β-Catenin–related protein WRM-1 is a multifunctional regulatory subunit of the LIT-1 MAPK complex

Significance

Vertebrate β-catenin has two functions: as a structural component in cell adhesion and as the T-cell factor coactivator in canonical Wnt (wingless-related integration site) signaling. These two functions are split between three β-catenin–related proteins in Caenorhabditis elegans. A fourth worm β-catenin–related protein, worm arm motif 1 (WRM-1), exhibits neither of these functions but is essential, together with loss of intestine 1 (LIT-1) MAPK, for transcriptional activation of Wnt target genes. Here we characterize previously unidentified mechanisms by which the LIT-1 kinase is activated and the different regulatory roles that WRM-1 plays in LIT-1 activation and LIT-1 subcellular localization. This study suggests the potential for novel and as yet undiscovered functions for vertebrate β-catenin.

Abstract

Vertebrate β-catenin has two functions, as a structural component of the adherens junction in cell adhesion and as the T-cell factor (TCF) transcriptional coactivator in canonical Wnt (wingless-related integration site) signaling. These two functions are split between three of the four β-catenin–related proteins present in the round worm Caenorhabditis elegans. The fourth β-catenin–related protein, WRM-1, exhibits neither of these functions. Instead, WRM-1 binds the MAPK loss of intestine 1 (LIT-1), and these two proteins have been shown to be essential for the transcription of Wnt target genes by phosphorylating and regulating the nuclear level of the sole worm TCF protein. We showed previously that WRM-1 binds to worm TCF and functions as the substrate-binding subunit for LIT-1. In this study, we show that phosphorylation of T220 in the activation loop is essential for LIT-1 kinase activity in vivo and in vitro. T220 can be phosphorylated either through LIT-1 autophosphorylation or directly by the upstream MAP3K MOM-4. Our data support a model in which WRM-1, which can undergo homotypic interaction, binds LIT-1 and thereby generates a kinase complex in which LIT-1 molecules are situated in a conformation enabling autophosphorylation as well as promoting phosphorylation of the T220 residue by MOM-4. In addition, we show that WRM-1 is essential for the translocation of the LIT-1 kinase complex to the nucleus, the site of its TCF substrate. To our knowledge, this is the first report of a MAP3K directly activating a MAPK by phosphorylation within the activation loop. This study should help uncover novel and as yet underappreciated functions of vertebrate β-catenin.

The catenins function in cell–cell adhesion by linking the transmembrane protein cadherin, which forms homotypic interactions through its extracellular domain, to the underlying actin cytoskeleton (1). The catenin gene family is composed of three subfamilies, namely p120 (seven members), beta (two members), and the structurally unrelated alpha (three members) (reviewed in ref. 2). p120 and beta family members all contain a central domain composed of 9–12 armadillo (Arm) repeats (3), which fold upon themselves to form a superhelical ARM domain with a positively charged central groove (4).

The beta subfamily consists of β-catenin and its paralog γ-catenin (also called “plakoglobin”), with diverged sequences and functions (2). β-Catenin functions in adherens junctions, whereas γ-catenin functions in desmosomes. In addition, β-catenin functions as a coactivator for the T-cell factor (TCF)/LEF transcription factors in the wingless-related integration site (Wnt) pathway.

The β-catenin gene also has undergone species-specific and phylum-specific duplications in nonvertebrates. For example, planaria have two β-catenins, one functioning in adhesion and the other in Wnt signaling (5). The Caenorhabditis elegans genome encodes four highly diverged β-catenin–related proteins (HMP-2, BAR-1, SYS-1, and WRM-1), each as dissimilar to each another in amino acid sequence as they are to Drosophila Armadillo and mammalian β-catenin (6–10). However, structure determination and modeling, along with phylogenetic clustering based on protein sequence, support these four proteins being bona fide 12 Arm repeat-containing homologs of vertebrate β-catenin (11). In addition, genetic analyses showed that three of these four β-catenin–related proteins exhibit one or the other of the two functions of vertebrate β-catenin (6–9, 12). HMP-2 binds cadherin (HMR-1), and α-catenin (HMP-1), localizes to adherens junctions, and the hmp-2 mutant phenotype is entirely consistent with a defect in cell–cell adhesion (6, 12, 13). Therefore, HMP-2 appears to have retained only the cell–cell adhesion function of β-catenin. Both BAR-1 and SYS-1 bind to the conserved N-terminal β-catenin–binding domain of the sole C. elegans TCF protein, POP-1, function with POP-1 to activate transcription from Wnt reporter constructs in tissue-culture cells and in worms, and do not bind to α-catenin or cadherin (8, 12–14). It appears that both BAR-1 and SYS-1 retain the function of β-catenin as coactivators for the TCF/POP-1 in a canonical Wnt pathway, but act at different developmental stages and in different developmental processes (7, 8, 15–17).

The fourth C. elegans β-catenin–related protein, WRM-1, does not exhibit either of these two well-known functions of vertebrate β-catenin. WRM-1 does not bind to α-catenin, cadherin, or the conserved N-terminal β-catenin–binding domain of TCF/POP-1, nor does it function as a coactivator of POP-1 in standard in vitro Wnt reporter assays (12–14). However, loss-of-function and reduction-of-function genetic analyses clearly showed that WRM-1 is required in early embryos in a Wnt signal-dependent process—the specification of the E blastomere in the eight-cell embryo as the sole intestinal precursor (10). Proper specification of E requires a canonical Wnt signaling event that occurs at the four-cell stage between EMS, the mother cell of E, and the adjacent posterior blastomere, P₂ (10, 18–20). The β-catenin SYS-1 functions as the coactivator for TCF/POP-1 in the transcriptional activation of Wnt target genes in E (15, 17). We have shown previously that the E blastomere exhibits a higher level of β-catenin SYS-1 and a lower nuclear level of TCF POP-1 than does its anterior sister, MS (15–17). The high SYS-1 and low POP-1 levels (a high β-catenin–to–TCF ratio) in E require the P₂-to-EMS signal and are crucial for the transcriptional activation of Wnt target genes in E by the SYS-1/POP-1 complex (10, 15, 17, 19, 20).

The main mechanism by which POP-1 nuclear levels are lowered in E in response to the P₂ signal is through nuclear export, which requires POP-1 to be phosphorylated at five specific sites (21). We have shown that the MAPK LIT-1 is responsible for POP-1 phosphorylation at these five sites, both in vivo and in vitro. In lit-1 mutant embryos, POP-1 nuclear levels remain high in both the MS and E nuclei, and the E intestinal fate is repressed in both blastomeres (10, 22). WRM-1 is required for LIT-1 phosphorylation of POP-1, both in C. elegans embryos and in a mammalian tissue-culture cell–based assay (14, 21, 23). Active LIT-1 kinase activity has been obtained in vitro only when coexpressed in mammalian cells along with WRM-1 (14, 21, 23). WRM-1 binds to LIT-1 in C. elegans embryos and mammalian tissue-culture cells (14) and in a yeast two-hybrid assay (23). These results all demonstrate that LIT-1 activity in C. elegans embryos is regulated through WRM-1 binding. Recent genetic analyses have suggested that in larvae CACN-1, a conserved protein of unknown function, also may regulate POP-1 levels partially through LIT-1 (24). We showed recently that one essential function of WRM-1 bound to the LIT-1 kinase is to bind to the kinase substrate, TCF/POP-1 (14). WRM-1 binds to a domain of POP-1 distinct from the conserved N-terminal β-catenin–binding domain to which both SYS-1 and BAR-1 bind. In addition, our analyses showed a second essential requirement for WRM-1 in LIT-1 phosphorylation of POP-1, separate from its substrate-binding function. The second function of WRM-1 in the LIT-1 kinase complex, required for LIT-1 activity, remained unknown.

LIT-1 is an evolutionarily conserved atypical MAPK most closely related to Drosophila Nemo and mammalian Nemo-like kinase (NLK) (23, 25–27). Typical MAPKs have a conserved motif, threonine-X-tyrosine (TXY, in which X can be any amino acid), in their activation loop (Fig. 1A). Activation of typical MAPKs requires phosphorylation of both the threonine and tyrosine residues by an upstream dual-specificity MAP2K (28, 29). However, LIT-1, NEMO, and NLK all have a glutamic acid residue (E) in the third position of this conserved activation motif (T₂₂₀HE, T₂₀₀QE, and T₂₈₆QE, respectively). No upstream MAP2K capable of phosphorylating the threonine in this motif has yet been identified. Multiple genetic screens for genes functioning in the specification of the intestinal precursor did not identify any MAP2K (10, 20, 30). Unlike C. elegans LIT-1, mammalian NLK, when expressed in tissue-culture cells, does not appear to require a binding partner for activation (25, 31, 32). Indeed, it was shown that NLK activation can be achieved through homodimerization. Upon homodimerization in vitro, NLK can autophosphorylate the T286 residue of its TQE activation motif (33).

Fig. 1.

WRM-1 is important for LIT-1 T220 phosphorylation in worm embryos and mammalian cells. (A) Sequences of the activation loop from mouse (Mm) ERK2, p38α, JNK1, and NLK, Drosophila (Dm) NEMO, and C. elegans (Ce) LIT-1, with identical amino acids highlighted in yellow. Conserved amino acids in the activation motifs, TXY or TXE, are in red. (B–D) Western blots using anti–T220-P and anti–LIT-1 antibodies on wild-type embryo extracts that had been treated (+) or not treated (−) with calf-intestinal alkaline phosphatase (CIP) (B), embryo extracts derived from wild-type, lit-1(t1512), or lit-1(or131) worms raised at 25 °C (C), and embryo extracts derived from wild-type or wrm-1–depleted worms (D). (E) GFP-tagged LIT-1 variants were coexpressed with Myc-tagged WRM-1 and Flag-tagged POP-1 in HeLa cells. Following anti-GFP (LIT-1) or anti-Flag (POP-1) pulldowns, Western blots were performed with the antibodies indicated. Input Myc, anti-Myc blot of extract before pulldowns; IP, immunoprecipitation.

Homeodomain-interacting protein kinase 2 (HIPK2) and MAP3K7 (previously TAK1) have been shown to phosphorylate NLK and promote its activation (34–37). However, it is not clear whether these kinases phosphorylate the activation loop of NLK or promote its autophosphorylation. Among the genes identified by genetic screens as functioning in the specification of the C. elegans intestinal precursor was mom-4, the C. elegans homolog of MAP3K7 (10, 20, 30). mom-4 was found to function upstream of lit-1 and wrm-1. However, although depletion of either lit-1 or wrm-1 results in 100% of the embryos lacking intestine (gutless phenotype), strong loss-of-function mom-4 mutations resulted in no more than 50% of the embryos exhibiting the gutless phenotype (10, 20, 22, 27, 30), indicating that activation of LIT-1 is not entirely dependent upon MOM-4. How MOM-4 functioned in the promotion of LIT-1 activity was not known.

In this study, we show that phosphorylation of T220 in the THE motif of the LIT-1 activation loop is essential for LIT-1 kinase function in vivo. T220 can be phosphorylated either through LIT-1 autophosphorylation or directly by the MAP3K MOM-4. WRM-1 is required for LIT-1 T220 autophosphorylation and promotes T220 phosphorylation by MOM-4. We show that WRM-1 normally forms multimers and by so doing promotes multimerization of the LIT-1/WRM-1 complex. This multimerization results in LIT-1 self-association that favors autophosphorylation of LIT-1 at T220.

Results

LIT-1 Is Phosphorylated at T220 in Wild-Type Embryos.

LIT-1 T220 is located in the putative activation loop within kinase domains VII and VIII and is equivalent to T286 of mouse NLK and T200 of Drosophila Nemo, corresponding to T180 of human and mouse p38 MAPK (Fig. 1A) (23, 30). It has been shown that T220 is required for LIT-1 kinase activity in vitro (30). We now show that T220 is required for LIT-1 activity in C. elegans embryos. The lit-1(t1512) temperature-sensitive (ts) mutation changes leucine 177 in the kinase domain of LIT-1, which also is conserved in both Nemo and NLK, to serine (L177S), resulting in 100% embryonic lethality at the nonpermissive temperature (25 °C), with all embryos lacking intestinal cells (100% gutless, n >300) (10, 22). A transgene that expresses wild-type GFP::LIT-1 specifically in the EMS blastomere fully rescues the intestinal defect (0% gutless, n = 50; Table 1). A transgene that expressed GFP::LIT-1 carrying either a T220-to-valine (T220V) or T220-to-aspartate (T220D) mutation failed to rescue (100% gutless, n = 50 and 54, respectively). This result demonstrates that T220 is absolutely essential for LIT-1 kinase activity in vivo.

Table 1.

Summary of transgenic GFP::LIT-1 proteins

Transgenic GFP::LIT-1 variant	In vivo rescue of lit-1(t1512) gutless phenotype*, % gutless (no. of embryos scored)	In vitro kinase activity using POP-1 as a substrate	Binding to WRM-1 as assayed by co-IP	Subcellular localization in embryos
None	100 (>300)	N.A.	N.A.	N.A.
Wild type	0 (50)	Yes	Yes	N>>C
T220V	100 (50)	No	Yes	N<<C
T220D	100 (63)	No	Yes	N = C
D186N	100 (30)	No	Yes	N ≥ C
K89M	100 (32)	No	Yes	N ≥ C
K89G	100 (43)	No	No	N<<C
L177S	100 (32)	No	No	N<<C (excluded in N)
C361Y	0 (21)	Yes, reduced	Yes	N ≥ C

Comparison of GFP::LIT-1 wild type versus seven GFP::LIT-1 variants for (i) the ability to rescue the gutless phenotype of lit-1(t1512) as assayed by presence or absence of gut granules; percent gutless transgenic embryos and number of embryos scored are indicated; (ii) the ability to phosphorylate POP-1 in an in vitro kinase assay, as measured by S117-P by Western blot; (iii) binding to Flag::WRM-1 by coimmunoprecipitation following coexpression in tissue-culture cells; and (iv) subcellular localization in transgenic embryos. C, cytoplasm; co-IP, coimmunoprecipitation; N, nucleus; N.A., not applicable.

^*Scored by the presence or absence of birefringent gut granules.

To elucidate the mechanism by which the LIT-1 kinase is activated in vivo, we raised an antibody against a portion of LIT-1 that included phosphorylated T220. We will refer to this antibody as “anti–T220-P.” A control antibody against amino acids 335–454 of LIT-1 (anti–LIT-1) was generated also. On Western blots, both anti–T220-P and anti–LIT-1 detected an expected band of ∼50 kD in wild-type embryo extracts (Fig. 1B). Pretreatment of the embryo lysates with phosphatase before electrophoresis abolished the anti–T220-P but not the anti–LIT-1 reactive band, indicating that the anti–T220-P antibody recognizes LIT-1 phosphorylated at T220 in vivo. We also performed Western blots using lysates prepared from lit-1(t1512) embryos, as well as from another ts mutant, lit-1(or131) (22, 27), raised at 25 °C. Although the anti–LIT-1 band was detected in embryo extracts from both mutant strains at levels comparable to those in wild type, the anti–T220-P band was not detected (Fig. 1C), showing that both ts lit-1 mutations do not affect stability of the LIT-1 protein but do greatly interfere with phosphorylation at T220.

WRM-1 Regulates LIT-1 Phosphorylation at T220.

Western blots using the anti–T220-P antibody and lysates prepared from embryos that had been depleted of WRM-1 were performed. The deletion mutation wrm-1(tm514) results in embryonic lethality (38) and cannot be maintained as a homozygous stock. We prepared lysates from embryos derived from wrm-1(tm514)/+ mothers that had been fed with wrm-1(RNAi) bacteria. The anti–T220-P reactivity was greatly reduced in these wrm-1− embryo extracts, compared with wild-type embryo extracts (Fig. 1D), demonstrating that in C. elegans embryos WRM-1 activity is required for LIT-1 phosphorylation at T220.

The WRM-1–dependent LIT-1 phosphorylation at T220 observed in C. elegans embryos was recapitulated in mammalian tissue-culture cells (Fig. 1E). LIT-1 is not phosphorylated at T220 when expressed by itself in HeLa cells. However, when coexpressed with WRM-1, LIT-1 phosphorylation at T220 is detected by Western blot. We also assayed LIT-1 activity by coexpressing POP-1 and measuring LIT-1–dependent POP-1 phosphorylation using an antibody specific for POP-1 phosphorylated at S117 (anti–S117-P). S117 is one of the five residues phosphorylated by LIT-1 that are required for POP-1 nuclear export (21). POP-1 S117 phosphorylation was detected only when POP-1 was coexpressed with wild-type LIT-1 and WRM-1 but not when wild-type LIT-1 was expressed without WRM-1 or when the LIT-1 T220V or T220D variants were coexpressed with WRM-1 (Fig. 1E and Table 1). These results demonstrate that T220 is essential for LIT-1 kinase activity and that WRM-1 promotes LIT-1 kinase activity by promoting its phosphorylation at T220.

LIT-1 Can Autophosphorylate at T220.

We have been unable to obtain active recombinant LIT-1 from bacteria. His-tagged LIT-1 purified from bacteria did not undergo autophosphorylation and was unable to phosphorylate POP-1, either by itself or when bacterially purified maltose-binding protein (MBP)-tagged WRM-1 was included. Therefore, to test whether LIT-1 can undergo autophosphorylation, we immunoprecipitated LIT-1 from mammalian cell lysates, incubated it with ATP in an in vitro kinase assay, and then evaluated phosphorylation at T220. When such an experiment was performed using lysates from tissue-culture cells expressing LIT-1 by itself, no phosphorylation of LIT-1 at T220 was observed, even after incubation with ATP in the kinase assay (Fig. 2A, lanes 1 and 2). However, LIT-1 pulled down from cell lysates in which WRM-1 had been coexpressed exhibited clear T220 phosphorylation, which was increased further following incubation with ATP (Fig. 2A, lanes 3 and 4). No increase in anti–T220-P level was observed following the addition of ATP to the in vitro kinase assay when a kinase-dead (KD) version of LIT-1, in which the catalytic aspartate was mutated to asparagine (D186N), was coexpressed with WRM-1 (Fig. 2A, lanes 5 and 6). This result demonstrates that the increased T220 phosphorylation is caused by LIT-1 autophosphorylation.

Fig. 2.

LIT-1 autophosphorylates at T220. (A–D) Western blot analyses of immunoprecipitated samples using the indicated antibodies. (A) GFP-tagged wild-type or D186N (D/N) LIT-1 was pulled down from lysates of tissue-culture cells expressing the indicated proteins, incubated in vitro with or without cold ATP, and assayed by Western blot for LIT-1 T220 phosphorylation. (B) Wild-type or D186N LIT-1 was coexpressed with WRM-1 in MAP3K7-KO or control (wild-type) MEFs. (C) MAP3K7 and TAB1 coexpression promotes WRM-1–dependent LIT-1 T220 phosphorylation. (D) WRM-1–dependent LIT-1 T220 phosphorylation by an endogenous mammalian kinase. Kinase-dead LIT-1(D186N) is phosphorylated at T220 in mammalian tissue-culture cells when WRM-1 is coexpressed. Phosphorylation at T220 does not happen when LIT-1(D186N) and WRM-1 are coexpressed in the absence of MAP3K7 (compare lane 4 in D with lane 4 in B).

Direct Phosphorylation of LIT-1 T220 by the MAP3K MOM-4.

In addition to autophosphorylation, we observed that LIT-1, when coexpressed with WRM-1, can be phosphorylated by an endogenous kinase present in many mammalian tissue-culture cell lines. KD LIT-1(D186N), when coexpressed with WRM-1 in HeLa or HEK293 cells, is still phosphorylated at T220 (Fig. 2A, lane 5, and Fig. 2D). We believe that the endogenous kinase that is capable of phosphorylating LIT-1 at T220 is mammalian MAP3K7 (TAK1). No T220 phosphorylation of either wild-type LIT-1 or KD LIT-1(D186N) was detected in mouse embryo fibroblasts (MEFs) in which the MAP3K7 gene was deleted (Fig. 2B). In addition, when mouse MAP3K7 or both MAP3K7 and its specific activator TAB1 were coexpressed with LIT-1 and WRM-1, we detected an increase in LIT-1 T220 phosphorylation (Fig. 2C). Interestingly, both phosphorylation of KD LIT-1(D186N) T220 by the endogenous MAP3K7 (Fig. 2D) and enhancement of wild-type LIT-1 T220 phosphorylation by cotransfected MAP3K7 (Fig. 2C) are dependent on WRM-1.

Mammalian MAP3K7 and C. elegans MAP3K MOM-4 not only share sequence similarity; they also appear to be at least partial functional homologs. Expression of wild-type MOM-4::GFP from the med-1 promoter, which drives transgene expression specifically in EMS (39), rescued the intestinal defect of mom-4(or39) mutant embryos, from 40% gutless (n >500) to 0% gutless (n = 98). Mouse MAP3K7 similarly expressed in EMS from a med-1 promoter-driven transgene was able to rescue the intestinal defects of mom-4(or39) embryos partially, from 40% to 16% gutless (n = 125).

Three lines of evidence support the conclusion that the C. elegans MAP3K, MOM-4, is a direct upstream kinase for LIT-1 T220 phosphorylation in vivo. First, MOM-4 activity is required for a wild-type level of LIT-1 T220 phosphorylation in embryos. Anti–T220-P levels were greatly reduced in the ts mom-4(ne1539) mutant extract, compared with wild-type extracts (Fig. 3A). Second, wild-type, but not kinase-dead (D176N), MOM-4 can phosphorylate bacterially expressed and purified HIS::LIT-1 at T220 in vitro (Fig. 3B). Active MOM-4 was obtained by coexpressing MOM-4 and the C. elegans TAB1 homolog, TAP-1, in mammalian cells. Phosphorylation of HIS::LIT-1 T220 by MOM-4/TAP-1 occurred without WRM-1 but reached a higher level when MBP::WRM-1 was added to the kinase reaction. Third, coexpression of MOM-4 and TAP-1 with LIT-1 in HeLa cells resulted in phosphorylation of LIT-1 at T220 (Fig. 3C). We performed this experiment using a kinase-dead version of LIT-1 mutated in the ATP binding site (K89M), because this LIT-1 mutant is not phosphorylated by endogenous mammalian MAP3K7. This MOM-4/TAP-1–dependent phosphorylation of LIT-1(K89M) is independent of WRM-1 but, again, is enhanced when WRM-1 is expressed also.

Fig. 3.

Western blot analyses to measure LIT-1 T220 phosphorylation using the anti–LIT-1 and anti–T220-P antibodies. (A) Western blots of C. elegans embryo extracts derived from wild-type or ts mom-4(ne1539) worms raised at 25 °C using the indicated antibodies. (B) MOM-4 phosphorylates bacterially expressed LIT-1 at T220. Wild-type or kinase-dead (D176N) MOM-4/TAP-1 complex was prepared from HEK293 cells in a kinase assay using bacterially expressed and purified LIT-1 and WRM-1. (C) Coexpression of either WRM-1 or MOM-4/TAP-1 promotes LIT-1 T220 phosphorylation. K89M is a kinase-dead version of LIT-1.

Taken together, these results demonstrate that MOM-4 phosphorylates LIT-1 at T220 directly in vitro and suggest that MOM-4 is likely to do so in vivo. Furthermore, phosphorylation of LIT-1 at T220 by MOM-4 does not require WRM-1 but is enhanced when WRM-1 is present.

WRM-1 Promotes LIT-1 T220 Phosphorylation by Promoting Homotypic LIT-1 Interaction.

Mammalian NLK activation by autophosphorylation requires homodimerization (33). Therefore we asked whether LIT-1 can undergo homotypic interaction, whether WRM-1 regulates such an interaction, and whether a homotypic interaction is required for LIT-1 activation. We expressed two separately tagged full-length versions of LIT-1, GFP::LIT-1 and Myc::LIT-1, together in HeLa cells. GFP::LIT-1 was pulled down, and coimmunoprecipitated Myc::LIT-1 was assayed by Western blot. A low but reproducibly detectable amount of Myc::LIT-1 was pulled down with GFP::LIT-1 (Fig. 4A). The amount of Myc::LIT-1 coimmunoprecipitated with GFP::LIT-1 increased when full-length WRM-1 was coexpressed (Fig. 4A). This result suggests that WRM-1 might promote LIT-1 autophosphorylation at T220 by facilitating LIT-1 homotypic interaction.

Fig. 4.

WRM-1 promotes LIT-1 T220 phosphorylation by facilitating LIT-1 homotypic interaction. Western blot analyses of input or immunoprecipitated samples from mammalian tissue-culture cells using the indicated antibodies. Proteins expressed in each lysate are indicated above each panel. (A) WRM-1 facilitates LIT-1 homotypic interaction. Asterisks denote Myc-LIT-1; the three arrowheads in Middle and Bottom indicate full-length Myc-WRM-1 (slowest migrating; top arrowhead), Myc-WRM-1(1-552) (middle arrowhead), and Myc-WRM-1(1-410) (fastest migrating; bottom arrowhead). (B) CC3- or CC4-fused LIT-1 exhibits T220 phosphorylation in the absence of WRM-1. FL, full-length. Arrowheads and asterisk denote LIT-1 and IgG heavy chain, respectively. (C) Schematic of WRM-1 proteins used in this study. The 12 Arm repeats are shown as open rectangles, and the N-terminal and C-terminal domains are shown as solid lines. (D) Domains of WRM-1 required for LIT-1 binding, T220 phosphorylation, and POP-1 S117 phosphorylation. (E) Domains of WRM-1 required for homotypic interaction. The Flag-tagged version of WRM-1 is indicated first, followed by the Myc-tagged version after the forward slash. (F) CC2 or CC3, but not CC4, fused to the C terminus of WRM-1(1–650) can promote LIT-1 T220 phosphorylation.

If WRM-1 promotes LIT-1 autophosphorylation by facilitating LIT-1 homotypic interaction, the question then arises: Can LIT-1 activation occur in the absence of WRM-1, if LIT-1 is made to undergo homotypic interaction? The 31-aa GCN4 leucine-zipper coiled-coil (CC) domain has been used as an artificial multimerization domain for CC-tagged proteins. Normally, this domain almost exclusively promotes the formation of dimers. Mutagenesis has been used to generate variant GCN4 CC domains that specifically form trimers or tetramers (40, 41). We will refer to the GCN4 CC domains that specifically di- tri-, or tetramerize as “CC2,” “CC3,” and “CC4” motifs, respectively (Fig. 4B). We fused the three variant GCN4 CC motifs separately to the C terminus of LIT-1 and expressed LIT-1::CC2, LIT-1::CC3, or LIT-1::CC4 individually in HeLa cells. By Western blot we could detect T220 phosphorylation with LIT-1::CC3 and LIT-1::CC4, albeit at a low level (Fig. 4B). Cotransfection of WRM-1 resulted in an increase of T220 phosphorylation for all three LIT-1::CC variants. The finding that WRM-1 is dispensable in promoting T220 phosphorylation as long as LIT-1 can form homotypic interactions via the GCN4 CC domains supports the model in which WRM-1 activates LIT-1, at least in part, by facilitating homotypic interaction of LIT-1.

Multiple Domains of WRM-1 Are Required for Stable LIT-1 Interaction.

To elucidate the mechanism by which WRM-1 promotes LIT-1 homotypic interaction, we mapped the domains of WRM-1 that form homotypic interaction and that interact with LIT-1. Like vertebrate β-catenin, WRM-1 consists primarily of 12 Arm repeats, flanked by a 150-aa N-terminal domain and an approximately 30-aa C-terminal domain (Fig. 4C) (23). The 148 N-terminal amino acids of WRM-1 have been shown to be both necessary and sufficient to bind LIT-1 in a yeast two-hybrid assay (23). In a pull-down assay using Myc-tagged WRM-1 and GFP-tagged LIT-1, we found that both the 148 N-terminal amino acids and the 145 C-terminal amino acids of WRM-1 are required for stable interaction with LIT-1 (Fig. 4D). Despite not forming a stable interaction with LIT-1, both WRM-1(149–796) and WRM-1(1–650) are sufficient to promote a low level of LIT-1 T220 and POP-1 S-117 phosphorylation in tissue-culture cells, suggesting that both WRM-1 truncations can form functional, albeit transient, interactions with LIT-1 in cells (Fig. 4D). However, deleting a larger fragment from the C terminus [WRM-1(1–552)] results in undetectable LIT-1 T220 and POP-1 S117 phosphorylation.

The 148 N-terminal amino acids of WRM-1 also are required for WRM-1 homotypic interaction. Flag-tagged full-length WRM-1 can pull down Myc-tagged full-length WRM-1 (Fig. 4E). Deleting 148 amino acids from the N terminus of WRM-1 results in no homotypic interaction in the pull down (Fig. 4E). The first 410 amino acids of WRM-1 are sufficient for homotypic interaction in the pull-down assay.

Artificial Multimerization Motifs Can Substitute for the Wrm-1 C-Terminal Domain.

The C terminus of WRM-1 contains putative CC domains (amino acids 651–682 and 751–780) (COILS: www.ch.embnet.org/software/COILS_form.html, accessed November 2011). Although the WRM-1 CC domains are not required for the WRM-1 homotypic interactions detected in a pull-down assay (Fig. 4E), they are required for stable WRM-1/LIT-1 interaction (Fig. 4D) and the ability of WRM-1 to promote LIT-1 homotypic interaction. WRM-1(1–552) and WRM-1(1–410) not only fail to promote, but even appear to interfere with, LIT-1 homotypic interaction (Fig. 4A). It is possible that the C-terminal domain of WRM-1 stabilizes the WRM-1/LIT-1 complex through weak homotypic interaction. We asked whether the CC2, CC3, or CC4 domain from the GCN4 protein could substitute for these CC domains at the WRM-1 C terminus. Specifically, could WRM-1(1–650) fused with any of the CC domains promote LIT-1 T220 phosphorylation? We did not use WRM-1(1–552) to test the rescue by the CC domains because of the concern that such a large deletion might yield negative results that would be more difficult to interpret. CC2, CC3, and CC4 were fused individually to WRM-1(1-650), and each fusion protein was coexpressed separately with LIT-1 in tissue-culture cells. We observed that coexpression of CC2 or CC3, but not CC4, fused WRM-1(1–650) with full-length LIT-1 resulted in enhanced LIT-1 phosphorylation at T220, as compared with the coexpression of WRM-1(1–650) lacking any CC domain. However, LIT-1 T220 phosphorylation levels were less than those observed following coexpression of LIT-1 with full-length WRM-1 (Fig. 4F). This result nonetheless supports the model that WRM-1 homotypic interaction (through both its N- and C-terminal domains) underlies the ability of WRM-1 to promote LIT-1 T220 autophosphorylation.

WRM-1 Regulates LIT-1 Nuclear Levels in Embryos.

GFP-tagged LIT-1 expressed in C. elegans embryos is predominantly nuclear, although cytoplasmic and cortical signals are observed also (21). The levels of GFP::LIT-1 in the nucleus are asymmetric between the MS and E blastomeres, with higher levels observed in the E nucleus (Fig. 5) (21). Depletion of wrm-1 by RNAi resulted in reduced nuclear enrichment of GFP::LIT-1 in all GFP⁺ cells examined during embryogenesis, suggesting that WRM-1 is involved in regulating nuclear LIT-1 levels (Fig. 5). WRM-1 may regulate LIT-1 nuclear levels indirectly by regulating its kinase activity, directly by promoting its nuclear entry or retention, or both. Consistent with the first possibility, we observed that all LIT-1 mutant proteins with reduced or abolished kinase activity also exhibited reduced overall nuclear levels, relative to cytoplasmic levels, when expressed as GFP fusion proteins (Fig. 5 and Table 1). However, differences in kinase activity alone do not explain the differences in LIT-1 subcellular localization. For example, both L177S and D186N mutations abolish LIT-1 kinase activity by all measures but exhibit different localization patterns; L177S is excluded completely from the nucleus, whereas D186N is only marginally defective in its nuclear localization. We find it intriguing that the two mutations that exhibit the most severe exclusion of LIT-1 from the nucleus, L177S and K89G, also are the only two LIT-1 variants among the ones we tested in our standard pull-down assay that fail to bind to WRM-1 (Figs. 1E, 2D, and 5 and Table 1). This result suggests that WRM-1 also may regulate LIT-1 nuclear levels directly via physical interaction. The Arm repeats found in β-catenins, along with the superhelical ARM domains formed by Arm repeat stacking, are very similar to the HEAT repeats and the superhelical structures formed by the Importin β family (42), which includes the exportins (43). It has been shown that β-catenin can undergo nuclear import independent of known importins and in so doing can translocate specific protein cargoes into the nucleus (44). It is an intriguing possibility that WRM-1 can function in a manner analogous to an importin β, transporting its cargo, LIT-1, from the cytoplasm to the nucleus.

Fig. 5.

Nucleo-cytoplasmic distribution of GFP::LIT-1 in embryos. Confocal micrographs of live embryos expressing wild-type LIT-1 or the indicated LIT-1 variants tagged at their N termini with GFP. (A and B) Embryos expressing wild-type GFP::LIT- 1 at the 1MS, 1E (A) or 2MS, 2E (B) stages. MS and E blastomeres are indicated. Sister blastomeres are connected by a line. Note that GFP is predominantly nuclear in all stages shown, and the nuclear level is higher in E than in MS. (C and D) GFP::LIT-1 in wild-type embryos that have not (C) or have (D) been depleted of wrm-1 by RNAi. Note that GFP is no longer enriched in the nucleus following wrm-1 RNAi (D). (E–K) Expression of GFP::LIT-1 carrying the indicated amino acid changes in wild-type embryos at the 2MS, 2E stage. (Scale bar: 10 μm.)

Localization of these LIT-1 GFP variants remains unchanged regardless of whether the endogenous, wild-type LIT-1 is present or depleted (Fig. S1). The observation that endogenous wild-type LIT-1 activity within the same cell does not rescue the defective subcellular localization of GFP::LIT-1 variants suggests that nuclear enrichment of GFP::LIT-1 is determined by an inherent characteristic of the GFP::LIT-1 protein variant and not by the overall LIT-1 protein level or kinase activity in the cell.

Discussion

We show in this study that phosphorylation of the MAPK LIT-1 at T220 in the conserved TXE motif of the activation loop is absolutely required for LIT-1 kinase activity. We find that phosphorylation of LIT-1 at T220 can be achieved by either of two independent pathways: first, by autophosphorylation, or second, via the upstream MAP3K, MOM-4. Furthermore, we show that the β-catenin–related protein WRM-1 is essential for LIT-1 autophosphorylation and that it enhances phosphorylation of LIT-1 T220 by MOM-4. Our data support a model in which WRM-1, which binds LIT-1 through multiple domains, undergoes homotypic interactions, generating a complex that provides a physical scaffold for LIT-1 homotypic interactions, enabling LIT-1 autophosphorylation as well as enhancing LIT-1 phosphorylation by MOM-4.

Three sets of experiments support direct phosphorylation of LIT-1 at T220 by MOM-4. First, in extracts derived from ts mom-4(ne1539) embryos, we observed a dramatic reduction of LIT-1 T220 phosphorylation. Second, we observed MOM-4 activity-dependent phosphorylation at T220 of a kinase-dead LIT-1 in mammalian tissue-culture cells. Third, we obtained MOM-4 activity-dependent phosphorylation at T220 of a bacterially expressed LIT-1 in vitro.

Activation of a MAPK by direct phosphorylation within the activation loop by a MAP3K has not been reported previously. Most MAPKs have a TXY motif in the activation loop that requires phosphorylation at both the threonine and tyrosine residues for activation, and this phosphorylation is carried out by a single MAP2K with dual serine/threonine and tyrosine specificity (28, 45). MAP3Ks exhibit only serine/threonine kinase activity and therefore, typically, are unable to activate MAPKs directly. However, the atypical MAPKs LIT-1, NEMO, and NLK, have a TXE motif, instead of TXY, in their activation loop (23, 25, 26). Therefore, the tyrosine kinase activity of a MAP2K is not required for phosphorylation of this activation loop motif, and presumably the glutamic acid residue mimics a phospho-tyrosine residue. Therefore it is very likely that phosphorylation of LIT-1 at T220, by whatever means, would activate LIT-1. Consistent with MOM-4 being capable of phosphorylating LIT-1 at T220 directly in vivo, no MAP2K that functions in the specification of C. elegans endoderm has been identified to date, either by extensive genetic screens or by candidate gene approaches.

We also show that phosphorylation of LIT-1 at T220 can result from autophosphorylation. This finding is consistent with genetic data that mom-4 activity is not absolutely required for LIT-1 activity in vivo. Although lit-1 depletion causes 100% gutless embryos, likely null or severe reduction-of-function mutations in mom-4 result in only ∼40% gutless embryos (20, 22, 27, 30). Although we cannot completely rule out the possibility that a second, yet to be identified upstream kinase exists that can activate LIT-1 in vivo, our results support autophosphorylation being a second mechanism for LIT-1 activation.

The LIT-1 requirement of WRM-1 for autophosphorylation is peculiar, given that NLK was shown to be active when expressed by itself in tissue-culture cells (25, 31, 32). This difference could reflect differences in the ability of NLK and LIT-1 to form productive homotypic interactions that promote autophosphorylation. NLK readily forms homodimers, and dimerization is essential for functional activation as well as for nuclear entry (33). We suggest that LIT-1 by itself does not form productive dimers/multimers through homotypic interaction (Fig. 6). By binding to LIT-1 through multiple regions of contact, WRM-1, which can undergo homotypic interaction, helps organize LIT-1 into a multimeric form that is conducive for autophosphorylation. In addition, this WRM-1–driven LIT-1 multimerization also presents T220 in a context that is favorable for phosphorylation by MOM-4. Our observation that CC domain-fused multimeric LIT-1 is phosphorylated at T220 in the absence of WRM-1 supports this model. Our data show that WRM-1 homotypic interactions occur through the 148 N-terminal amino acids. However, the C-terminal region of WRM-1 (amino acids 651–769) also appears to form weak homotypic interactions that help stabilize the WRM-1–LIT-1 interaction and which can be replaced by artificial multimerization domains.

Fig. 6.

Model for LIT-1/WRM-1 interaction and LIT-1 T220 phosphorylation. Schematic of full-length WRM-1 (dark blue; N and C termini are indicated) and LIT-1 (aqua). (A) Multiple interactions between WRM-1 and LIT-1 stabilize the complex, allowing autophosphorylation (red P) at T220 and facilitating phosphorylation by MAP3K MOM-4. (B and C) WRM-1 lacking either the 148 N-terminal or the 224 C-terminal amino acids (indicated by the dashed lines) cannot form a stable complex with LIT-1. (D) Artificial multimerization domains added to LIT-1 or to C-terminal–truncated WRM-1 partially restore the conformation required for LIT-1 T220 phosphorylation.

In vivo, LIT-1–dependent phosphorylation is tightly regulated by signals from the P₂ blastomere. However, if LIT-1 can be activated simply by WRM-1 binding, how does P₂ signaling regulate LIT-1–dependent events in embryos? We suggest the following scenarios whereby LIT-1–dependent events can be regulated. First, activation of the MOM-4 MAP3K is dependent upon P₂ signaling, which will ensure that maximum LIT-1 activity occurs only after the reception of the P₂ signal. Second, binding between WRM-1 and LIT-1 could be regulated by P₂ signaling. Although LIT-1 and WRM-1 exhibit a nearly identical subcellular localization pattern in early embryos (21, 38, 46), their binding could be promoted by P₂ signaling. Third, P₂ signaling promotes a higher nuclear level of both LIT-1 and WRM-1 in E, localizing the kinase complex in the same subcellular domain as its substrate, TCF/POP-1. Fourth, binding of WRM-1 to POP-1 could be regulated by P₂ signaling (14).

Many kinases do not become fully activated until bound to a partner protein. In almost all cases examined, the binding partners activate their cognate kinases by forming crucial interactions that “complete” the proper folding of the kinase domain (reviewed in ref. 47). One of the best-known examples is the activation of cyclin-dependent kinases (CDKs) by their binding partners, the cyclins (48–51). Cyclin A forms an extensive interface with unphosphorylated Cdk2, which releases the Cdk2 activation segment from its autoinhibitory conformation. Cdk2 then adopts a partially active form whereby a particular threonine in the activation loop becomes accessible for phosphorylation by the upstream kinase. Furthermore, cyclin binding also determines the substrate specificity of CDKs by either binding directly to substrate or regulating the localization of the CDK complex to the proper subcellular site where the substrate is located (49). Although crystal structures for LIT-1 or WRM-1 are unavailable, our current analyses here reveal a striking parallel between the multifunctional nature of WRM-1 in the LIT-1 kinase complex and that of cyclin for a CDK. WRM-1 promotes the presentation of the substrate, POP-1, to LIT-1 in two independent ways: first, directly, by binding to the substrate, and second, indirectly, by elevating LIT-1 levels in the nucleus, where the substrate is localized. In addition, we propose that, as a result of WRM-1 binding, LIT-1 adopts a conformation that makes T220 in the activation loop accessible for autophosphorylation as well as for phosphorylation by MOM-4, the upstream MAP3K.

Although WRM-1 does not exhibit the two major functions typically associated with β-catenin, it does exhibit tantalizing structural characteristics that resemble β-catenin as well as lesser-known functions that recently have been attributed to β-catenin. For example, Xenopus β-catenin has been shown to facilitate phosphorylation of TCF3 by the HIPK2 complex by functioning as a scaffold for multiple components (52). This finding is analogous to our showing that one function of WRM-1 in the LIT-1 kinase complex is to bind to the substrate TCF/POP-1 (14). WRM-1 amino acid sequence analysis predicts 12 contiguous Arm repeats, typical of β-catenins, and phylogenetic analysis places WRM-1 within the β-catenin clade (11). In addition, structural modeling predicted distinct parallels between the binding to the conserved N-terminal domain of TCF/POP-1 by β-catenin/SYS-1 and binding by WRM-1 to the POP-1 C-terminal domain (11, 14). Finally, our data suggest that WRM-1 may facilitate LIT-1 nuclear localization through direct physical interaction. This observation is consistent with the importin-independent nuclear localization displayed by β-catenin. We suggest that all four β-catenin–related proteins in C. elegans arose from two rounds of duplication from one single β-catenin. During evolution, each of the four proteins has diverged and retains only a subset of the functions carried out by vertebrate β-catenin (9). Although not retaining either of the primary functions associated with the vertebrate β-catenin, WRM-1 may play a key role in establishing additional vertebrate β-catenin functions.

Materials and Methods

Strains.

N2 was used as the wild-type strain. Genetic markers used for LGI were mom-4(ne1539ts), mom-4(or39)/hT2 and for LGIII were lit-1(t1512ts), lit-1(or131ts), wrm-1(tm514)/hT2. All transgenic strains used were generated by injection and are nonintegrated lines. For every transgenic construct, multiple transgenic lines were obtained. The GFP fluorescence of GFP::LIT-1 variants was compared with that of wild-type GFP::LIT-1 in 12- to 28-cell–stage embryos. All rescue experiments used transgenic lines that exhibit fluorescence levels similar to that of wild-type GFP::LIT-1. Similarly, the levels of GFP fluorescence of GFP:: MAP3K7 and GFP::MOM-4 in respective transgenic strains used for rescue assays are comparable.

Plasmid Construction.

All the expression clones used in this study were generated using the Gateway cloning technology (Invitrogen). Mutations were generated with the QuikChange II Site-Directed Mutagenesis Kit (Stratagene/Agilent Technologies) and confirmed by sequencing. All clones expressed in HeLa cells were driven by the CMV promoter; those expressed in embryos were driven by the EMS-specific med-1 promoter (39). Unless specified, all tags were added at the N terminus, although fusion to the C terminus resulted in the same results for both WRM-1 and LIT-1. Expression clones used in this study are listed in Table S1.

Analysis of Embryos and Imaging.

Images of GFP::LIT-1 variants in live embryos were collected as 3D stacks using an LSM700 confocal microscope (Zeiss). Assay for rescue of the E defect in lit-1(t1512) and mom-4(or39) mutant embryos was performed as described (21), with modifications, using an Axioplan microscope (Zeiss) equipped with epifluorescence, polarizing, and differential interference contrast (DIC) optics and a MicroMax-512EBFT CCD camera (Princeton Instruments) controlled by MetaMorph acquisition software (Molecular Devices) (15, 19). Production of MS-derived pharyngeal tissues was scored by DIC optics. Formation of intestine was assayed with both DIC and polarizing optics.

Antibody Generation.

Anti–T220-P, raised against the peptide QRDRLNMT[p]HEVVTQYY, and anti–LIT-1, raised against a bacterially expressed protein corresponding to amino acids 335–454 of LIT-1e, were generated in rabbits at Genemed Synthesis, Inc. and Proteintech Group, respectively, and purified according to the company’s protocol. Anti–T220-P antiserum was subjected to dual selection at Genemed Synthesis, first for binding to the phosphopeptide and then for lack of binding to the nonphosphorylated peptide.

Lysate Preparation, Immunoprecipitation, and Western Blots.

Embryos were collected from gravid adults, and lysates were prepared as described (14). Transfections were performed into mammalian tissue-culture cells, and lysates were prepared as described (14). All transfections were performed using HeLa cells, except (i) transfection of MOM-4, TAP-1, MAP3K7, or TAB1 into HEK293 cells, and (ii) transfection of wild-type and mutant LIT-1 into wild-type and MAP3K7-KO MEF cells (a gift from Zhijian J. Chen, University of Texas Southwestern Medical Center, Dallas).

Antibodies used for immunoprecipitations were camelid anti-GFP antibody (GFP-Trap; ChromoTek) and anti-FLAG agarose beads (M2; Sigma). Antibodies used in Western blots include anti–S117-P at 1:500 (21), anti–LIT-1 (this study) at 1:2,000, anti-T220P (this study) at 1:2,000, anti–c-Myc (9E10; Santa Cruz Biotechnology) at 1:2,000, and anti-FLAG (M2; Sigma) at 1:2,000. Secondary antibodies used were donkey anti-rabbit IgG-HRP (GE Healthcare) and goat anti-mouse IgG1-HRP (Santa Cruz Biotechnology), both at 1:20,000.

Kinase Assays.

In vitro kinase assays were performed as described (14). GFP-LIT-1 variants were expressed alone or together with WRM-1 in HeLa cells and were immunoprecipitated with GFP-Trap beads. Immunoprecipitates were washed three times with wash buffer (50 mM Hepes, 250 mM NaCl, 0.1% Nonidet P-40), rinsed twice with PBS, rinsed one time with kinase buffer minus ATP, followed by incubation in 20 μL kinase buffer [20 mM Hepes (pH 7.4), 5 mM MgCl₂, 1 mM DTT, 2 mM ATP] at 30 °C for 60 min. For MOM-4 phosphorylation of bacterially purified LIT-1, Flag-tagged wild-type MOM-4 and kinase-dead mutant MOM-4 D176N were expressed separately in HEK293 cells, immunoprecipitated with anti-Flag M2 beads, washed as described above, and incubated with 1 μg bacterially purified LIT-1 in 20 μL kinase buffer [20 mM Tris⋅HCl (pH 7.5), 5 mM MgCl₂, 1 mM DTT, 2 mM ATP] at 30 °C for 30 min. After incubation, reactions were stopped with SDS gel-loading buffer, and phosphorylation levels were assayed by Western blot using anti–T220-P antibody. Kinase assays were performed initially at 20 °C, 25 °C, and 30 °C, with no qualitative differences noted between the three temperatures. Kinase assays were performed at 30 °C because reactions were more robust at this temperature.

LIT-1 expressed in E. coli was found to be inactive in all our assays, even if WRM-1 was coexpressed and copurified with LIT-1. In addition, for reasons that remain unclear, phosphatase treatment of the LIT-1/WRM-1 complex pulled down from mammalian cells resulted in dissociation of WRM-1 from LIT-1 and therefore in the loss of LIT-1 activity. Therefore autophosphorylation (Fig. 2A) was assayed using LIT-1 pulled down from tissue-culture lysates for LIT-1 kinase activity-dependent incorporation of ATP at T220.

Recombinant Proteins.

His-tagged full-length LIT-1 was expressed alone or together with MBP-tagged full-length WRM-1 in E. coli at 16 °C and was purified using TALON metal affinity resin (BD Biosciences) according to the manufacturer’s protocol. The purity of the eluted proteins was evaluated on SDS/PAGE gels stained with Coomassie Brilliant Blue, and the coimmunoprecipitation of WRM-1 along with LIT-1 was confirmed by Western blot.