SUMO-1 modification of the C-terminal KVEKVD of Axin is required for JNK activation but has no effect on Wnt signaling.

Axin is a multifunctional protein, regulating Wnt signaling and the c-Jun N-terminal/stress-activated protein kinase (JNK/SAPK) pathway as well as tumorigenesis. In the present study, we found that Axin interacts with three SUMO-1 (small ubiquitin-related modifier) conjugating enzymes 3 (E3), PIAS1, PIASxbeta, and PIASy. The extreme C-terminal six amino acid residues of Axin are critical for the Axin/E3 interaction as deletion of the six residues (AxinDeltaC6) completely abolished the ability of Axin to interact with E3 enzymes. AxinDeltaC6 also failed to activate JNK, although it was intact in both its interaction with MEKK1 and homodimerization. Consistent with the presence of a doublet of the KV(E/D) sumoylation consensus motif at the C-terminal end (KVEKVD), we found that Axin is heavily sumoylated. Deletion of the C-terminal six amino acids drastically reduced sumoylation, indicating that the C-terminal six amino acids stretch is the main sumoylation site for Axin. Sumoylation-defective mutants failed to activate JNK but effectively destabilized beta-catenin and attenuated LEF1 transcriptional activity. In addition, we show that dominant negative Axin mutants blocked PIAS-mediated JNK activation, in accordance with the requirement of sumoylation for Axin-mediated JNK activation. Taken together, we demonstrate that sumoylation plays a role for Axin to function in the JNK pathway.

We found previously that Axin robustly activates JNK by binding to the upstream kinase MEKK1 via a distinct domain (2). In addition to the MEKK1-interacting domain (MID), the C-terminal of Axin is required for its JNK activation (2). Based on our deletion mapping analysis, we have observed that sequence downstream of the ␤-catenin-binding site cannot be removed for JNK activation. 2 The carboxyl region contains binding sites for PP2A, Axam, casein kinases, and Dishevelled (10 -15). Casein kinase I and Dishevelled exert inhibitory effects on Axin-regulated JNK signaling (10,23). To understand how the Axin C-terminal region (including what structural requirements therein) exerts its positive role in JNK activation, we performed a yeast two-hybrid screen using the extreme C-terminal 145 aa region of Axin as bait and found that it interacts with multiple proteins, three of which separately encode the protein inhibitor of activated STAT (PIAS) PIAS1, PIASx␤, and PIASy. These proteins have been shown to be SUMO-1-conjugating enzymes (24 -27, 46), although PIAS was initially cloned as a negative regulator for the JAK/STAT signaling-mediated gene transcription (24). The PIAS protein family consists of at least five members, PIAS1, PIAS3, two splicing variants of PIASx (PIASx␣, and PIASx␤, also known as Miz1), and PIASy (28). PIASy markedly increases sumoylation of LEF1 that is a Wnt-responsive transcription factor in association with ␤-catenin and represses its transcriptional activity (29). When reconstituted in vitro, PIASy acts as a SUMO E3 ligase for LEF1. In addition, PIAS1 has also been shown to activate JNK and apoptosis (30).
Sumoylation represents one of the most important posttranslational modifications, modulating a wide spectrum of proteins that participate in protein translocation, transcriptional regulation, signal transduction, and cell growth control (27, 31-33, 40, 41). The first described sumoylated protein is the nucleocytoplasmic transport factor RanGAP1 (37). Many additional SUMO substrates have since been found, including the tumor suppressor p53 (33,47), transcription factor c-Jun (27), and homeodomain-interacting protein kinase 2 (HIPK2) (45). Sumoylation of p53 and c-Jun strongly represses their transcriptional activity. In yeast, SUMO has been found to conjugate many different proteins, including the bud neck Septins. An E3-like factor that promotes the Septin sumoylation has been identified (35). In contrast to ubiquination, sumoylation does not tag proteins for degradation but seems to enhance their stability or modulate their subcellular compartmentalization (34). Sumoylation occurs on the lysine residue in the KX(E/D) consensus sequence (42). Here, we identified two contiguous sumoylation sites (KVEKVD) within the extreme carboxyl six amino acid residue sequence. We have found the evidence that Axin is a sumoylated protein that requires the extreme C-terminal six amino acids. Removal of either site diminished JNK activation by Axin with virtually no activating activity in the deletion mutant lacking both of the sites. Interestingly, deletion of the sumoylation sites has no effect on Wnt signaling, adding to the emerging picture that Axin can be dedicated to at least two independent pathways, the Wnt signaling and the JNK MAP kinase pathway.

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid System-Screening for interacting proteins of the C-terminal region of mouse Axin was carried out with MATCHMAKER GAL4 Two-Hybrid System 3 (Clontech) according to the user manual. Axin C-terminal 145 aa was used as bait. It was expressed as fusion protein to the GAL4 DNA-binding domain in pGBKT7 plasmid in yeast strain AH109. A MATCHMAKER mouse brain cDNA library (complexity: 1.2 ϫ 10 8 , Clontech) expressed as fusion proteins with GAL4 activation domain in the pACT2 plasmid pretransformed into yeast strain Y187 was screened. Positive interactors were selected based on their ability to transactivate GAL4-operator-HIS3, ADE2, and LacZ genes, which enabled the yeast to grow in the absence of histidine and adenine and to form blue colonies in the presence of X-Gal, respectively.
Construction of Plasmids-Expression vectors for wild-type mouse HA-Axin, Myc-Axin, Axin⌬MID, HA-MEKK1, and FLAG-tagged JNK were as described previously (2). Deletion mutations were performed using PCR-based point mutagenesis. To avoid unwanted mutations resulting from PCR reactions, we used only the 3Ј-end EcoRI fragment as template; after mutagenesis by PCR, the entire fragment was verified by sequencing and fused to the wild-type 5Ј-fragment. Full-length cDNA encoding human SUMO-1 was obtained by PCR using human pituitary cDNA and was cloned into pCMV5; the expression plasmid for PIAS1 in pCMV5 was created by fusing the insert from the yeast two-hybrid screen to a PCR fragment of the 5Ј-end fragment using mouse testis cDNA as template. All PCR cDNA products were verified by sequencing.
Coimmunoprecipitation for Sumoylation-Detection of sumoylation was performed as described (33,34). Briefly, cells were collected in phosphate-buffered saline, centrifuged, and lysed in an SDS-containing buffer (one part of SDS sample buffer: 5% SDS, 0.15 M Tris-HCl, pH 6.7, 30% glycerol plus three parts of radioimmune precipitation buffer: 25 mM Tris, pH 8.2, 50 mM NaCl, 0.5% Nonidet P40, 0.5% sodium deoxycholate, 0.1% SDS, 0.1% sodium azide) containing 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, and 10 g/ml leupeptin (Sigma). The lysates were then diluted 10 times with phosphate-buffered saline containing 0.5% Nonidet P-40 with protease inhibitors and were incubated on ice for 15 min and centrifuged at 12,000 ϫ g for 10 min at 4°C. Supernatants were incubated with 1 mg of SUMO-1 antibodies (Santa Cruz Biotechnology, Inc.) and Protein A/G Plus-agarose beads (Santa Cruz Biotechnology, Inc.) for 3 h and washed four times with ice-cold phosphate-buffered saline containing 0.5% Nonidet P-40 and protease inhibitors. Proteins were eluted from agarose beads by boiling in SDS sample buffer. All steps were carried out at 4°C with rocking. Immunoprecipitates or total cell lysates were analyzed by Western blotting as described previously (48).
Transient Transfection and Immunokinase Assays-Human embry-onic kidney 293T cells were maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum, 100 IU of penicillin, 100 g/ml streptomycin, and 2 mM glutamine. Transfections were performed in 60-mm dishes using Dosper liposomal transfection reagent (Roche Molecular Biochemicals) according to the manufacturer's instructions. The total amount of transfected DNA was adjusted to 4.5 g with the empty vector pCMV5 where necessary. Cells were harvested at 40 h after transfection and lysed in a lysis buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton, 2.5 mM sodium pyrophosphate, 1 mM ␤-glycerolphosphate, 1 mM sodium orthovanadate, 1 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride). FLAGtagged JNK1 was immunoprecipitated using mouse monoclonal anti-FLAG M2 beads (Sigma), and the JNK activity was determined as described previously using 1 g of GST-c-Jun-(1-79) (Stratagene) as substrate followed by Western blotting using Phospho-c-Jun antibody (Cell Signaling Inc.) to examine the phosphorylation of c-Jun. Fold activation of the kinase was determined by an imaging analyzer (Molecular Dynamics model 425E) and normalized to their expression levels. Data are expressed as fold kinase activation compared with that in vector-transfected cells with the values representing the mean Ϯ S.E. from three separate experiments.
Coimmunoprecipitation and Western Blotting-Transiently transfected 293T cells in 60-mm dishes were lysed in the same lysis buffer as described, sonicated 10 times for 1 s each, and centrifuged at 13,000 rpm for 30 min at 4°C. HA-or Myc-tagged Axin proteins or MEKK1 proteins were immunoprecipitated from the cell lysate with anti-HA (Roche Applied Science), anti-Myc (9E10), or anti-MEKK1 (C-22, Santa Cruz Biotechnology, Inc.) antibodies and Protein A/G Plus-agarose beads (Santa Cruz Biotechnology, Inc.). Immunoprecipitates or total cell lysates were analyzed by Western blotting as described previously (48). The boiled samples were separated on 10% SDS-polyacrylamide gels and transferred to Immobilon-P membranes (Millipore). After blocking with 5% skim milk in phosphate-buffered saline with 0.1% Tween 20 for 1 h, the membranes were probed with anti-Myc (9E10), anti-MEKK1 (C-22), anti-HA, or anti-FLAG antibodies. Bound antibodies were visualized by enhanced chemiluminescence (Amersham Biosciences) using horseradish peroxidase-conjugated antibodies.
LEF1-luciferase Reporter Gene Assay-293T cells were transfected with 0.5 g of pGL3-fos-7LEF-luciferase (provided by L. Williams), 0.1 g of pCMV-␤-galactosidase, 1 g of ␤-catenin, and 1.5 g of vector or of each of the Axin constructs indicated using Dosper liposomal transfection reagent (Roche Applied Science) according to the manufacturer's instructions. Luciferase activities were measured as described previously. At 32 h after transfection, cells were lysed, divided into two portions, and measured for luciferase and ␤-galactosidase activities (Promega). The ratio of luciferase activity to ␤-galactosidase activity varied less than 10% among the samples. Data are presented as means from three separate experiments performed in triplicate.

Axin Interacts with SUMO-1-conjugating Enzymes E3 PIAS
Family Members-We were intrigued by our previous observation that although removal of the Axin C-terminal region was fully capable of binding to the JNK upstream kinase MEKK1, it failed to activate JNK. We reasoned that Axin might involve other cellular factors for it to activate JNK. We employed a yeast two-hybrid screen using the EcoRI cDNA fragment encoding the C-terminal 145 aa as bait and a mouse brain pACT2 library. Upon mating with Y187 yeast cells pretransformed with mouse brain cDNA in the vector of pACT2, a total of 1.4 ϫ 10 6 colonies survived the selective medium lacking Leu and Trp. After stringent selection procedures, 12 colonies grew on Ade Ϫ /Leu Ϫ /His Ϫ /Ade Ϫ medium and turned blue on X-gal-containing plates. Three of the colonies (designated as AIP4, -5, and -12 for Axin-interacting proteins) separately carried a cDNA insert coding for PIAS1 (Fig. 1A), PIASx␤ (Fig. 1C), and PIASy (data not shown). PIAS1, PIASx␤, and PIASy belong to the novel protein family of protein inhibitors of activated STAT, which have now been recognized as E3 ligase for SUMO-1 conjugation. To test for interaction of AIP with Axin in mammalian 293T cells, the inserts with the HA tag at the N terminus were removed from pACT2 and were then ligated into pCMV5 mammalian expression vector. When co-expressed with Myc-Axin in 293T cells, PIAS1 and PIASx␤ were found to strongly interact with Axin (Fig. 1, B and D). Interestingly, removal of the KV(E/D) motifs in the extreme C-terminal six amino acid residues of Axin completely abolished its interaction with PIAS1 or PIASx␤, indicating that high affinity interaction between PIAS members and Axin requires an intact C terminus.
Axin Is SUMO-1-modified-The finding that Axin interacted with three sumoylation E3 enzymes, together with the fact that the Axin C-terminal contains two conserved SUMO-1-conjugating sites, suggests that Axin may be a sumoylated protein. We assayed for the presence of SUMO-1 binding to Axin by transfecting HA-tagged Axin into 293T cells, either alone or with SUMO-1, and immunoprecipitated either Axin or SUMO-1. Western blotting analysis revealed that the anti-SUMO-1 immunoprecipitates contained Axin as detected by anti-HA antibody but not in the empty vector-transfected cells (Fig. 2). Deletion of either KVD (pKVE) or KVE (pKVD) decreased the extent of sumoylation; removal of both of the two sumoylation sites (pK, pAVE, p⌬C6) drastically reduced sumoylation. Deletion of the MEKK1 binding MID domain had little effect on Axin sumoylation (data not shown). To further analyze the structural requirements for sumoylation in the carboxyl six aa sequence, we created single and double mutations of the Lys residues by substituting with arginine, confirming that the lysine residues are critical for the sumoylation (data not shown). The remaining weak but readily detectable sumoylation in Axin⌬C6 may derive from other potential sumoylation sites, a total of eight other conserved KX(E/D) sites, throughout the Axin protein.
The C-terminal KVEKVD Sequence Is Critical for JNK Activation-We observed previously that Axin requires both the MID domain and the C-terminal region downstream of the ␤-catenin binding site for JNK activation. We have been interested in how large a role the C-terminal fragment plays in the regulation of the JNK pathway. Fine deletions of the Axin C-terminal region were created and were transfected into 293T cells with FLAG-tagged JNK. Immunokinase assays showed that mere removal of the extreme C-terminal six amino acids completely abolished the Axin-mediated JNK activation. Deletion of either of the KV(E/D) motifs (pKVE, pKVD) reduced JNK activity by ϳ40%. Therefore, sumoylation of Axin appears to be required for its JNK activation (Fig. 3).
Axin Interaction with MEKK1 Was Not Affected by Sumoylation-As removal of the sumoylation sites in the C-terminal six aa of Axin abolished its JNK activation, we asked whether sumoylation has any effect on MEKK1 binding to Axin, which  (1). The EcoRI to the 3Ј-end cDNA fragment encoding C-terminal aa 688 -832 of mouse Axin was fused to pGBKT7 vector as bait. The resulting pGBKT7-Axin-CT vector was transformed into yeast strain AH109. The C-terminal mutants KVE, KVD, K, AVE, and ⌬C6, indicated beneath the gray bar, refer to different mutations at the C terminus of Axin. A and B, one of the identified clones, AIP4, that is able to grow on Ade Ϫ /Leu Ϫ /His Ϫ /Trp Ϫ medium when reconstituted with pGBKT7-Axin-CT but not with pGBKT7 vector alone or pGBKT7-Axin⌬C6 (left panel). Results of co-immunoprecipitation of HA-AIP4 with different Myc-Axin constructs expressed in 293T cells are shown on the right panel. IP, immunoprecipitation; IB, immunoblotting. C and D, as in the case of AIP4, AIP5 can only interact with the wild-type (WT) Axin C-terminal (CT). After sequencing analysis, AIP4 and AIP5 are revealed to encode PIAS1 and PIASx␤, respectively. is a prerequisite for Axin to activate JNK and highly regulated by differential complex formations. We separately co-transfected different HA-Axin constructs with MEKK1-C and immunoprecipitated with anti-HA-Axin. As expected, MEKK1 was coimmunoprecipitated with the wild-type Axin. All the C-terminal Axin mutants co-precipitated with MEKK1-C as effectively as the wild type, indicating that the C-terminal of Axin is not required for MEKK1 interaction (Fig. 4A).
Axin Dimerization Is Intact with the C-terminal Six aa Deleted-Homodimerization is another feature associated with Axin regulation of the JNK pathway. We then tested whether sumoylation affects Axin dimerization. We cotransfected HAtagged Axin with Myc-tagged Axin or Myc-Axin⌬C6 and carried out co-immunoprecipitation. As shown in Fig. 4B, HA-Axin could coprecipitate Myc-Axin or Axin⌬C6. These results suggest that the defect in JNK activation is neither due to a lack of MEKK1 binding nor due to homodimerization of Axin in the C-terminal deletion mutants.
Effect of Sumoylation Status of the C-terminal Region of Axin on ␤-catenin Stability and LEF1 Activity-Axin plays a central role in the regulation of ␤-catenin levels and has a dual function in Wnt signaling and the JNK pathway. We were therefore interested in checking whether sumoylation of Axin plays a role in its regulation of Wnt signaling. A LEF1-luciferase reporter was co-transfected separately with different constructs of Axin, and luciferase activities in differently transfected cells were determined. As shown in Fig. 5A, all the sumoylation-deficient Axin mutants, Axin-KVE, KVD, K, AVE, and ⌬C6, attenuated LEF1 reporter activity as effectively as the wild-type Axin. To further our assertion, we also checked cellular ␤-catenin levels in cells transfected with different Axin constructs. As expected, Axin drastically reduced endogenous ␤-catenin levels as detected by Western blotting with anti-␤-catenin (BD Transduction Laboratories). In contrast to their loss of the ability to activate JNK, the sumoylation-deficient Axin mutants (including pK, pAVE, and p⌬C6) retained the ability to down-regulate ␤-catenin levels (Fig. 5B).

PIAS-mediated JNK Activation Is Blocked by Dominant Negative Axin
Mutants-It was reported previously that PIAS could indirectly activate JNK (30). Since Axin is a sumoylated protein and PIAS is a SUMO-1-conjugating E3, we tested whether PIAS and Axin have any functional interaction in the activation of JNK. As shown in Fig. 6, PIAS activated JNK as expected. We took advantage of previous observations that Axin requires homodimerization for its JNK activation and that Axin⌬C6 and Axin⌬MID still retain their ability to form an oligomer with wild-type Axin. These mutants serve as dominant negative forms to interfere with wild-type Axin in the JNK pathway. We cotransfected PIAS with either Axin⌬C6 or Axin⌬MID and assayed for JNK activities in the cotransfected cells. Results showed that the PIAS-mediated JNK activation was attenuated by Axin⌬C6 and Axin⌬MID, indicating that PIAS activates JNK via Axin (Fig. 6). DISCUSSION We have demonstrated here that Axin is a sumoylated protein. Consistent with its sumoylation, it was found to interact with PIAS1, PIASx␤, and PIASy, which are now recognized as SUMO-1-conjugating enzyme E3. These E3s interact with an Axin C-terminal region including the extreme carboxyl six amino acid residues KVDKVE that harbor two contiguous conserved sumoylation sites. Both sites seem to be SUMO-conjugated as removal of either of the two sites reduced the sumoylation signal. Removal of both of the sumoylation motifs of Axin (Axin⌬C6) abolished the ability of Axin to activate JNK but has no effect on its ability to facilitate ␤-catenin degradation or its ability to down-regulate Wnt signaling. Loss of the JNK-activating ability in Axin⌬C6 is not caused by its failure to form homodimer or to bind to MEKK1, suggesting that sumoylation at the C terminus is required for its JNK activation. The PIAS family members contain a conserved RING finger that is responsible for their E3-conjugating activity (32). The PIASx␤ sequence in the two-hybrid clone contains only the C-terminal region (aa 424 -623), suggesting that the PIAS family members do not require the conserved RING finger for Axin binding.
Sumoylation has been shown to play regulatory roles in protein stability, protein subcellular localization, protein-protein interaction, and transcriptional regulation (31, 32, 36). The best characterized examples include RanGAP1 and promy- FIG. 2. Axin is a sumoylated protein. 293T Cells were co-transfected with 1.5 g of different HA-tagged Axin mutants and 2.5 g of pCMV5-SUMO-1. At 36 h after transfection, cell lysates were immunoprecipitated with anti-SUMO-1 (Santa Cruz Biotechnology, Inc.). The immunoprecipitates and total lysates were then analyzed by Western blotting with anti-HA for Axin. The mutant constructs are the same as described in the legend for Fig. 1. IP, immunoprecipitation; IB, immunoblotting.   FIG. 3. The C-terminal six amino acid residues of Axin are critical for JNK activation. HA-tagged Axin mutants (2.0 g of each) were cotransfected into 293T cells with 1 g of FLAG-tagged JNK. Following immunoprecipitation of FLAG-JNK, its kinase activity was determined using GST-c-Jun as substrate and is presented as means Ϯ S.E. from at least three separate experiments. The amount of JNK in each immunoprecipitate was quantified by immunoblotting (IB) with anti-FLAG; expression levels of different Axin proteins were determined by immunoblotting with anti-HA. elocytic leukemia protein (PML). In the case of PML, SUMO-1 modification is required for PML-mediated complex formation with other proteins (49). It is unclear how sumoylation plays a part in the Axin-mediated JNK activation. Axin requires many domains for its JNK activation, including the region flanked by the ␤-catenin binding site and the DIX domain, in addition to the MID domain and the currently characterized C-terminal six aa downstream of the DIX domain. In light of that, it is conceivable that many other as yet unidentified factors participate with MEKK1 to activate JNK. It is possible that sumoylation helps to regulate protein-protein interaction among the many factors. There seem to be multiple potential sumoylation sites in Axin. However, the main sites responsible for Axin SUMO modification reside in the extreme C-terminal KVEKVD sequence. Removal of either site results in signifi- FIG. 5. Effect of C-terminal mutations on ␤-catenin signaling. As shown in A, C-terminal mutations retain their ability to downregulate LEF1 reporter activity. Axin and mutants (2 g each) as indicated were separately cotransfected into 293T cells with 0.5 g of pGL3-fos-7LEF-luciferase (provided by L. Williams), 0.1 g of pCMV-␤-galactosidase, and 1 g of ␤-catenin, as described previously. As shown in B, mutant Axins retain the ability to deregulate ␤-catenin levels. Axin and mutants (3 g of each) were separately transfected into 293T cells; blank vector pCMV5 was also transfected as a control. After 36 h, cells were directly lysed with 0.5 ml of 1.2ϫ SDS sample buffer and sonicated. An equal volume (10 l) of each sample was subjected to SDS-PAGE electrophoresis for Western blotting with anti-␤-catenin (BD Transduction Laboratories). Expression of Axin and mutants in their respective transfected cells was quantified with anti-HA. IB, immunoblotting.
FIG. 6. Axin⌬MID and Axin⌬C6 exert dominant negative effects on PIAS-mediated JNK activation. FLAG-tagged JNK was cotransfected with Myc-tagged PIAS1 plus Axin⌬MID or Axin⌬C6. Following immunoprecipitation of FLAG-JNK, JNK activity from different transfections was determined as described in the legend for Fig.  3. Expression levels of Axin and PIAS1 in total lysates were determined with anti-HA and anti-Myc, respectively. IB, immunoblotting.
FIG. 4. Axin mutants with alteration in the C-terminal six aa sequence are intact in MEKK1 interaction and homodimerization. As shown in A, MEKK1-C (2 g) was cotransfected separately with 2 g each of C-terminal mutants of HA-tagged Axin. After 36 h, immunoprecipitation was performed with anti-MEKK1. The immunoprecipitates were analyzed by Western blotting with anti-HA for the presence of Axin (middle panel) and with anti-MEKK1 to detect MEKK1 (bottom panel). Axin proteins in total lysates were also quantified with anti-HA to determine their expression levels in each transfection. As a negative control, Axin⌬MID that lacks the MEKK1-interacting domain was included in the experiment. IP, immunoprecipitation; IB, immunoblotting. As shown in B, HA-tagged Axin (2 g) was cotransfected into 293T cells with 2 g of Myc-tagged Axin or Myc-tagged Axin⌬C6. Immunoprecipitation was performed with anti-HA. The co-presence of Myc-tagged Axin or ⌬C6 was determined by Western blotting with anti-Myc. cant loss of sumoylation, whereas deletion of both sites almost abolishes the sumoylation signal. In addition to the two consensus sites, there are several other potential sites in different regions of the axin molecule, KSE 58 -60 , KQE 112-114 ,  KLE 131-133 , KSD 208 -210 , KCD 253-255 , KLE 414 -416 , KLD 526 -528 ,  KAE 609 -611 (the numbers refer to the amino acid positions in  the short form of mouse Axin). Nevertheless, these additional sites represent only a small fraction of sumoylation activity in Axin.
It is interesting to note that both PIAS and Axin are capable of activating JNK and causing apoptosis in a JNK-dependent manner. Remarkably, Axin mutants act dominant negatively on the PIAS-mediated JNK activation. Therefore, PIAS acts either upstream or at a position parallel to Axin. Our current finding thus links PIAS to Axin and to apoptosis.
Axam, a recently identified novel Axin-interacting protein, has been shown to contain SUMO-specific protease activity toward sumoylated proteins such as p53. The catalytic activity of Axam is required for its down-regulation of ␤-catenin (43,50). This is consistent with the observation that sumoylation represses LEF1 activity (29). Our finding that the sumoylationdeficient Axin mutants retain their ability to attenuate LEF1 transcriptional activity indicates that Axam does not act on Axin, at least as far as the sumoylation of Axin at the Cterminal sites is concerned. This notion is further supported by our observation that Axam has no effect on Axin activation of JNK, which requires sumoylation. 3 However, it is not clear whether sumoylated Axin is a direct substrate of Axam on other KV(D/E) sites of Axin. It is conceivable that Axin serves as a platform for the desumoylating enzyme to regulate other protein(s) acting in the Wnt pathway. It is also reasonable to speculate that Axam serves as a SUMO-specific protease to keep Axin desumoylated on other potential sumoylation sites for Axin to function as an inhibitor of Wnt signaling. These findings all point to the possibility that sumoylation plays a role in the Axin-based signaling regulations. Besides sumoylation, other post-translational modification of Axin has been reported. Axin is phosphorylated by GSK-3␤ and is dephosphorylated in response to Wnt signals (38,39). The phosphorylation of Axin regulates its stability and its affinity for ␤-catenin. GSK-3␤ also participates in Axin function in the JNK pathway. However, it seems to exclude MEKK1 from binding to the Axin complex, in contrast to the requirement of its kinase activity in the regulation of Wnt signaling.
Taken together, we have delineated one more biochemical element for the complex mechanism of activation of the JNK pathway by Axin, in addition to our previous findings that Axin utilizes distinct domains from those for Wnt signaling and requires homodimerization for JNK activation. Questions remain as to how sumoylation plays a role in the activation of JNK. What might be the role of sumoylation in the coordination of Axin with MEKK1, dimerization, and eventual activation of JNK? Why is the sequence flanked by the ␤-catenin-binding domain and the DIX domain critical for the Axin-mediated JNK activation? Studies attempting to address these issues are underway.