEX-3C pull-downs, HEK293T cells were transfected with pQE empty vector (EV), WT and RINGless Strep-His-MEX-3C, lysed in 1% NP-40 buffer. Lysates were incubated with 20 U ml À 1 RNase-I for 3 min at 37°C and IP on Streptactin beads as described above. Co-immunoprecipitated proteins were digested with trypsin using the filter-aided sample preparation protocol and analysed by LC-MSMS. Raw spectra were processed using Proteome Discoverer 1.2 and searched against a Uniprot Human database using Mascot Daemon 2.3.2. A false-discovery rate for peptides of 0.05 was applied and reported proteins required a minimum of two peptides and a score higher than 35.
For detection of ubiquitination on CNOT7, cell lysates from 5 Â 10 6 HEK293T cells at 72 h post transfection were lysed for 30 min in 1% SDS (in TBS þ 1 mM ZnCl 2 , 0.5 mM PMSF, 10 mM IAA, Roche complete protease inhibitor and benzonase nuclease (Sigma)) and heated for 10 min at 85°C to remove noncovalently bound ubiquitination before IP. Samples were then diluted tenfold in 0.1% Triton X-100 buffer (in TBS þ 1 mM ZnCl 2 , 0.5 mM PMSF, 10 mM IAA, Roche complete protease inhibitor) and IP for 2 h using rabbit anti-CNOT7 þ protein-A sepharose beads or anti-HA beads (EZview Red Anti-HA Affinity Gel, Sigma). For the blotting of ubiquitinated species, polyvinylidene difluoride membranes were incubated in 0.5% glutaraldehyde before probing with VU-1 antibody (LifeSensors) following the manufacture's guidelines.
Poly(A) tail-length (PAT) assay. The PCR-based poly (A) tail assay was conducted using the Poly(A) Tail-Length Assay Kit from Affymetrix and following the manufacturer's conditions. Total RNA was isolated from FACS-sorted MEX-3Cexpressing cells 3 using RNeasy kit (Qiagen). The specific upstream primer sequence for HLA-A2 mRNA is PAT-Fwd2: 5 0 -TGCATGTGTCTGTGTTCGTG-3 0 , and the downstream primer A2-3UTR-Rev 5 0 -ATCTTCTAGATTTAATAGG GAAGGAAGAAGTTACAGC-3. The universal reverse and actin 3 0 UTR primers were provided by the kit. PCR reactions were performed in 20 ml containing 1 Â PCR buffer, 0.4 mM each primer, 0.5 U Taq DNA polymerase and 100 ng of cDNA. The amplification protocol was: 2 min at 95°C, followed by 33 cycles of 15 s at 95°C, 30 s at 58°C and 30 s at 72°C, and was completed by a final extension of 5 min at 72°C. PCR products were electrophoresed on 2.0% agarose gel, stained with ethidium bromide and visualized by exposure to ultraviolet light.
RNA extraction and qRT-PCR analysis. Total RNA was extracted using the RNeasy Plus kit (Qiagen). Total RNA (2 mg) was reverse transcribed into cDNA using a poly(d)T primer and Super RT reverse transcriptase (HT Biotechnology Ltd.) following the manufacturer's instructions. Real-time qRT-PCR was performed using the ABI Prism 7700HT Sequence Detector Systems (Applied Biosystems) and SYBR Green Master mix kit (Applied Biosystems). Briefly, all reactions were performed with 120 ng of cDNA, 12.5 ml of SYBR GREEN PCR master mix and 0.2 mM forward and reverse primers in a final reaction volume of 25 ml. Cycling parameters were 95°C for 10 min, followed by 40 cycles of 94°C for 30 s, 58°C for 1 min.
Firefly and Renilla luciferase primers and PCR conditions were as described in Flow cytometry. Cells were stained with mAb BB7.2 (anti-HLA-A2) primary antibody in PBS þ 5% FCS and visualized with goat anti-mouse Cy5-conjugated secondary antibody (Jackson ImmunoResearch Laboratories). Cells were fixed in PBS with 1% paraformaldehide (PFA), read on a FACSCalibur (BD) and analysed in FlowJo.