The P-body component USP52/PAN2 is a novel regulator of HIF1A mRNA stability

Cells respond to reduced oxygen tension by executing a transcriptional programme that is principally orchestrated by HIF1A (hypoxia-inducible factor 1α) [1]. HIF1A protein is synthesized continually, but degraded rapidly by the ubiquitin–proteasome system under normal oxygen concentrations (normoxia) [2,3]. This is a result of oxygen-dependent proline hydroxylation mediated by a family of PHDs (prolyl hydroxylases). HIF1A containing this hydroxyproline modification is a substrate for the VHL (von Hippel–Lindau) E3 ubiquitin ligase complex which targets the protein for ubiquitin-mediated proteolysis [4,5]. Upon decreased oxygen concentration (hypoxia), such as that observed in solid tumours, HIF1A escapes proline hydroxylation and degradation to bind its constitutively expressed partner HIF1B (hypoxia-inducible factor 1β) and drive the expression of many genes involved in glycolysis, angiogenesis, cell survival and cancer progression [3].

Whereas the regulation of HIF1A protein is well documented, little is known about the regulation and turnover of HIF1A mRNA. The presence of multiple AREs (AU-rich elements) in the 3′-UTR (untranslated region) of HIF1A and the observation that HuR binds this 3′-UTR suggested regulation of the HIF1A transcript via AMD (ARE-mediated degradation) [6]. In support of this, the presence of AREs in the HIF1A 3′-UTR has been reported to be necessary for TTP (tristetrapolin)-mediated degradation of HIF1A mRNA during prolonged hypoxia [7,8]. In addition, the existence of an aHIF (antisense hypoxia-inducible factor) complementary to 1027 bases in the HIF1A 3′-UTR has led to the proposal that HIF1A mRNA is targeted for degradation by binding of aHIF to its 3′-UTR and exposing AREs to TTP [9]. Indeed, aHIF was shown to be up-regulated by prolonged hypoxia and correlated with a reduction in HIF1A mRNA stability [10].

USP52 (ubiquitin-specific protease 52)/PAN2 [poly(A) nuclease 2] belongs to the ubiquitin-specific protease superfamily, but exhibits no deubiquitylating activity owing to the lack of an active-site cysteine residue [11]. It also contains a C-terminal exonuclease III domain and has been well characterized in its role as a poly(A) nuclease as part of the PAN2–PAN3 deadenylation complex [12,13]. PABPC1 [poly(A)-binding protein C1] recruits the PAN2–PAN3 complex to poly(A) tails through binding PAN3 and stimulating USP52/PAN2 poly(A) nuclease activity [14,15]. However, whereas Saccharomyces cerevisiae USP52/PAN2 deletion mutants accumulate longer poly(A) tails [12], they are viable because of the CCR4–NOT1 complex providing the major cellular deadenylation activity [16], and this is also similarly the case in mammalian cells [17]. USP52/PAN2 and PAN3 have also been reported to be components of cytoplasmic P-bodies (processing bodies) [18]. Interestingly, as PABPC1 is not present in P-bodies, but is required for USP52/PAN2 nuclease activity, it is likely that USP52/PAN2 has additional functions within P-bodies [18].

In the present study, screening has identified USP52/PAN2 as an important regulator of the HIF1A-mediated hypoxic response. USP52 was required for HIF1A mRNA stability, and we provide evidence that this acts through HIF1A's AU-rich 3′-UTR, but is independent of poly(A) tail length regulation. Disrupting P-bodies by GW182 depletion displaces USP52 from P-bodies and reduces HIF1A mRNA levels. These data thereby reveal USP52 as a key component of P-bodies required for HIF1A mRNA stability.

U2OS, HeLa, HEK (human embryonic kidney)-293 and RCC4 cells were maintained in DMEM (Dulbecco's modified Eagle's medium) (Gibco) supplemented with 10% (v/v) FBS (fetal bovine serum). 786-O cells were maintained in RPMI 1640 medium (Gibco) supplemented with 10% (v/v) FBS. U2OS-HRE cells [19] were maintained in 0.5 μg/ml puromycin (Sigma). U2OS-HRE cells stably expressing YFP (yellow fluorescent protein)–USP52 were generated by transfecting U2OS-HRE cells with pEFIRES-B-eYFP-USP52 and selecting with 10 μg/ml blasticidin 48 h after transfection then were pooled and maintained with 10 μg/ml blasticidin and 0.5 μg/ml puromycin. Tetracycline-inducible FLAG–USP52 HEK-293 cells were generated using the T-REx system (Invitrogen) according to the manufacturer's instructions and maintained in the presence of 5 μg/ml blasticidin and 100 μg/ml hygromycin. Induction was carried out using 1 μg/ml tetracycline. MLN4924 was used at 3 μM final concentration for 3 h and MG132 treatment was at 20 μM for 4 h. CHX (cycloheximide) (Sigma) treatment was performed for 2 h at 5 μg/ml and puromycin treatment was for 1 h at 100 μg/ml to modify P-bodies.

Hypoxia experiments were performed by placing cells under 1% oxygen in an Invivo₂ 300 workstation (Ruskinn) for 24 h. Cell extracts for protein and RNA were taken inside the workstation to avoid reoxygenation.

Luciferase assays were performed by lysing cells in luciferase buffer [25 mM Tris/phosphate (pH 7.8), 8 mM MgCl₂, 1 mM DTT (dithiothreitol), 1% (w/v) Triton X-100, 15% (v/v) glycerol, 0.5 mM ATP, 0.5% BSA, 0.125 mM luciferin and 4 μM sodium pyrophosphate] and reading counts on an Envision 2104 plate reader (PerkinElmer). Dual-luciferase assays were carried out on HEK-293 and U2OS cells co-transfected with either Renilla luciferase–HIF1A-3′-UTR or Renilla luciferase–HIF2A-3′-UTR and pcDNA3.1+-firefly luciferase using the Dual Luciferase Reporter assay system according to the manufacturer's instructions (Promega). Renilla luciferase counts were normalized to firefly luciferase to control for transfection efficiency. All luciferase assays were carried out in triplicate and represent at least two independent experiments.

ON-TARGETplus Non-Targeting (NT), USP52, GW182, LSM1 and PAN3 (Dharmacon) siRNAs were used at a final concentration of 20 nM. siRNA transfections were carried out using Lipofectamine™ RNAi Max (Invitrogen) according to the manufacturer's instructions. USP52 and LSM1 siRNA treatments were for 48 h, GW182 siRNA treatment was for 72 h and PAN3 treatment was for 96 h. Sequences of individual siRNA duplexes are given in Supplementary Table S1 at http://www.biochemj.org/bj/451/bj4510185add.htm.

Cells were lysed in NP-40 (Nonidet P40) buffer [50 mM Hepes/KOH (pH 7.2), 400 mM NaCl, 1% NP-40, 0.2 mM EDTA and 10% (v/v) glycerol with protease inhibitor cocktail (Sigma)] and Western blotting was carried out using the following antibodies: anti-HIF1A (R&D Systems #MAB1536, mouse monoclonal, 1:1000 dilution) used in Figure 1(A) then anti-HIF1A (Novus #NB100-134, rabbit polyclonal, 1:1000 dilution) was used in all subsequent experiments, anti-tubulin (Sigma #T0198, mouse monoclonal, 1:2000 dilution), anti-USP52 [17] (rabbit polyclonal, 1:1000 dilution), anti-GFP (green fluorescent protein) (Roche #11 814 460 001, mouse monoclonal, 1:2000 dilution), anti-HIF1B (Cell Signaling Technology #5537, rabbit polyclonal, 1:1000 dilution), anti-GLUT1 (glucose transporter 1) (Thermo Scientific #RB-9052-P, rabbit polyclonal, 1:1000 dilution), anti-LDHA (lactate dehydrogenase A) (Cell Signaling Technology #2012, rabbit monoclonal, 1:1000 dilution), anti-CUL2 (cullin 2) (Invitrogen #51-1800, rabbit polyclonal, 1:2000 dilution) and anti-FLAG (Sigma #F7425, rabbit polyclonal, 1:1000 dilution).

Further methods can be found in the Supplementary Online Data at http://www.biochemj.org/bj/451/bj4510185add.htm.

To identify new mediators of the HIF1A-mediated hypoxia pathway, we used U2OS osteosarcoma cells stably expressing a firefly luciferase reporter construct fused to three tandem copies of the iNOS (inducible nitric oxide synthase) HRE (hypoxia-response element) [19] (Figure 1A). U2OS-HRE cells responded to hypoxia (1% oxygen) by stabilization of HIF1A protein and concomitantly displayed an approximately 10-fold induction of luciferase activity (Figure 1A). The increase in luciferase activity was largely dependent on HIF1A, as siRNA-mediated depletion of HIF1A reduced luciferase activity to near-normoxic levels (Figure 1A). U2OS-HRE cells were screened with our custom-assembled ‘ubiquitome’ siRNA library targeting all known and assumed components of the ubiquitin and ubiquitin-like systems. On this basis, we identified a pool of siRNAs against USP52 that decrease hypoxia-dependent HRE-response activity. Deconvolution analysis revealed that two out of four individual siRNA duplexes targeting USP52 (si1 and si3) caused over a 40% reduction in luciferase activity, and this correlated closely with the reduction in USP52 protein by these siRNAs (Figure 1B). Interestingly, we were only able to deplete USP52 by approximately 50% as judged by immunoblot (Figure 1B) and real-time PCR (Supplementary Figure S1A at http://www.biochemj.org/bj/451/bj4510185add.htm) analysis. This is consistent with previous reports in mouse NIH 3T3 cells, where siRNA was reported to reduce USP52 by a maximum of ~65% compared with control siRNA [17]. USP52 knockdown specifically impaired the hypoxia response, as it caused no such impairment to the NF-κB (nuclear factor κB) response after TNFα (tumour necrosis factor α) stimulation of the HeLa C57A cell line [20] (Supplementary Figure S1B). Next we generated U2OS-HRE cell lines stably expressing YFP-tagged USP52 resistant to either USP52 siRNA si1 or si3 (Figure 1C) and assessed the ability of siRNA-resistant USP52 to rescue the impaired hypoxic response upon endogenous USP52 knockdown. Hypoxia-induced luciferase activity was rescued to approximately 80% and 95% in si1- and si3-resistant cells respectively, thus validating USP52 as a novel activator of the HIF1A-mediated hypoxic response (Figure 1D).

To investigate further the role of USP52 in the HIF1A-mediated hypoxic response, U2OS cells were treated with individual USP52 siRNA duplexes 1 and 3 and exposed to hypoxia. Knockdown of USP52 correlated with a specific decrease in HIF1A, but not HIF1B, protein levels (Figure 2A), suggesting that decreased HIF1A protein is responsible for the impaired hypoxia response. GLUT1 and LDHA are both hypoxia-responsive genes which accumulate in U2OS cells in a HIF1A-dependent manner (Supplementary Figure S1C). Examination of GLUT1 and LDHA levels by immunoblot analysis of hypoxic U2OS cells revealed that USP52 depletion impairs the accumulation of GLUT1 and LDHA in response to hypoxia (Figure 2A), thus demonstrating that USP52 is required to potentiate the HIF1A-mediated hypoxic response. Similar results were obtained in mouse cells, indicating that USP52 has a conserved role in the hypoxia response (results not shown). Furthermore, real-time RT (reverse transcription)–PCR analysis from USP52-depleted U2OS cells demonstrated that the accumulation of the HIF1A target genes CA9 (carbonic anhydrase IX), PHD2 and VEGF (vascular endothelial growth factor) in response to hypoxia were all reduced (Figure 2B), confirming the role of USP52 in the HIF1A pathway. Interestingly, it has been reported previously that USP52/KIAA0710 mRNA itself was induced 3.1-fold in response to hypoxia determined by microarray analysis of human renal cancer 786-O cells [21]. Whereas examination of the USP52 promoter region revealed the presence of an HRE consensus motif (RCGTG) (Supplementary Figure S1D), no hypoxic increase in USP52 protein was detected in either 786-O or RCC4 cells (Supplementary Figure S1D), or in U2OS cells, HeLa cells or MEFs (mouse embryonic fibroblasts) (results not shown).

To gain insight into how USP52 may alter HIF1A protein levels, we performed protein network analysis of known USP52-interacting partners (Supplementary Figure S2A at http://www.biochemj.org/bj/451/bj4510185add.htm). Interestingly, USP52 was reported previously to immunoprecipitate TCE (transcription elongation factor) B1/elongin C and TCEB2/elongin B [22], which constitute part of the VHL E3 ligase complex required for the bulk of HIF1A degradation. The VHL complex is a cullin-type E3 ligase which requires CUL2 NEDD8ylation for ligase activity. To address whether USP52 negatively regulates HIF1A protein through interactions with the VHL complex, U2OS cells were treated with MLN4924 [23] to inhibit the NEDD8 (neural-precursor-cell-expressed developmentally down-regulated 8)-interacting enzyme and block VHL activity. Non-ubiquitylated HIF1A accumulated in normoxia upon MLN4924 treatment concomitant with the blockage of CUL2 NEDD8ylation (Figure 3A). However, depletion of USP52 caused a reduction in HIF1A protein even in the presence of MLN4924, suggesting that USP52 affects the HIF1A pathway independently of the VHL complex (Figure 3A). RCC4 cells lack functional VHL and therefore constitutively express HIF1A in normoxia [4]. In support of a VHL-independent role of USP52 in the HIF1A pathway, knockdown of USP52 in RCC4 cells caused a subsequent decrease in HIF1A levels (Figure 3B). In addition to VHL, several other proteins have been reported to be involved in targeting HIF1A for proteasomal degradation, such as RACK1 (receptor for activated C-kinase 1) [24], HAF (hypoxia-associated factor) [25] and CHIP [C-terminus of the Hsc (heat-shock cognate) 70-interacting protein] [26]. In order to exclude USP52 knockdown promoting HIF1A depletion through alternative degradation pathways, U2OS cells depleted of USP52 were treated with the proteasome inhibitor MG132. Proteasome inhibition caused the accumulation of ubiquitylated HIF1A (Supplementary Figure S2B) in NT siRNA-treated, but not USP52-depleted, cells, confirming that regulation of HIF1A by USP52 does not occur at the level of proteasomal degradation (Supplementary Figure S2B). To determine whether USP52 knockdown alters HIF1A mRNA levels, real-time RT–PCR was performed on cDNA prepared from U2OS cells. USP52 depletion caused an approximately 60% reduction in the abundance of HIF1A mRNA (Figure 3C). As a control, HIF1A siRNA depleted HIF1A mRNA by over 90% (Figure 3C). To assess whether depletion of USP52 also affected HIF2A mRNA steady-state levels, HeLa cells were used as they express higher levels of HIF2A than U2OS cells. Whereas USP52 depletion caused a comparable reduction in HIF1A mRNA in HeLa cells, the levels of HIF2A actually increased approximately 2-fold (Figure 3D). This demonstrates that USP52 knockdown specifically reduces HIF1A mRNA abundance, and supports the previous observation that decreased HIF1A mRNA is associated with up-regulation of HIF2A [19]. We also assessed the effect of USP52 depletion on mRNA levels of the unrelated genes ERG (Ets-related gene) and CTNNB1 (β-catenin). Transcript levels of both ERG and CTNNB1 were increased ~1.5-fold upon USP52 depletion (Figure 3D), consistent with the expected role of USP52 in mRNA degradation-dependent deadenylation. These results show that the role of USP52 in maintaining HIF1A mRNA levels displays a certain level of specificity.

USP52 depletion could modify the steady-state levels of HIF1A mRNA through modulating its transcription or stability. To distinguish between these possibilities, we performed actinomycin D chase experiments to measure the half-life of HIF1A mRNA after USP52 knockdown. Actinomycin D inhibits RNA polymerase II to block transcription and therefore allows measurement of mRNA decay. Depletion of USP52 in U2OS cells dramatically enhanced the degradation of HIF1A mRNA (Figure 4A), where the half-life of HIF1A in NT-treated cells was 214 min (Supplementary Figure S3A at http://www.biochemj.org/bj/451/bj4510185add.htm) and reduced to 35 min upon USP52 depletion (Supplementary Figure S3B). In order to determine whether the effect of USP52 depletion on HIF1A mRNA stability was associated with a change in HIF1A poly(A) tail length, we analysed poly(A) tail length in U2OS cells treated with NT or USP52 siRNA. Surprisingly, there was no apparent difference in the length of the HIF1A poly(A) tail after USP52 depletion, indicating that USP52 stabilizes HIF1A mRNA in a poly(A)-tail-length-independent manner (Figure 4B). The presence of multiple AREs in HIF1A's 3′-UTR results in the regulation of HIF1A by AMD [7,8]. To determine whether USP52 regulates HIF1A stability through stabilizing its 3′-UTR, we used a Renilla luciferase reporter construct fused to the HIF1A 3′-UTR. Depletion of USP52 in HEK-293 cells inhibited HIF1A 3′-UTR activity by approximately 50% (Figure 4B), supporting a role for USP52 in HIF1A stabilization by preventing its AMD. Similar results were obtained upon USP52 knockdown in U2OS cells (Supplementary Figure S3C). Interestingly, USP52 knockdown resulted in an increase in the levels of HIF2A Renilla 3′-UTR in HEK-293 cells (Figure 4C), suggesting that the observed antagonism between HIF1A and HIF2A levels [19] (Figure 3D) may be regulated in a manner depending on the HIF2A 3′-UTR. To rule out any effects of USP52 on HIF1A translation, we performed polysome profiling on U2OS cells treated with either NT or USP52 siRNA. There were no clear differences in the distribution of HIF1A mRNA between the non-translating and translating fractions (Supplementary Figure S4 at http://www.biochemj.org/bj/451/bj4510185add.htm), suggesting the effects of USP52 depletion are not mediated at the translation level. Collectively, these results demonstrate that USP52 regulates HIF1A mRNA stability in a 3′-UTR-dependent but poly(A)-tail-length-independent manner.

To obtain insight into the observed USP52-mediated stabilization of HIF1A mRNA, we performed proteomic analysis of USP52 immunoprecipitates to identify new USP52-associated proteins. HEK-293 cells which stably express FLAG-tagged USP52 in a tetracycline-responsive manner were generated (Figure 5A). FLAG–USP52 expression was induced for 24 h before anti-FLAG-conjugated beads were used to immunoprecipitate FLAG–USP52 complexes. Immunoprecipitated material from induced and uninduced control cells was separated by denaturing gel electrophoresis and silver-stained to reveal the presence of specific USP52-associated proteins of various sizes (Figure 5B). The remaining material was subject to in-solution tryptic digestion and LC (liquid chromatography)–MS/MS (tandem MS) analysis to identify interacting proteins (Figure 5B). The full dataset is listed in Supplementary Table S4 at http://www.biochemj.org/bj/451/bj4510185add.htm. We were interested by the identification of the P-body components TRIM21 (tripartite motif-containing 21), PCBP1 and PAN3 in USP52 immunoprecipitates (Supplementary Table S4), given the role of P-bodies in mRNA storage and/or degradation [27]. To confirm that USP52 is a component of P-bodies [18], immunostaining of U2OS cells with antibodies against USP52 and the essential P-body component GW182 demonstrated that USP52 co-localizes to P-bodies (Figure 5D). Quantification revealed that 92.5% of GW182 P-bodies were also USP52-positive (Figure 5D). Treatment of cells with CHX or puromycin is known to abolish or increase respectively the number of P-bodies [27]. We confirmed that USP52 P-body foci were likewise abolished by CHX treatment, while the number increased upon puromycin treatment (Supplementary Figure S5A at http://www.biochemj.org/bj/451/bj4510185add.htm). This confirmed that USP52 is a genuine P-body component and raised the possibility that USP52 may serve a role in regulating HIF1A mRNA through its association with P-bodies.

The co-localization of USP52 with P-bodies led us to hypothesize that P-bodies might be important for HIF1A mRNA stability. GW182 is essential for P-body integrity, as its depletion abolishes the presence of visible P-bodies [28]. To determine whether P-body disruption correlated with changes in HIF1A mRNA levels, we depleted GW182 from U2OS cells. GW182 siRNA caused a decrease in the proportion of cells containing P-bodies, whereas any remaining P-bodies were visibly smaller (Figure 6A). USP52 foci were dispersed upon GW182 depletion, confirming that the presence of USP52-positive P-bodies is dependent on GW182 (Figure 6A) and quantification of P-bodies revealed that the percentage of cells containing at least one P-body decreased from 80% in NT siRNA-treated cells to approximately 20% in GW182 siRNA-treated cells (Supplementary Figure 5B). Importantly, we observed a reduction in HIF1A mRNA levels of almost 80% upon GW182 depletion (Figure 6B), suggesting that P-body integrity plays an important role in maintaining HIF1A mRNA levels. Whereas GW182 depletion dispersed USP52-positive P-bodies, it had no effect on steady-state levels of USP52 protein, suggesting that HIF1A mRNA levels were not affected by simply reducing the amount of USP52 protein (Supplementary Figure S5C). As GW182 has also been proposed to be involved in regulating the miRNA (microRNA) pathway, it was important to confirm that dispersing P-bodies by depleting another P-body component independent of the miRNA pathway had the same effect on HIF1A mRNA. We therefore depleted LSM1, which dispersed P-bodies (Supplementary Figure S5C) while also resulting in a reduction of HIF1A mRNA to approximately 20% (Figure 6C). PAN3 has been shown to recruit USP52 to P-bodies, and its depletion was also shown to reduce P-body number as well as destabilize AMD substrates [18]. We therefore depleted PAN3 in U2OS cells and observed that HIF1A mRNA levels were decreased to 60% compared with NT siRNA (Figure 6D). Altogether, these results suggest that P-bodies are important regulators of steady-state HIF1A mRNA levels, and that USP52 is a key component of P-bodies required for HIF1A mRNA stability.

The finding that USP52 is a P-body component required for HIF1A mRNA stability coupled to our finding that reducing the number of cellular P-bodies depletes HIF1A mRNA raised the possibility that HIF1A mRNA might be present within P-bodies. To test this directly, we performed FISH (fluorescent in situ hybridization) with antisense Texas-Red-X-labelled probes specific to HIF1A mRNA on U2OS cells transfected with the P-body marker GFP–DCP1A (decapping enzyme 1A). A GFP-tagged P-body marker was used to circumvent problems associated with the destruction of endogenous epitopes upon harsh denaturing conditions required for FISH. HIF1A antisense probes formed discrete cytoplasmic foci which co-localized with GFP–DCP1A-positive P-bodies (Figure 7A), suggesting that HIF1A mRNA is a P-body constituent. Non-specific fluorescence was deduced by using HIF1A sense control probes, which did not show an obvious co-localization with GFP–DCP1A (Figure 7A). The proportion of GFP–DCP1A P-bodies co-localizing with HIF1A antisense probes was calculated to be 52%, whereas antisense probes showed a 10% co-localization rate (Figure 7B), probably due to background fluorescence showing coincidental overlap with GFP–DCP1A. Therefore a proportion of HIF1A mRNA is found within P-bodies, suggesting that these may act as storage sites for HIF1A mRNA.

HIF1A is the master regulator of the cellular response to hypoxia which is responsible for executing a transcriptional programme endowing cells with the capacity to deal with hypoxic stress [3]. Whereas the oxygen-dependent regulation of HIF1A protein has been well described, it is becoming increasingly apparent that regulation of HIF1A mRNA also plays an important role in regulating HIF1A-dependent processes. In the present study, we have identified the P-body component USP52/PAN2 as an important stability factor for HIF1A mRNA, and have demonstrated that it prevents degradation of HIF1A mRNA in a 3′-UTR-dependent, but poly(A)-tail-length-independent, manner.

P-bodies are dynamic structures which are important in both mRNA decay and miRNA-mediated translational silencing, acting as a store for mRNAs before their release back into active polysomes [27]. It is clear that miRNA pathways are important regulators of the hypoxic response and that several miRNAs are up-regulated in response to hypoxia [29]. In addition, a Drosophila genome-wide siRNA screen identified several components of the miRNA machinery that were required for a complete hypoxic response, including the P-body components Ago1 and GW182 [30]. The fact that P-bodies are reported to be a site of ARE-containing mRNA storage and AMD [31] suggested that HIF1A mRNA may be stored and/or degraded in P-bodies. Our finding that depletion of the P-body component USP52 destabilizes HIF1A mRNA via its AU-rich 3′-UTR suggests that USP52 plays a key role in preventing AMD of HIF1A mRNA in P-bodies. This may occur through preventing the association of HIF1A 3′-UTR with AMD-promoting proteins such as TTP, which is present in P-bodies and known to promote the degradation of ARE-containing mRNAs including HIF1A [7,8,31]. Although we cannot rule out the possibility of indirect effects of USP52 on HIF1A mRNA stability, our finding that HIF1A mRNA and USP52 protein both reside in P-bodies coupled with the fact that abolishing visible P-bodies also destabilizes HIF1A mRNA supports a more direct role for USP52 in regulating HIF1A mRNA.

Deadenylation of mRNA occurs through the concerted actions of the PAN2–PAN3 and CCR4–CAF1 complexes [17], and is a necessary first step in all major mRNA decay pathways including AMD and miRMD (miRNA-mediated decay) [32]. Although USP52/PAN2 is dispensable in yeast [12], consistent with the fact that CCR4 provides the major cellular deadenylation activity [16], our finding that USP52 is required for HIF1A mRNA stability was surprising. However, this apparent paradox may be explained by USP52 playing additional roles to deadenylation when localized in P-bodies, which is supported by the finding that PABPC1 (which is required for USP52 nuclease activity) is absent from P-bodies [18,33]. Indeed, we did not observe any effect of USP52 depletion on the length of the HIF1A poly(A) tail (Figure 4B), and it is consistent with our finding that, whereas USP52 depletion destabilized HIF1A mRNA, overexpression of a nuclease-dead USP52 was as efficient as wild-type USP52 in enhancing the HIF1A-mediated hypoxic response (results not shown). Interestingly, it has also been reported that PAN3 depletion causes a reduction in P-body number concomitant with a destabilization of ARE-containing transcripts [18]. Furthermore, these authors showed that PAN3 was able to recruit USP52/PAN2 to P-bodies [18]. It is therefore a possibility that PAN3 depletion prevents USP52 localization to P-bodies and thus contributes to the destabilization of AMD substrates, as we have observed in the case of HIF1A. In this scenario, USP52 may regulate the stability of other ARE-containing transcripts in addition to HIF1A.

USP52 has also been implicated in miRNA pathways, as it was shown to be recruited by GW182 through interactions with PAN3 in humans, Drosophila and Caenhorhabditis elegans [34–37]. It is therefore possible that GW182 recruits USP52 to P-bodies through interactions with PAN3 [18]. We observed that P-body dispersal correlates with decreased levels of HIF1A mRNA, suggesting that the degradation of HIF1A mRNA is prevented by USP52 within P-bodies. This is consistent with previous observations that ARE-containing transcripts are recruited to and stored in P-bodies [31], and we confirm further that HIF1A mRNA is present within P-bodies [38,39]. In addition, it was shown that genetic disruption of GW182 in Drosophila dramatically reduced the mRNA levels of the HIF1A homologue Sima in response to hypoxia [30], suggesting that there is a highly conserved role for P-bodies in maintaining HIF1A mRNA levels.

USP52 exhibits homology with the ubiquitin-specific protease family, but is not an active DUB (deubiquitinating enzyme) owing to mutations in two of the three residues of the catalytic triad [11]. We therefore were initially interested in the possible role of USP52 as a pseudo-DUB. Pseudo-DUBs may be predicted to act in a dominant-negative manner preventing other DUBs from gaining access to their substrate, or alternatively they may bind ubiquitylated proteins and play a scaffolding role in protein–protein interactions. Interestingly, we found several peptides corresponding to ubiquitin in our USP52 immunoprecipitates, and transiently overexpressing USP52 harbouring a Cys-box deletion mutation in the UCH (ubiquitin C-terminal hydrolase) domain resulted in a decrease in HIF1A protein (results not shown), suggesting that this domain may be important in protecting HIF1A mRNA in P-bodies. Interestingly, it has recently been reported that the RNA-binding E3 ubiquitin ligase MEX-3C is required for the RING-dependent 3′-UTR-mediated degradation of HLA-A2 mRNA, suggesting that there is a wider-role for ubiquitin-handling proteins in mRNA stability pathways [40]. Intriguingly, we identified the E3 ubiquitin ligase and P-body component TRIM21 as a USP52-associated protein, opening the possibility that USP52 may be involved in regulating the function of TRIM21 or its substrates. Future studies exploring the potential role of USP52 as a pseudo-DUB in its role in P-body function will be useful in expanding the understanding of ubiquitin-mediated regulation of RNA metabolism.

aHIF

antisense hypoxia-inducible factor

ARE

AU-rich element

AMD

ARE-mediated degradation

CA9

carbonic anhydrase IX

CHX

cycloheximide

CTNNB1

β-catenin

CUL2

cullin 2

DCP1A

decapping enzyme 1A

DUB

deubiquitinating enzyme

ERG

Ets-related gene

FBS

fetal bovine serum

FISH

fluorescent in situ hybridization

GFP

green fluorescent protein

GLUT1

glucose transporter 1

HEK

human embryonic kidney

HIF1A

hypoxia-inducible factor 1α

HIF1B

hypoxia-inducible factor 1β

HRE

hypoxia-response element

LC

liquid chromatography

LDHA

lactate dehydrogenase A

miRNA

microRNA

MS/MS

tandem MS

NEDD8

neural-precursor-cell-expressed developmentally down-regulated 8

NP-40

Nonidet P40

NT

Non-Targeting

PABPC1

poly(A)-binding protein C1

PAN2

poly(A) nuclease 2

P-body

processing body

PHD

prolyl hydroxylase

RT

reverse transcription

siRNA

short interfering RNA

TCE

transcription elongation factor

TRIM21

tripartite motif-containing 21

TTP

tristetrapolin

USP52

ubiquitin-specific protease 52

UTR

untranslated region

VEGF

vascular endothelial growth factor

VHL

von Hippel–Lindau

YFP

yellow fluorescent protein

John Bett performed all of the experiments in the study, except those shown in Supplementary Figure S1(B), which were performed by Adel Ibrahim. Amit Garg prepared Supplementary Figure 2(A). Adel Ibrahim and Amit Garg performed preliminary experiments, initiated the screen and analysed the data. Van Kelly and Patrick Pedrioli performed the MS analysis. Sonia Rocha contributed reagents, equipment and expert advice. John Bett, Sonia Rocha and Ronald Hay designed the experiments and analysed the data. John Bett and Ronald Hay wrote the paper.

We thank Dr Ann-Bin Shyu (Department of Biochemistry and Molecular Biology, University of Texas Health Science Centre, Houston, TX, U.S.A.) for proving the anti-USP52/PAN2 antibody and Professor Marvin Fritzler (Department of Biochemistry and Molecular Biology, University of Calgary, Calgary, AB, Canada) for providing the anti-GW182 antibody. We are grateful to Professor Peter Ratcliffe (Nuffield Department of Medicine, University of Oxford, Oxford, U.K.) for providing RCC4 cells and Mr Ellis Jaffray (College of Life Sciences, University of Dundee) for advice on P-body staining and quantification. We also thank Nikki Wood, Mel Wightman and Rachel Toth (all at the Scottish Institute for Cell Signalling) for constructs and to Gabriella Alexandru, Thimo Kurz and Arno Alpi laboratories of the Scottish Institute for Cell Signalling for provision of reagents.

The work was funded by the Scottish Institute for Cell Signalling, the Wellcome trust and GlaxoSmithKline.