The Deubiquitinase USP47 Stabilizes MAPK by Counteracting the Function of the N-end Rule ligase POE/UBR4 in Drosophila

RAS-induced MAPK signaling is a central driver of the cell proliferation apparatus. Disruption of this pathway is widely observed in cancer and other pathologies. Consequently, considerable effort has been devoted to understanding the mechanistic aspects of RAS-MAPK signal transmission and regulation. While much information has been garnered on the steps leading up to the activation and inactivation of core pathway components, comparatively little is known on the mechanisms controlling their expression and turnover. We recently identified several factors that dictate Drosophila MAPK levels. Here, we describe the function of one of these, the deubiquitinase (DUB) USP47. We found that USP47 acts post-translationally to counteract a proteasome-mediated event that reduces MAPK half-life and thereby dampens signaling output. Using an RNAi-based genetic interaction screening strategy, we identified UBC6, POE/UBR4, and UFD4, respectively, as E2 and E3 enzymes that oppose USP47 activity. Further characterization of POE-associated factors uncovered KCMF1 as another key component modulating MAPK levels. Together, these results identify a novel protein degradation module that governs MAPK levels. Given the role of UBR4 as an N-recognin ubiquitin ligase, our findings suggest that RAS-MAPK signaling in Drosophila is controlled by the N-end rule pathway and that USP47 counteracts its activity.


Introduction
The RAS-MAPK pathway is one of the principal signaling conduits controlling proliferation and differentiation in metazoans. Perturbations in pathway activity are closely associated to oncogenesis and developmental disorders [1,2]. Signaling through the small GTPase RAS is typically initiated by signal-receiving transmembrane receptors. Active RAS triggers the successive activation of the three pathway kinases: RAF, MEK, and MAPK/ERK. A complex network of factors is now known to regulate the steps leading up to MAPK activation [3,4]. To date, much effort has been devoted to identifying and characterizing the post-translational mechanisms that govern pathway signaling dynamics [3][4][5]. In the case of MAPK, MEK is the principal activator that operates by phosphorylating MAPK's activation loop, and MAPK-specific phosphatases dephosphorylate these residues to inactivate the kinase. Other sources of regulation include scaffold proteins and factors that control subcellular localization [6]. As is the case for most other RAS-MAPK pathway components, very little is known about the processes that act to regulate MAPK abundance.
Recently, we reported the identification of a series of factors in a genome-wide RNAi screen that impacted RAS-MAPK signaling at a pre-translational level [7,8]. Unexpectedly, these factors were in large part associated with either the transcription or splicing of mapk pre-mRNA transcripts, indicating that this might be an important source of regulatory input into this pathway. Here, we focus on USP47 (also commonly referred to as UBP64E; FBgn0016756), a factor that was also identified in the aforementioned RNAi screen. USP47 is a deubiquitinase (DUB) of the ubiquitin specific protease (USP) family ( Fig 1A) [9]. It has previously been associated with the regulation of the transcription factors TTK (tramtrack; FBgn0003870) and SLBO (FBgn0005638) in Drosophila [10,11]. Its human orthologue, USP47 interacts with the beta-TRCP E3 ligase complex [12] and has been found to regulate base-excision repair (BER) by controlling the levels of Polymerase β [13,14]. USP47 has also been linked to axonal growth by working antagonistically to the E3 ligase CHIP in regulating the microtubule severing protein katanin-p60 [15]. Additionally, it has been associated with cell-cell adhesion through its ability to prevent E-cadherin degradation, which stabilizes adherens junction formation [16]. Finally, USP47 has recently been shown to control Wnt signaling by controlling β-catenin stability both in human and Drosophila cells [17].
In the present study, we show that USP47 acts post-translationally to stabilize Drosophila MAPK (rl; FBgn0003256) levels both in cell culture and in vivo, and that this mechanism is independent of MAPK activity. Moreover, this event appears to be specific to MAPK, as no other core components of the pathway nor other Drosophila MAPK-like proteins, JNK (bsk; FBgn0000229) and p38B (FBgn0015765), are modulated by USP47 function. To identify putative MAPK destabilizing components that might act in opposition to USP47, we conducted a targeted RNAi screen for genetic interactors of Usp47 depletion. Interestingly, this approach led to the identification of three factors, namely, the E2 ubiquitin conjugating enzyme UBC6 (FBgn0004436) and the putative E3 ligases POE (FBgn0011230; the fly UBR4 ortholog) and UFD4 (FBgn0032208), which restored MAPK levels when individually codepleted with USP47. Moreover, in a search for additional factors that act similarly to the  [8]. Phosphorylated MAPK was measured by quantitative microscopy and normalized to a GFP dsRNA control [7]. (C) MAPK activation is induced by three different receptor tyrosine kinases (RTKs) acting upstream: InR through insulin stimulation; EGFR-expressing cells stimulated with the Spitz ligand; or a heat-shock inducible constitutively activated form of Sevenless (Sev S11 ). JNK pathway activation induced by RAC1 V12 is used as a negative control. (B-C) A dsRNA targeting the Exon Junction Complex (EJC) component mago, a factor also positioned downstream of MEK and known to reduce MAPK expression [7], is shown for comparison. The cnk dsRNA is a control for a factor known to act at the level of RAF [24]. (D) The sal EPv -Gal4 drives expression in the wing pouch, which corresponds to a segment of the wing blade extending from the L2 wing vein to the L4-5 intervein with a weaker expression area extending in the periphery of this region [25,26]. A mapk gain of function Sevenmaker (Sem) mutant induced the production of extra wing vein material. Two different Usp47 RNAi lines suppressing the extra wing vein material generated by mapk Sem are shown (these correspond to the following fly lines: Usp47 (1), VDRC line GD26027; Usp47 (2), NIG line 5486R-3). Raw data for (B and C) can be found in S1 Data.
Remarkably, UBC6, UBR4, and UFD4 are all linked to a mechanism called the N-end rule pathway, which has mostly been studied in yeast and whose relevance to metazoan cells remains largely unknown. The N-end rule pathway is a ubiquitin/proteasome-dependent protein degradation process that hinges on the recognition of a degron located at the N-terminal extremity (N-degron) of peptides and whose principal determinant is the identity of the N-terminal amino acid. This N-terminal amino acid is recognized by the ubiquitin protein ligase E3 component n-recognin (UBR) box domain of a class of E3 ligases called N-recognins, which includes POE/UBR4 [18][19][20][21]. UBC6 is the fly ortholog of RAD6, which is the main E2 enzyme that functions with UBR box N-recognins [18,22], whereas UFD4 is a poorly characterized HECT domain E3 ligase whose yeast homolog, Ufd4, has also been linked to N-end rule function [23].
These new findings suggest that the N-end rule pathway plays a role in establishing MAPK levels and, together with previously characterized pre-translational mechanisms, fine-tunes MAPK signaling output during development.

USP47 Modulates RAS-MAPK Signaling Downstream of MEK
We initially identified Usp47 in a genome-wide RNAi screen for factors modulating RAS V12induced (Ras85D; FBgn0003205) pMAPK activity in Drosophila S2 cells [8]. Like most of the factors identified in this screen, Usp47 was found to act downstream of MEK (Dsor1; FBgn0010269). Indeed, signaling by constitutively active RAS, RAF, or MEK were all equally suppressed by a dsRNA targeting Usp47 to a degree that was similar to that induced by mago (FBgn0002736), a factor known to act downstream of MEK ( Fig 1B) [7]. In contrast, cnk (FBgn0021818), a well characterized regulator of RAF activation [24], reduced pMAPK signaling in the active RAS assay, but had no impact in active RAF (FBgn0003079) or MEK assays ( Fig 1B). Usp47 also demonstrated a broad capacity to modulate MAPK signaling in different contexts; signal induced by the Insulin-like Receptor (InR; FBgn0283499), Sevenless (sev; FBgn0003366), or Epidermal Growth Factor Receptor (Egfr; FBgn0003731) was suppressed to the same extent by Usp47 RNAi (Fig 1C). JNK activity induced by RAC1 V12 (Rac1; FBgn0010333) was, however, not modified by Usp47 depletion (Fig 1C).
We next conducted genetic interaction experiments to validate Usp47's newfound function in vivo. In flies, RAS-MAPK signaling is important in many aspects of development, including the induction of photoreceptor and cone cell differentiation in the eye [27] as well as in promoting the formation of veins during wing development [28,29]. Overexpression of a mapk gain-of-function allele (mapk Sem ) [30] in Drosophila wings using a sal EPv -Gal4 driver led to the production of extra wing vein material (Fig 1D) [31]. The severity of this ectopic wing vein phenotype was significantly reduced by co-expressing Usp47 RNAi (Fig 1D). In the eye, genetic lesions that lower pathway activity cause a readily observable rough eye phenotype in adults that is due to missing photoreceptor cells (S1A Fig) [32,33]. Two alleles of Usp47 (Usp47 Δ1 and Usp47 Δ2 ) that reduce USP47 protein expression [10] were found to increase the severity of the rough eye phenotype of rl 1 homozygous flies carrying a hypomorphic allele of mapk (S1A Fig). Lastly, similar observations were also made in hemizygotes for csw lf (S1B Fig), a dominant negative form of the csw/shp-2 (FBgn0000382) phosphatase that regulates RAS activity [34]. In sum, these experiments support our cell culture-based observations pointing towards a positive regulatory role for USP47 in MAPK signaling.

USP47 Regulates MAPK Levels Post-translationally
Like many of the candidates identified in our initial genome-wide screen, Usp47 RNAi was found to cause a decrease in MAPK protein levels (Fig 2A) [8], whereas it did not visibly alter the expression levels of JNK (FBgn0000229) and p38b (FBgn0024846), two other members of the MAPK family, nor of other RAS-MAPK pathway proteins (Fig 2B). This effect was also observed in vivo, where knocking down Usp47 in both fly larval wing and eye imaginal discs caused a reduction in MAPK protein levels (Fig 2C and 2D). Moreover, lysates prepared from Usp47 Δ1 homozygotes were found to contain less MAPK protein than WT controls (Fig 2E). A clear distinction between Usp47 and other candidates from our initial screen arose when examining their impact on mapk mRNA. Indeed, while the other factors caused changes in mapk mRNA levels or splicing, Usp47 had no impact at these levels ( Fig 2F and 2G). Moreover, it did not cause a change in the rate of mapk translation as measured by polysomal loading of mapk mRNA transcripts (Fig 2H). Thus, regulation of MAPK levels by USP47 does not appear to be pre-translational.
In order to determine if USP47 is acting at a post-translational level, we resorted to pulsechase metabolic labeling with [35S]-methionine. We found labeled MAPK to be degraded more rapidly in cells treated with Usp47 RNAi (Fig 3A and 3B). Indeed, in comparison to a half-life of 13.68 h in control conditions, the half-life of MAPK went down to 10.34 h upon USP47 depletion, which is consistent with the~50% reduction of steady-state levels of MAPK typically seen by western blot following 4 d of dsRNA treatment (Fig 2A). Importantly, an exogenous HA-MAPK cDNA, but not HA-JNK or HA-p38B, stably expressed in S2 cells also displayed sensitivity to Usp47 dsRNA (S2A Fig). Taken together, these experiments strongly suggest that USP47 functions post-translationally to specifically regulate MAPK protein levels.
One likely mechanism for post-translational control of MAPK levels by a deubiquitinase would be the removal of ubiquitin moieties from specific lysine residues on MAPK. As no ubiquitin ligases are associated to the regulation of MAPK expression in Drosophila, and as we had not identified any such factors in our genome-wide RNAi screen, we next sought to determine if the ubiquitin proteasome system (UPS) was required for the destabilization of MAPK following Usp47 knockdown. We found that co-depletion of Uba1-the sole ubiquitin-activating enzyme (E1) in Drosophila-and Usp47 by RNAi rescued MAPK levels to near baseline ( Fig 3D). Proteasome inhibition using epoxomicin had a similar effect ( Fig 3E). Thus, these results indicate that the ubiquitin-proteasome system acts in opposition to USP47 to control MAPK protein levels.
Despite these indicators of UPS involvement, we did not, however, detect any polyubiquitination of MAPK when co-expressing HA-tagged ubiquitin in S2 cells (S3 Fig), which contrasted with the readily observable ubiquitination of controls such as RAF and KSR (FBgn0015402) (S3 Fig). Since ubiquitin might likely be attached to an exposed lysine at the surface of MAPK, we also mutated all predicted surface-exposed lysine residues. Unexpectedly, MAPK modified in this manner still responded to Usp47 RNAi to a degree comparable to WT (Fig 3C). In contrast, mutation of all lysine residues did abrogate the effect of Usp47 ( Fig 3C). However, this latter mutant was not phosphorylatable by MEK (unlike the WT or surface-exposed lysine mutant; S2B Fig), thus suggesting that its structure is significantly altered. Accordingly, mutations predicted to disrupt the N-or C-lobe of the kinase domain of MAPK also displayed a reduced sensitivity to Usp47 RNAi (S2C Fig). Moreover, deletion of MAPK's C-terminal α L16 helix, which interacts with the N-lobe and stabilizes it [6,35,36], also led to a loss of sensitivity to Usp47 knockdown (S2D Fig). Thus, while we could not identify specific MAPK lysine residues as potential ubiquitination sites to explain the requirement in USP47 activity, our data strongly suggest that a properly folded MAPK protein is required. Usp47 regulates MAPK levels. (A) Usp47 depletion causes a reduction in MAPK levels probed by immunoblot. S2 cells were treated with the indicated dsRNAs. The EJC component eIF4AIII was used as a control for a factor that reduces MAPK expression [7]. ACT5C (FBgn0000042), α-TUBULIN (FBgn0003884), and AKT (FBgn0010379) levels (loading controls) were not impacted by Usp47 depletion. Densitometry quantifications are provided for MAPK levels normalized to the AKT loading controls (similar quantification results are obtained by using either the ACTIN or the TUBULIN controls for normalization (see S1 Data). Two AKT isoforms (short and long) are detected in S2 cells by western blot. We exclusively used the shorter variant as loading control for quantification purpose, as its levels are in the same detection range as MAPK from the same sample lysate. The levels of the long AKT isoform were slightly variable across experiments and thus were not used for quantification. Their experimental variation was likely caused by their lower abundance, which in turn would produce signals outside the linear range of the film. This experiment was performed in seven replicates, and pvalues were calculated (paired two tailed Student's t test) comparing values to the Rluc control (*: p < 0.05; **: p < 0.01). (B) Usp47 depletion did not alter the levels of other RAS-MAPK pathway components (RAS, MEK, and CNK), nor of two other MAPK family kinases (JNK and p38). As AKT, ACTIN, and TUBULIN were found to be interchangeable loading controls in experiments that address the impact of USP47 on MAPK levels (A), we decided for convenience to monitor only AKT levels as loading control here and in most subsequent experiments. AKT was selected over ACTIN or TUBULIN because its transcript levels, according to modENCODE, are similar to those of mapk (unlike for actin or tubulin, which are 200-500-fold more highly expressed), and thus allowed the comparison of similarly expressed genes. Importantly, AKT levels are not known to be modulated by MAPK signaling. For instance, we never observed variations in AKT levels following perturbations in RAS-MAPK signaling [7,8], which is also reflected here in (B). Finally, the other monitored proteins in (B) also serve as points of comparison to infer that AKT levels are not impacted by Usp47 depletion. (C-D) This conclusion implies that turnover of misfolded or unstructured MAPK protein products is unlikely to account for the observed impact of USP47. Still, it is possible that MAPK is ubiquitinated on one of its buried lysines. However, USP47's post-translational access to buried lysines would have to occur in a manner that does not involve a prior significant disruption of MAPK structure.
One alternative way in which buried lysines might be readily targeted for ubiquitination is during co-translational quality control that occurs when the nascent peptide is not yet fully folded [37]. To investigate this possibility, we again used [35S]-methionine metabolic labeling, but, unlike the previously described pulse-chase experiment (Fig 3A), labeling was performed for a shorter period (4 h), after which MAPK levels were immediately assayed. The newly synthesized MAPK proteins marked in this manner did not show any changes when we compared Usp47 RNAi treated samples to negative controls (S4 Fig). Therefore, this indicates that USP47 is not acting co-translationally on newly synthesized MAPK peptides during a hypothetical ribosomal quality control step.
We next sought to examine the importance of two key regions of MAPK that are linked to its function and regulation. First, we mutated the tyrosine and threonine residues of the activation segment of MAPK to see if phosphorylation by MEK, which is required for MAPK activation, could have an impact on regulation by USP47. However, we found that such a TEY!AEF MAPK mutant was still sensitive to Usp47 depletion (S2E Fig). We then mutated MAPK's D-site recruitment site (also known as "DRS") that binds to MAPK D-site docking domains, which are present on some MAPK substrates and regulators [6]. The DRS mutant still responded to Usp47 depletion (S2F Fig). This is also in line with our in vivo data showing that mapk Sem , a MAPK gain-of-function point mutation that disrupts the DRS [30], is also sensitive to Usp47 knockdown ( Fig 1D). In addition to the DRS, MAPK comprises an F-site recruitment site (FRS) that is involved in binding to substrates such as Elk1 and c-Fos, and which is exposed only following MAPK activation [6]. Since the non-activatable AEF mutant still responds to USP47 depletion, this implies that the FRS also does not play a role in determining sensitivity to USP47. Finally, in parallel to this, we tested the impact of Usp47 RNAi on mammalian ERK1 and ERK2 stably expressed in S2 cells. Interestingly, we found them to respond to the same extent as Drosophila MAPK (S2G Fig). Together, these results indicate that post-translational regulation of MAPK activity by dynamic phosphorylation is likely uncoupled from regulation by USP47. Also, because the evolutionary distant mammalian ERK1/2 also respond to Usp47 RNAi, sensitivity to USP47 is most likely determined by an MAPK levels are also reduced by Usp47 knockdown in vivo. Tissue expressing Usp47 RNAi (using the VDRC RNAi line: GD26027) is marked with GFP. (C) The expression of Usp47 RNAi was induced using an engrailed-Gal4 promoter that drives expression in the posterior segment of wing discs. (D) Heat shock-induced flip-out clones expressing Usp47 RNAi in eye imaginal discs also show reduced MAPK levels. (E) MAPK protein levels are reduced in flies bearing the Usp47 Δ1 mutation. Lysates were prepared from adult flies homozygous for either the hypomorphic Usp47 Δ1 allele or a Usp47 rev control (a WT Usp47 line generated by a precise excision of the P element used to generate the Usp47 Δ1 allele). (F-G) Usp47 depletion does not impact the mapk transcript levels. (F) mapk transcript levels were measured following dsRNA treatment using an RT-PCR assay targeting the 5 0 and 3 0 UTRs that amplifies the RD and RE splice isoforms of mapk. eIF4AIII (FBgn0037573) dsRNA is used as a control that alters mapk splicing. Act5C and γ-Tubulin (γTub23C; FBgn0260639) loading controls are also shown. (G) mapk transcript levels measured by qPCR (normalized to GFP dsRNA control) are not impacted by two different dsRNA reagents targeting Usp47 (p > 0.05) compared to an untreated control, while Usp47 levels are decreased in both cases (**: p < 0.01). "ns" denotes not significant. (H) Polysomal loading of mapk transcripts is not altered by Usp47 depletion. The polysome and messenger ribonucleoprotein (mRNP)/ monosome fractions were separated on a sucrose gradient. qPCR was used to assay mapk and jnk mRNA transcript levels in both fractions. A slight decrease of mapk transcript abundance (**: p < 0.01) was observed in the mRNP/ monosome fractions upon Usp47 depletion. However, this was observed for jnk transcripts and was not accompanied by a concomitant variation in the polysome fraction. "ns" denotes not significant. Raw data for (A, G, and H) can be found in S1 Data.

Identification of Factors that Antagonize USP47
Given that proteasome inhibition and Uba1 knockdown counteracted Usp47 depletion by restoring MAPK levels, we reasoned that at least one E2 ubiquitin conjugating enzyme and E3 ligase might also have similar properties. Previous studies have shown that RNAi co-depletion can be used as an effective means to identify related factors and establish epistasis relationships [38][39][40]. Thus, in order to identify such specific regulators involved in counteracting USP47 activity, we sought to isolate candidates that, like Uba1, interacted genetically with Usp47. Genetic interaction is defined here as a dual depletion effect that differs significantly from the sum of the individual depletion effects (S5 Fig) [39,41,42]. We used an immunofluorescence assay to quantify MAPK level variations that allowed us to obtain both a robust reduction following Usp47 RNAi as well as a rescue effect induced by co-depletion of Uba1 ( Fig 4A). We employed this assay to conduct a targeted RNAi screen using a focused dsRNA library encompassing factors linked to ubiquitin-proteasome function (S1 Table; S1 Text) that might interact genetically with Usp47. RNAi reagents were tested in combination with a GFP control dsRNA or in co-depletion with Usp47 dsRNA (Fig 4B and 4C, S6A Fig and S1 Table).
In order to distinguish specific modifiers of Usp47 from other candidates that might simply alter MAPK levels in an unrelated process (such as transcription or splicing regulation), we derived a genetic interaction score (Δm) that was used for primary hit selection (S5A Fig and S1 Text). The Δm indicates the difference between the expected and observed outcomes of the combined depletion of Usp47 and gene x. Absence of genetic interaction occurs when the expected and observed outcomes are equivalent (Δm = 0), such as in the case of a purely additive effect when combining two dsRNAs. An aggravating genetic interaction is said to occur when the observed effect of combined Usp47 and x depletion is greater than the expected sum of the individual depletion effects (Δm > 0; x is an enhancer of Usp47). Conversely, an alleviating interaction (Δm < 0; suppressor) occurs when the observed depletion effect of Usp47 and x is lesser than the expected sum (S5B and S5C Fig). Using Uba1 as a reference, we selected a Δm cutoff to guide our choice of primary hits for follow-up validation (Fig 4D, S6B and S6E Fig  and S1 Text). A number of components from the basal UPS displayed genetic interaction with Usp47, and, as might be expected, many of these factors also caused a reduction in cell count indicative of an impact on cell viability (S1 Table). We thus used a more stringent cutoff for factors that caused an appreciable reduction in cell number (S6E Fig and S1 Text), thus removing some UPS factors from our primary hit selection in order to focus on non-cell lethal candidates.
A total of 55 primary hits were recovered from the screen and further selected for validation using independent dsRNA reagents (S2 Table). In our validation experiments, only candidates alleviating the impact of Usp47 dsRNA on MAPK levels were confirmed (Δm < 0) (S2 Table). The absence of aggravating genetic interactions (Δm > 0) could signify that this type of regulation is not operating on USP47/MAPK. However, as Usp47 RNAi only partially depletes MAPK levels compared to the effect of mapk dsRNA itself (Fig 2A), it might be expected that E1 ligase, Uba1 (FBgn0023143; see S5 Table for qPCR validation of RNAi reagent), rescues the impact of Usp47 RNAi on MAPK levels. (E) Treating S2 cells with the epoxomicin proteasome inhibitor restored MAPK levels in a Usp47 depleted context. (C-E) Densitometry quantifications are provided for MAPK levels (normalized to the AKT loading controls). All immunoblots were performed in triplicate or more, and p-values were calculated (paired two tailed Student's t test) comparing values to the Rluc control (*: p < 0.05; **: p < 0.01). t tests performed on other samples are shown in red. Raw data for (B-E) can be found in S1 Data.
The USP47 Deubiquitinase Regulates MAPK Stability another DUB is acting redundantly to protect MAPK from complete degradation, in which case this factor should display an aggravating genetic interaction with Usp47 RNAi (S5 Fig). In order to further explore this hypothesis-and the possibility that a redundant DUB might have been missed in our primary screen-we re-screened the other 41 predicted DUBs in the fly genome using a set of distinct dsRNAs. This second experiment confirmed our primary screen data, yielding no other DUBs that significantly impacted MAPK levels, either alone or when co-depleted with Usp47 ( S7 Fig and S3 Table). Taken together, these results suggest that Usp47 The rescue effect of Uba1 depletion can be robustly measured in a plate-based quantitative microscopy assay suitable for large-scale screening. Usp47 and Uba1 co-depletion significantly restored MAPK levels (***: p < 0.001) compared to a Usp47 single depletion (**: p < 0.01). Uba1 single depletion did not significantly alter MAPK levels compared to GFP dsRNA treated control cells (used for normalization). In this experiment, S2 cells were preincubated with either GFP dsRNA (first and third sample) or with Usp47 dsRNA (second and fourth sample) for 3 d. The cells were then distributed in 384 well plates containing the indicated dsRNAs and incubated for another 3 d. Following dsRNA treatment, anti-MAPK stained cells were imaged and analyzed by high-content microscopy. (B) Screening strategy for a large-scale RNAi screen focused on ubiquitin-proteasome associated factors. RNAi treatment and MAPK quantification for the ubiquitin-proteasome dsRNA set was performed as in A. Each condition was tested in quadruplicate. (C) Histogram showing the distribution of results from the Usp47 RNAi screen. The frequency (number of dsRNAs) is displayed on the y-axis. MAPK levels (x-axis) were normalized to GFP dsRNA treated cells. The Usp47 co-depleted cells show a clear shift towards a reduction in MAPK levels. The Uba1 dsRNA control stands out from the lot, as it completely counteracts the Usp47 dsRNA. (D) Distribution of Usp47 genetic interaction scores (Δm; x-axis) for the candidate dsRNAs (frequency displayed on y-axis) tested in the Usp47 RNAi screen. The negative Δm obtained for Uba1 is consistent with the observed alleviation of the MAPK level reduction. Raw data for (A) can be found in S1 Data. Numerical data for (C and D) can be found in S1 Table. doi:10.1371/journal.pbio.1002539.g004 The USP47 Deubiquitinase Regulates MAPK Stability is not acting redundantly with another DUB and that the partial impact on MAPK levels must have another cause. For instance, dsRNA may not allow for knockdown below a certain threshold, and the degradation machinery may not be able to access the entire pool of MAPK or de novo synthesis of MAPK compensates for Usp47 RNAi-induced destabilization.
We next further narrowed down our list to 13 candidates that displayed a significant rescue effect that was consistent with what we had observed in our primary screen (S4 Table and S1 Text). Of particular interest, Ubc6, an E2 ubiquitin-conjugating enzyme, and two E3 ligases, poe (pushover or purity of essence) and CG5604 (hereafter referred to as Ufd4 or Ubiquitin fusion-degradation 4-like, after the name of its closest yeast counterpart) were identified as the sole E2/E3 enzymes in this set (Fig 5A and 5B). All three factors displayed a capacity to restore normal MAPK levels following Usp47 RNAi treatment, although Ufd4 was the weakest ( Fig  5A). We were also able to confirm the rescue effects of poe and Ubc6 on endogenous MAPK levels by western blot from S2 cell lysates (Fig 5C) as well as by immunofluorescence in Drosophila wing imaginal discs (Fig 5D and S8 Fig). Additionally, poe RNAi was able to counteract Usp47 and restore the ectopic wing vein phenotype induced by mapk Sem expression in wings ( Fig 5E). The ability of POE depletion to restore MAPK levels appeared specific to Usp47 codepletion, as demonstrated by poe RNAi's inability to rescue MAPK levels following the depletion of eIF4AIII, a regulator of mapk splicing ( Fig 5F). Moreover, unlike Uba1, both Ubc6 and poe did not otherwise seem to have a generalized impact on protein ubiquitination in S2 cells (S9A and S9B Fig), indicating that their activity is restricted to a subset of substrates. Finally, depletion of poe and Ubc6 RNAi also displayed an ability to rescue exogenously expressed HAtagged MAPK (Fig 5G), which is consistent with a post-translational rescue mechanism.
Among the two putative E3 ligases recovered in the screen, poe RNAi was the strongest in terms of its capacity to rescue MAPK levels ( Fig 5A). In comparison, Ufd4 depletion produced only a partial rescue effect on MAPK levels (Fig 5A), and this was the case for both dsRNAs tested (S2 Table). However, POE does not contain a readily identifiable E3 ligase domain [21] and is only putatively designated as an E3 because it belongs to the UBR family of E3 ligases, with which it shares a UBR box domain. Because of this, we reasoned that another yet unidentified factor might be acting in conjunction with POE and UFD4. Interestingly, recent work found that UBR4, the human orthologue of POE, physically associates with the zinc-finger protein Potassium Channel Modulatory Factor 1 (KCMF1) [43]. This factor was shown to contain an atypical C6H2-type RING finger domain that may function as a E3 ligase domain [44]. We tested four predicted Drosophila orthologues of Kcmf1, which were absent from our ubiquitome dsRNA library and thus had not been previously tested for their capacity to rescue the impact of Usp47 dsRNA on MAPK levels. Strikingly, one of these factors, CG11984 (hereafter referred to as Kcmf1 as it is also the closest predicted orthologue to its human counterpart), was found to rescue MAPK levels in a manner that was comparable with poe (Fig 6A and 6B). Like poe and Ubc6, Kcmf1 knockdown did not have a major impact on global protein ubiquitination in S2 cells (S9A Fig), indicating a similar degree of specificity to MAPK. Moreover, exogenous KCMF1 was found to bind to at least two distinct portions of POE (Fig 6C). Interestingly, knockdown of both poe and Kcmf1 also caused an increase in UFD4 levels (Fig 6A), which provides another indication that these three factors are functionally interrelated. Furthermore, the human orthologues of KCMF1, POE, and UBC6 have recently been shown to co-localize, and the KCMF1 N-terminus was found to bind to UBR4, while the C-terminus associated with human RAD6 (UBC6 counterpart) [43], strengthening the notion that they are functioning as part of an E2-E3 complex. When we stably expressed N-and C-terminal truncations of KCMF1 in S2 cells, we observed a dominant negative suppression of Usp47 RNAi ( Fig  6D), presumably due to the truncated protein segregating its binding partners. In sum, these  Table). After validation with independent dsRNA reagents, our set of candidates included one E2 conjugating enzyme, Ubc6, and two E3 ligases, poe and Ufd4. The set also included proteasome components such as Rpn2 (FBgn0028692) and ubiquitin genes such as Ubi-p5E (FBgn0086558). Ufd1-like, a proteasome-associated factor also linked to endoplasmicreticulum-associated protein degradation (ERAD) was also present in our set (see S2 Table for full validation screen data). Data shown here is from the RNAi validation experiment. All MAPK levels are normalized to GFP dsRNA controls. The numerical data presented in this panel can be found in S2 One striking commonality shared by the three factors identified in the Usp47 genetic interaction screen is that they are all linked to the N-end rule ubiquitin-dependent protein degradation process. The identification of these three factors together thus suggests that MAPK might be the target of N-end rule regulation. This prompted us to test additional N-end rule related factors. Experiments in which we knocked down a series of known and predicted Drosophila N-end rule factors by RNAi and assayed their ability to counteract Usp47 knockdown did not bear fruit, as none of these displayed a capacity to restore MAPK levels (S6 Table). However, upon testing the different Drosophila UBR family orthologues by RNAi, we found that the UBR1/2 orthologue also showed a capacity to restore MAPK levels (S10A Fig and S6 Table). Unlike POE/UBR4, the UBR1/2 orthologue caused a significant reduction in cell count (S6 Table), explaining why it was not retained in our Usp47 genetic interaction screen. Thus, this additional result lends further support to the notion that MAPK might be the target of an Nend rule mechanism.
Since a majority of proteins undergo co-translational cleavage of their initiator methionine by methionine aminopeptidases [46], it is possible that the glutamate following the methionine of MAPK might be exposed, which could constitute an N-degron. Most of our experiments in which MAPK is expressed exogenously make use of an N-terminally tagged HA-MAPK. As HA-MAPK still responds to Usp47 RNAi, this suggests that the glutamate residue that follows the N-terminal methionine on WT MAPK is not involved in regulation by USP47. To further investigate this possibility, we employed the ubiquitin-fusion technique, which allows for selective exposure of a residue at the N-terminus of a peptide [47], and used this method to expose the penultimate glutamate residue on MAPK (S10B Fig). However, both the N-terminally exposed glutamate (WT) and glycine (stabilizing mutant) forms were sensitive to Usp47 depletion, indicating that Usp47 does not intervene on the N-terminal penultimate residue of MAPK (S10C and S10D Fig). This also precludes the possibility that the initiator methionine itself might constitute an N-degron (recent work has shown that when the initiator methionine is not cleaved, N-acetylation of the methionine can also constitute an N-degron [19]) as MAPK, without its N-terminal methionine, still responded to Usp47 RNAi (S10D Fig). Thus, our results suggest that MAPK turnover is regulated by an N-end rule-mediated process and that USP47 counteracts the destabilizing impact of these factors on MAPK. However, conducted in third instar Drosophila wing discs show that poe and Ubc6 can rescue Usp47 depletion in vivo. The engrailed-gal4 driver was used to drive expression of RNAi in the posterior segment (GFP-positive) of the disc. Ubc6 RNAi expression caused extensive larval lethality. Those wing discs that could be recovered were of reduced size (cell lethality was also problematic in confirming Ubc6 depletion by qPCR (see S5  Table). The RNAi lines used in this experiment correspond to the following VDRC lines: Usp47, GD26027; poe (1), KK108296; poe (2), GD17648; Ubc6, GD23229. The depletion of the Usp47, poe, and Ubc6 transcripts were also measured by qPCR (see S5 Table). (E) The extra wing vein material induced by mapk Sem expression under the control of sal EPv -Gal4 is suppressed by Usp47 RNAi (see Fig 1D). Coexpression of poe dsRNA counteracts Usp47 RNAi and restores the extra wing vein phenotype. The RNAi lines used here are the same as in C. (F) Pre-translational downregulation of mapk expression is not rescued by poe RNAi. Knocking down the EJC component eIF4AIII reduces MAPK levels due to altered splicing of the mapk transcript. This, unlike the depletion produced by Usp47 dsRNA, is not rescued by co-depleting POE. (G) Exogenous MAPK levels measured by immunoblot are rescued by co-depletion of Usp47 with poe or Ubc6. An HA-mapk stable cell line was treated with the indicated dsRNA for 4 d. A rescue effect was observed on both endogenous and exogenous MAPK. (C, F, and G) Densitometry quantifications are provided for MAPK levels (normalized to the AKT loading controls). All experiments were performed in triplicate or more, and p-values were calculated (paired two tailed Student's t test) comparing values to the Rluc control (*: p < 0.05; **: p < 0.01). t tests performed on other samples are shown in red. "ns" denotes not significant. Numerical data for (A) can be found in S2 Table. Raw data for (C, F, and G) can be found in S1 Data. doi:10.1371/journal.pbio.1002539.g005 The USP47 Deubiquitinase Regulates MAPK Stability Notably, Ufd4 levels are seen to increase upon poe depletion, and this is also observed upon Kcmf1 depletion (validated by qPCR in S5 Table). (B) Schematic representation of the KCMF1 protein product with the amino acid length and positions of the C6H2-type RING domain ("really interesting new gene") and predicted coiled-coil region (CC). The C6H2 RING domain is an uncommon zinc-finger subclass that is present in a restricted set of proteins [44] and bears some resemblance to the zinc finger domain of MetAP1 methionine aminopeptidase [45]. (C) Different FLAG-tagged portions of POE (presented in the schematic below the blot with their respective amino acids (a.a.) positions) were expressed in S2 cells and tested for interaction with V5-tagged KCMF1, WTS, and MAPK by immunoprecipitation followed by western blotting. WTS, a member of the HPO (hippo) pathway, was used as a negative control. Both truncations containing the UBR4 Conserved Region (UBR4 CR) as well as one truncation comprising the UBR domain and part of POE's mid region were detected in the KCMF1 IP. By comparison, no interaction was observed between POE and MAPK, while only a faint nonspecific band (marked with an asterisk) was observed in the WTS IP. The IgG heavy chain position is also indicated on the α-V5 IP blot. (D) S2 cell lines stably expressing V5-tagged KCMF1 (a.a. 1-599), KCMF1 N-terminal (KCMF1-NT; a.a. position 1 to 239), and C-terminal (KCMF1-CT; a.a. position 241 to 599). KCMF1-NT and CT act as dominant negative proteins and partially rescue MAPK levels following Usp47 depletion compared to WT KCMF1. KCMF1-NT had a slightly weaker rescue effect than KCMF1-CT. The cell lines were probed by immunoblot following treatment with the indicated dsRNAs. (A and D) Densitometry quantifications are provided for MAPK levels (normalized to the AKT loading controls). All experiments were performed in triplicate or more, and p-values were calculated (paired twotailed Student's t test) comparing values to the Rluc control (*: p < 0.05; **: p < 0.01). t tests performed on other samples are shown in red. "ns" denotes not significant. Raw data for (A and D) can be found in S1 Data. it remains unclear whether or not MAPK is a direct target of this process, and, in the event that it is a direct target, it appears that neither the N-terminal methionine nor the glutamate that follows it is involved in recognition by the degradation machinery.

Discussion
In this study, we show that USP47, which was previously identified in a genome-wide screen for factors acting downstream of RAS V12 [8], acts to stabilize MAPK post-translationally through a mechanism involving the UPS. We also present results from a new candidate-based RNAi screen for genetic interactors of Usp47 that led us to identify three new factors that act to destabilize MAPK and thus counteract USP47 activity. Moreover, we present evidence for in vivo regulation of MAPK levels and signaling by USP47, POE, and UBC6 that corroborates our cell culture data. Lastly, we identify a fourth destabilizing factor, the putative E3 ligase KCMF1, and demonstrate that it acts in a similar manner to POE. Together, these results present strong evidence for a new means of post-translational control of MAPK levels (Fig 7), which unravel a completely novel aspect of MAPK regulation.
The screening approach we employed in this study differs from classical techniques in that it was based on a genetic interaction approach by RNAi co-depletion. This approach is reminiscent of synthetic genetic interaction screens commonly conducted in yeast, but has only recently been introduced to the RNAi screening field [39,40]. Our results further support this approach as a viable technique in identifying functionally related factors, i.e. factors counteracting Usp47, in this case. The advantage of such co-depletion-based approaches is highlighted by the fact that the ubiquitin-proteasome components and ligases identified here were not identified in the initial RAS-MAPK pathway screen that uncovered Usp47. This can be readily explained by our observation that these factors only have a sizeable impact on MAPK levels in a context in which Usp47 is depleted. Thus, while the regulatory landscape of the RAS-MAPK pathway has now been extensively explored through single-depletion RNAi screening [8,48], it is likely that other regulatory components would be uncovered by further co-depletion/genetic interaction RNAi screens.
The most striking result from our Usp47 co-depletion RNAi screen was the identification of the three N-end rule related factors, poe, Ubc6, and Ufd4. In contrast to the canonical N-recognins UBR1 and UBR2, non-canonical N-recognins such as UBR4 remain relatively poorly studied. UBR4 in particular is characterized by the absence of a readily identifiable ubiquitin ligase domain, which may indicate that they function by associating with catalytically factors such as UBR1/2, UFD4 or possibly KCMF1. CG5604 is referenced as Ufd4 in this study because it is the fly gene with highest sequence homology to Saccharomyces cerevisiae UFD4, a HECT domain ubiquitin ligase that is a member of the ubiquitin fusion degradation (UFD) family (Fig 5B). The UFD pathway is specifically linked to the recognition and ubiquitination of fusion proteins that incorporate a N-terminal ubiquitin moiety [49]. UFD4 in particular has been shown to recognize the N-terminal ubiquitin moiety of UFD substrates via its armadillo repeats [50]. Interestingly, UFD4 (and other UFD pathway factors) has also been shown to work synergistically with the N-end rule pathway [23] and to interact physically and functionally with UBR1 and contribute to the ubiquitination and degradation of N-end rule substrates [51]. Intriguingly, UBC4 (the main E2 associated to UFD4) was not found to act on MAPK levels, indicating that if there is such a UBR4-UFD4 complex, it most likely would operate differently from the UBR1/RAD6-UFD4/UBC4 complex described in yeast.
One alternative to the N-end rule pathway hypothesis is that these factors instead control lysosomal proteolysis of MAPK. Indeed, recent studies have linked RAD6, UBR4, and KCMF1 to lysosomal degradation and autophagy. UBR4 has both been shown to be degraded through autophagy and to have an impact on autophagic flux in mice [52]. RAD6 has been shown to promote mitochondrial turnover (mitophagy) in both flies and mammals [53] and also to promote the autophagic degradation of HP1, a heterochromatin factor involved in DNA repair [54]. More recently, RAD6, UBR4, and KCMF1 have been found to interact physically and to bind lysosomal/mitophagy factors as well [43]. However, based on our data, a hypothetical autophagy-based mechanism for MAPK degradation would also have to incorporate proteasomal regulation, invoking a model involving both systems. To test this possibility, we assayed Indirect regulation of MAPK stability. In this alternative model, UBC6, the E3s, and USP47 act on a yet unidentified factor whose proposed role is to stabilize MAPK. In this scenario, the unidentified factor is the direct target of (de)ubiquitination, whereas MAPK is destabilized through ubiquitin-independent means. The unidentified factor may be degraded by the proteasome following its ubiquitination. MAPK might be degraded through a process that does not implicate the proteasome. Alternatively, even though it is not directly ubiquitinated, MAPK might be degraded by the proteasome through ubiquitin-independent degradation. One possibility in this latter case would be that the unidentified factor acts as a chaperone, bringing MAPK to the proteasome. (A,B) Arrow colors are used to indicate positive (green) or negative (red) regulatory impact on MAPK. doi:10.1371/journal.pbio.1002539.g007 The USP47 Deubiquitinase Regulates MAPK Stability some of the principal autophagy-related factors by RNAi, but observed no impact of MAPK levels (S6 Table), and this was also the case for predicted RAD6-associated factors from Hong et al. (S6 Table) [43]. Thus, despite previously published links, our data does not support an autophagy/lysosome-based degradation mechanism for MAPK.
USP47 does not uniquely work on MAPK, as it has been found to stabilize other proteins. For instance, both the human and Drosophila forms have recently been found to stabilize βcatenin [17]. Human Usp47 has also been shown to be transported to adherens junctions by the KIFC3 kinesin, where it then acts to stabilize E-cadherin [16]. In flies, Usp47 has also been previously described as a negative regulator acting downstream of MAPK in the context of eye development, in which it stabilizes the transcriptional repressor TTK [10]; our description of Usp47 as a positive regulator seemingly opposes this previously described role and implies that Usp47 may play a dual role in the RTK/MAPK pathway with respect to eye development. Importantly, another E3 ligase, SINA, was implicated in the context of TTK, indicating a difference in the way these two components are regulated. This would explain the different outcomes we obtained in our genetic interaction experiments (Fig 1D and S1 Fig  and [10]); specific genetic lesions in the MAPK pathway may be more sensitive to the positive or negative impact of regulation by USP47. One plausible model is that USP47 inactivation leads to a rapid degradation of TTK followed by a slower degradation of MAPK. Thus, inactivating USP47 could initially provide a positive input into the pathway as TTK levels drop rapidly, which would then transition into a negative input as MAPK levels gradually decrease.
Our findings indicate that USP47, together with UBC6, POE, UFD4, KCMF1, and the UPS, is part of a regulatory process that acts to control MAPK levels post-translationally. This discovery was unexpected given that MAPK was not generally known to be regulated by the UPS. Two notable exceptions to this deserve mention. Firstly, human ERK1/2 has been observed to undergo proteasomal degradation in a hyperosmotic condition induced by a prolonged treatment with high levels of sorbitol [55]. Secondly, ERK1c, an alternatively spliced isoform of ERK1, has been shown to accumulate in the Golgi apparatus and undergo monoubiquitination in conditions of elevated cell density [56]. However, to our knowledge, the UPS was not previously known to regulate the stability of the main isoforms of MAPK (ERK1/2) in physiological conditions. The impact of the UPS on MAPK may have eluded previous investigation attempts due to the presence of USP47, which stabilizes MAPK in basal conditions. Indeed, it is only in USP47-depleted conditions that we were able to observe the effect of ubiquitin/proteasome components on MAPK. Furthermore, our inability to observe direct ubiquitination of MAPK may constitute another reason why this phenomenon had eluded prior investigation. It is possible that MAPK is subjected to a non-canonical form of ubiquitination [57] that is more difficult to detect or that it is altogether not the direct target for ubiquitination and is instead degraded through a mechanism that relies on unstructured regions or a ubiquitinated chaperoning protein for targeting to the proteasome (Fig 7B) [57,58].
Future work examining the regulation of USP47 and the associated ubiquitin ligases will be needed to determine in what physiological context this system is used to control MAPK levels. In terms of our understanding of pathway function, these findings add another layer of regulation to the set of transcriptional and splicing regulators we have described previously [7,8], thereby positioning MAPK at the center of an intricate web of regulators acting on its expression. Further exploration of the environmental or developmental cues that control these factors will no doubt have important implications for our understanding of RAS-MAPK signaling dynamics.

Materials and Methods Nomenclature
Following the recommended Drosophila nomenclature, gene names are italicized. Protein product names are not italicized and, with the exception of p38, are capitalized to distinguish from gene names. CG5604 is referred to here as Ufd4 based on the name of its predicted yeast orthologue and the naming scheme used for other UFD family members in Drosophila such as Ufd1-like (FBgn0036136). CG11984 (FBgn0037655) is referred to as Kcmf1 based on the name of its human counterpart. Upon first mention of a specific Drosophila gene or protein, the Fly-Base gene ID number is provided. The FlyBase gene symbol is also listed when it differs from the symbol used in this text.

RNAi
All dsRNAs were generated by in vitro transcription using T7 RNA polymerase. Following NaOAc ethanol precipitation, dsRNA concentration was assessed by gel dosage. Individual dsRNA probes were added to cells to a final concentration of 10 μg/ml. RNAi in S2 cells was conducted using dsRNA following previously described procedures [8]. For the Usp47 codepletion screen and other quantitative microscopy-based experiments, cells were pre-incubated with Usp47 or GFP control dsRNA for 3 d and then distributed in clear-bottom 384 well plates (Greiner) containing the dsRNA sets for a second 3 d depletion. This was done as Usp47's impact on MAPK levels was strongest at incubation times greater than 3 d, whereas other dsRNAs can cause lethality at longer depletion times. The full list of dsRNA reagents and primer sequences used in this study is presented in S7 Table. Subsequent to the Usp47 candidate screen, we selected one of the two dsRNAs from the validation step to use for follow-up experimentation. See S1 Text for further information on the targeted USP RNAi library.

RT-qPCR and RT-PCR
RT-qPCR was conducted using TaqMan Gene-specific assays (primer sets and TaqMan hydrolysis probes), which were designed using the Universal Probe Library assay design center (Roche Applied Science). Primer sequences and the Universal Probe Library probe number are listed in S7 Table. Assays were designed such that the amplified regions did not overlap with sequences targeted by dsRNA. Usp47, Uba1, Kcmf1, and hit validation qPCR experiments were performed on S2 cell lysates obtained using the Cells-to-cDNA (Ambion) lysis buffer following treatment with the indicated dsRNAs. Samples were biological triplicates (validation and Usp47), quadruplicates (Uba1 and Kcmf1), or six biological replicates (polysome RNA extraction). The in vivo experiments were biological triplicate samples prepared from the brain and eye/antennal disc tissue of four L3 larvae per sample. Samples from the polysome fractionation experiment were prepared as previously described [7], after which RNA extraction was performed using TRIzol reagent (Invitrogen).
With the exception of the polysome fractionation experiment in which Ct values are presented, all other qPCR values were log normalized to negative control samples (GFP dsRNA for the cell culture dsRNA experiments, or mCherry RNAi controls in the case of the in vivo experiments). Act5C and RpL32 (FBgn0002626) were used as reference genes for standardization. Student's t tests (homoscedastic, two-tailed) were performed by comparing the experimental values to those of the negative control samples.
For RT-PCR, 5 μg of total RNA extracted from S2 cells (RNeasy, Qiagen) was primed with random primers followed by reverse transcription (RT) with SuperScript II Reverse Transcriptase (Invitrogen). Primers are also listed in S7 Table.

High-Content Microscopy
RNAi-treated S2 cells in 384 well plates were fixed in 4% paraformaldehyde and stained overnight with an α-MAPK antibody (1/1000, Cell Signaling #4695), phalloïdin, and DAPI. Plates were imaged on an Operetta (PerkinElmer) imaging platform. Acquired images were segmented and analyzed using Harmony (PerkinElmer) to derive the average MAPK and actin levels per cell as well as the average cell count per well.
Cell Culture, Western Blots, Immunoprecipitation, and Metabolic Labeling S2 cell culture, transfection, lysates, and western blotting were performed as previously described [59]. Pulse-chase radioactive labeling was performed by incubating S2 cells in ESF medium containing [35S]-methionine for 6 h. Cells were then harvested at different time points (0-36 h) following this incubation period, after which an α-HA immunoprecipitation was performed to enrich for HA-MAPK. Fluorographic reagent (NAMP100, GE Healthcare) was used to amplify [35S] signal. To visualize newly translated MAPK, we used a shorter labelling period of 4 h, after which cells were immediately lysed and submitted to immunoprecipitation. In both metabolic labeling experiments, RNAi treatment was performed for 5 d at a concentration of 15 μg/ml. Epoxomicin treatment was performed at a concentration of 1 μM (DMSO used for controls) 18 h prior to lysis of S2 cells. Whole fly lysates were prepared from 20 adults homogenized in 500 μl of RIPA buffer.

Fly Genetics and Microscopy
Fly husbandry was conducted according to standard procedures. All crosses were performed at 25°C. The Usp47 Δ1 , Usp47 Δ2 , and Usp47 rev lines were initially described in [10]. The rl 1 (FBst0000386) and UAS-rl Sem (FBst0059006) lines were obtained from Bloomington. The sal EPv -Gal4 line was a kind gift from J.F. de Celis, and the csw lf line was originally obtained from L. Perkins. RNAi clones were generated using a line carrying a heat shock inducible flipout actin promoter driving the expression of GAL4 and GFP in clonal tissues (hs-flp;; Act5C. CD2.Gal4, UAS-GFP). L1 larvae were heat shocked for 15 min at 37°C and later collected for dissection upon reaching late L3 (wandering) stage. RNAi expression in wing discs was carried out using an engrailed-Gal4 driver line also carrying UAS-dcr2 to enhance RNAi depletion effectiveness. Staining was performed as previously described [8], and immunofluorescence confocal microscopy was conducted using a Zeiss LSM 700. For the in vivo qPCR experiments, RNAi expression was induced by a prolonged 30' heat shock using hs-flp;; Act5C.CD2.Gal4, UAS-GFP flies to express the indicated RNAi constructs in large clones that encompassed most of the larval tissue. Ubc6 RNAi expression was more problematic as the RNAi caused larval lethality, and clonal expression was not as prominent in tissue recovered from escapers. Accordingly, qPCR results for Ubc6 showed a weaker and more variable depletion.
Supporting Information S1 Data. Underlying data and quantitative observations presented in figures.
(XLSX) S1 Fig. CKDN and EGFR genetic interaction experiments. Two hypomorphic alleles of Usp47 (Usp47 Δ1 and Usp47 Δ2 ) were tested for genetic interaction with bona fide RAS/MAPK pathway mutants. Both the Usp47 Δ1 and Usp47 Δ2 alleles are caused by imprecise excisions of the P(EP)GE28938 P element and the Usp47 rev (revertant) control was generated by precise excision of the same P element [10]. (A) Usp47 Δ1 and Usp47 Δ2 were found to enhance the severity of the rough eye phenotype in flies homozygous for rl 1 , a hypomorphic allele of mapk. MAPK signaling is impaired in rl 1 homozygotes, leading to a mild rough eye phenotype that is due to a lack of photoreceptor cells. Flies homozygous for both Usp47 alleles (only Usp47 Δ2 is shown here) also present a slight rough eye phenotype as has been previously reported [10]. Flies homozygous for both the mapk and the Usp47 mutations display an increase in the severity of the rough eye phenotype, which is consistent with a positive role of USP47 in MAPK signaling. (B) The corkscrew (csw) phosphatase is a positive regulator of RAS/MAPK signaling acting upstream of RAS. Like mapk 1 flies, hemizygotes for a csw loss of function (csw lf ) mutation also have an impaired RAS/MAPK pathway, which translates itself in a visible rough eye phenotype. In flies bearing both csw lf and homozygous for either Usp47 Δ1 and Usp47 Δ2 , the severity of the rough eye is also increased, while this collaboration was not observed for the Stably expressed MAPK mutants altering the proper folding of MAPK partially negate the impact of USP47 depletion. The L41E mutation is located in the kinase's N-lobe, in the β1 strand preceding the glycine-rich ATP-phosphate-binding loop. The W225E is located in the αF helix that follows the activation segment and which forms the core of the C-lobe. Introduction of a charge residue (Glu) within these two key structural elements disrupts hydrophobic interactions and, hence, alters the structural integrity of the kinase domain. (D) Truncation of MAPK C-terminal tail also partially abrogates its sensitivity to USP47 depletion. Truncated MAPK was stably expressed in S2 cells and treated with the indicated dsRNAs. The Y329 Ã truncation removes the C-terminal extension of MAPK, which includes helix α L16 that associates with the N-lobe. (E) The T and Y residues of MAPK's activation segment are not required for regulation by Usp47. The TEY residues of MAPK's activation segment at position 198-200 were replaced with AEF and stably expressed in S2 cells. Stably expressed HA-tagged wild-type MAPK is used as a control. (F) The ERK1/2 family D-site recruitment site that interacts with the docking motif of substrates and other ERK/MAPK interactors is not required for regulation by Usp47. The D-site recruitment site of Drosophila MAPK was inactivated by changing the D331 and D334 residues to asparagines. Exogenous MAPK modified in this manner was stably expressed and submitted to treatment with Usp47 dsRNA. (G) Samples prepared from stable Drosophila S2 cell lines expressing HA-tagged mouse ERK1 and human ERK2 or Drosophila HA-MAPK were treated with the indicated dsRNA. Western blotting shows that, like Drosophila MAPK, ERK1/2 also decrease following Usp47 knockdown. AKT is used as a loading control. (A, C-E, and G) Densitometry quantifications are provided for MAPK levels (normalized to the AKT loading controls). All experiments were performed in four replicates or more, and p-values were calculated (paired two-tailed Student's t test) comparing values to the Rluc control ( Ã : p < 0.05; ÃÃ : p < 0.01). t tests performed on other samples are shown in red. "ns" denotes not significant. Raw data for (A, C-E, and G) can be found in S1 Data. ). An aggravating genetic interaction takes place when the combined depletion value increases the severity of the phenotype beyond the expected additive value (this is also called positive epistasis; Δm < 0). Conversely, an alleviating effect takes place when the combined depletion is less severe than the expected additive value (negative epistasis; Δm > 0). The Δm is further explained in S1 Text. (B and C) Hypothetical genetic interaction examples in which a is replaced by Usp47 RNAi. Since Usp47 KD has a negative impact on MAPK levels, the Δm for alleviating and aggravating interactions will be of the opposite sign to those associated with gene a KD in (A). (B) In this example, an RNAi targeting gene x reduces MAPK levels. A purely additive result for Usp47 + x indicates that no genetic interaction is taking place and that gene x will slightly reduce MAPK levels irrespective of whether Usp47 is depleted or not. This might occur if a transcription or splicing factor that acts on mapk were to be co-depleted with Usp47; like Usp47, the factor also reduces MAPK levels, but functions in parallel, acting at a completely different regulatory step. However, if x aggravates the impact of Usp47 RNAi by decreasing MAPK levels beyond the expected additive value, then x interacts genetically with Usp47. An example of this would be a putative deubiquitinase acting redundantly with Usp47. On the contrary, an alleviating genetic interaction takes place if Usp47+x is lesser than that expected additive value. This might occur in the case of a factor acting in the same "pathway" as Usp47, for instance, a transcription factor promoting Usp47 expression or a factor stabilizing USP47 protein levels. (C) An RNAi targeting gene y slightly increases MAPK levels. The additive value of Usp47 + y (marked with an arrowhead) is represented here by subtracting the empty y bar (dashed line contour) from the full black Usp47 bar. If y alleviates the impact of Usp47 RNAi by restoring MAPK levels beyond this expected additive value, then y interacts genetically with Usp47. An example of this would be a putative ubiquitin ligase or UPS component. This category of hit was most frequently observed in our screen. An aggravating effect in this example is more counterintuitive, invoking a mechanism whereby RNAi of y has a positive impact on MAPK when depleted alone, but has an opposite effect-causing an enhanced reduction of MAPK levels-when combined with Usp47 dsRNA. A factor that targets MAPK to a cellular compartment where it is protected from both USP47 and any potential DUBs might be a good example of an aggravating genetic interactor in this case. samples against the corresponding Usp47 co-depletion samples. X-axis values are normalized to GFP dsRNA controls, while y-axis values are normalized to Usp47 dsRNA controls. In general, MAPK levels from Usp47 co-depletion samples generally mirror those of the GFP codepletion controls (following the dashed diagonal). This suggests that most candidates do not differ from the expected neutral phenotype of a merely additive co-depletion effect. Genetic interactors of Usp47 are situated either above the diagonal (alleviating: green shaded area) or below the diagonal (aggravating: red shaded area). Ubiquitin genes, poe, Ubc6, and Ufd4, are highlighted as examples of factors that alleviate Usp47 RNAi. (B) Volcano plot of Δm scores plotted against Δm p-values. The two values are generally well correlated, prompting us to use the p-value to guide candidate selection. (C) Average cells per well imaged for both GFP and Usp47 co-depletion samples throughout the screen. The absence of a significant difference between the two suggests that there was not a widespread synthetic lethality effect caused by co-depletion of Usp47 and other ubiquitin-proteasome factors. (D) A reduction in cell count is correlated with a non-neutral Δm ( ÃÃÃ : p < 0.001). We therefore used cell count as a secondary criterion to increase the stringency of hit selection. (E) Graphical representation of the hit selection criteria. The false discovery rate (FDR) of Δm is plotted against the FDR of the cell count values. A global Δm cutoff (blue line) is used to select hits irrespective of their impact on viability (blue shaded area). A second less stringent Δm cutoff (red line) is used to select candidates that had little or no impact on cell count (red shaded area). Raw data for (B-D) can be found in S1 Data. Data related to (A and E) is contained in S1 Table. (TIF) S7 Fig. DUB redundancy screen. Results from the DUB co-depletion screen. Usp47 RNAi depletion has a partial impact on MAPK levels, even at prolonged depletion times. An independent set of dsRNAs targeting the predicted DUBs in Drosophila was designed to specifically rescreen these genes to look for any potential redundant factors that might act synergistically with Usp47. However, as was the case in the primary screen, no other DUB was found to cause an appreciable MAPK reduction, either on its own or in conjunction with Usp47 depletion. Rpn11 does, however, rescue MAPK levels in co-depletion with Usp47. This is most likely attributable to its function as a proteasome component. The data used to prepare this figure is contained in S3 Table. (TIF) S8 Fig. In vivo depletion of poe and Ubc6 alone has no impact on MAPK levels. As a control for the dual RNAi depletions presented in Fig 5D, we depleted poe and Ubc6 alone, using the same RNAi lines. The engrailed-gal4 driver was used to drive expression of RNAi in the posterior segment (GFP-positive) of the disc. Ubc6 RNAi expression caused extensive larval lethality. Those wing discs that could be recovered were of reduced size. The RNAi lines used in this experiment correspond to the following VDRC lines: mapk, KK109108; poe (1), KK108296; poe (2), GD17648; Ubc6, GD23229. Canton-S flies were crossed to engrailed-gal4 flies as a negative control. We retested the Drosophila UBR family factors (N-recognins) to verify whether members of this family other than UBR4/POE also suppress the impact of Usp47 dsRNA on MAPK levels. Independent dsRNAs were designed for UBR1/2 (CG9086), UBR3 (CG42593), UBR4 (poe), UBR5 (hyd), UBR6 (FBX011), and UBR7 (CG15141). Besides UBR4/poe, only UBR1/2 dsRNA had a robust alleviating impact on MAPK levels when co-depleted with Usp47 ( ÃÃÃ : p < 0.001; ÃÃÃÃ : p < 0.0001). However, unlike UBR4, UBR1/2 RNAi caused a sharp reduction in cell count (see experimental data in S6 Table). (B) Diagram of the ubiquitin-fusion construct used to expose the glutamate residue following the initiator methionine (penultimate residue). Ubiquitin is cleaved by DUBs (following the G76 residue), exposing the penultimate residue. (C) The Ub-fusion MAPK protein is cleaved to expose the penultimate residue (the glutamate residue following the initiator methionine on WT MAPK). An uncleavable G76V Ub mutant is used as a control for co-translational Ub cleavage. The E2G MAPK mutant has the penultimate glutamate replaced with a glycine (a stabilizing residue). For this experiment, S2 cells were transiently transfected with the Ub-fusion constructs prior to treatment with epoxomicin or DMSO (controls). (D) Following cleavage of the N-terminal ubiquitin, both the WT (glutamate) and E2G mutant forms of MAPK are sensitive to Usp47 depletion. S2 cells were stably transfected with a Ub-fusion construct containing either the WT MAPK sequence (WT) or the E2G mutant form. These cell lines were then treated with Usp47 dsRNA for 4 d. Two exposures of the MAPK blot are presented as differences in expression levels of the Ub-MAPK fusions between both cell lines make it difficult to observe the Usp47 depletion effect on a single exposure. (TIF) S1 providing us with the Usp47 mutant flies and α-UBP64E (USP47) antibodies. We are grateful to J. F. de Celis for providing us with the sal EPv -Gal4 driver fly line and Liz Perkins for the csw lf flies. We would also like to thank S. Meloche for providing us with mouse ERK1 and human ERK2 plasmids. We thank the IRIC HTS facility for their help with the use of the Operetta HCS device. DAB was recipient of a Cole Foundation studentship. CL was recipient of a postdoctoral fellowship from the Fonds de Recherche du Québec -Santé. SR is recipient of a studentship from the Natural Sciences and Engineering Research Council of Canada.

Author Contributions
Conceptualization: DAB CL CMU SR MT.