Directed E3 Ligase That Regulates Axin Degradation and Wnt Signalling

LETTERS

RNF146 is a poly(ADP-ribose)-directed E3 ligase that regulates axin degradation and Wnt signalling

Yue Zhang1, Shanming Liu1, Craig Mickanin1, Yan Feng1, Olga Charlat1, Gregory A. Michaud1, Markus Schirle1, Xiaoying Shi1, Marc Hild1, Andreas Bauer1, Vic E. Myer1, Peter M. Finan1, Jeffery A. Porter1, Shih-Min A. Huang1,2 and Feng Cong1,3

The Wnt/β-catenin signalling pathway plays essential roles thus representing the concentration-limiting factor for complex in embryonic development and adult tissue homeostasis, and assembly3. As a key node of the Wnt pathway, the concentration of deregulation of this pathway has been linked to cancer. Axin is a axin needs to be tightly regulated. Indeed, axin2 is a major target concentration-limiting component of the β-catenin destruction gene of β-catenin4 and activation of the Wnt pathway itself leads to complex, and its stability is regulated by tankyrase. However, degradation of axin5. Using a chemical genetics approach, we have the molecular mechanism by which tankyrase-dependent recently discovered that tankyrases (TNKS1 and TNKS2) regulate axin poly(ADP-ribosyl)ation (PARsylation) is coupled abundance, and that tankyrase inhibitor XAV939 potently inhibits to ubiquitylation and degradation of axin remains undefined. Wnt signalling through stabilization of axin6. Tankyrases belong to Here, we identify RNF146, a RING-domain E3 ubiquitin the poly(ADP-ribose) polymerase (PARP) family of proteins, which ligase, as a positive regulator of Wnt signalling. RNF146 function by synthesizing ADP-ribose polymers onto protein acceptors7. promotes Wnt signalling by mediating tankyrase-dependent This modification, called poly(ADP-ribosyl)ation or PARsylation, degradation of axin. Mechanistically, RNF146 directly interacts is emerging as an important regulatory mechanism and is gaining with poly(ADP-ribose) through its WWE domain, and promotes increasing attention8. Tankyrase is implicated in many important degradation of PARsylated proteins. Using proteomics cellular functions, such as telomere homeostasis, mitosis and vesicle approaches, we have identified BLZF1 and CASC3 as further trafficking7. Tankyrase promotes ubiquitylation and degradation of substrates targeted by tankyrase and RNF146 for degradation. its substrates, such as axin, TRF1 and tankyrase itself through an Thus, identification of RNF146 as a PARsylation-directed unknown mechanism, as no PARsylation-directed E3 ligase has ever E3 ligase establishes a molecular paradigm been identified. A better understanding of PARsylation-dependent that links tankyrase-dependent PARsylation to ubiquitylation. ubiquitylation would provide further insights into axin homeostasis RNF146-dependent protein degradation may emerge and may yield further targets for modulating Wnt signalling. as a major mechanism by which tankyrase exerts its function. To identify the E3 ligase that mediates tankyrase-dependent axin degradation, we carried out a short interfering RNA (siRNA) screen The evolutionarily conserved Wnt/β-catenin signalling pathway plays against 258 genes of the ubiquitin conjugation system using a Wnt3a- essential roles during embryonic development and adult tissue induced Super TOPFlash (STF) luciferase reporter assay in HEK293 homeostasis, and it is often aberrantly activated in cancers1,2. The main cells. In this screen, two independent siRNAs against RNF146, which function of this pathway is to regulate proteolysis of β-catenin. In the encodes a putative RING-domain E3 ligase, significantly inhibited the absence of Wnt ligands, β-catenin is associated with the multiprotein STF reporter. Both RNF146 siRNAs strongly inhibited the Wnt3a- β-catenin destruction complex that contains axin, glycogen synthase induced STF reporter without inhibiting (the TNF-α tumour-necrosis kinase 3 (GSK3) and adenomatous polyposis coli (APC). In this factor-α)-induced (NF-κB nuclear factor-κB) reporter (Fig. 1a), complex, β-catenin is constitutively phosphorylated and degraded by whereas siRNAs against NEDD4, which encodes a control E3 ligase, had the ubiquitin–proteasome pathway. Wnt ligands induce dissociation no effect on the STF reporter (Supplementary Fig. S1a). Depletion of of the β-catenin degradation complex, which leads to stabilization and RNF146 also blocked Wnt3a-induced β-catenin accumulation (Fig. 1b) subsequent nuclear translocation of β-catenin. Within this complex, and axin2 expression (Fig. 1c). We next tested whether RNF146 siRNA the concentration of axin is much lower than that of other components, inhibits Wnt signalling through stabilizing axin. Indeed, depletion of

1Developmental and Molecular Pathways, Novartis Institutes for Biomedical Research, Cambridge, Massachusetts 02139, USA. 2Present address: Sanoﬁ-Aventis Oncology, Cambridge, Massachusetts 02139, USA. 3Correspondence should be addressed to F.C. (e-mail: [email protected])

Received 10 May 2010; accepted 4 February 2011; published online 10 April 2011; DOI: 10.1038/ncb2222

LETTERS

a bc300 3.0 NF-κB (TNF-α) 250 2.5 STF (Wnt3a conditioned medium) 200 axin2

2.0 -catenin

siRNA: pGL2 pGL2 RNF146-A RNF146-B TNKS1+2 β 150 1.5 Wnt3a –+++++ Mr (K) 100 mRNA level activity 1.0 115 Relative β-catenin 50

Relative luciferase 0.5 0 Wnt3a –++ + + 0 Tubulin 64 siRNA: siRNA: pGL2 pGL2 pGL2 pGL2 RNF146-ARNF146-B RNF146-ARNF146-B -catenin

β RNF146-A deRNF146-B

siRNA: -catenin

pGL2 RNF146-A RNF146-B TNKS1+2 β M (K) r dsRNA: Axin1 115 White White DRNF146-A DRNF146-A DRNF146-B DRNF146-B Mr (K) HA Daxin–3×HA 182 TNKS1 82 TNKS ∗ TNKS2 115 64 Tubulin Tubulin 64

Figure 1 RNF146 positively regulates Wnt signalling by affecting the upregulation. Error bars denote the s.d. between triplicates. (d) Depletion of protein level of axin. (a) Depletion of RNF146 speciﬁcally inhibits the RNF146 increases the protein level of axin and TNKS1/2 in HEK293 cells. Wnt3a-induced STF reporter, but not the TNF-α-induced NF-κB reporter The asterisk indicates a background band. (e) Knockdown of Drosophila in HEK293 cells. Error bars denote the s.d. between four replicates. RNF146 (CG8786) using independent dsRNAs increases the protein level of (b) Depletion of RNF146 blocks Wnt3a-induced accumulation of cytosolic HA-tagged Drosophila axin in Drosophila S2 cells. dsRNA against white was β-catenin in HEK293 cells. Co-depletion of TNKS1 and TNKS2 was used used as a control. Uncropped images of blots are shown in Supplementary as a control. (c) Depletion of RNF146 abolishes Wnt3a-induced axin2 Fig. S7.

RNF146 markedly increased the protein level, but not the messenger that it may interact with poly(ADP-ribose) (PAR) and function as a RNA level, of axin1 in HEK293 cells (Fig. 1d and Supplementary Fig. substrate recognition module. Indeed, overexpressed RNF146, but not S1b). Similar results were also obtained in PA-1 cells (Supplementary RNF1461WWE, was immunoprecipitated by anti-PAR antibodies, Fig. S1c,d). In addition, knockdown of the Drosophila orthologue indicating that RNF146 may interact with PARsylated proteins through of RNF146 (CG8786) using double-stranded RNAs (dsRNAs) in the WWE domain (Fig. 2c). Sequence alignment revealed several S2 cells increased the protein level, but not the mRNA level, of positively charged amino-acid residues at the carboxy terminus of the exogenously expressed Drosophila axin (Fig. 1e and Supplementary WWE domain (Supplementary Fig. S2). As PAR is negatively charged, Fig. S1e), supporting an evolutionarily conserved role for RNF146 these residues may be critical for PAR interaction. We mutated these in the regulation of axin. It has been shown that autoPARsylation of residues and found that RNF146R163A was no longer precipitated by tankyrase leads to its degradation9. Interestingly, depletion of RNF146 anti-PAR antibodies, whereas RNF146R161A behaved similarly to the increased the protein levels, but not mRNA levels, of TNKS1 and wild-type protein (Fig. 2c).We next carried out a dot-blot experiment TNKS2 (Fig. 1d and Supplementary Fig. S1b–d). These results indicate using recombinant glutathione S-transferase (GST)–WWE proteins that RNF146 may target both tankyrase and axin for degradation and purified PAR, and showed that GST–WWE, but not BSA, bound through the same mechanism. to PAR in vitro (Fig. 2d). Importantly, the R163A mutation, but not RNF146 has two distinct domains, a RING domain, which is the R161A mutation, completely abolished the interaction between predicted to interact with an E2, and a Trp–Trp–Glu (WWE) domain the WWE domain and PAR (Fig. 2d). Surface plasmon resonance with unknown function (Fig. 2a). A complementary DNA rescue analysis further demonstrated that GST–WWE and GST–RNF146, but experiment indicated that both domains are critical for the function of not GST–WWER163A or GST–RNF146R163A, bound to PAR efficiently RNF146. Expression of siRNA-resistant full-length RNF146 completely (Fig. 2e and Supplementary Fig. S3), indicating that RNF146 binds to rescued the effect of RNF146 siRNA on axin1 and tankyrase, whereas PAR through its WWE domain. expression of either RNF1461RING or RNF1461WWE failed to do We next examined the interaction between axin and RNF146 using so (Fig. 2b). Note that the protein level of RNF1461RING was much a co-immunoprecipitation assay. As shown in Fig. 2f, axin1 interacted higher than that of full-length RNF146, consistent with a critical role of with RNF146, but not RNF146R163A, and this interaction was abolished the RING domain in auto-ubiquitylation and degradation of RNF146. by tankyrase inhibitor XAV939. Furthermore, axin1 was PARsylated The protein level of axin1 was modestly increased on overexpression in vivo and its PARsylation was abolished by XAV939 (Fig. 2f). of RNF1461RING, and it was further enhanced by RNF146 siRNA These results indicate that the interaction between RNF146 and axin (Fig. 2b), indicating that RNF1461RING has a dominant negative is mediated by the WWE domain and PAR moiety. Importantly, effect. The WWE domain is a conserved globular domain found siRNA-resistant RNF146R163A failed to rescue the effect of RNF146 in multiple PARPs and E3 ligases10. As it is the only recognizable siRNA in a cDNA rescue experiment, whereas RNF146R161A behaved domain within RNF146 apart from the RING domain, we speculated similarly to the wild-type protein (Fig. 2g). Together, these results

LETTERS

abWild c RNF146–HA – type ΔRING ΔWWE RNF146 RING WWE –HA –HA WWE–HA Δ siRNA: R161A R163A pGL2 RNF146-B pGL2 RNF146-B pGL2 RNF146-B pGL2 RNF146-B Mr (K) Axin1 115 RNF146–HA RNF146 RNF146 Mr (K) RNF146 d 182 TNKS1 TNKS TNKS2 IP: PAR 64

R161A R163A 115 WB: HA

64 TCL 64 HA WB: HA 49

BSA GST–WWE GST–WWE GST–WWE 49 PAR 64 Tubulin

GST

e fgDMSO XAV939 Wild 140 RNF146–HA – type R161A R163A 120 R163A R163A 100 WWE

80 RNF146 RNF146 RNF146 RNF146 siRNA:

Flag–axin1: –+–+–+–+ pGL2 RNF146-B pGL2 RNF146-B pGL2 RNF146-B pGL2 RNF146-B WWER163A M (K) 60 M (K) r r Axin1 115 40 IP: Flag 64 RNF146–HA WB: HA 49 TNKS1

Resonance units 20 TNKS TNKS2 0 182 ∗ 115 WB: PAR PARsylated 115 –axin1 0 200 400 600 800 HA 64 Time (s) 82 TCL 64 182 Tubulin 115 Flag–axin1 WB: Flag 82 WB: HA 64 RNF146–HA

Figure 2 Interaction between the WWE domain and PAR is essential total cell lysates. (d) Dot-blot analysis of PAR-binding activities of for RNF146-dependent regulation of axin in vivo.(a) Schematic GST–WWE proteins with BSA used as a control. (e) Binding of PAR representation of the domain structure of RNF146. (b) Expression with wild-type WWE, but not WWER163A, as shown by surface plasmon of siRNA-resistant wild-type RNF146, but not RNF146 mutants resonance. (f) Axin1 and RNF146 interact with each other in a with either the RING domain or the WWE domain deleted (1RING PARsylation-dependent and WWE-domain-dependent manner. The and 1WWE), prevents RNF146-siRNA-induced stabilization of axin1 asterisk indicates a nonspeciﬁc band. (g) Expression of RNF146R161A, and TNKS1/2 in HEK293 cells. (c) Immunoprecipitation (IP) of but not RNF146R163A, prevents RNF146-siRNA-induced stabilization RNF146 and RNF146R161A, but not RNF1461WWE or RNF146R163A, of axin1 and TNKS1/2 in HEK293 cells. Uncropped images of blots by anti-PAR antibodies in HEK293 cells. WB, western blot; TCL, are shown in Supplementary Fig. S7. indicate that RNF146 recognizes PARsylated proteins for degradation treatment (Fig. 3b,c), indicating that endogenous axin is PARsylated through a direct binding between the WWE domain and PAR moiety. in a tankyrase-dependent manner in vivo. We next reasoned that if Our previous study demonstrated that tankyrase promotes ubiq- PARsylated proteins were targeted for degradation they should greatly uitylation and degradation of axin6. However, whether endogenous accumulate when the E3 ligase responsible for their degradation is axin is PARsylated and subsequently degraded in vivo is less clear. inhibited. Indeed, depletion of RNF146 using a doxycycline (DOX)- To address this question, we took advantage of the strong binding induced RNF146 short hairpin RNA (shRNA) markedly increased the between the WWE domain and PAR, and used GST–WWE as a tool level of axin1 or tankyrase pulled down by GST–WWE, although the to isolate PARsylated proteins from cells. Lysates of cells expressing total level was only moderately increased (Fig. 3b,c). By comparing Flag–axin1 were pulled down by GST–WWE, eluted and recaptured the amounts of proteins in the input and GST–WWE precipitates, it by anti-Flag antibody. As a control, fractions of cell lysates were is clear that only a very small fraction of axin1 was pulled down by directly immunoprecipitated with anti-Flag antibody. Compared with GST–WWE (Supplementary Fig. S4a), whereas the fraction of tankyrase straight immunoprecipitation, sequential pulldown brought down pulled down by GST–WWE was significantly higher (Supplementary much more PARsylated axin1 (Fig. 3a), indicating that GST–WWE Fig. S4b). Presumably, autoPARsylation of tankyrase is more efficient binds and enriches PARsylated axin1. We then used GST–WWE than tankyrase-dependent PARsylation of axin. Taken together, these to purify PARsylated proteins from cells treated with tankyrase data strongly indicate that RNF146 is responsible for degradation of inhibitor or RNF146 RNA interference (RNAi). As reported earlier6, PARsylated axin and tankyrase in cells. treatment of XAV939 led to stabilization of endogenous axin1 and We then investigated whether RNF146 is required for ubiquitylation tankyrase (Fig. 3b,c). However, the amount of axin1 or tankyrase of axin in vivo. To enhance the detection of ubiquitylation of pulled down with GST–WWE was significantly decreased on XAV939 endogenous axin, we pretreated SW480 cells with XAV939 to increase

LETTERS

ab dWash-off – XAV939 DMSO MG132 IP: Flag RNF146 shRNA

siRNA: pGL2 RNF146-B pGL2 RNF146-B pGL2 RNF146-B pGL2 RNF146-B

Mr (K) DMSO XAV939 –DOX +DOX TCL 115 Axin1 IP: axin2 M (K) M (K) 2nd Flag IP r r 1st GST–WWE Straight Flag IP WB: ubiquitin 182 GST–WWE WB: axin1 115 182 115 pulldown Axin1 115 182 GST–WWER163A 82 115 Axin1 WB: PAR 115 pulldown 82 IP: ubiquitin 182 ceRNF146 WB: axin2 115 shRNA 82 siRNA pGL2 RNF146-B h01.01.5 2.0 01.0 1.5 2.0 M (K) M 115K 182 r DMSO XAV939 –DOX +DOX r Axin2 182 115 TCL TNKS S35 axin2 115 82 100 TCL 182 TNKS GST–WWE 80 Ubiquitin pulldown 115 182 RNF146-B 182 60 pGL2 115 GST–WWER163A TNKS

115 Percentage of 82 pulldown axin2 remaining 40 0 0.51.0 1.5 2.0 Tubulin 64 Time (h)

Figure 3 RNF146 is required for PARsylation-dependent degradation depletion of RNF146 markedly increases, the amount of axin (b) or of axin and tankyrase in vivo.(a) Enrichment of PARsylated tankyrase (c) pulled down by GST–WWE. GST–WWER163A was used axin1 using GST–WWE in a sequential pulldown analysis. Straight as a control. DMSO, dimethylsulphoxide. (d) RNF146 is required immunoprecipitation with anti-Flag antibody was used as a control. for tankyrase-dependent ubiquitylation and degradation of axin in a (b,c) Inhibition of tankyrase by XAV939 or depletion of RNF146 using compound wash-off experiment. (e) RNF146 siRNA stabilizes axin2 in DOX-inducible RNF146 shRNA increases the protein levels of axin SW480 cells in a pulse-chase analysis. Uncropped images of blots are (b) and tankyrase (c) in HEK293 cells. XAV939 reduces, whereas shown in Supplementary Fig. S7. the protein level of axin2. Axin2 was quickly degraded after XAV939 was not observed with RNF146R163A (Fig. 4b). As a control, PAR was washed off, and the proteasome inhibitor MG132 blocked axin2 had no effect on auto-ubiquitylation of the ubiquitin ligases MDM2 degradation and induced accumulation of polyubiquitylated axin2 (Mdm2 p53 binding protein homologue) and SMURF1 (SMAD (Fig. 3d). Significantly, depletion of RNF146 suppressed compound specific E3 ubiquitin protein ligase 1; Supplementary Fig. S5b), wash-off-induced degradation of axin2 and completely blocked demonstrating the specificity of this effect. We next tested the effect MG132-dependent accumulation of polyubiquitylated axin2 (Fig. 3d). of PARsylated TNKS2 on auto-ubiquitylation of RNF146. His-tagged These results indicate that RNF146 is required for PARsylation- TNKS2 was subjected to autoPARsylation followed by pulldown with dependent ubiquitylation of axin. A pulse-chase experiment further Ni–nitrilotriacetic acid beads and extensive washes. TNKS2-coated demonstrated that depletion of RNF146 significantly increased the beads were subsequently added to the in vitro ubiquitylation assay. half-life of axin2 protein (Fig. 3e). Note that the half-life of newly Compared with unmodified TNKS2, PARsylated TNKS2 significantly synthesized axin2 (Fig. 3e) is longer than that of total axin2 (Fig. 3d). increased auto-ubiquitylation of RNF146, but not RNF146R163A It is possible that pre-existing total axin2 forms a tight complex (Supplementary Fig. S5c,d). Consistent with these in vitro results, with stabilized tankyrase and therefore is quickly PARsylated on XAV939 decreased auto-ubiquitylation of exogenously expressed compound wash-off, whereas newly synthesized axin2 would need RNF146 in HEK293 cells (Supplementary Fig. S5e). Together, these more time to bind to tankyrase tightly, leading to a lower efficiency results indicate that interaction between RNF146 and PAR enhances of PARsylation and degradation. Together, these results indicate that auto-ubiquitylation of RNF146. This observation can be explained by RNF146 promotes ubiquitylation and degradation of axin in vivo. the ‘glue’ effect of PAR. PAR contains multiple ADP-ribose units, and it To gain further insights into the mechanism of PARsylation- can bind to multiple RNF146 molecules at the same time. Presumably, dependent ubiquitylation, we established an in vitro ubiquitylation PAR holds multiple RNF146 molecules together, increases their local assay using GST–RNF146, E1, the E2 ubiquitin-conjugating enzyme concentration and promotes cross-ubiquitylation of RNF146. UbcH5a and haemagglutinin (HA)–ubiquitin (Supplementary Fig. Direct examination of PARsylation-dependent ubiquitylation of S5a). Using this ubiquitylation assay, we showed that RNF1461RING TNKS2 is not practical, as both PARsylation and polyubiquitylation no longer possessed E3 ligase activity, whereas RNF146R163A, which is of TNKS2 led to the formation of indistinct, high-molecular-weight deficient in PAR interaction, had similar E3 ligase activity to the wild- species as assessed by SDS gel electrophoresis. To overcome this type protein (Fig. 4a). We then examined the link between PARsylation problem, we decreased the concentration of NAD+ in the PARsylation and ubiquitylation. We found that purified PAR markedly increased reaction to prevent the build-up of poly(ADP-ribose) chains on TNKS2. auto-ubiquitylation of RNF146 (Fig. 4b,c) and this enhancement A significant increase of high-molecular-weight species of TNKS2

LETTERS

a – E1+E2 b – PAR c GST–RNF146 d His–TNKS2– His–TNKS2 PAR GST– GST– – – – – PAR PAR PAR PAR RING RING

RNF146: – Time (min)10 10 20 20 30 30 60 60 WT WT R163A R163A Δ Δ RNF146: – – WT WT R163A R163A GST–

RNF146: – – WT WT R163A R163A M (K) M (K) r M (K) r 220 r 182 182 115 120 Mr (K) 100 115 82 182 Ub–HA 80 82 Ub–HA 64

Ub–HA 115 64

TNKS2 82

182 220 115 120 182 GST 82 100 115 64 GST

80 GST 82 50 64

Figure 4 PARsylation-dependent ubiquitylation in vitro. auto-ubiquitylation of RNF146. (d) Ubiquitylation of (a) Auto-ubiquitylation of RNF146 in an in vitro ubiquitylation TNKS2 by wild-type RNF146, but not RNF146R163A, in a assay. (b) PAR increases auto-ubiquitylation of RNF146, PARsylation-dependent manner. Uncropped images of blots are but not RNF146R163A.(c) Time course of PAR-enhanced shown in Supplementary Fig. S7. was observed when partially PARsylated TNKS2 was subjected to the for degradation. The protein level of GFP–CASC3 was increased by ubiquitylation assay with wild-type RNF146 (Fig. 4d), which probably XAV939 (Supplementary Fig. S6a) or siRNA against TNKS1/2 or represents the formation of polyubiquitylated TNKS2. Importantly, this RNF146 (Supplementary Fig. S6b). Deletion of the TBD of CASC3 increase was not observed with RNF146R163A (Fig. 4d). These results (Supplementary Fig. S6c) blocked the interaction between CASC3 provided direct evidence that RNF146 recognizes PARsylated TNKS2 and tankyrase (Supplementary Fig. S6d), and abolished tankyrase and and mediates its subsequent ubiquitylation. RNF146-dependent regulation of CASC3 (Supplementary Fig. S6e,f). To further understand the physiological roles of tankyrase and Together, our study shows that BLZF1 and CASC3 are also degraded RNF146, we sought to identify their other cellular substrates using by tankyrase and RNF146. BLZF1, also called Golgin-45, is located a quantitative affinity proteomics approach. We reasoned that on the surface of the medial cisternae of the Golgi complex, and is depletion of RNF146 should stabilize PARsylated proteins targeted important for the maintenance of Golgi complex structure13. Notably, for degradation, and these PARsylated proteins can be pulled down TNKS1 has been shown to be a Golgi-associated protein14. CASC3 is by GST–WWE and analysed by quantitative mass spectrometry. We a component of the exon junction complex, which is assembled on carried out the experiment as described in Fig. 5a, and identified spliced mRNAs and plays important roles in post-splicing events15. Our multiple proteins that were significantly enriched in cells depleted findings therefore indicate that tankyrase regulates many biological of RNF146 (Fig. 5b, highlighted). Identification of TNKS1, a known processes that have not previously been associated with tankyrase. substrate of RNF146, validated the experimental design. After fusing Various post-translational modifications can serve as a signal for candidate proteins with GFP and examining their expression on the ubiquitin conjugation system, thus coupling protein turnover treatment with XAV939, we identified BLZF1 (basic leucine zipper with cell signalling16–19. Our previous study revealed an unexpected nuclear factor 1) as the only validated hit (data not shown). As seen role of tankyrase in regulating axin stability, indicating that in Fig. 5c, tankyrase inhibitors such as XAV939 and IWR-1 (ref. 11) PARsylation can also earmark proteins for degradation. Here, we have significantly increased the protein level of GFP–BLZF1, whereas discovered RNF146 as the E3 ligase that mediates tankyrase-dependent ABT-888, a PARP inhibitor with minimal activity on tankyrase6, had degradation of axin and two further tankyrase substrates. Our study has no effect. Furthermore, depletion of TNKS1/2 or RNF146 increased identified the first PAR-directed E3 ligase and provided mechanistic the protein level of GFP–BLZF1 (Fig. 5d). Significantly, the amount insight into how PARsylation leads to ubiquitylation. of PARsylated BLZF1 pulled down by GST–WWE was decreased on The biological functions of PARsylation are diverse and XAV939 treatment, and greatly increased on depletion of RNF146 growing8,20–22. Various PAR-binding domains may exist to read the (Fig. 5e). On the basis of similarity to known tankyrase-binding PAR modification. So far, three PAR-binding domains have been motifs12, we identified an evolutionarily conserved tankyrase-binding described: the macro domain23; eight-amino-acid motifs that have domain (TBD) at the C terminus of BLZF1 (Fig. 5f). Deletion of the been identified within histones, the DNA repair protein XRCC1 (X-ray TBD abolished the interaction between BLZF1 and tankyrase (Fig. 5g), repair cross-complementing protein 1) and p53 (refs 22,24); and the and led to stabilization of BLZF1 (Fig. 5h). These results strongly PAR-binding zinc-finger (PBZ) domain25. Each of these domains has indicate that BLZF1 is targeted for degradation by tankyrase and important roles in the DNA-damage response. Here we demonstrate RNF146 in a PARsylation-dependent manner. that the WWE domain of RNF146 is another PAR-binding domain. As an alternative way to identify substrates of tankyrase and RNF146, Interestingly, the WWE domain is identified in multiple PARPs we mined a yeast two-hybrid database for tankyrase interactors, and (ref. 20). These PARPs may interact with the PAR moiety on their tested the effect of XAV939 on their expression (data not shown). substrates through this domain, which may regulate their activities, Using this approach, we identified CASC3 (cancer susceptibility such as chain elongation. In addition, the WWE domain is found in candidate 3) as another protein targeted by tankyrase and RNF146 several other E3 ligases, such as DTX1–4 (deltex homologues 1–4),

LETTERS

a HEK293 RNF146 shRNA b c 1.0 BLZF1 +DOX –DOX TOMM70A 3 days 0.8 M (K) ABT-888 GMDS r DMSO XAV939 IWR-1 ABT-888 treatment 0.6 SSSCA1 82 MYH4 GFP GFP–BLZF1 Lyse cells 0.4 64 GST–WWE pulldown 0.2 TNKS1 Tubulin 0 Wash d

fold change DOX –0.2

SDS–PAGE versus no DOX 10 –0.4 Quantitative mass spectrometry Log siRNA: –0.6 Mr (K) pGL2 TNKS1+ 2 RNF146-B e –0.8 82 GFP GFP–BLZF1

siRNA 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 64 P value Tubulin

Mr (K) DMSO XAV939 pGL2 RNF146-B siRNA Input 82 GFP–BLZF1 g h Wild WB: GFP BLZF1: type ΔTBD 64 D GST–WWE ild type TB pulldown 82 GFP–BLZF1 BLZF1: W Δ

WB: GFP Mr (K) DMSO XAV939 DMSO XAV939 64 IP: Flag M (K) 82 GFP–BLZF1 r WB: GFP GFP 82 GFP–BLZF1 f Human VTSSPIRGAGDGMETEEP 115 Flag–TNKS1 Mouse LTSTPIRGAGDGMETEEP WB: Flag 82 64 Xenopus ATSTPLRGTGDGMETEAP 82 Tubulin Drosophila VSVVAVRGPGDGMETDKP Input: GFP GFP–BLZF1 64

Figure 5 BLZF1 is identified as a substrate of tankyrase and RNF146 using or co-depletion of TNKS1 and TNKS2 increases the protein level of quantitative mass spectrometry. (a) A quantitative proteomics approach to GFP–BLZF1. (e) XAV939 decreases, whereas RNF146 siRNA markedly identify substrates of RNF146 through combining RNF146 knockdown and increases, the amount of GFP–BLZF1 pulled down by GST–WWE, which is GST–WWE pulldown. (b) A scatter plot depicting proteins identified and presumably PARsylated. (f) The TBD of BLZF1 is evolutionarily conserved. quantified in a quantitative proteomics experiment. Proteins significantly (g) Deletion of the TBD abolishes the interaction between TNKS1 and enriched in the DOX-treated condition are highlighted as potential substrates BLZF1 in a co-immunoprecipitation assay. (h) Deletion of the TBD stabilizes of RNF146. (c) Inhibition of tankyrase by either XAV939 or IWR-1 increases GFP–BLZF1 in SW480 cells. Uncropped images of blots are shown in the protein level of GFP–BLZF1 in SW480 cells. (d) Depletion of RNF146 Supplementary Fig. S7.

HUWE1 (HECT, UBA and WWE domain containing 1, also known ACKNOWLEDGEMENTS as MULE) and TRIP12 (thyroid hormone receptor interactor 12; We thank D. Patel, C. Xin, E. McWhinnie, S. Zhao, J. Murphy, Y. Mishina and J. Klekota for technical assistance and W. Shao, F. Stegmeier, J. Tallarico, ref. 10). These E3 ligases may also regulate protein turnover in a T. Bouwmeester and M. Kirschner for comments and advice. PARsylation-dependent manner. β Axin is the concentration-limiting factor in the -catenin AUTHOR CONTRIBUTIONS degradation complex. The discovery of RNF146 as the E3 ligase Y.Z., C.M., Y.F., G.A.M., M.S., M.H., A.B., V.E.M, P.M.F., J.A.P., S-M.A.H and F.C. responsible for axin degradation hints at a potential role for RNF146 conceived and designed the study. Y.Z., S.L., C.M., Y.F., O.C., G.A.M., M.S., X.S. and in Wnt-related diseases. An association between Wnt signalling and F.C. designed and implemented experiments. Y.Z. and F.C. wrote the paper. breast cancer is known. Wnt1 itself was first identified as a mammary COMPETING FINANCIAL INTERESTS oncogene26, and forced expression of Wnt1 in transgenic mice caused The authors declare no competing financial interests. the development of mammary tumours27. Accumulating literature has also indicated a role of aberrant Wnt signalling in human Published online at http://www.nature.com/naturecellbiology Reprints and permissions information is available online at http://npg.nature.com/ 28,29 breast cancer . Intriguingly, a genome-wide gene-association study reprintsandpermissions/ identified a breast-cancer locus at 6q22.33, which contains only two 1. Logan, C. Y. & Nusse, R. The Wnt signalling pathway in development and disease. genes, RNF146 and ECHD1 (enoyl CoA hydratase domain containing Annu. Rev. Cell Dev. Biol. 20, 781–810 (2004). 1; ref. 30). The major haplotype of the locus confers protection from 2. Clevers, H. Wnt/β-catenin signalling in development and disease. Cell 127, 469–480 (2006). disease, whereas the minor haplotype confers risks. It is possible 3. Lee, E., Salic, A., Kruger, R., Heinrich, R. & Kirschner, M. W. The roles of APC and that these two haplotypes confers cancer protection or risk through Axin derived from experimental and theoretical analysis of the Wnt pathway. PLoS. Biol. 1, E10 (2003). affecting RNF146 and Wnt signalling. A potential role of RNF146 4. Leung, J. Y. et al. Activation of AXIN2 expression by β-catenin-T cell factor. in mammary-gland development and mammary-tumour formation A feedback repressor pathway regulating Wnt signalling. J. Biol. Chem. 277, should be examined in future studies. 21657–21665 (2002). 5. Willert, K., Shibamoto, S. & Nusse, R. Wnt-induced dephosphorylation of axin releases β-catenin from the axin complex. Genes Dev. 13, 1768–1773 (1999). 6. Huang, S. M. et al. Tankyrase inhibition stabilizes axin and antagonizes Wnt METHODS signalling. Nature 461, 614–620 (2009). Methods and any associated references are available in the online 7. Hsiao, S. J. & Smith, S. Tankyrase function at telomeres, spindle poles, and beyond. Biochimie 90, 83–92 (2008). version of the paper at http://www.nature.com/naturecellbiology/ 8. Gagne, J. P., Hendzel, M. J., Droit, A. & Poirier, G. G. The expanding role of poly(ADP-ribose) metabolism: current challenges and new perspectives. Curr. Opin. Note: Supplementary Information is available on the Nature Cell Biology website Cell Biol. 18, 145–151 (2006).

LETTERS

9. Yeh, T. Y. et al. Tankyrase recruitment to the lateral membrane in polarized epithelial 20. Schreiber, V., Dantzer, F., Ame, J. C. & de, M. G. Poly(ADP-ribose): novel functions cells: regulation by cell–cell contact and protein poly(ADP-ribosyl)ation. Biochem. J. for an old molecule. Nat. Rev. Mol. Cell Biol. 7, 517–528 (2006). 399, 415–425 (2006). 21. Hassa, P. O. & Hottiger, M. O. The diverse biological roles of mammalian PARPS, 10. Aravind, L. The WWE domain: a common interaction module in protein ubiquitination a small but powerful family of poly-ADP-ribose polymerases. Front. Biosci. 13, and ADP ribosylation. Trends Biochem. Sci. 26, 273–275 (2001). 3046–3082 (2008). 11. Chen, B. et al. Small molecule-mediated disruption of Wnt-dependent signalling in 22. Scovassi, A. I. The poly(ADP-ribosylation) story: a long route from Cinderella to tissue regeneration and cancer. Nat. Chem. Biol. 5, 100–107 (2009). Princess. Riv. Biol. 100, 351–360 (2007). 12. Sbodio, J. I. & Chi, N. W. Identification of a tankyrase-binding motif shared 23. Karras, G. I. et al. The macro domain is an ADP-ribose binding module. EMBO J. 24, by IRAP, TAB182, and human TRF1 but not mouse TRF1. NuMA contains 1911–1920 (2005). this RXXPDG motif and is a novel tankyrase partner. J. Biol. Chem. 277, 24. Pleschke, J. M., Kleczkowska, H. E., Strohm, M. & Althaus, F. R. Poly(ADP-ribose) 31887–31892 (2002). binds to specific domains in DNA damage checkpoint proteins. J. Biol. Chem. 275, 13. Short, B. et al. A GRASP55-rab2 effector complex linking Golgi structure to 40974–40980 (2000). membrane traffic. J. Cell Biol. 155, 877–883 (2001). 25. Ahel, I. et al. Poly(ADP-ribose)-binding zinc finger motifs in DNA repair/checkpoint 14. Chi, N. W. & Lodish, H. F. Tankyrase is a golgi-associated mitogen-activated protein proteins. Nature 451, 81–85 (2008). kinase substrate that interacts with IRAP in GLUT4 vesicles. J. Biol. Chem. 275, 26. Nusse, R., van, O. A., Cox, D., Fung, Y. K. & Varmus, H. Mode of proviral activation 38437–38444 (2000). of a putative mammary oncogene (int-1) on mouse chromosome 15. Nature 307, 15. Palacios, I. M., Gatfield, D., St, J. D. & Izaurralde, E. An eIF4AIII-containing complex 131–136 (1984). required for mRNA localization and nonsense-mediated mRNA decay. Nature 427, 27. Tsukamoto, A. S., Grosschedl, R., Guzman, R. C., Parslow, T. & Varmus, H. E. 753–757 (2004). Expression of the int-1 gene in transgenic mice is associated with mammary 16. Hunter, T. The age of crosstalk: phosphorylation, ubiquitination, and beyond. gland hyperplasia and adenocarcinomas in male and female mice. Cell 55, Mol. Cell 28, 730–738 (2007). 619–625 (1988). 17. Min, J. H. et al. Structure of an HIF-1 α-pVHL complex: hydroxyproline recognition 28. Mohinta, S., Wu, H., Chaurasia, P. & Watabe, K. Wnt pathway and breast cancer. in signalling. Science 296, 1886–1889 (2002). Front. Biosci. 12, 4020–4033 (2007). 18. Ikura, T. et al. DNA damage-dependent acetylation and ubiquitination of H2AX 29. Howe, L. R. & Brown, A. M. Wnt signalling and breast cancer. Cancer Biol. Ther. 3, enhances chromatin dynamics. Mol. Cell Biol. 27, 7028–7040 (2007). 36–41 (2004). 19. Yoshida, Y. et al. E3 ubiquitin ligase that recognizes sugar chains. Nature 418, 30. Gold, B. et al. Genome-wide association study provides evidence for a breast cancer 438–442 (2002). risk locus at 6q22.33. Proc. Natl Acad. Sci. USA 105, 4340–4345 (2008).

METHODS DOI: 10.1038/ncb2222

METHODS TNKS (Hs00186671_m1), TNKS2 (Hs00228829_m1), CTNNB1 (Hs00170025_m1), Plasmids. STF reporter, 3×HA-tagged Drosophila axin, Flag-tagged TNKS11PARP RNF146 (Hs00258475_s1). and His-tagged human TNKS2 were generated as previously described6. siRNA- resistant full-length RNF146, RNF1461WWE (missing amino acids 100–175), Immunoblotting, immunoprecipitation and GST pulldown assay. Total cell RNF1461RING (missing amino acids 36–73), RNF146R161A and RNF146R163A were lysates were prepared in RIPA buffer (50 mM Tris-HCl at pH 7.4, 150 mM NaCl, tagged with an HA epitope at the C termini and cloned into pcDNA4-TO or pCMV6. 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS and 1 mM EDTA). Lysates BLZF1, BLZF11TBD (missing amino acids 18–23), CASC3 and CASC31TBD were normalized for protein concentration, resolved by SDS–polyacrylamide gel (missing amino acids 146–151) were fused with GFP epitope at the amino termini electrophoresis (PAGE), transferred onto nitrocellulose membranes and probed and cloned into a lentiviral vector under the control of the metallothionein promoter. with the indicated antibodies. N-terminal GST-tagged RNF146, RNF1461RING, RNF146R163A, WWE (amino For co-immunoprecipitation experiments, cells were lysed in either RIPA buffer acids 100–175), WWER161A and WWER163A bacterium expression constructs were or EBC buffer (50 mM Tris-HCl at pH 7.4, 150 mM NaCl, 0.5% NP-40 and generated by cloning into pGEX-6p. 1 mM EDTA). Cleared cell lysates were incubated with the indicated antibodies and protein G-Sepharose beads overnight at 4 ◦C. Beads were washed six times RNA interference, transfection, luciferase assay and inducible shRNA cell- with lysis buffer. The bound proteins were dissolved in SDS sample buffer, line generation. HEK293 and SW480 cells were grown in DMEM supplemented resolved by SDS–PAGE and immunoblotted with the indicated antibodies. For the with 10% FCS. Plasmid or siRNA transfection was done using Fugene 6 (Roche) or RNF146–axin1 co-immunoprecipitation experiment, cells were treated with MG132 Dharmafect 1 (Dharmacon). to block protein degradation. Sequences of siRNAs used are listed as follows: TNKS1, sense, 50- The compound wash-off assay was carried out as previously described6. Briefly, GCAUGGAGCUUGUGUUAAUUU-30, antisense 50-AUUAACACAAGCUCCAU SW480 cells transfected with the indicated siRNA were pretreated overnight GCUU-30 (Dharmacon); TNKS2, sense, 50-GGAAAGACGUAGUUGAAUAUU-30, with 3 µM XAV939. The compound was then washed and incubated with antisense, 50-UAUUCAACUACGUCUUUCCUU-30 (Dharmacon); CTNNB1, medium containing indicated reagents for 1 h. Cells were lysed with RIPA sense, 50-UGUGGUCACCUGUGCAGCUdTdT-30, antisense, 50-AGCUGCACAG- buffer supplemented with 5 mM N -ethylmaleimide and 5 µM ADP-HPD (ADP- GUGACCACAdTdT-30 (Qiagen); RNF146-A, sense, 50-GGCUAGACUGUGAUGC- (hydroxymethyl)pyrrolidinediol; Alexis) to block the activities of deubiquitylases UAAdTdT-30, antisense, 50-UUAGCAUCACAGUCUAGCCdTdA-30 (Qiagen); and PARG (poly(ADP-ribose) glycohydrolase) and immunoprecipitated with the RNF146-B, sense, 50-GCACGUUUUCUGCUAUCUAdTdT-30, antisense, 50- indicated antibodies. UAGAUAGCAGAAAACGUGCdTdT-30 (Qiagen); pGL2, sense, 50-CGUACGCGG- For the GST pulldown experiments, GST–WWE and GST–WWER163A re- AAUACUUCGAdTdT-30, antisense, 50-UCGAAGUAUUCCGCGUACGdTdT-30 combinant proteins were produced in Escherichia coli and purified using glu- (Dharmacon); NEDD4-A, sense, 5-GGAGGGAACAUACAAAGUAUU-30, anti- tathione–agarose beads (GE Healthcare). HEK293 cells or SW480 cells expressing sense, 50-TACTTTGTATGTTCCCTCCUU-30 (Dharmacon); NEDD4-B, sense, GFP–BLZF1 were subjected to the indicated treatments, lysed in RIPA buffer 50-GAACUAGAGCUUCUUAUGUUU-30, antisense, 50-ACATAAGAAGCTCTAG- supplemented with 5 µM ADP-HPD and incubated with glutathione–agarose beads TTCUU-30 (Dharmacon). charged with GST fusion proteins for 4 h at 4 ◦C. The beads were then washed six To test the effect of RNF146 depletion in Drosophila cells, S2R cells stably trans- times with lysis buffer. Bound materials were resolved by SDS–PAGE and blotted fected with Daxin-3×HA were seeded in 24-well plates and treated with the indicated with the indicated antibodies. dsRNA for 5 d. DsRNAs were produced from the polymerase chain reaction prod- In all experiments, 1× protease inhibitor cocktail (Sigma) and 1× phosphatase ucts using T7-linked primers (for White, forward, 50-ACCTGTGGACGCCAAGG- inhibitor cocktail (Upstate) were added to the lysis buffers. Commercial antibodies 30, reverse, 50-AAAAGAAGTCGACGGCTTC-30; for DRNF146-A, forward, 50- used in this study include goat anti-axin1 (1:1,000; R&D Systems), mouse anti- TGTCGCTGGTCACCTGGTT-30, reverse, 50-CTAGCCACCATACCCAAAAAG- TNKS (1:1,000; Abcam), rat anti-HA (1:1,000; Roche), mouse anti-GST (1:1,000; G-30; for DRNF146-B, forward, 50-TGTCGCTGGTCACCTGGTT-30, reverse, Upstate), rabbit anti-axin2 (76G6; 1:1,000 for WB; 1:100 for IP), rabbit anti-GFP 50-CTAGCCACCATACCCAAAAAGG-30) using a MEGAscript high-yield tran- (1:1,000), mouse anti-ubiquitin (1:1,000; Cell Signaling Technology), mouse anti- scription kit (Ambion). β-catenin and rabbit anti-poly(ADP-ribose) (1:1,000; BD Pharmingen), mouse STF luciferase assays were carried out using a Dual Luciferase Assay kit (Promega) anti-tubulin (1:1,000) and anti-Flag (M2; 1:1,000; Sigma), EZview Red anti-Flag according to the manufacturer’s instructions. M2 Affinity Gel (1:100 for IP, Sigma), mouse anti-poly(ADP-ribose) (1:100 for IP, RNF146 shRNA cell lines were generated by infecting HEK293 or PA-1 cells Trevigen), mouse anti-mdm2 (D-12; 1:1,000; Santa Cruz) and agarose-conjugated with lentivirus containing DOX-inducible shRNA targeting RNF146 followed by anti-multi-ubiquitin antibodies (1:100 for IP, MBL). selection with puromycin. The targeting sequence of RNF146 shRNA is the following (amino acids 2046–2066): 50-GCTTTGCTGTCTAGTCTTATA-30. In vitro auto-PARsylation, ubiquitylation assay. Recombinant GST–RNF146 proteins were produced in E. coli and purified using glutathione–agarose beads. The Surface plasmon resonance. GST-tagged proteins were coupled to a Biacore CM5 in vitro auto-ubiquitylation assay of RNF146 was carried out in 1× ubiquitylation sensor chip coated with anti-GST antibody. 625 nM PAR (Trevigen) was then buffer (50 mM Tris-HCl at pH 8.0, 100 mM NaCl, 5 mM MgCl2-ATP and 1 mM profiled at a flow rate of 30 ml min−1 for 300 s, followed by 600 s flow of wash dithiothreitol). Various forms of GST–RNF146 at 0.6 µM were incubated with buffer. After analysis in BiaEvalution (Biacore), the normalized resonance units were 125 nM E1, 2 µM E2 (UbcH5a) and HA–ubiquitin (Boston Biochem) at 37 ◦C for plotted over time with the assumption of one-to-one binding. 6 h. Reduced amounts of enzymes were used for PAR-enhanced auto-ubiquitylation assay (E1, 62.5 nM; E2, 250 nM; E3, 256 nM).The reaction was stopped by Poly(ADP-ribose) dot-blot analysis. Purified GST–WWE proteins or BSA were addition of 2× SDS sample loading buffer and subject to SDS–PAGE followed by blotted onto nitrocellulose membrane (Invitrogen). The nitrocellulose membrane immunoblotting with indicated antibodies. was rinsed with TBST buffer (10 mM Tris-HCl at pH 7.4, 150 mM NaCl and 0.05% To examine PARsylation-dependent ubiquitylation, recombinant His-tagged Tween 20) three times. The membrane was incubated with purified PAR (Trevigen) TNKS2 was first subjected to an in vitro auto-PARsylation assay in the presence for 1 h at room temperature. After extensive washes with TBST and TBST containing or absence of 1 mM NAD+, as described previously6. PARsylated or unmodified 1 M NaCl, the membrane was blocked with 5% milk followed by immunoblotting TNKS2 was then pulled down by Ni–nitrilotriacetic acid beads (Qiagen), extensively with mouse anti-PAR antibody (Trevigen). washed and subjected to an in vitro ubiquitylation assay as described above. To block the build-up of poly(ADP-ribose) chains, the concentration of NAD+ was reduced Quantitative polymerase chain reaction with reverse transcription. Total to 100 µM and the reaction was shortened to 30 min. RNA from siRNA-treated cells was extracted using the RNeasy Plus Mini Kit (Qiagen) and reverse transcribed with Taqman reverse-transcription reagents Pulse-chase analysis. SW480 cells were pretreated with XAV939. The com- (Applied Biosystems) according to the manufacturer’s instructions. Transcript pound was then washed off and pulse-chase analysis was carried out as de- levels were assessed using the ABI PRISM 7900HT sequence detection system. scribed previously6. Real-time polymerase chain reaction was carried out in 12 µl reactions consisting of 0.6 µl of 20× Assay-on-Demand mix (premixed concentration of 18 µM for Statistical analysis. Results are expressed as the means ± s.d. from an appropriate each primer and 5 µM probe), 6 µl Taqman Universal PCR Master Mix and number of experiments as indicated in the figure legends. The statistical analysis was 5.4 µl cDNA template. The thermocycling conditions used were 2 min at 50 ◦C done using an unpaired Student t-test and P < 0.05 was considered significant. and 10 min at 95 ◦C, followed by 40 cycles of 15 s at 95 ◦C and 1 min at 60 ◦C. All experiments were carried out in triplicate. Gene expression analysis was GST pulldown and mass spectrometry. HEK293 cells containing an inducible carried out using the comparative cycle threshold method with the housekeeping shRNA targeting RNF146 were grown in medium with or without DOX for three gene GUSB (glucuronidase, β) for normalization. The Assay-on-Demand reagents days followed by overnight treatment with 1 µM ABT-888, a PARP1/2 inhibitor, to used were as follows: AXIN1 (Hs00394718_m1), AXIN2 (Hs00610344_m1), suppress PARsylation mediated by PARP1/2. Cells were then washed with ice-cold

DOI: 10.1038/ncb2222 METHODS

PBS twice and lysed in RIPA buffer supplemented with 5 µM ADP-HPD (Alexis). dissociation were searched against the IPI database using Mascot (Matrix Science). For each sample, cell lysates from 30×15 cm dishes of cells were used for pulldown For each peptide sequence and modification state, reporter-ion signal intensities with glutathione–agarose beads charged with 30 µg of GST–WWE. The mixture from all spectral matches were summed for each reporter-ion type and corrected was incubated for 6 h at 4 ◦C and washed six times with 1× PBS containing 0.05% according to the isotope correction factors given by the manufacturer. Only peptides Tween 20. The bound materials were resolved by SDS–PAGE. Complete gel lanes unique to a given protein within the total data set of identified proteins were were excised and samples were subjected to in-gel tryptic digestion. Peptide extracts used for relative protein quantification. Peptide fold changes over the control of controls (no RNF146 knockdown) were labelled with TMT Reagents 131 and (TMT 131) were calculated and subsequently renormalized using the median fold combined with extracts from corresponding samples of RNF146-knockdown lanes change of all quantified peptides to compensate for differences in total protein labelled with TMT Reagents 129. Peptide sequencing was carried out by liquid yield for each affinity purification. Protein fold changes were derived as median chromatography–tandem mass spectrometry on an Eksigent 1D+ high-pressure peptide fold change, P values were calculated using a one-way t-test and data were liquid chromatography system coupled to an LTQ-Orbitrap XL mass spectrometer visualized for further analysis using Spotfire DXP. All identified proteins are shown (Thermo Scientific). Peptide mass and fragmentation data acquired by pulsed-Q in Supplementary Table S1.

SUPPLEMENTARY INFORMATION

DOI: 10.1038/ncb2222

a 1.2 b c STF(Wnt3a CM) 140 1 120 100 0.8 80 β -catenin sh -DOX β -catenin sh +DOX RNF146sh -DOX RNF146sh +DOX 0.6 60 115 40 axin1 82 0.4 20

Relative mRNA level Relative mRNA 0 182 0.2 TNKS1 TNKS 115 TNKS2 pGL2 pGL2 pGL2 pGL2 Relative luciferase activity 0 siRNA 64 TNKS1+2 TNKS1+2 TNKS1+2 TNKS1+2 RNF146-B RNF146-B RNF146-B RNF146-A RNF146-A RNF146-A RNF146-B RNF146-A Tubulin siRNA: β-cat pGL2 Probe axin1 RNF146 TNKS1 TNKS2 49 RNF146-ARNF146-BNEDD4-ANEDD4-B PA1

d 140 140 e 120 120 100 100 80 80 60 40 60 20 40 0

Relative mRNA level Relative mRNA 20 Relative mRNA level Relative mRNA 0 shRNA dsRNA White White β -catsh -DOX β -catsh -DOX β -catsh -DOX β -catsh -DOX β -catsh -DOX β -catsh +DOX β -catsh +DOX β -catsh +DOX β -catsh +DOX β -catsh +DOX DRNF146-B DRNF146-B DRNF146-A DRNF146-A RNF146sh -DOX RNF146sh -DOX RNF146sh -DOX RNF146sh -DOX RNF146sh -DOX RNF146sh +DOX RNF146sh +DOX RNF146sh +DOX RNF146sh +DOX RNF146sh +DOX Daxin Probe axin1 β-cat RNF146 TNKS TNKS2 Probe DRNF146 -3XHA

Figure S1 Depletion of RNF146 increased the protein level, but not mRNA (c) Depletion of RNF146 by inducible shRNA increases the protein level of level, of axin1 and TNKS1/2. (a) Knockdown of NEDD4 does not affect axin1 and TNKS1/2 in PA1cells. Inducible b-catenin shRNA was used as a Wnt3a-induced STF in 293 cells. Error bars denote the standard deviation control. (d) Depletion of RNF146 does not affect the mRNA level of axin1 (s.d.) between four replicates. (b) Depletion of RNF146 does not have or TNKS1/2 in PA1 cells. (e) Knockdown of Drosophila RNF146 (CG8786) a significant effect on the mRNA level of axin1 or TNKS1/2 in HEK293 does not affect the mRNA level of exogenously expressed Drosophila axin. cells. Error bars denote the standard deviation (s.d.) between triplicates. Error bars represent the standard deviation (s.d.) between four replicates.

SUPPLEMENTARY INFORMATION

161 163 WW E hRNF146 EELKAASRGNGEYA W YYEGRNG--WWQYDERTSR------ELEDAFSKGKKN------TEMLIAGFLYVADLENMVQYRRNEH--GRRRKIKR dRNF146 EDICTTRATEDGFQ W YYEGRNG--WWQYDDRT------SQDIEDAFKKGDK------SCTILVAGYVYVVDLEQLVQQRQNEP--TRCRRVKR hDTX4 NYYDPSSAPGKGVV W EWENDNGS-WTPYDMEV------GITIQHAYEKQHP------WIDLTSIGFSYVIDFNTMGQINRQTQ---RQRRVRR hDTX1 NFYDPSSAPGKGIV W EWENDGGA-WTAYDMDI------CITIQNAYEKQHP------WLDLSSLGFCYLIYFNSMSQMNRQTR---RRRRLRR hDTX2 HLFPQHSAPGRGVV W EWLSDDGS-WTAYEASV------CDYLEQQVARGNQ------LVDLAPLGYNYTVNYTTHTQTNKTSS---FCRSVRR hMULE1 RAQMTKYLQSNSNN W RWFDDRSGRWCSYSASN------NSTIDSAWKSGET------SVRFTAGRRRYTVQFTTMVQVNEETG---NRRPVML hTRIP12 MLKKGNAQNTDGAI W QWRDDRGL-WHPYNRIDS------RIIEQINEDTGTAR------AIQRKPNPLANSNTSGYSESKKDDARAQLMKEDPEL hPARP13 SVTKPANSVFTTKWI W YWKNESGT-WIQYGEEKDKRKNS NVDSSYLESLYQSCPRG------VVPFQAGSRNYELSFQGMIQTNIASKT--QKDVIRR hPARP7 STPPSSNVNSIYHTV W KFFCRDHFGWREYPESVI------RLIEEANSRGLK------EVRFMMWNNHYILHNSFFRREIKRRP---LFRSCFI hPARP11 NEVDDMDTSDTQWG W FYLAECGK-WHMFQPDTNSQC--SVSSEDIEKSFKTNPCGS------ISFTTSKFSYKIDFAEMKQMNLTTG---KQRLIKR hPARP12 SVTKPPHFILTTDWI W YWSDEFGS-WQEYGRQGTVHPVTTVSSSDV EKAYLAYCTPGSDGQAATLK FQAGKHNYELDFKAFVQKNLVYGT--TKKVCRR hPARP14 EQESRADCISEFIE W QYNDNNTS--HCFNKMT------NLKLEDARREKKK------TVDVKINHRHYTVNLNTYTATDTKGHSLSVQRLTKS

Figure S2 Sequence alignment of WWE domains. Three highly conserved amino positively charged amino acid residues at the carboxyl terminal of the WWE acid residues, which WWE domain is named after, are highlighted in red. Note domain. Arg161 and Arg163 of the WWE domain of RNF146 are labeled.

SUPPLEMENTARY INFORMATION

18 16 14 12 10 8 6 RNF146 4 RNF146R163A 2 Resonance Units 0 -2 -4

0 100 200 300 400 500 600 700 800 900 Time (s)

Figure S3 Full-length RNF146 interacts with PAR. Binding of PAR with GST-RNF146, but not GST-RNF146R163A, as shown by surface plasmon resonance. The experiment was performed as in Fig. 2e.

SUPPLEMENTARY INFORMATION

0.1% GST- GST- cell lysate R163A a b WWE WWE 0.1% GST- GST- cell lysate WWE WWER163A RNF146 shRNA: -DOX +DOX Marker -DOX +DOX Marker -DOX +DOX 182 short exposure RNF146 shRNA: 115 -DOX +DOX Marker -DOX +DOX Marker -DOX +DOX WB:TNKS WB:axin1 116 182 long exposure 115

Figure S4 Depletion of RNF146 increases the amount of axin1 and TNKS pulled down by GST-WWE. The experiment was performed as in Fig. 3b and 3c. Note that only a very small fraction of axin1 was pulled down by GST-WWE.

SUPPLEMENTARY INFORMATION

GST-RNF146 a c His-TNKS2 His-TNKS2-PAR UbcH1 UbcH1 UbcH5a - - E2: UbcH2 UbcH3 UbcH5b UbcH5c UbcH6 UbcH7 UbcH8 UbcH9 UbcH10 UbcH12 UbcH13 GST- - - - - RNF146: WT R163A WT R163A 182 182 115 115 Ub-HA 82 82 Ub-HA 64 64 4926 37 49

- GST- GST-SMURF1- b MDM2 HECT d GST- GST- RNF146 RNF146R163A - - - PAR PAR PAR e DMSO XAV939 - - Ub-HA His-TNKS2 His-TNKS2-PAR His-TNKS2 His-TNKS2-PAR 182 IP:Ubiquitin 116 poly-Ub-RNF146 182 82 WB:HA 64 115 GST 82 64 182 TCL 49 RNF146-HA MDM2 GST 37 WB:HA 116 26 82 64

Figure S5 PARsylated proteins increase autoubiquitination of RNF146 in subjected to in vitro ubiquitination assay. Western blot analysis with anti-HA vitro and in vivo. (a) A screen for compatible E2s in an in vitro RNF146 antibody revealed that PARsylated TNKS2 enhances total ubiquitination autoubiquitination assay reveals UbcH5a as a functional partner for RNF146. induced by RNF146 but not RNF146R163A. (d) PARsylated TNKS2 enhances (b) PAR does not affect autoubiquitination of MDM2 and SMURF1. (c) autoubiquitination of RNF146 but not RNF146R163A. The experiment was PARsylated or unmodified His-tagged TNKS2 (TNKS2-PAR or TNKS2) was performed as described in (c), and western blot analysis was performed using generated in an in vitro PARsylation reaction by addition of ATP and NAD+, anti-GST antibody. (e) XAV939 significantly reduced autoubiquitination of recovered by Ni-NTA beads, extensively washed to remove residual NAD+, and exogenously expressed RNF146 in HEK293 cells.

SUPPLEMENTARY INFORMATION

a b c

DMSO XAV939 siRNA: 182 pGL2 TNKS1+2 RNF146-B Human TVTGERQSGDGQESTE GFP 182 GFP-CASC3 Mouse TVTGERQSGDGQESTE 116 GFP GFP-CASC3 116 Xenopus AVTGERQSGDGQESTE Drosophila NLAGERQSGDGQESTE Tubulin 64 64 49 Tubulin 49 f d e CACS3: WT ∆TBD CACS3: WT ∆TBD

CACS3: WT ∆ TBD

IP:FLAG 182 GFP-CASC3 WB:GFP DMSO XAV939 DMSO XAV939 siRNA: 116 pGL2 TNKS1+2 RNF146-B pGL2 TNKS1+2 RNF146-B 182 182 GFP-CASC3 GFP 116 GFP GFP-CASC3 WB:FLAG FLAG-TNKS1 116 116 82 64 64 182 Tubulin Tubulin Input:FLAG GFP-CASC3 49 49 116

Figure S6 CASC3 is degraded by tankyrase and RNF146. (a) Inhibition TBD abolishes the interaction between TNKS1and CASC3 in a of tankyrase by XAV939 increases the protein level of GFP-CASC3 in co-immunoprecipitation assay. (e) The TBD domain is required for SW480 cells. (b) Depletion of RNF146 or co-depletion of TNKS1 and XAV939-induced stabilization of CASC3 in SW480 cells. (f) Deletion of TNKS2 increases the protein level of GFP-CASC3 in SW480 cells. (c) TBD abolishes tankyrase and RNF146-dependent degradation of The TBD domain of CASC3 is evolutionary conserved. (d) Deletion of CASC3.

SUPPLEMENTARY INFORMATION

WB: TNKS WB: Tubulin WB: axin1 WB: Tubulin WB: β-catenin WB: HA WB: Tubulin 115 182 82 115 64 82 82 115 64 82 64 64

Fig. 1b Fig. 1d Fig. 1d WB: HA WB: axin1 WB: PAR 182 WB: HA 115 64 64 182 49 WB: HA WB: HA 49 115 82 WB: FLAG WB: TNKS 82 WB: Tubulin 64 182 49 64 182 115 49 WB: HA 64 115 64 82 82 49 Fig. 2c

Fig. 2b Fig. 2f WB: axin1 WB: TNKS WB: axin1 WB: PAR WB: GST 182 182 WB: TNKS WB: TNKS 115 115 82 182 182 182 182 115 115 115 115 82 82 82 82 WB: PAR WB: HA WB: Tubulin

64 49 64 Fig. 2d 49 Fig. 3a Fig. 3c

S35 labeled Fig. 2g WB: Ub 182 WB: axin1 WB: axin1 WB: axin1 WB: axin2 WB: axin2 115 182 182 115 182 115 82 182 182 182 82 115 82 82 115 115 115 82 82 82

Fig. 3e Fig. 3b Fig. 3d

WB: GST WB: HA WB: GST WB: HA WB: Ub WB: Tubulin 220 182 182 220 182 115 100 115 64 82 100 115 82 49 50 82 64 50 64

Fig. 3d Fig. 4a Fig. 4b WB:TNKS WB: HA WB: GST WB: Tubulin

WB: GFP 182 182 182 182 182 115 115 82 115 115 82 82 82 82 64 64 64 64 64 49

Fig. 4c Fig. 4d Fig. 5c and d

WB: GFP WB:GFP WB:FLAG 182 182 115 WB:Tubulin 115 115 82 115 WB:GFP 82 82 82 64 64 64 182 64 49 64 115 82 64

Fig. 5e Fig. 5g Fig. 5h

Figure S7 Full scans

SUPPLEMENTARY INFORMATION

Table S1 Quantitative data for all 70 identified proteins (ProteinProphet-derived False Positive Rate <1%) with fold changes for DOX (TMT129) over control (No DOX, TMT131). The following information is provided: Representative IPI accession number, Entrez gene name, ProteinProphet-derived protein probability, Number of total spectra matched to the protein; Number of identified peptides unique to this protein in this dataset and considered for protein quantitation; Number of spectra matched to unique peptides; Log10 fold changes and associated p-values for DOX (TMT 129) with respect to No DOX (TMT131). P-values are arbitrarily set to 1 for non-significant single peptide quantitations.