Phosphorylation by tyrosine and serine/threonine kinases regulate the interactions between components of the cadherin–catenin cell-adhesion complex and thus can influence the dynamic modulation of cell adhesion under normal and disease conditions. Previous mutational analysis and localization experiments suggested an involvement of single members of the family of PAKs (p21-activated kinases) in the regulation of cadherin-mediated cell adhesion, but the molecular mechanism remained elusive. In the present study, we address this question using the Drosophila PAK protein Mbt, which is most similar to vertebrate PAK4. Previous phenotypic analysis showed that Mbt has a function to maintain adherens junctions during eye development and indicated a requirement of the protein in regulation of the actin cytoskeleton and the cadherin–catenin complex. Here we show that activation of Mbt leads to destabilization of the interaction of the Drosophila β-catenin homologue Armadillo with DE-cadherin resulting in a decrease in DE-cadherin-mediated adhesion. Two conserved phosphorylation sites in Armadillo were identified that mediate this effect. The findings of the present study support the previous observation that activation of the human Mbt homologue PAK4 leads to anchorage-independent growth and provide a functional link between a PAK protein and the cadherin–catenin complex.
- cadherin–catenin complex
- kinase activity
- p21-activated kinase (PAK)
The establishment, maintenance and modulation of cell–cell contacts is of crucial importance during the development of multicellular organisms to build complex tissues by means of cell-sorting processes, cell movement and changes in cell morphology. Consequently, aberrant changes in the cells' adhesive properties can result in pathological conditions such as cancer formation. Adherens junctions provide the mechanical strength to cell–cell contacts and consist of classic cadherins, which undergo homophilic, Ca2+-dependent connections between cells [1–3]. The cytoplasmic tail of cadherins is anchored to the underlying actin cytoskeleton via members of the catenin family. β-Catenin and its Drosophila melanogaster homologue Armadillo, in addition to their role in the Wnt (Wingless in Drosophila) signalling pathway, serve as molecular adaptors between cadherins and α-catenin, which is linked directly or indirectly to the actin cytoskeleton. Recent data show that α-catenin acts as a molecular switch: α-catenin in its monomeric form is a structural component of the CCC (cadherin–catenin complex) and does not interact with actin, whereas α-catenin homodimers regulate actin filament organization . On the other hand, β-catenin/Armadillo is subject to a complex (and not completely deciphered) pattern of phosphorylation by tyrosine and serine/threonine kinases to regulate its diverse function in cell adhesion and Wnt/Wingless signalling [5,6]. Degradation of cytoplasmic β-catenin/Armadillo is induced by phosphorylation of four conserved N-terminal serine/threonine residues through the combined activities of GSK3β (glycogen synthase kinase-3β)/Zw3 (Zeste-White 3) and CK1 (casein kinase 1). Phosphorylation of Tyr654 in β-catenin by Src kinases leads to a loss of binding to cadherin, whereas phosphorylation of Tyr142 by Fer and Fyn kinases disrupts binding to α-catenin but promotes binding to the co-factor BCL9-2 to induce gene transcription. PKA (protein kinase A) and Akt kinase induced phosphorylation of Ser552 and Ser675 promotes the transcriptional activity of β-catenin. Phosphorylation of Tyr489 by the Abl tyrosine kinase upon stimulation of the axon guidance receptor Robo decreases the affinity of β-catenin for N-cadherin . Beside GSK3β and CK1, CK2 (casein kinase 2) phosphorylates β-catenin at several sites in the N-terminal region and thereby influences the stability of the protein and strengthens the binding to α-catenin . Mutational analysis in Drosophila support the idea that Armadillo localized at adherens junctions can be phosphorylated by a serine/threonine kinase other than GSK3β/Zw3 . In the present study, we addressed the question whether the Drosophila PAK (p21-activated kinase) Mbt can fulfil this function.
PAK proteins are a family of serine/threonine kinases that are regulated by binding of RhoGTPases such as Cdc42 (cell division cycle 42) or Rac1. Based on structural features, PAK proteins are classified into two subgroups, each representing three members in human (group 1, PAK1–3; group 2, PAK4–6) . In Drosophila melanogaster, two PAK proteins (DmPAK and DmPAK3) belong to group 1, whereas Mbt (DmPAK2) is the only group 2 PAK. Cadherin-mediated cell adhesion influences the activity of the RhoGTPases Cdc42, Rac and Rho, which in turn can also modulate the assembly and stability of the cell-adhesion complex [3,11,12]. Several lines of evidence indicated a function of single group 2 PAK proteins in regulation of adhesive properties of cells. Constitutive activation of human PAK4 leads to changes in cell morphology, decreased cell adhesion and induction of anchorage-independent growth [13,14]. In Xenopus laevis, the group 2 PAK protein X-Pak5 is expressed in regions of the embryo that undergo extensive cell movements during gastrulation. X-Pak5 localizes to cell–cell contact sites and regulates cell adhesion in a calcium-dependent manner . Similarly, Drosophila Mbt localizes at adherens junctions and is required for morphogenetic processes during eye development [16,17]. In the present study, we provide first evidence on the molecular function of Mbt in cadherin-mediated cell adhesion.
MATERIALS AND METHODS
DE-cadherin (D. melanogaster E-cadherin) was cloned as an Asp718/BamHI fragment from pRm-DE-cadherin (kindly provided by P. Rorth, Developmental Biology Unit, European Molecular Biology Laboratory, Heidelberg, Germany) into pAc5.1 and pIZ/V5 (Invitrogen). To generate a chimaeric IgG-fusion protein, the coding region of the extracellular domain of DE-cadherin (1–3987 bp) was amplified by linker-PCR, cut with EcoRI/BamHI and inserted into the pCR3 vector containing the hinge and Fc-region from human IgG1 (kindly provided by Dr P. Schneider, Department of Biochemistry, University of Lausanne, Switzerland). The generation of stable S2R cell lines expressing Myc–Mbt, Myc–MbtT525A and Myc–MbtS492N/S521E are described in . MbtT525A and MbtS492N/S521E were further cloned into the Asp718 site of pEBG, a GST (glutathione transferase)-expression vector for HEK (human embryonic kidney)-293 cells (a gift from Dr O. Bernard, Walter and Eliza Hall Institute of Medical Research, Parkville, Australia). For expression of a full-length GST–Armadillo fusion protein in Escherichia coli, the coding region of Armadillo (cDNA clone LD33342, obtained from the Drosophila Genome Resource Center) was amplified by linker PCR and first cloned into the XhoI/XbaI sites of pcDNA3.1Myc. For generation of the N-terminally truncated Armadillo protein (amino acids 76–843), first an additional XhoI site was introduced by in vitro mutagenesis at base pair position 229, followed by XhoI digestion and selfligation. Both constructs were then cloned as BamHI/XbaI fragments into the pGEX vector (GE Healthcare). For expression of HA (haemagglutinin) epitope-tagged variants of Armadillo in S2R cells, the HA-epitope encoding sequence was introduced into the KpnI site of the pAC5.1 vector to generate pAC5.1-HA. Armadillo was subcloned with BamHI/XbaI from pcDNA3.1Myc into the pBS vector (Stratagene). The QuikChange® site-directed mutagenesis protocol (Stratagene) was used to change the codons of Ser561 and Ser688 to alanine. Wild-type and mutated Armadillo constructs were then cloned with XhoI/XbaI into pAC5.1-HA. All constructs were verified by sequencing.
Cell culture and cell fractionation
Drosophila Schneider S2R and S2 cells  were cultured at 25 °C in Schneider medium (Biowest) supplemented with 10% fetal calf serum, 1% L-glutamine, 1% penicillin/streptomycin (Gibco) and if necessary with 100 μg/ml hygromycin (Calbiochem) and 400 μg/ml zeocin (Invitrogen) for establishment and maintenance of stable cell lines. Transfections of S2R and S2 cells were performed with the calcium phosphate method. Membrane fractions were prepared as described in . HEK-293 cells were cultured in DMEM (Dulbecco's modified Eagle's medium) with 10% fetal calf serum. Transient transfections of HEK-293 cells were carried out using the JetPEI Polyplus reagent (Biomol) according to the manufacturer's protocol.
Immunohistochemistry and microscopy
S2R cells were grown on coverslips (if necessary induced with 0.05 μM CuSO4), fixed with 4% (w/v) paraformaldehyde for 20 min, permeabilized with 0.2% (v/v) Triton X-100 in PBS for 10 min and incubated with rat anti-DE-cadherin [1:30, hybridoma cell line supernatant obtained from the DSHB (Developmental Studies Hybridoma Bank)], mouse anti-Armadillo (1:100, concentrate from the DSHB), rat anti-α-catenin (1:100, concentrate from the DSHB) or rabbit anti-Mbt (1:500, ) in PBS containing 1% (v/v) normal goat or normal donkey serum. After washing in PBS, cells were incubated with Cy3-, Cy5- or Alexa Fluor® 488-conjugated secondary antibodies (Dianova and Molecular Probes) diluted in PBS. In some cases, a ready-to-use Hoechst stain solution (Sigma–Aldrich) was added to stain DNA. Cells were embedded in Mowiol and analysed with a Leica DM5000B microscope using a 63×HCX PL APO oil objective. Images were taken with a Leica DFC350FX digital camera and false-colour coded with the Amira software (Mercury Computer Systems). Control stainings with single primary antibodies were performed to exclude bleed-through in other detection channels (see Supplementary Figure S1 at http://www.BiochemJ.org/bj/416/bj4160231add.htm).
Protein expression and purification, immunoprecipitation and immunoblot analysis
Expression of Myc–Mbt variants in Schneider S2 and S2R cells was induced by addition of 0.8 mM CuSO4. Immunoprecipitations, SDS/PAGE and Western blot analysis were performed as described previously  with mouse anti-Myc (Santa Cruz Biotechnology), mouse anti-HA (Santa Cruz Biotechnology), mouse anti-α-tubulin (Sigma–Aldrich), rat anti-DE-cadherin, mouse anti-Armadillo, rat anti-α-catenin (DSHB) or anti-GST (Novagen) antibodies. Expression and purification of bacterially expressed GST–Armadillo proteins were performed as described previously . GST–MbtT525A and GST–MbtS492N/S521E were purified from transiently transfected HEK-293 cells after lysis in lysis buffer [25 mM Tris/HCl, 150 mM NaCl, 2 mM EDTA, 2 mM EGTA, 10% glycerol, 0.1% Nonidet P40 plus protease inhibitor cocktail tablets (pH 7.5)] with gluthatione–agarose beads (GE Healthcare). After washing with lysis buffer and PBS, bound proteins were eluted with elution buffer [10 mM glutathione, 150 mM NaCl and 50 mM Tris/HCl (pH 7.5)] and dialysed against 20 mM Tris/HCl (pH 7.5). For expression and purification of (DE-cadherin-Fc)2, nine 1.5 cm diameter dishes were seeded with HEK-293 cells in DMEM supplemented with IgG-depleted fetal bovine serum (Gibco). Each dish was transfected with 6 μg of DNA. After 48 h, the cell culture supernatant was collected, filled up to 50 ml with HBSS buffer (Hank's balanced salt solution; Roth) and stored at −80 °C until further use. Next, 30 μl slurry of beads (Dynabeads Protein A; Invitrogen) were incubated overnight with 5 ml of (DE-cadherin-Fc)2 in HBSS buffer. The beads were washed in HBSS (with 300 mM NaCl and 0.1% Nonidet P40) and taken up in 100 μl of HBSS buffer.
Kinase assays (radioactive and non-radioactive) were performed as described previously .
Laser tweezer experiment
The set-up for the laser tweezer experiment has been described previously . Beads were considered tightly bound when resisting laser displacement at 42 mW settings. A slurry of 5 μl of (DE-cadherin-Fc)2-coated microbeads in HBSS were allowed to settle on to the surface of cells in 200 μl of Schneider medium for 30 min. Always in a single experiment 100 beads were probed by the laser tweezer. For each cell line, five independent experiments were performed. Statistical analysis was performed with Statistica. To deplete Ca2+, EGTA was added to the culture medium for 20 min. Different EGTA concentrations were tested. Consistent results were obtained for EGTA concentrations equal or higher than 20 mM. After the experiment, EGTA-treated cells were cultured in normal Schneider medium for several days. We did not observe a negative effect on cell viability or obvious changes in cell morphology.
Identification of phosphorylation sites
Reduction and carbamidomethylation of the Armadillo protein in vitro phosphorylated by Mbt was performed in-gel, followed by trypsin digestion (trypsin gold from porcine pancreas, MS grade; Promega) and an additional chymotryptic digest (chymotrypsin from bovine pancreas, sequencing grade; Roche). Peptides were extracted from the gel with 1% aqueous trifluoroacetic acid.
The samples were analysed by nanoflow chromatography coupled to a hybrid triple quadrupole/linear ion trap mass spectrometer (4000 Q TRAP® LC/MS/MS system; Applied Biosystems) equipped with a Flow NanoSpray® source and a heated desolvation chamber interface set to 150 °C . The protein digests were separated on a Tempo™ one-dimensional nanoLC System (Applied Biosystems) using a PepMap C18 column (75 μm×15 cm; Dionex) and a 45 min linear gradient of 5–40% acetonitrile versus water with 0.1% formic acid as the modifier. Samples were concentrated and desalted on a PepMap C18 capillary trap (300 μm×5 mm; Dionex) using aqueous 2% acetonitrile/0.5% formic acid at a flow rate of 20 μl/min for 5 min prior to the analytical separation.
For standard LC/MS/MS (liquid chromatography tandem MS) analysis, a linear ion trap single MS scan over a mass range of m/z 400–1400 at 4000 amu (atomic mass unit)/s was used to detect eluting peptides. Up to five precursors per experimental cycle were examined by an enhanced resolution scan at 250 amu/s to establish charge state and molecular mass. Suitable precursors were then subjected to enhanced product ion scans at 4000 amu/s, with the collision energies automatically determined based on charge state and molecular mass.
For selective detection of phosphopeptides by LC/MS/MS, precursor ion scanning for precursors of m/z 79 (PO3−) was performed in negative-ion mode over a mass range of m/z 400–1400 at 333 amu/s, with the Q1 set to low resolution [2.0 amu FWHH (full width at half height)] and the Q3 set to unit resolution (0.6 amu FWHH). Precursors were collisionally activated in Q2 with nitrogen at a collision energy ramp of −65 to −110 V across the m/z range. Up to two precursors per experimental cycle were examined by an enhanced resolution scan at 250 amu/s in negative-ion mode. On detection of suitable precursors the polarity was switched to positive-ion mode, and up to two precursors subjected to enhanced product ion scans at 4000 amu/s. At the end of the cycle, polarity was switched back to negative-ion mode.
The acquired LC/MS/MS data were submitted for a database search against an NCBInr protein sequence database using the MASCOT® V2.0 search engine (Matrixscience) to identify (phospho-)peptides and proteins.
Expression of DE-cadherin and Mbt in Drosophila S2R cells
Establishment of the final retinal pattern of the Drosophila eye involves complex morphological rearrangements and cell-sorting processes, which depend on several functional interdependent cell-adhesion systems [23,24]. To address the question of whether Mbt specifically modulates DE-cadherin-mediated cell adhesion and to minimize possible secondary effects due to interactions with other cell-adhesion molecules, a Drosophila S2R cell-culture system was established. Schneider S2R cells are derived from S2 cells . They express α-catenin, whereas DE-cadherin and Armadillo are only present at low amounts (Figure 1A). Previous experiments have shown that expression of DE-cadherin in S2 cells resulted in stabilization of Armadillo at the plasma membrane . Similarly, expression of DE-cadherin under the actin promotor in S2R cells resulted in high Armadillo protein levels (Figure 1A) and accumulation of DE-cadherin, Armadillo and α-catenin at the plasma membrane mainly at cell contact sites (Figure 1B) indicating that the proteins can form a functional CCC. Identical localization patterns were observed after transient transfection of the DE-cadherin construct in S2R cells (Figure 1B) or upon generation of a stable DE-cadherin-expressing S2R cell line (S2R_DEcad; results not shown). Assembly of the CCC was also verified by co-immunoprecipitation experiments with an anti-DE-cadherin antibody. Armadillo and α-catenin were co-purified with DE-cadherin from S2R_DEcad cells (see Figure 5A).
In the next step, we combined expression of the CCC and variants of Mbt in S2R cells. Stable S2R cell lines expressing Myc epitope tagged wild-type Mbt (S2R_Myc–Mbt), a kinase dead version (S2R_Myc–MbtT525A) and a constitutively activated Mbt protein (S2R_Myc–MbtS492N/S521E) under the control of the inducible metallothionine promotor  were transiently transfected with the DE-cadherin construct to allow formation of the CCC at the plasma membrane, followed 12 h later by the addition of Cu2+ to induce Mbt expression. Western blot analysis confirmed expression of the corresponding proteins (Figure 2A). Immunocytochemical analysis revealed the presence of Myc–Mbt and Myc–MbtT525A in the cytoplasm and accumulation of the proteins at the plasma membrane, where they co-localize with DE-cadherin and Armadillo (Figure 2B). This corresponds to the analysis in flies, where Mbt co-localizes with Armadillo at adherens junctions of the developing photoreceptor cells . In the presence of the constitutively active Myc–MbtS492N/S521E protein, only in a fraction of cells, clear localization of Mbt, DE-cadherin and Armadillo at cell contacts could be detected (Figure 2B). Instead, we frequently observed accumulation of DE-cadherin and Armadillo in large vesicle-like structures within cells and enrichment of Myc–MbtS492N/S521E in the cytoplasm (Figure 2C). The identity of these vesicles remained obscure, because staining with markers for the endocytic and exocytic pathways revealed only partial overlap (results not shown).
Activation of Mbt induces destabilization of the DE-cadherin–Armadillo complex
To investigate whether Mbt influences the interaction of DE-cadherin with Armadillo, we used three different approaches. In the first assay, the CCC was purified from S2R_DEcad cell lysates with the anti-DE-cadherin antibody coupled to Protein G–agarose beads. The beads with bound proteins were equally distributed and subjected to non-radioactive in vitro kinase assays in the presence of GST–MbtT525A or GST–MbtS492N/S521E proteins in solution, which were purified from transiently transfected HEK-293 cells. After the kinase reaction, the bead fraction was directly analysed for DE-cadherin and associated Armadillo, whereas the complete supernatant fraction was used for immunoprecipitation with the anti-Armadillo antibody to quantitatively purify Armadillo protein released from DE-cadherin during the kinase reaction (Figure 3A). Compared with the control reaction (Figure 3A, lane 1), the presence of kinase inactive GST–MbtT525A protein had no influence on the amount of Armadillo bound to DE-cadherin (Figure 3A, lane 2). In contrast, incubation with the constitutively activated GST–MbtS492N/S521E protein released Armadillo from DE-cadherin into the supernatant (Figure 3A, lane 4). This effect is dependent on a phosphotransferase reaction, because the same experiment performed in the absence of ATP resulted in no obvious reduction in DE-cadherin-bound Armadillo (Figure 3A, lane 3). This experiment also argues against the possibility that MbtS492N/S521E simply competes out DE-cadherin binding to Armadillo.
To provide in vivo evidence for a role of activated Mbt on the DE-cadherin–Armadillo interaction, the stable S2R_Myc–MbtT525A and S2R_Myc–MbtS492N/S521E cell lines described above were first transfected with a DE-cadherin expression vector to allow formation of the CCC followed 12 h later by induction of Mbt expression. After immunoprecipitation with the anti-DE-cadherin antibody, the amount of DE-cadherin-bound Armadillo was analysed by Western blot. The presence of the Myc–MbtS492N/S521E protein resulted in a markedly reduced amount of co-immunoprecipitated Armadillo protein compared with cells expressing the Myc–MbtT525A protein (Figure 3B). One caveat of this experiment is that it was done with total cell lysates and this might not reflect the situation found at the plasma membrane. Therefore we repeated the experiment but isolated membrane fractions at various time points upon induction of Myc–MbtS492N/S521E or Myc–MbtT525A and analysed them for Armadillo, DE-cadherin and Mbt proteins. Although Myc–MbtT525A and Myc–MbtS492N/S521E show similar expression levels in total cell lysates (Figure 3B), significantly less Myc–MbtS492N/S521E protein is found in the membrane fraction (Figure 3C). The Myc–MbtT525A protein had no effect on the amount of membrane-associated DE-cadherin or Armadillo. In contrast, the amount of Armadillo and DE-cadherin in the membrane fraction was reduced when the Myc–MbtS492N/S521E protein was expressed (Figure 3C). In summary, the experiments support the idea that activation of Mbt disrupts the DE-cadherin–Armadillo complex.
Activation of Mbt reduces DE-cadherin-dependent adhesiveness of cells
In order to directly assess changes in the adhesive properties of cells upon activation of Mbt, we applied the laser tweezer technique . Paramagnetic microbeads were coated with a recombinant chimaeric protein comprising the extracellular domain of DE-cadherin fused to the hinge-Fc domain of human IgG1 [(DE-cadherin-Fc)2; Figure 4A]. In the case of vertebrate N-, VE- and E-cadherin it has been shown that the corresponding chimaeric proteins can assemble as cis dimers through intermolecular disulfide bridges and are able to form adhesive (trans) complexes with cadherins expressed at the cell surface [21,26,27]. For cellular expression of DE-cadherin alone or in combination with Mbt variants, we chose Schneider S2 cells because they are semi-adherent and have, in contrast with S2R cells, a lower tendency to self aggregate, which makes them more suitable for adhesion assays. To apply the laser tweezer technique it was necessary to identify single cells expressing the appropriate Mbt constructs, which turned out to be very inefficient by transient transfections. Therefore a stable S2 cell line was established, which constitutively expresses DE-cadherin (S2_DEcad). By using a double-selection strategy, this cell line was further modified to allow additional inducible expression of Myc-tagged MbtS492N/S521E (S2_DEcad_Myc–MbtS492N/S521E) or MbtT525A (S2_DEcad_Myc–MbtT525A). Expression of DE-cadherin and the Mbt variants in the different cell lines was verified by Western blot analysis (Figure 4B). DE-cadherin-Fc-coated beads were allowed to settle for 30 min on the cell surface before they were probed for displacement with a laser beam (Figure 4C). Beads were considered tightly bound when resisting laser displacement at 42 mW. In a single experiment, 100 beads were tested. On average 64 beads (±1.67; n=5) resisted laser displacement when S2_DEcad cells were used, whereas in the parental S2 cell line on average only 38 beads (±5.54; n=5) remain attached to the cell surface (Figure 4D). Specificity and Ca2+ dependency of binding of (DE-cadherin-Fc)2-coated beads to cellular DE-cadherin were tested by depletion of Ca2+ from the medium with different concentrations of EGTA. With 30 mM EGTA, the number of beads resisting displacement from S2_DEcad cells dropped to 33 (±5.45; n=5), which corresponds to the value determined with S2 cells. The effect of kinase dead Myc–MbtT525A and activated Myc–MbtS492N/S521E was tested 12 h after induction with Cu2+, when Mbt proteins accumulated at high levels (Figures 4B and 3C). Compared with S2_DEcad cells, the number of (DE-cadherin-Fc)2 coated beads that remained bound on S2_DEcad_Myc–MbtT525A cells was not significantly changed (64±5.10; n=5). In contrast, a significant reduction was observed upon expression of activated Mbt in S2_DEcad_Myc–MbtS492N/S521E cells. The number of beads that remained bound (45±4.39; n=5) was slightly higher than in the S2 and the S2_DEcad+EGTA control experiments indicating that activated Mbt strongly decreases, but not completely eliminates, DE-cadherin-dependent cell adhesion.
Activation of Mbt results in phosphorylation of Armadillo
The experiments described so far are compatible with a model that activated Mbt mediates its effect on the DE-cadherin–Armadillo interaction by phosphorylation of one of the components. Two different approaches were applied to verify this assumption. In the first assay, the CCC was immunopurified from S2R_DEcad cells with the anti-DE-cadherin antibody and subjected to in vitro kinase reactions with GST–MbtS492N/S521E or GST–MbtT525A proteins purified from transiently transfected HEK-293 cells. The presence of DE-cadherin, Armadillo and α-catenin in the immunoprecipitates was verified by Western blot analysis (Figure 5A). Only in the presence of the GST–MbtS492N/S521E protein, a phosphorylation signal became visible that corresponded in size to the Armadillo protein, whereas no 32P-incorporation into DE-cadherin and α-catenin was detected (Figure 5A). From this experiment we concluded that Armadillo is accessible to Mbt phosphorylation in its native conformation and in complex with DE-cadherin and α-catenin. To verify Armadillo phosphorylation by Mbt, in vitro kinase reactions were performed with bacterially expressed GST–Armadillo fusion proteins. In addition to a full-length Armadillo protein (GST–Myc–Arm), a N-terminal truncated Armadillo protein (GST–Myc–Arm [76–843]) was used. The activated Myc–MbtS492N/S521E protein, but not the kinase dead Myc–MbtT525A protein, was able to phosphorylate both Armadillo proteins to a similar degree (Figure 5B). This indicated that Mbt phosphorylated Armadillo at sites distinct from the four N-terminal phosphorylation sites for Zw3 and CK1, which are important to target cytoplasmic Armadillo for degradation [28–30].
Identification and characterization of phosphorylation sites in Armadillo
A standard LC/MS/MS analysis of a tryptic/chymotryptic digest of Armadillo (approx. 250 fmol) resulted in a sequence coverage of 91% (Figure 6A). There was first indication for a phosphorylation at Ser561 (results not shown). Further phosphopeptides could not be detected. Therefore a precursor ion-scanning experiment was performed in negative-ion mode for m/z 79 Da (PO3−). This allowed selective detection of phosphopeptides at high sensitivity. Intact phosphorylated precursors were then selected for MS/MS fragmentation in the positive-ion mode. This experiment revealed the presence of six phosphopeptides (Figure 6B) corresponding to two phosphorylation sites. MS/MS fragment spectra and assignment of y- and b-type ions allowed the localization of the respective phosphorylated residues. One phosphorylation site could be assigned to Ser561 by MS/MS experiments from three phosphopeptides resulting from missed proteolytic cleavages (Figure 6C), and the other to Ser688 by MS/MS experiments of three phosphopeptides resulting from missed proteolytic cleaveages and the formation of pyroglutamate from glutamate (Figure 6D) respectively.
Both serine residues are conserved between Drosophila Armadillo and vertebrate β-catenins and are preceded by an arginine residue at the −2 position (Figure 7A). PAK proteins strongly favour serine as a phosphoacceptor and have a strong bias for peptides containing an arginine residue at the −2 position [31,32]. To determine whether Ser561 and Ser688 are the only sites for Mbt phosphorylation, a bacterially expressed Armadillo protein carrying the corresponding serine to alanine substitutions (GST–Myc-ArmS561A/S688A) was used for in vitro kinase assays and compared with the non-mutated Armadillo protein. As shown in Figure 7(B), the two substitutions are sufficient to reduce Myc–MbtS492N/S521E-mediated phosphorylation of Armadillo to background levels as seen with the kinase dead Myc–MbtT525A protein. Finally we tested whether mutation of Ser561 and Ser688 abolish the Mbt-induced disruption of the DE-cadherin–Armadillo interaction. HA-tagged ArmadilloS561A/S688A or HA–Armadillo as a control were co-expressed together with DE-cadherin and Mbt in S2R cells. Equal amounts of HA–Armadillo and HA–ArmadilloS561A/S688A were co-purified with DE-cadherin in Myc–MbtT525A-expressing cells (Figure 7C). This shows that both Armadillo proteins become integrated into the CCC and argues against the possibility that mutation of both serine residues has a general influence on binding to DE-cadherin. Similar to endogenous Armadillo protein (Figure 3B), binding of HA–Armadillo to DE-cadherin was significantly reduced in the presence of Myc–MbtS492N/S521E when compared with Myc–MbtT525A (Figure 7C). In contrast, HA–ArmadilloS561A/S688A resisted Myc–MbtS492N/S521E-induced release from DE-cadherin. No difference in binding of HA–ArmadilloS561A/S688A to DE-cadherin was observed in cells expressing Myc–MbtS492N/S521E or Myc–MbtT525A. We also tested the corresponding single mutants HA–ArmadilloS561A and HA–ArmadilloS688A in the same assay to evaluate the importance of each phosphorylation site. Unlike the HA–Armadillo and HA–ArmadilloS561A/S688A proteins, which behaved consistent in independent experiments, both single mutant Armadillo proteins produced variable results upon co-expression with Myc–MbtS492N/S521E in several independent experiments (results not shown). Sometimes, we observed a reduction in binding, sometimes the proteins largely resisted Myc–MbtS492N/S521E induced release from DE-cadherin. Therefore we were unable to determine the relative contribution of each serine residue for the Armadillo–DE-cadherin interaction. In summary, the results provide evidence that destabilization of the DE-cadherin–Armadillo complex can be induced by phosphorylation of two conserved serine residues in Armadillo.
Recent studies have implicated single group 2 PAK proteins in regulation of the adhesive properties of cells under normal and disease conditions [13,14,33]. The human PAK4 transcription unit localizes to a genomic region on chromosome 19 amplified in several tumours, and increased PAK4 activity can be detected in cell lines derived from many different cancers [14,33]. One of the key features during the transition from a benign to a metastatic behaviour of cells is the breakdown of adhesive bonds to allow migration to, and invasion of, foreign tissues. It was shown that changes in cell morphology and adhesion play an important role in the ability of an activated version of PAK4 to transform cells [13,14]. Xenopus laevis X-Pak5 is expressed in regions of the embryo that undergo extensive cell movements during gastrulation. X-Pak5 localizes to cell–cell contact sites and regulates cell adhesion in a calcium-dependent manner . Similarly, Drosophila Mbt localizes at adherens junctions and is required for morphogenetic processes during eye development [16,17]. Although these studies focused on the effects of PAK activation on cell morphology and in the context of developmental processes, the underlying molecular mechanisms remained elusive. We established a cell-culture system, which allowed on one hand the assembly and purification of a functional CCC upon expression of DE-cadherin and on the other hand inducible expression of variants of Mbt. We could show that in the presence of activated Mbt, Armadillo becomes phosphorylated as a free substrate but also as part of the CCC. The results of the present study are in agreement with previous studies postulating the existence of an adherens-junction-localized kinase, which phosphorylates Armadillo  and support the idea that Mbt can fulfil such a function. We further provided evidence that it is not merely the physical presence of Mbt that weakens the DE-cadherin–Armadillo interaction and consequently DE-cadherin-mediated cell adhesion, but the kinase activity of Mbt is essential to induce this effect. In this way, Mbt would differ from proteins such as IQGAP1 (IQ motif containing GTPase-activating protein 1), which binds to E-cadherin and β-catenin and thereby causes disruption of the CCC . One important question is whether Mbt directly binds to Armadillo. Although we could co-immunprecipitate Armadillo and Mbt from S2R cells and also upon heterologous co-expression in human HEK-293 cells, we were unable so far to reconstitute the Mbt–Armadillo interaction in cell-free systems (N. Menzel and H. Wecklein, unpublished work).
Previously we have shown that Mbt, like other PAK proteins, binds RhoGTPases in the active, GTP-bound conformation. The RhoGTPase-binding domain of Mbt is essential for recruitment to adherens junctions . Because cadherin-mediated cell adhesion influences the activity of the RhoGTPases [3,11,12], one can envisage a model in which activation of RhoGTPases recruits Mbt to the CCC where Mbt can phosphorylate Armadillo. Studies in vertebrates indicate that a primary regulatory point for cell adhesion is at the level of cadherin distribution at the cell membrane and in intracellular compartments [35–37]. Our observations that activated Mbt leads to accumulation of DE-cadherin and Armadillo in vesicle-like structures (Figure 2C), reduces the amount of membrane-bound DE-cadherin and Armadillo (Figure 3C) and decreases cell adhesiveness (Figure 4D) support the view that Mbt interferes with DE-cadherin dynamics by disruption of the DE-cadherin–Armadillo interaction. Internalization of ectopically expressed mouse E-cadherin in Drosophila cl-8 cells and redistribution of membrane-bound Armadillo to the cytoplasm has also been observed upon activation of the Wingless signalling pathway .
How can phosphorylation of Armadillo weaken the association with DE-cadherin? The Armadillo repeat domain contains twelve copies of an imperfect sequence repeat (Arm repeats), each of which consists of three helices. Structural analysis of the murine E-cadherin–β-catenin complex revealed that the Arm repeats form a superhelix, which provides in its entire length the interaction surface for the cytoplasmic domain of E-cadherin . It has been suggested that this elongated interface is suited for integration of multiple signal inputs that can modulate the overall affinity of the interaction. In this way, the mechanical properties and dynamics of adherens junctions could be modulated during morphogenetic processes without necessarily destroying them completely. In support of this idea, CK2- and GSK-3β-mediated phosphorylation of E-cadherin increases the β-catenin–E-cadherin interaction . On the other hand, Src kinase induced phosphorylation of mouse β-catenin at Tyr654, which is located in helix 3 of Arm repeat 12, weakens the β-catenin–E-cadherin association . Also in our experiments, phosphorylation of Armadillo by Mbt does not cause a complete elimination of the Armadillo–DE-cadherin interaction. One of the identified phosphorylation sites (Ser561) lies within Arm repeat 10 of Armadillo, whereas Ser688 localizes in an α helix next to Arm repeat 12. Both serine residues are conserved in vertebrate β-catenins (Ser552 and Ser675). Importantly, phosphorylation of Ser552 by the protein kinase Akt also leads to dissociation of β-catenin from cell contacts and translocation to the nucleus. Mutation of Ser552 to alanine was sufficient to largely inhibit dissociation of β-catenin from the membrane . We did not observe enrichment of Armadillo in the nucleus in the presence of activated Mbt. In the case of Armadillo, mutation of Ser561 and Ser688 to alanine was necessary to efficiently block Mbt-induced release of Armadillo from DE-cadherin (Figure 7C), whereas experiments with the corresponding single mutants gave equivocal results (results not shown). This indicates that phosphorylation of either serine residue alone only slightly interferes with the DE-cadherin–Armadillo interaction and again emphasizes the fine-tuning of the whole system. The crystal structure of β-catenin does not immediately suggest a role of these serine residues in the β-catenin–E-cadherin interaction [38,42]. However, the structural data were obtained with unmodified β-catenin and phosphorylation could lead to conformational changes that influence the binding properties of β-catenin–Armadillo. Ser675 and to some extent Ser552 were also identified as target sites for PKA, which leads to enhanced transcriptional activity of β-catenin [43,44]. It appears that β-catenin–Armadillo possess common phosphorylation sites for different protein kinases, which allows integration of various signalling pathways.
Although these studies provide a conceptual framework, many details of this intricate network remain elusive. In particular, the requirement of a specific phosphorylation site for regulation of diverse cadherin-dependent cellular processes such as cell migration, cell-shape changes and neurite extension must be verified under physiological conditions. This can be exemplified in the Drosophila embryo where Armadillo is the major tyrosine phosphorylated protein in the CCC . In accordance with studies in vertebrates, tyrosine hyperphosphorylation causes loss of epithelial integrity. However, no significant dissociation of Armadillo from DE-cadherin was observed and it has therefore been suggested that tyrosine hyperphosphorylation weakens the link of the CCC to the actin cytoskeleton . Another example is seen during Drosophila oogenesis, where a specialized group of somatic follicle cells (border cells) delaminate from the follicular epithelium, invade the nurse cell cluster and migrate to the oocyte in a DE-cadherin-dependent manner. However, mutation of Tyr667 in Armadillo, which corresponds to Tyr654 of β-catenin and should therefore abrogate regulation of the DE-cadherin–Armadillo interaction, had no effect on border cell migration . In the case of regulation of DE-cadherin activity by Mbt during eye development, one must take into account expression of N-cadherin in some, but not all, cells of the developing eye [47,48] and the cross-talk between different adhesion systems and signalling molecules expressed in different subsets of cells [23,49,50].
In summary, Mbt adds to an increasing list of kinases that can regulate the function and integrity of the CCC. Armadillo/β-catenin can serve as convergence points for many kinase-mediated signals. This allows combinatorial regulation and the integration of multiple signals depending on the cellular context. Therefore we consider Mbt as only one component that mediates fine-tuning of DE-cadherin-mediated cell adhesion in morphogenetic processes.
We thank Dr Ora Bernard, Dr Pernille Rorth and Dr Pascal Schneider for sending DNA constructs. The monoclonal antibodies against DE-cadherin (developed by T. Uemura), Armadillo (developed by E. Wischaus) and α-catenin (developed by M. Takeichi) were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD (National Institute of Child Health and Human development) and maintained by The University of Iowa, Iowa City, IA, U.S.A. The work was supported by grants from the Deutsche Forschungsgemeinschaft (SFB487) to D. D. and T. R.
Abbreviations: amu, atomic mass unit; CCC, cadherin–catenin complex; Cdc42, cell division cycle 42; CK1(2), casein kinase 1(2); DSHB, Developmental Studies Hybridoma Bank; DMEM, Dulbecco's modified Eagle's medium; FWHH, full width at half height; GSK3β, glycogen synthase kinase-3β; GST, glutathione transferase; HA, haemagglutinin; HBSS, Hank's balanced salt solution; HEK, human embryonic kidney; LC/MS/MS, liquid chromatography tandem MS; PAK, p21-activated kinase; PKA, protein kinase A; Zw3, Zeste-White 3
- © The Authors Journal compilation © 2008 Biochemical Society