Regulation of Homeodomain-interacting Protein Kinase 2 (HIPK2) Effector Function through Dynamic Small Ubiquitin-related Modifier-1 (SUMO-1) Modification*

Homeodomain-interacting protein kinase 2 (HIPK2) is involved in transcriptional regulation, growth suppression, and apoptosis. Previous reports showed that HIPK2 can signal cell death via p53, and independently of p53 by activating the c-Jun NH2-terminal kinase (JNK) pathway or mediating CtBP degradation. Here we demonstrate that human HIPK2 is small ubiquitin-related modifier-1 (SUMO-1)-modified in vitro and in vivo at lysine residue 25, a SUMO consensus modification motif conserved in human and mouse HIPK family proteins. SUMO modification of HIPK2 altered neither its nuclear body localization nor its recruitment to promyelocytic leukemia-nuclear bodies. However, SUMO-1 modification inhibited HIPK2-induced JNK activation and p53-independent antiproliferative function. HIPK2 with a mutated SUMO acceptor lysine residue was refractory to inhibition of HIPK2-mediated JNK activation by SUMO-1. Furthermore, we demonstrate that SUMO protease SuPr-1 interacts with HIPK2, and both proteins predominantly colocalize in promyelocytic leukemia-nuclear bodies. SuPr-1 deconjugates SUMO-1 from HIPK2 in vitro and in vivo, which results in modestly increased HIPK2-induced JNK activity. Thus, our data demonstrate that HIPK2 effector function on JNK is modulated through dynamic SUMO-1 modification.

Post-translational protein modification is a crucial mechanism to regulate protein function in eukaryotic cells. Small ubiquitin-related modifier-1 (SUMO-1) 1 is covalently linked as a 97-amino acid long polypeptide to specific target lysine residues (1). SUMO modification usually takes place at consensus modification motifs matching the sequence (I/L/V/F)KXE (2), and is catalyzed by an enzymatic machinery analogous to that of ubiquitin modification (1). Conjugation of SUMO depends on an activating SAE1/SAE2 E1-enzyme heterodimer, the SUMO conjugating enzyme Ubc9, and the recently identified E3 SUMO ligases (3). Beside SUMO-1, mammalian cells express two further SUMO-1-related proteins, SUMO-2/Smt3A and SUMO-3/Smt3B, which are conjugated through the same enzymatic machinery but apparently differ functionally from SUMO-1 (4,5).
A growing number of proteins has been reported to be covalently modified by SUMO, including signal transduction components, transcriptional cofactors, and transcription factors (3). Similar to other post-translational modifications, SUMO conjugation exerts pleiotropic effects on its target proteins ranging from protein stabilization (6,7), altered subcellular localization, to changed activity or function (8 -11). The identification of numerous SUMO-specific proteases in mammals indicates that SUMO modification is a reversible and dynamic process in vivo (12)(13)(14)(15)(16)(17)(18)(19). SUMO proteases vary in their subcellular distribution and localize to the cytoplasm, nuclear envelope, cell nucleus, and nuclear bodies, suggesting their target protein specificity is in part regulated through their subcellular localization. SUMO-specific protease-1 (SuPr-1), a specific isoform of SUMO-specific protease Axam2/SENP2, which localizes to the cell nucleus and nuclear bodies, deconjugates SUMO from promyelocytic leukemia protein (PML), thereby regulating PML-nuclear body (PML-NB) disintegration and transcription (12). Recently, the transcriptional activity of the transcriptional regulators Sp3 and p300/CBP have been demonstrated to be regulated through reversible SUMO conjugation (10,11,20), demonstrating that dynamic SUMO conjugation and deconjugation is an important mechanism to regulate protein function in vivo.
Homeodomain-interacting protein kinase 2 (HIPK2) is a member of the HIPK family of conserved nuclear serine/threonine protein kinases and is involved in transcriptional regulation, cell differentiation, growth suppression, and apoptosis (21)(22)(23)(24)(25)(26). HIPK2 was originally identified as a corepressor for homeodomain transcription factors, and shown to interact with a transcriptional repressor complex consisting of Groucho and histone deacetylase 1 (24,27). In addition, HIPK2 functions also in apoptosis regulation. For example, in response to UV damage HIPK2 interacts with and phosphorylates tumor suppressor p53 at serine 46, thus facilitating its acetylation at lysine 382 resulting in p53-dependent apoptosis (21,22). HIPK2 colocalizes with p53 and CBP in part in PML-NBs (22), nuclear domains implicated in regulation of p53-dependent apoptosis (28,29). Furthermore, HIPK2 also regulates apoptosis independent of p53, either through direct phosphorylating and thus targeting the anti-apoptotic transcriptional corepressor CtBP for proteasomal degradation (26), or through activating the c-Jun NH 2 -terminal kinase (JNK) signaling pathway and participating in TGF-␤-induced JNK activation and apoptosis (30).
A previous report revealed that murine HIPK2 interacts with the SUMO-conjugating enzyme Ubc9, and it has been suggested that HIPK2 is SUMO modified at an atypical, nonconsensus SUMO acceptor site within its very C terminus (31). However, the functional impact of HIPK2 SUMO modification remained yet unclear.
In this study we examined the site and the functional consequences of human HIPK2 SUMO modification. Unexpectedly, our results show that HIPK2 is SUMO modified in vitro and in vivo at a single, consensus modification motif at acceptor lysine 25. Interestingly, we found that SUMO-1 modification negatively regulates HIPK2 effector function on JNK and p53independent apoptosis. Furthermore, we found that SUMO protease SuPr-1 interacts with HIPK2 and deconjugates SUMO from HIPK2, thereby antagonizing the inhibitory effect of SUMO-1 on HIPK2-mediated JNK activation. Thus, our findings indicate that HIPK2-induced JNK activation is modulated through dynamic SUMO-1 conjugation.
In Vitro Kinase Assays-JNK activity was determined as previously published (30). In brief, 293 cells were transfected with HA-JNK1 expression vector and the indicated FLAG-HIPK2 constructs. Total DNA amount was kept equal in all transfections using empty pcDNA3 vector. 24 h post-transfection, the cells were serum starved in Dulbecco's modified Eagle's medium, 0.5% heat-inactivated fetal calf serum overnight, and subsequently harvested. HA-JNK1 was immunoprecipitated using anti-HA antibodies (12CA5), and JNK assays were performed using [␥-32 P]ATP and 2 g of bacterially expressed GST-c-Jun-(5-89) as a substrate as published previously (30).
Luciferase Reporter Assays-SRE-luciferase reporter assays were done as published previously (30). In brief, cells were seeded in 6-well dishes and transfected with the indicated expression vectors by using Superfect reagent. Total DNA amounts were kept equal in all transfections by adding empty pcDNA3 vector. For the SRE-Luc assays, transfected 293 cells were incubated 24 h post-transfection overnight in Dulbecco's modified Eagle's medium, 0.5% heat-inactivated fetal calf serum, harvested in 1ϫ reporter lysis buffer (Promega), and luciferase activity was measured as published previously (30).
Coimmunoprecipitation-Transfected 293T cells were lysed in lysis buffer (1% Nonidet P-40, 0.1% SDS, 150 mM NaCl, 20% glycerol, 1 mM EGTA, 1 mM EDTA, 25 mM NaF, Complete protease inhibitor mixture) for 30 min on ice, and cell debris was spun down. Proteins were immunoprecipitated from the supernatants using 2 g of anti-GFP antibodies and 25 l of protein A/G-Sepharose beads (Santa Cruz Biotechnologies) for 2 h and subsequently washed three times using 750 l of lysis buffer. Precipitates were separated on 8.5 or 10% SDS-PAGE and analyzed by immunoblotting with the antibodies indicated.
Colony Formation Assays-HeLa cells plated in 10-cm dishes were transfected with 5 g of the pcDNA3, pcDNA3-HA-SUMO-1, or pcDNA3-HA-Ubc9 in combination with 8 g of either pcDNA3-Zeo or pcDNA3-Zeo-FLAG-HIPK2. One day post-transfection the cells were maintained for 14 days in double selection medium containing 50 g/ml zeocin and 850 g/ml G418. Cells were then fixed and stained with crystal violet as described previously (22).
In Vivo SUMO Conjugation-293 cells were transfected with the indicated FLAG-tagged HIPK2 expression constructs along with HA-SUMO-1 and HA-Ubc9 as indicated. Cells were lysed in TOTEX ϩ buffer, and equal amounts of proteins were analyzed by immunoblotting as indicated. For immunoprecipitation, cell lysates were diluted with SDS-free buffer to a final concentration of 0.1% SDS and then precipitated by using 4 g of anti-FLAG antibodies. Washed precipitates were analyzed by immunoblotting. SUMO-1 modification of endogenous HIPK2 was analyzed by lysing U2OS or HT1080 cells in TOTEX ϩ buffer, diluting the lysates in SDS-free buffer to a final concentration of 0.1% SDS, and subsequent immunoprecipitation with 10 g of anti-rabbit IgG antibodies or 10 g of anti-HIPK2 antibodies.
For the His 6 -SUMO-1 pull-down, 293 cells were transfected with a His 6 -SUMO-1 expression plasmid and either FLAG-HIPK2 or FLAG-HIPK2 K25R , respectively. Cells were lysed under denaturing conditions and His 6 -SUMO-1-linked proteins were isolated by Ni-NTA-agarose affinity chromatography as described (32). Bound proteins were analyzed by immunoblotting with anti-FLAG antibodies.
In Vitro SUMO Conjugation and Deconjugation Assays-For in vitro SUMO conjugation assays, SUMO-1, SUMO-2, GST-SUMO-1, Ubc9, and SAE1/SAE2 were expressed in and purified from Escherichia coli B834 as previously described (20). [ 35 S]Methionine-labeled forms of HIPK2 were generated in a wheat germ-coupled transcription/translation system according to the manufacturer's instructions (Promega). SUMO conjugation assays were carried out essentially as described (20) and reaction products were analyzed by phosphorimaging.
For SUMO deconjugation assays, 10 l of SUMO-conjugated HIPK2 was mixed with 1 l of 100 mM iodoacetamide and incubated at room temperature for 30 min. Subsequently, 1 l of 150 mM ␤-mercaptoethanol was added and the mixture was incubated at room temperature for 15 min. Finally, 2 g of SuPr-1 solved in 8 l of 50 mM Tris (pH 7.5), 2 mM MgCl 2 , 5 mM ␤-mercaptoethanol was added, and subsequently incubated at 37°C for 1 h. Reaction products were analyzed by phosphorimaging.

HIPK2 Is SUMO-1 Modified at Lysine 25 in Vitro-Using
human HIPK2 (amino acids 565-1191) in a yeast two-hybrid screen as bait, we isolated three full-length clones of the SUMO-conjugating enzyme Ubc9 as the interaction partner from a human total brain cDNA library (Fig. 1A, and data not shown). This suggests that human HIPK2 is subject to SUMO modification. To provide evidence for SUMO modification of HIPK2, we made use of an in vitro SUMO-1 modification assay. Incubation of [ 35 S]methionine-labeled, full-length HIPK2 with purified SAE1/2 (E1 enzyme), Ubc9, and SUMO-1 or SUMO-2 resulted in the appearance of a more slowly migrating species consistent with SUMO modification (Fig. 1B, lane 2, upper  panel). Substitution of GST-SUMO-1 for SUMO-1 caused a further retardation of the modified species (Fig. 1B, lane 3,  upper panel), and omission of SAE1/SAE2, Ubc9, or SUMO abrogated modification (Fig. 1B, lanes 7-9, upper panel).
To map the SUMO acceptor site we generated various trun- is SUMO-1 modified. Endogenous HIPK2 was immunoprecipitated with an affinity purified rabbit anti-HIPK2 polyclonal antibody from lysates of HT1080 cells. As a control, a polyclonal rabbit IgG antibody was used. Immunoprecipitates were washed, separated by SDS-PAGE, and analyzed by immunoblotting with mouse monoclonal anti-SUMO-1-specific antibodies and rabbit anti-HIPK2 antibodies. cated HIPK2 proteins and performed in vitro SUMO modification analyses. Interestingly, the results from the in vitro modification assays revealed modification of the HIPK2⌬C protein ( Fig. 1, lanes 9 and 10) but not of HIPK2⌬N (Fig. 1, lanes 1 and  2). Next, we generated two smaller deletions spanning the HIPK2⌬C protein (HIPK2-1-188) and HIPK2-KD) and performed in vitro SUMO modification assays. This approach enabled us to narrow down the SUMO acceptor site to the NH 2terminal 188 amino acids (Fig. 1C, lanes 3 and 4, and 7 and 8). Next, we searched this amino acid stretch for potential consensus SUMO modification motifs (2) and identified a single site that conformed to the ⌿KXE motif.
To test whether lysine residue 25 within the identified consensus motif serves as the SUMO acceptor site, we changed this residue of arginine (K25R) using site-directed mutagenesis. The K25R substitution abolished in vitro SUMO-1 modification of full-length HIPK2 protein (Fig. 1B, lower panel) and of HIPK2⌬C K25R mutant protein (Fig. 1C, lanes 5 and 6). Thus, our data demonstrate that human HIPK2 is modified by SUMO in vitro at acceptor lysine residue 25.
Strikingly, the SUMO modification motif identified in human HIPK2 was found to be conserved among all murine and human HIPK family members (Fig. 1D). Consistent with the results of the in vitro SUMO modification assays (Fig. 1, B and C), no additional consensus motifs were identified in the entire HIPK2 amino acid sequence, suggesting that HIPK2 is modified at a single site.
In Vivo SUMO-1 Modification of HIPK2 at Lysine 25-To establish that lysine 25 serves as a SUMO acceptor site also in vivo, we coexpressed FLAG-HIPK2 and FLAG-HIPK2 K25R along with His 6 -SUMO-1 in 293 cells. Cells were lysed in guanidine hydrochloride, and His 6 -tagged proteins were isolated by Ni-NTA-agarose chromatography. Immunoblot analysis of the eluted proteins indicated that HIPK2 was modified with His 6 -SUMO-1 in vivo (Fig. 2B). In contrast, HIPK2 K25R was not found to be modified with His 6 -SUMO-1 (Fig. 2B), demon-strating that HIPK2 is SUMO-1 modified in vivo at lysine residue 25.
To examine whether endogenous HIPK2 is SUMO-1 modified we immunoprecipitated HIPK2 from HT1080 fibrosarcoma cell lysates using conditions that preserve SUMO-1 modification. Immunoblot analysis of the immunocomplexes with a SUMO-1-specific mouse monoclonal antibody revealed a slightly upshifted SUMO-1-conjugated HIPK2 species (Fig. 2,  C and D). Thus, this result indicates that a fraction of endogenous HIPK2 protein is modified with SUMO-1.
SUMO-1 Modification Does Not Influence HIPK2 Localization-SUMO modification has been implicated in regulating the subcellular distribution of some of its target proteins, including PML (8,33) and CtBP (34). To determine whether SUMO modification may affect HIPK2 localization, we expressed FLAG-HIPK2 and FLAG-HIPK2 K25R in HeLa cells and analyzed their subcellular distribution using immunofluorescence microscopy (Fig. 3, A and B). HIPK2 K25R showed a similar subcellular distribution to wild type HIPK2 when expressed in HeLa cells or other cell lines tested (U2OS, H1299; data not shown), indicating that SUMO modification does not change the subcellular distribution of HIPK2.
HIPK2 is recruited to PML-NBs by the PML isoform PML-IV (21,22). Immunofluorescence stainings revealed that HIPK2 K25R was as efficiently recruited to PML-NBs as HIPK2 WT (Fig. 3, C and D), demonstrating that SUMO conjugation is not required for PML-NB recruitment of HIPK2.
SUMO-1 Regulates HIPK2-mediated JNK Activation and p53-independent Growth Suppression-Recently, we demonstrated that HIPK2 activates the JNK pathway (30). To determine whether SUMO-1 modification would regulate HIPK2 effector function on JNK, we coexpressed HIPK2 with HA-JNK1 and GFP-SUMO-1 in 293 cells, immunoprecipitated HA-JNK1 from the cell lysates, and monitored JNK activation by in vitro kinase assays using recombinant GSTc-Jun as a substrate. Interestingly, SUMO-1 expression re-

FIG. 3. SUMO modification is dispensable for nuclear body formation and PML-NB recruitment of HIPK2.
HeLa cells were transiently transfected on coverslips with expression vectors for FLAG-tagged HIPK2 or HIPK2 K25R expression vectors alone (A and B), and in combination with a PML-IV expression construct (C and D). 20 h later cells were fixed, and proteins were detected by indirect immunofluorescence staining with mouse anti-FLAG and rat anti-PML-N antibodies as primary antibodies. FLAGtagged HIPK2 proteins (green) were stained using goat anti-mouse Alexa 488 secondary antibody. PML-IV (red) was stained using goat anti-rat Alexa 594 antibodies. Overlapping localization is shown in yellow. DNA was visualized by staining with 4,6,diamidino-2-phenylindole (DAPI).
sulted in a dose-dependent decrease in HIPK2-induced JNK activation (Fig. 4A). Notably, this effect was not because of decreased HIPK2 protein levels (Fig. 4A, lower panel). Importantly, expression of SUMO-1 and Ubc9 showed no effect on JNK activity (Fig. 4D, lane 2, and data not shown), indicating that the effect of SUMO-1 on HIPK2 is specific. Similar to SUMO-1, enforced expression of the SUMO-conjugating enzyme Ubc9 also resulted in decreased HIPK2-mediated JNK activation (Fig. 4B), and coexpression of SUMO-1 and Ubc9 decreased HIPK2-induced JNK activation in an additive manner (Fig. 4C). In contrast, JNK activation by the mitogen-activated protein kinase kinase kinase MLK-3 (35,36) was not affected by coexpression of SUMO-1 and Ubc9 (Fig. 4D), indicating that the effect of SUMO-1 on HIPK2 is specific and not because of a general repressive effect of SUMO-1 or Ubc9 on JNK activation. JNK activates the serum responsive element (SRE) by phosphorylating ternary complex factor (37), and consistently HIPK2 was shown to activate SRE-regulated transcription (30). As expected, a dose-dependent decrease in HIPK2-mediated activation of a SRE-containing luciferase reporter construct was observed upon coexpression of SUMO-1 (Fig. 4E).
Interestingly, consistent with a very recent report (38), no significant impact of SUMO-1 was observed on the p53 activating function of HIPK2 (data not shown). Thus, these findings indicate that SUMO-1 modification of HIPK2 specifically inhibits its effector function on JNK.
HIPK2 acts in TGF-␤-induced JNK activation and apoptosis in p53-deficient Hep3B hepatoma cells (30). Therefore, we addressed the p53-independent apoptotic function of HIPK2 and HIPK2 K25R in the absence or presence of SUMO-1 in this cell system. As expected, HIPK2 or HIPK2 K25R expression potentiated TGF-␤-induced cell death in a comparable manner (Fig.  4F). Coexpression of SUMO-1 significantly inhibited the potentiating function of HIPK2, whereas SUMO-1 expression showed no effect on HIPK2 K25R function (Fig. 4F). These data indicate that SUMO-1 modification of HIPK2 negatively regulates its potentiating effect on TGF-␤-induced JNK activation.
SUMO Acceptor Site Mutant HIPK2 K25R Is Refractory Toward Functional Inhibition by SUMO-1-Next we wanted to analyze the effect of SUMO-1 expression on the function of the SUMO acceptor lysine mutant HIPK2 K25R . Of note, HIPK2 K25R was found to be refractory toward SUMO-1-mediated inhibition of HIPK2-induced JNK activation (Fig. 5A). Similar re- Aliquots of the cell lysates were analyzed by immunoblotting with the indicated antibodies. E, 293 cells were transfected with the indicated expression plasmids along with a 2ϫSRE containing luciferase reporter plasmid (2ϫSRE-Luc). Luciferase activity was measured as described under "Materials and Methods." Standard deviation of three independent experiments is shown. F, p53-deficient Hep3B cells transfected with the indicated expression vectors were stimulated using 2 ng/ml TGF-␤. After 48 h cell death was determined as specified under "Materials and Methods." sults were obtained using the SRE-containing luciferase reporter construct (Fig. 5B), demonstrating that SUMO-1 failed to down-regulate HIPK2-induced JNK activation. These results demonstrate that the inhibition of HIPK2 effector function can be ascribed to its covalent SUMO-1 modification, which is absent in HIPK2 K25R .
HIPK2 and SUMO Protease SuPr-1 Colocalize in PML-NBs-Recently, SUMO protease SuPr-1 has been described to localize to PML-NBs (12). Because HIPK2 can also distribute to PML-NBs (22,30), we were prompted to investigate a possible colocalization of HIPK2 and SuPr-1 in mammalian cells. Confocal laser scanning microscopy revealed that ectopically expressed SuPr-1 colocalized with HIPK2 in NBs (Fig. 6A). However, HIPK2 only colocalized with a fraction of the SuPr-1 nuclear bodies. As enhanced expression of PML-IV leads to recruitment of endogenous HIPK2 to PML-NBs (22, 39), we and SuPr-1 (green) localization was analyzed by confocal microscopy. B, U2OS cells were transfected with PML-VI and GFP-SuPr-1 expression constructs as indicated. Subcellular localization of endogenous HIPK2 (red), GFP-SuPr-1 (green), and PML-VI (blue) was analyzed by confocal microscopy. C, 293T cells were transfected with GFP-SuPr-1 plasmids and FLAG-HIPK2 constructs as indicated. FLAG-HIPK2 proteins were immunoprecipitated from the cell lysates with an anti-FLAG (M2) antibody, and coprecipitated GFP-SuPr-1 proteins were detected by immunoblotting (Western blot) with a GFP antibody. Lower panels, immunoprecipitation (IP) as well as expression of the expressed protein was verified by immunoblotting with the antibodies indicated. D, 293T cells were transfected with GFP-SuPr-1 and the indicated FLAG-HIPK2 expression vectors. SuPr-1 was immunoprecipitated from cell lysates with anti-GFP antibodies, and coprecipitated FLAG-HIPK2 proteins were detected by immunoblotting with the antibodies indicated. A schematic representation of the HIPK2 SuPr-1-binding region is shown at the bottom.
coexpressed SuPr-1 and PML-IV in U2OS cells and analyzed the localization by immunofluorescence staining and confocal microscopy. Notably, endogenous HIPK2 colocalized with SuPr-1 in PML-NBs (Fig. 6B) under these conditions, indicating that SuPr-1 and HIPK2 in fact colocalize in PML-NBs.
To map the interaction domain on the HIPK2 protein we coexpressed HIPK2 deletion mutants with SuPr-1 and performed coimmunoprecipitation assays. These experiments demonstrated that the amino-terminal part of HIPK2 (HIPK2⌬C, amino acids 1-520), but not the carboxyl-terminal region (HIPK2⌬N, amino acids 551-1191), mediates the interaction with SuPr-1 (Fig. 6D). Thus, these data indicate that SuPr-1 and HIPK2 interact in vivo, and suggest SuPr-1 as a FIG. 7. Effect of SuPr-1 on SUMO modification and HIPK2 function. A, in vitro SUMO conjugation and deconjugation of SUMO-1-modified HIPK2. 35 S-Labeled in vitro SUMO-1-modified HIPK2⌬C was incubated with SuPr-1, and reactions were analyzed by autoradiography. B, in vivo SUMO deconjugation of SUMO-1-modified HIPK2. 293T cells were transfected with FLAG-HIPK2, HA-SUMO-1, HA-Ubc9, and GFP-SuPr-1 or the inactive protease GFP-SuPr-1 C466S as indicated. Immunoblotting of total cell lysates was performed with the antibodies indicated. C, SuPr-1 increases HIPK2-induced JNK activation. 293T cells were transfected with FLAG-HIPK2, HA-JNK1, and increasing amounts of GFP-SuPr-1. In vitro JNK kinase assays using GST-c-Jun-(5-89) as substrate were performed as specified under "Materials and Methods." Aliquots of the cell lysates were analyzed by immunoblotting with the antibodies indicated. D, quantification of the kinase assays shown in C. Standard deviation of three independent experiments is shown. candidate SUMO protease for HIPK2.
In Vitro and in Vivo SUMO Deconjugation by SuPr-1 Deconjugation SUMO in Vitro and in Vivo from HIPK2-SUMOspecific proteases play an important role in reversible, dynamic regulation of SUMO modification in vivo. Because SuPr-1 colocalizes and interacts with HIPK2, we next analyzed whether it operates as a SUMO protease on HIPK2. To this end, we in vitro modified wild type HIPK2 and HIPK2⌬C with SUMO-1, and analyzed SUMO-1 deconjugation by SuPr-1/SENP2. SuPr-1 efficiently deconjugated SUMO-1 from modified HIPK2⌬C and wild type HIPK2 (Fig. 7A). In comparison to HIPK2⌬C, wild type HIPK2 is rather inefficiently SUMO-1 modified, which is presumably because of its high molecular weight and the lack of the E3 SUMO ligase in our in vitro modification system. Nonetheless, SuPr-1 comparably deconjugated SUMO-1 from modified HIPK2 as from the modified Sp100 control protein (Fig. 7A).
Next, we tested whether SuPr-1 functions as a SUMO protease for HIPK2 in vivo. To this end, we expressed either SuPr-1 or the inactive point mutant SuPr-1 C466S along with HIPK2 in 293T cells, and analyzed SUMO-1 modification of HIPK2 by immunoblotting. Ectopic expression of SuPr-1 dramatically reduced SUMO-1 modification of HIPK2 (Fig. 7B). In contrast, inactive SuPr-1 C466S did not (Fig. 7B). Thus, these findings demonstrate that SuPr-1 can regulate SUMO-1 deconjugation of HIPK2 in vitro and in vivo.
SuPr-1 Modulates HIPK2 Effector Function-SuPr-1 was previously described to increase the transactivating function of the transcription factor c-Jun without affecting c-Jun phosphorylation by JNK (12). Consistent with this report, we did not observe JNK activation by ectopic expression of GFP-SuPr-1 (data not shown). This allowed us to determine the effect of SuPr-1 expression on SUMO-1-mediated reduction of HIPK2induced JNK activation. To test whether SuPr-1 could reverse the inhibitory effect of SUMO on HIPK2 effector function, we inhibited HIPK2 SUMO modification in vivo by enforced expression of SuPr-1, and investigated the impact on HIPK2mediated JNK activation by using in vitro kinase assays. SuPr-1 expression modestly increased HIPK2-mediated JNK activation by SUMO-1 (Fig. 7, C and D). Taken together, these data provide further evidence for the regulation of HIPK2 effector function on JNK through dynamic SUMO-1 modification in vivo.

DISCUSSION
It is now established that post-translational modification by SUMO-1 is used to regulate protein functions in eukaryotic cells. Many proteins are known to be modified by SUMO, and the biological consequences are substrate specific. Therefore, we attempted to define a function for SUMO-1-modified HIPK2, a serine/threonine protein kinase implicated in growth suppression and apoptosis. Here we demonstrate that HIPK2 effector function on JNK and its p53-independent growth suppressing activity is regulated through reversible SUMO-1 conjugation and deconjugation.
In contrast to a previous report, which suggested that murine HIPK2 is SUMO-1 modified at a non-consensus modification site at its very C terminus (31), our in vitro and in vivo SUMO modification assays unambiguously demonstrate that human HIPK2 is modified at a single SUMO modification site within its NH 2 terminus at lysine 25. Our additional evidence that the single SUMO consensus modification motif is conserved at the NH 2 terminus of all three human and mouse HIPK protein family members (HIPK1, HIPK2 and HIPK3, Fig. 1D) strongly argues for a general mechanism used to regulate HIPKs through dynamic SUMO conjugation and deconjugation at this particular site. In addition, a very recent report also revealed SUMO modification of HIPK2 at lysine 25 (38).
HIPK2 can mediate apoptosis both dependently and independently of p53 (21,22,26,30). Previously, we showed that HIPK2 activates the JNK mitogen-activated protein kinase signaling pathway and acts in TGF-␤-induced JNK activation and apoptosis independently of p53 (30). Our experiments here show that covalent SUMO-1 modification of HIPK2 negatively regulates its effector function on JNK. Because SUMO-1 modification of HIPK2 showed no significant effect on its p53-activating function and its kinase activity (data not shown), which is consistent with a very recent study (38), our data suggest that SUMO-1 confers specificity on the HIPK2 response.
In a very recent study (38), a non-SUMOlatable HIPK2 point mutant (HIPK2 K25A) was shown to have an increased halflife when compared with wild type HIPK2. From this finding a regulatory effect of SUMO-1 modification at position 25 on HIPK2 stability was deduced. However, this interpretation is not supported by the observation in this and our study showing similar steady state levels of wild type HIPK2 with little and high degrees of SUMOlation, the latter induced by coexpression with a combination Ubc9 and SUMO-1. Because lysine residues are also subject to other post-translational modifications, including ubiquitination, acetylation, or methylation, it is well conceivable that the lysine residue at position 25 of HIPK2 is not only SUMOlated but also ubiquitinated in wild type HIPK2. If this site-specific ubiquination does indeed occur, it may be involved in proteasomal degradation of wild type HIPK2 and offers an alternative explanation for the increased half-life of the HIPK2 K25A mutant protein (38). This interpretation is fully compatible with the findings of both studies. Further experiments are required to establish whether or not the lysine residue at position 25 is post-translationally modified by several moieties competitively.
This study identified SUMO-specific protease SuPr-1 as a SUMO protease for HIPK2. SuPr-1-deconjugated SUMO-1 from SUMO-modified HIPK2, both in vitro and in vivo, and modulated its JNK activating function. SuPr-1 was previously identified by means of its interaction with the transcription factor c-Jun, and has been shown to increase c-Jun transactivating activity independent of its phosphorylation by JNK (12). Our findings here indicate that SuPr-1 can also modulate the JNK pathway through regulation of the SUMO-1 modification status of the interacting HIPK2.
Our data indicate that balancing the ratio between SUMO-1-modified and non-modified HIPK2 in vivo through dynamic SUMO conjugation and deconjugation by SuPr-1 modulated HIPK2-mediated JNK activation. Because SUMO-1 modification did not affect HIPK2 localization, and SUMO-1 has been demonstrated previously to modulate protein-protein interaction (20), SUMO modification presumably leads to a changed HIPK2 protein complex composition that abrogates JNK activation. Future studies will be necessary to clarify the exact molecular mechanism of how SUMO conjugation negatively regulates HIPK2 effector function on JNK.
Previous reports showed that SUMO-1 modification of PML is required for proper formation of PML-NBs as distinct nuclear multiprotein complexes (8,33). Disintegration of PML-NBs is accompanied by PML phosphorylation and concomitant SUMO-1 deconjugation (40,41). Interestingly, ectopic expression of HIPK2 leads to its PML-NB targeting and subsequently to delocalization of PML-NBs (30), a process accompanied by PML phosphorylation and SUMO deconjugation (40). SuPr-1 was previously described to localize to PML-NBs, and to trigger PML-NB disintegration through SUMO deconjugation from PML (12). Because we found here that HIPK2 and SuPr-1 colocalize and interact in PML-NBs, it is reasonable to assume that both enzymes cooperate to alter the post-translational modification status of PML, thereby inducing PML-NB disintegration. Consistent with this interpretation, coexpression of SuPr-1 and HIPK2 efficiently induces PML-NBs disintegration. 2 Although the amounts of SUMO-1-modified HIPK2 were modest (which is the case for most SUMO target proteins), the effects of SUMO-1 on HIPK2 function were robust. This apparent contradiction was discussed in a recent review by us (42) that provides a possible explanation for this phenomenon. SUMO modification may play an important role during the initial steps to lock a protein in a distinct conformational state (e.g. active or inactive). Once the SUMO-modified protein is incorporated in a specific protein complex that can keep the protein in its locked conformation, SUMO modification may be dispensable and thus removed (42). This model is also consistent with our data presented here.
PML and several PML-NB components, including HIPK2 and p300/CBP, play an important role in regulating the activity of tumor suppressor p53 (28,29), which like PML, HIPK2, or p300/CBP, is also SUMO modified (32,43,44). Our experiments (data not shown) and results from another group (38) revealed that SUMO-1 modification did not significantly influence the p53-activating function of HIPK2. However, we cannot exclude that SUMO-1 might exert a fine modulation of HIPK2 effector function on p53. As p53 functions are regulated in part through association with PML-NBs (28,29), and SuPr-1 can interfere with PML-NB integrity (12), this argues for a role of SuPr-1 in modulating p53 function. Future studies may clarify if and how SuPr-1 can regulate p53 function.
Consistent with its known coactivator and corepressor function, HIPK2 interacts with both HDAC complexes (27) and CBP (22), factors regulating transcriptional repression or activation, respectively. Similar to our results demonstrated here, recent reports showed that transcriptional cofactors Sp3 and the acetyltransferase p300/CBP are functionally regulated through dynamic SUMO modification in vivo (10,11,20), indicating a broader role of SUMO for the regulation of contextspecific transcriptional cofactors.
In summary, our data show that HIPK2 effector function on JNK is regulated in vivo through dynamic SUMO-1 modification. Although the exact mechanism of how SUMO affects HIPK2 effector function on JNK remains to be elucidated, our data revealed a novel regulatory mechanism of how specificity is conferred on the HIPK2 response.