A novel non-canonical mechanism of regulation of MST3 (mammalian Sterile20-related kinase 3)

The human genome encodes 28 mammalian Ste20 (Sterile20)- related protein kinases (so-called because of their homology with Saccharomyces cerevisiae Ste20p or Sps1p) comprising 22 GCKs (germinal centre kinases) and six PAKs (p21-activated kinases) [1,2]. GCKs possess an N-terminal ‘catalytic’ domain of approximately 275 residues and a C-terminal ‘regulatory’ domain, whereas the domain order is reversed in PAKs. GCKs are subdivided into eight subgroups (GCKI–VIII) on the basis of sequence motifs in the regulatory domains and the identity of a residue (cysteine, phenylalanine or tyrosine) in the P+1 loop of subdomain VIII. In addition, there are two STRAD (Ste20-related adapter) ‘pseudokinases’ which are most closely related to the GCKVI subgroup [3,4]. The GCKIII subgroup [5] is the simplest in terms of the complexity of the regulatory domain and comprises MST (mammalian Ste20-related protein kinase) 3 (also known as Stk24) [6], SOK1 (Ste20/oxidant stress response kinase 1)/YSK1 (yeast Sps1/Ste20-related kinase 1; also known as Stk25) [7,8], and MST4 [also known as MASK (MST3 and SOK1-related kinase)] [9,10]. MST3 transcripts are most abundant in the heart, placenta, skeletal muscle and pancreas [6], SOK1/YSK1 transcripts are most abundant in the testis (and possibly the brain) [7,8], and MST4 transcripts are most abundant in placenta [9,10], although all of the tested tissues probably express each GCKIII to a degree. Interest in GCKIII members is likely to increase because of their putative roles in pathologies such as cerebral cavernous malformations (see [11] for a review of this disease) and those associated with cardiac development [12].

The 431 residue MST3 ‘short’ isoform (also called MST3a by Irwin et al. [13] to avoid confusion with the longer MST3b isoform [14]) was first cloned from a HeLa cell library [6]. For simplicity, we refer to this isoform as ‘MST3’. Subsequently, brain-restricted MST3b (443 amino acid residues) was identified [14]. As shown in Supplementary Figure S1 (at http://www.BiochemJ.org/bj/442/bj4420595add.htm), there is considerable confusion about MST3 nomenclature in the literature with the nomenclature of the 431- and 443-residue isoforms frequently being reversed (see [15] for a recent example). MST3b may be identical with a protein kinase (confusingly known as protein kinase N) first identified in 1987 [16–18]. The two MST3 isoforms arise from translation of alternative 5′ coding exons: exon 1 (42 nt) encodes the 14 N-terminal residues in MST3, whereas exon 2 (78 nt) encodes the 26 N-terminal residues in MST3b. Otherwise, the sequences of MST3 and MST3b are identical. The 431 residue MST3 is activated by autophosphorylation of the activation loop Thr¹⁷⁸ [19]. MST3b is additionally phosphorylated at Thr¹⁸ (absent from MST3) by PKA (cAMP-dependent protein kinase) [14], but the role of this phosphorylation in the regulation of MST3b activity is ambiguous. The only established MST3 ‘activator’ [6] is the PP (protein serine/threonine phosphatase) 1/PP2A inhibitor calyculin A [20], which presumably simply inhibits the dephosphorylation step of a constitutively active phosphorylation/dephosphorylation cycle. In contrast, ‘physiological regulators’ of MST3b have been identified. In PC12 cells, MST3b is activated by nerve growth factor or inosine and is inhibited by 6-thioguanine [13]. In addition to its possible involvement in systemic pathologies [12], MST3 may be involved in the regulation of cell motility [19], in the regulation of other protein kinases [14,21–23] and caspase-dependent [24] or -independent apoptosis [24,25]. The biological functions of MST3b may include stimulation of axon outgrowth [13,26], and its activation by inosine is potentially of therapeutic benefit in restoring limb usage following cerebral focal ischaemia (‘stroke’) [27].

Only one pathway of MST3 activation (the ‘canonical’ pathway) has thus far been characterized in detail and this is involved in Jurkat cell apoptosis [24]. Anti-Fas antibodies or staurosporine induce caspase-dependent cleavage of MST3 at Asp³¹³ at the boundary between the catalytic and regulatory domains [24]. This reveals a bipartite NLS (nuclear localization signal) in the C-terminus of the catalytic domain, as well as possibly removing a nuclear export signal in the regulatory domain [28]. Cleaved MST3 translocates into the nucleus and is at least partly responsible for subsequent apoptosis [24,28], although whether the cleaved MST3 is autophosphorylated and activated seems unclear. In contrast, on the basis of incomplete inhibition of staurosporine-induced apoptosis by the caspase inhibitor Z-DEVD-FMK (benzyloxycarbonyl-DL-Asp-Glu-Val-DL-Asp-fluoromethylketone), a caspase-independent involvement of MST3 in HeLa cell apoptosis has been identified [25]. This involves loss of mitochondrial membrane potential, and the subsequent nuclear import of apoptosis-inducing factor and endonuclease G released from the mitochondria [25]. Although the canonical pathway of MST3 regulation is widely accepted, in the present paper we describe a second non-canonical pathway of MST3 activation in cardiac myocytes which does not involve caspase-dependent cleavage.

Antibodies were used at a 1:1000 dilution except when stated otherwise. Mouse mAbs (monoclonal antibodies) against human MST3 (residues 275–393 as immunogen) were from BD Transduction Laboratories (catalogue number 611056). A variety of FLAG/OctA antibodies were used: rabbit polyclonal antibodies were from Sigma–Aldrich (catalogue number F7425) or Santa Cruz Biotechnology (catalogue number sc-807, lot K1607), and mouse mAbs were from Santa Cruz Biotechnology (catalogue number sc-166355). Anti-HA (haemagglutinin) antibodies (rabbit polyclonal catalogue number sc-805, mouse mAb catalogue number sc-7392) were from Santa Cruz Biotechnology. Mouse monoclonal EZview™ red anti-FLAG M2 affinity gel was from Sigma–Aldrich (catalogue number F2426). Rabbit polyclonal antibodies against GST (glutathione transferase) were from Abcam (catalogue number ab9085-200). Rabbit polyclonal antibodies against GOLGA2/gm130 (golgin A2/Golgi matrix protein 130; catalogue number 11308-1-AP) and CCM3/PDCD10 (cerebral cavernous malformation 3/programmed cell death 10; catalogue number 10294-2-AP) were from Proteintech Group. Rabbit mAbs against MO25α (catalogue number 2716) and antibodies against cleaved (activated) caspase 3(Asp¹⁷⁵) (catalogue number 9661) were from Cell Signaling Technology. Sheep polyclonal antibodies against MO25β were gifts from Professor Dario Alessi and Dr Paola de los Heros (MRC Protein Phosphorylation Unit, University of Dundee, Dundee, U.K.) and were used at a 1:400 dilution.

Horseradish peroxidase-conjugated anti-mouse (catalogue number P0260) and anti-rabbit (catalogue number P0448) secondary antibodies were from Dako, and Pierce horseradish peroxidase-conjugated anti-sheep antibodies were from ThermoFisher (catalogue number 31627). The Alexa Fluor® 488-conjugated anti-mouse secondary antibody (catalogue number A11001) and Texas Red–phalloidin (catalogue number T7471) were from Invitrogen.

The catalytic domain of human recombinant PP2A was from Cayman Chemicals (catalogue number CAY10011237) and was obtained via Cambridge Bioscience. MBP (myelin basic protein) was from Millipore (catalogue number 13-104).

Rats were housed and work was under taken in accordance with the local institutional animal care committee and the U.K. Animals (Scientific Procedures) Act 1986. Primary cultures of ventricular cardiac myocytes were prepared from 1–2-day-old Sprague–Dawley rats by collagenase and pancreatin digestion of ventricles using an adaptation of the method of Iwaki et al. [29] as described recently in detail [30]. Except for immunofluorescent staining experiments, cardiac myocytes were plated on gelatin-precoated Primaria™ surface-modified polystyrene culture dishes [BD Biosciences, catalogue numbers 353801 (35 mm) and 353802 (60 mm)] at densities of 2×10⁶ myocytes per 35 mm dish or 4×10⁶ myocytes per 60 mm dish. After 18 h, myocytes were confluent and beating spontaneously. Unless they were to be infected with AdV (adenoviral) vectors, serum was then withdrawn and myocytes were incubated in maintenance medium {Dulbecco's modified Eagle's medium/medium 199 [4:1 (v/v)] and penicillin and streptomycin (100 units/ml for each)} for a further 24 h. For AdV infection, serum was withdrawn 6 h beforehand and the infected myocytes were used within 48 h.

The MST3 coding sequence was amplified from neonatal rat cardiac myocyte cDNA with Pfu polymerase using the forward primer 5′-GACGACAAGGGTACCATGGCCCACTCCCCGGTGCAG-3′ and the reverse primer 5′-CTTATCTAGAAGCTTTCAGTGCGCTGAGGCTCCTCCA-3′ to produce a MST3 product (containing the initiator methionine residue) flanked by the KpnI and HindIII sites (underlined). These restriction sites were then used for insertion into the corresponding sites in FLAG-pShuttle-CMV [31]. The product was sequenced and submitted to GenBank® (accession number EU371958.1). MST3 C-terminal deletion mutants were constructed similarly, but using the reverse primers 5′-CTTATCTAGAAGCTTTCAATCTGTCTCCACATCTGA-3′ [for MST3(Δ314–431)], 5′-CTTATCTAGAAGCTTTCATCCGTTCTCCAGATTCTTGG-3′ [for MST3(Δ341–431)] and 5′-CTTATCTAGAAGCTTTCATTTCAGCTCCGCAAACAGAG-3′ [for MST3(Δ377–431)]. Point mutations were introduced into MST3 by PCR using forward and reverse complementary primer pairs that contained the desired base change(s) flanked by 15 bases on either side. Two separate PCR amplifications were performed using the pShuttle-CMV forward primer (5′-GGTCTATATAAGCAGAGCTG-3′) in combination with the MST3 mutant reverse primer and the pShuttle-CMV reverse primer (5′-GTGGTATGGCTGATTATGATCAG-3′) in combination with the mutant MST3 forward primer, with wild-type MST3 as a template. Subsequently, a third PCR was performed using pShuttle-CMV forward and reverse primers and a mixture of the two purified mutant products from the first round of PCRs. The mutant PCR product generated was purified and cloned into the Kpn1 and HindIII sites of FLAG-pShuttle-CMV. Where multiple mutations were required, they were introduced together on the same primer pair if close together or, if further apart, were introduced sequentially. All constructs were sequenced to verify that the desired mutation(s) was present and that no further mutations had been introduced during the PCR procedure. AdVs were generated from the shuttle plasmids using the AdEasy™ XL Adenoviral Vector System (Stratagene), as described previously [30,31].

Following treatments, myocytes were washed twice with ice-cold PBS and scraped into Buffer A [20 mM 2-glycerophosphate (pH 7.5), 50 mM NaF, 5 mM dithiothreitol, 2 mM EDTA, 0.2 mM Na₃VO₄, 2 μM microcystin LR and 1% (v/v) Triton X-100] containing protease inhibitors {10 mM benzamidine, 300 μM PMSF, 200 μM leupeptin and 10 μM E-64 [trans-epoxysuccinyl-L-leucylamido-(4-guanidino)butane]} (100 μl per 35 mm dish and 150 μl per 60 mm dish). Samples were incubated on ice for 10 min and then centrifuged (10000 g for 5 min at 4°C).

Supernatants were denatured by heating at 100°C with 0.33 vol. of 4×SDS/PAGE sample buffer as described previously [30]. Non-specific binding sites were blocked with 5% (w/v) non-fat milk powder in TBST [Tris-buffered saline plus Tween 20; 20 mM Tris/HCl (pH 7.5), 137 mM NaCl and 0.1% Tween 20] for 30 min. Blots were incubated with primary antibodies in TBST containing 5% (w/v) BSA overnight at 4°C. The blots were washed three times with TBST [for 5 min at room temperature (21°C)], incubated with secondary antibodies [1:5000 dilution in TBST containing 1% (w/v) non-fat milk powder for 1 h at room temperature] and then washed again three times with TBST (for 5 min at room temperature). Bands were detected by ECL (enhanced chemiluminescence; Santa Cruz Biotechnology) using Hyperfilm MP and were quantified by scanning densitometry. Latterly, bands were detected with GE Healthcare ECL Plus and a GE ImageQuant 350 imager with one-dimensional gel analysis software. Unless divided by a vertical bar, all individual immunoblots shown are images of single gels, although the lane order has been altered by image manipulation in some cases to improve comprehensibility.

Myocytes were washed twice with ice-cold PBS and extracted into Buffer B [20 mM Tris/HCl (pH 7.5), 300 mM KCl, 5 mM MgCl₂, 5 mM NaF, 1 mM EDTA, 0.2 mM Na₃VO₄, 2 μM microcystin LR, 10% (v/v) glycerol, 1% (v/v) Triton X-100 and 0.05% 2-mercaptoethanol containing the protease inhibitors as for Buffer A]. When myocytes had been pre-exposed to 100 nM calyculin A, calyculin A (100 nM) was also included in the extraction buffer. Samples were incubated on ice for 10 min and then centrifuged (10000 g for 5 min at 4°C). The supernatants were retained and samples were taken for immunoblotting to assess the input of the proteins subsequently immunoprecipitated. For immunokinase assays, supernatants (27.5 μl) were diluted to 150 μl with Buffer B and added to 15 μl of EZview™ red anti-FLAG M2 affinity gel (1:1 slurry in Buffer B), followed by overnight incubation at 4°C with rotation. The pellets were collected by centrifugation (10000 g for 1 min at 4°C) and washed three times in Buffer B (0.7 ml), once with kinase assay buffer [20 mM Tris/HCl (pH 7.5), 1 mM dithiothreitol, 1 μM microcystin LR, 4 μM leupeptin and 1 mM benzamidine; 150 μl) and finally resuspended in 45 μl of kinase assay buffer containing 0.5 mg/ml MBP and 1 μM PKA inhibitor (TTYADFIASGRTGRRNAIHD; Bachem).

FLAG–MST3 immunokinase assays were initiated by the addition of 5 μl of 1 mM [γ-³²P]ATP (0.5 μCi/assay)/200 mM MgCl₂ followed by incubation at 30°C for 30 min. Reactions were terminated by the addition of 5 μl of 0.5 M EDTA (pH 7.5), followed by cooling on ice. Samples were centrifuged (10000 g for 1 min at 4°C) and duplicate samples (20 μl) of supernatants were spotted on to Whatman P81 paper squares, which were then washed four times in 75 mM H₃PO₄ (for 15 min) on a rocking platform. Radioactivity was measured by Čerenkov radiation counting. Blanks (one with kinase assay buffer and [γ-³²P]ATP/MgCl₂, and one involving the assay of MBP kinase activity with immunoprecipitates from uninfected myocytes) were routinely included.

FLAG–MST3 in samples taken from pooled extracts of myocytes which had or had not been exposed to 100 nM calyculin A for 10 min was prepared essentially as described in the Immunoprecipitations and immunokinase assays section. Immunoprecipitates were washed three times with Buffer C [50mM Tris/HCl (pH 7.0), 0.15 mM EGTA, 0.1% 2-mercaptoethanol and 0.1% BSA] and resuspended in 25 μl of Buffer C. GST–MST3 was similarly precipitated using glutathione–Sepharose. Dephosphorylation was then initiated by the addition of PP2A (25 μl, 10 units/μl in Buffer C) in the absence or presence of 2 μM okadaic acid. Incubations were for up to 30 min at 30°C, and PP2A activity was terminated by the addition of 0.1 vol. of 50 μM okadaic acid/2 μM calyculin A. Immunoprecipitates were washed with Buffer C, centrifuged (10000 g for 5 min at 4°C) and the pellets extracted into 50 μl of 2×sample buffer for subsequent immunoblotting.

Supernatants (prepared in Buffer B from 60 mm dishes of monocytes as described in the Immunoblotting section) of AdV–MST3-infected myocytes were applied to columns of Sephacryl S200HR (1.4 cm×85 cm, 131 ml inclusion volume measured using ³H₂O) or Sephacryl S300HR (1.6 cm×95 cm, 185 ml inclusion volume) equilibrated with 50 mM Tris/HCl (pH 7.5), 250 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol and 0.03% Brij-35 which had been standardized with proteins of known Stokes' radii. When extracts of calyculin A-treated myocytes were examined, the elution buffer additionally contained 50 nM calyculin A. Flow rates were 0.2–0.3 ml/min and fractions of 0.6–1.3 ml as required were collected. In some cases, fractions were centrifugally concentrated at 4°C by up to 10–15-fold in Amicon Ultra-2 10 kDa nominal molecular mass limit centrifugal filter devices (Millipore).

Myocytes (1.5×10⁶ cells per 35 mm Primaria culture dish) were cultured on to glass coverslips precoated with 1% (w/v) gelatin followed by laminin (20 μg/ml in PBS). Myocytes were cultured as described in the Cardiac myocyte isolation and culture section, and were infected with AdV–FLAG–MST3 for 16 h in serum-free medium. After treatment, myocytes were washed with ice-cold PBS (3×0.5 ml), fixed in 4% (v/v) formaldehyde (10 min at room temperature), permeabilized with 0.1% Triton X-100 in PBS (10 min at room temperature) and incubated with 1% (w/v) BSA in PBS containing 0.1% Triton X-100 (10 min at room temperature) to block non-specific antibody binding. FLAG–MST3 was detected by incubation of cells with Santa Cruz Biotechnology mouse anti-FLAG mAbs (1:1000 dilution for 1 h at 37°C), then incubated with goat anti-(mouse IgG) coupled to Alexa Fluor® 488 (1:200 dilution for 1 h at 37°C) followed by Texas Red®-X phalloidin (Invitrogen; 1:200 dilution for 1 h at 37°C). Nuclei were stained with Hoechst 33258 (Sigma; 10 μg/ml for 15 min at 37°C). Coverslips were washed with PBS (3×0.5 ml) between each incubation, mounted with fluorescence mounting medium (Dako), viewed under a Zeiss Axioskop fluorescence microscope with a ×40 objective, and photographed using a digital camera.

The disorder probability plot was produced using the DISOclust method (http://www.reading.ac.uk/bioinf/DISOclust/) for the prediction of intrinsically unstructured regions in proteins [32]. Predicted secondary structure was obtained using the PSIPRED server (http://bioinf.cs.ucl.ac.uk/psipred/) [33]. The three-dimensional model was produced using the IntFOLD-TS method (http://www.reading.ac.uk/bioinf/IntFOLD/) [34,35] and was colour coded from blue (high accuracy) through green, yellow and orange to red (low accuracy) using the ModFOLDclust2 model quality assessment method (http://www.reading.ac.uk/bioinf/ModFOLD/) [36]. The image was rendered using PyMol (http://www.pymol.org). Protein sequences were taken from the NCBI (National Center for Biotechnology Information) protein database (http://www.ncbi.nlm.nih.gov/protein/). Alignments were carried out using ClustalW2 (http://www.ebi.ac.uk/Tools/msa/clustalw2/) with manual curation if necessary.

We cloned and sequenced rat MST3 (GenBank® accession number EU371958.1, submitted 23 January 2008). The predicted 431 residue sequence (Supplementary Figure S1) has 98.6% and 96.8% identity with the (431 residue) MST3 of mouse (GenBank® accession number NP_663440.1) and Homo sapiens (GenBank® accession number NP_001027467.2; cloned by Schinkmann and Blenis [6]) respectively, with mostly conservative or semi-conservative substitutions elsewhere. The ATP-binding site [GXGXXGX_(14–23)K] in rat MST3 lies between residues 31 and 53, the activation loop serine/threonine is Thr¹⁷⁸, and the ‘caspase cleavage’ site is V³¹⁰ETD³¹³ (AETD in H. sapiens). The residue in the 314 position is glycine in H. sapiens and mouse, but is serine in rat. Thus, although Gly³¹⁴ has been identified as a possible N-myristoylation site that becomes available following caspase cleavage of MST3 [37,38], this cannot occur in the rat and N-myristoylation of MST3(314–431) may not be of universal importance.

MST3(275–393) mAbs detected MST3 as a single band of ~52 kDa in cardiac myocyte extracts, and exposure of myocytes to calyculin A (100–200 nM) resulted in its rapid (complete in <10 min) loss (Figure 1A). To investigate this further, N-terminally FLAG-tagged rat full-length (431 residue) MST3 [FLAG–MST3(FL)] was expressed in cardiac myocytes using an AdV vector. In unstimulated myocytes, FLAG–MST3(FL) was detected as a band of ~54 kDa with anti-MST3(275–393) mAbs which predominated over endogenous MST3 (Figure 1B, upper panel). Exposure of myocytes to 100 nM calyculin A resulted in the loss of both endogenous MST3 and FLAG–MST3(FL) (Figure 1B, upper panel). Immunoblotting with anti-FLAG antibodies revealed that FLAG–MST3(FL) was not lost following calyculin A exposure, but instead migrated at ~58 kDa (Figure 1B, lower panel). Loss of the 54 kDa band and appearance of 58 kDa FLAG–MST3(FL) were essentially complete by 5–10 min (Figure 1B, lower panel), and there were no increases in bands of <58 kDa (Supplementary Figure S2 at http://www.BiochemJ.org/bj/442/bj4420595add.htm) indicating absence of detectable cleavage. The broad-spectrum serine/threonine protein kinase inhibitor staurosporine reduced the rate of the appearance of 58 kDa FLAG–MST3(FL) in calyculin A-treated myocytes (Figure 1C), suggesting that the band shift represents serine/threonine phosphorylation. Additional evidence that the 58 kDa band resulted from phosphorylation was obtained using phos-tag™ SDS/PAGE immunoblots [39]. Anti-FLAG antibodies detected multiple FLAG–MST3(FL) bands even in unstimulated myocytes (Figure 1D, left-hand panel). Following exposure of myocytes to calyculin A, the overall mobility of the bands was reduced, indicating that FLAG–MST3(FL) was increasingly phosphorylated. A ‘control’ SDS/PAGE immunoblot from the same experiment is shown for comparison (Figure 1D, right-hand panel).

We confirmed that serine/threonine phosphorylation of FLAG–MST3(FL) caused the shift from 54 kDa to 58 kDa (Figure 1B, lower panel) by treating FLAG–MST3(FL) immunoprecipitates from calyculin A-treated myocytes with recombinant PP2A. This resulted in the disappearance of the 58 kDa FLAG–MST3 band with the reappearance of the 54 kDa band (Figure 1E). This was prevented by inclusion of the PP1/PP2A inhibitor okadaic acid in the PP2A incubation (Figure 1F), indicating that proteolysis, for example, was not responsible. Furthermore, in glutathione–Sepharose precipitates of myocytes expressing GST–MST3(FL) which had been exposed to calyculin A, the band detected by anti-MST3(275–393) mAbs reappeared with PP2A treatment detected using anti-GST antibodies (Figure 1G, upper panel) with simultaneous reappearance of the more rapidly migrating dephosphorylated band (Figure 1G, lower panel). Analogous experiments using phos-tag™ SDS/PAGE gels and anti-FLAG antibodies confirmed that FLAG–MST3(FL) reverted to the more rapidly migrating forms and this was prevented by okadaic acid (Figure 1H). However, even in PP2A-treated immunoprecipitates, multiple FLAG–MST3 bands remained clearly visible in phos-tag™ SDS/PAGE gels (Figure 1H). We interpret this as indicating that either residues other than serine/threonine (e.g. tyrosine) are phosphorylated or that MST3 contains phosphorylated serine/threonine residues that are resistant to PP2A. There are at least two precedents for the latter interpretation [40,41].

The only other agent that resulted in the appearance of the 58 kDa FLAG–MST3 in cardiac myocytes was okadaic acid (results not shown). A very recent paper [15] has also shown that okadaic acid reduced the mobility of endogenous MST3 in HEK (human embryonic kidney)-293 cells [detected using mAbs to an epitope in the C-terminus of MST3(FL)], which was reversed by incubation with exogenous purified PP2A. The list of agonists/stimuli that do not activate (i.e. reduce the mobility to 58 kDa, as explained below) MST3 includes G_q-protein-coupled receptor agonists (e.g. endothelin-1 and α-adrenergic agonists), G_s-protein-coupled receptor agonists (e.g. isoprenaline), peptide growth factors (e.g. epidermal growth factor and insulin) and stress stimuli such as tunicamycin (to induce endoplasmic reticulum stress), hyperosmotic shock and oxidative stress (results not shown).

In summary, we conclude that exposure of myocytes to calyculin A induces serine/threonine phosphorylation of MST3(FL). Because anti-MST3(275–393) mAbs fail to recognize FLAG–MST3(FL) following calyculin A exposure (Figure 1B, upper panel), the phosphorylation site(s) probably lies between residues 275 and 393 and is distinct from the previously identified phosphorylation of MST3 on Thr¹⁷⁸ of its activation loop [19].

To identify the potential phosphorylation site(s) additional to MST3(Thr¹⁷⁸) deletion mutants were generated and expressed in myocytes: FLAG–MST3(Δ314–431) with deletion of Ser³¹⁴–His⁴³¹, FLAG–MST3(Δ341–431) with deletion of Thr³⁴¹–His⁴³¹ and FLAG–MST3(Δ377–431) with deletion of Glu³⁷⁷–His⁴³¹. Incubation of myocytes with calyculin A reduced the mobility of FLAG–MST3(FL) or FLAG–MST3(Δ377–431) on SDS/PAGE, but not that of FLAG–MST3(Δ314–431) or MST3(Δ341–431) (Figure 2A). Thus the site(s) of phosphorylation responsible for reduced migration probably lies between Ser³¹⁴ and Lys³⁷⁶. In addition, although the anti-MST3(275–393) mAbs recognized FLAG–MST3(FL) and MST3(Δ341–431), they did not recognize FLAG–MST3(Δ314–431) (Figure 2B). This suggests that the epitope recognized by the MST3(275–393) mAbs probably lies between Ser³¹⁴ and Gly³⁴⁰.

We used alanine scanning to identify the phosphorylation site on MST3, focusing first on the Thr³⁴¹–Lys³⁷⁶ region because of the differing behaviour of FLAG–MST3(Δ314–431) and FLAG–MST3(Δ341–431) to FLAG–MST3(Δ377–431) following calyculin A exposure (Figure 2A). In this regard, a proteomics study had shown that MST3(Thr³⁴¹) was phosphorylated in a lymphoma cell line [42]. All six serine/threonine residues in the Thr³⁴¹–Lys³⁷⁶ region of FLAG–MST3(FL) were mutated to alanine either individually or all together, and the proteins expressed in myocytes. On SDS/PAGE, the mobilities of all seven mutants were still reduced to 58 kDa by calyculin A (Supplementary Figure S3A at http://www.BiochemJ.org/bj/442/bj4420595add.htm). Thus, although the MST3(341–376) region is required for the phosphorylation(s) resulting in reduced mobility on SDS/PAGE (Figure 2A), it does not contain the phosphorylation site(s) and presumably represents a docking domain. We next examined the Ser³¹⁴–Gly³⁴⁰ region. We had separately examined FLAG–MST3(S314G) earlier because of the difference at residue 314 between rat (serine) and H. sapiens or mouse (glycine; Supplementary Figure S1), and found that its migration was reduced to the same extent by exposure to calyculin A as that of FLAG–MST3(FL) (results not shown). The appearance of 58 kDa FLAG–MST3 when Ser³¹⁷, Ser³²⁰, Ser³²² and Thr³²⁸ were all mutated to alanine [FLAG–MST3(4A)] was not reduced to the same extent as FLAG–MST(FL), although a slight reduction was apparent (Figure 2C, left-hand panel). Furthermore, anti–MST3(275–393) mAbs failed to recognize FLAG–MST3(4A) with or without exposure of myocytes to calyculin A (Figure 2C, right-hand panel). With individual single alanine substitutions, although FLAG–MST3(T328A) remained detectable using FLAG antibodies, this mutation alone caused the loss of the anti–MST3(275–393) mAb epitope even in the absence of exposure to calyculin A (Figure 2D), and the finding that the mobility of FLAG–MST3(T328A) was not reduced to the same extent as FLAG–MST3(FL) by calyculin A indicated that phosphorylation of FLAG–MST3(Thr³²⁸) was responsible for the major reduction in FLAG–MST3(FL) mobility (Figure 2E). The mobility of FLAG–MST3(T328A) was still slightly reduced in myocytes exposed to calyculin A, presumably resulting from phosphorylation of another residue [potentially FLAG-MST3(Thr¹⁷⁸)]. It is noteworthy that endogenous MST3 was still detectable by anti-MST3(275–393) mAbs in FLAG–MST3(T328A)-infected myocytes that had not been exposed to calyculin A (fourth lane from the right-hand side of Figure 2D), showing that the failure of these mAbs to recognize FLAG–MST3(T328A) did not result from technical problems.

PhosphSitePlus® (http://www.phosphosite.org) lists a further four potentially phosphorylatable serine residues in the MST3(275–393) region (A²⁹⁶EQpSHDDpSpSpSEDSDAETDGQASGGSDSGDWIFTIR³³⁰; where pS represents potentially phosphorylatable serine residues) identified by proteomics approaches [43,44]. (Readers should rely on sequence information rather than numerical annotation because of the confusion in MST3(a)/MST3b terminology.) However, serine-to-alanine mutation of each of these serine residues (Ser²⁹⁹, Ser³⁰³, Ser³⁰⁴ and Ser³⁰⁵) and Ser³⁰⁸ in addition, or mutation of all residues in FLAG–MST3 produced proteins that were indistinguishable from FLAG—MST3 (FL) in response to calyculin A (Supplementary Figure S3B). Our overall conclusions from the experiments shown in Figure 2 are as follows. First, that the MST3(341–376) region is required for the reduced migration of FLAG–MST3(FL) resulting from its phosphorylation, but that this region does not contain the phosphorylation site responsible for the reduced migration. Secondly, cis-autophosphorylation of MST3(Thr³²⁸) is responsible for the greater proportion of the calyculin A-induced reduction in FLAG–MST3(FL) migration. Thirdly, the identity and the phosphorylation of MST3(Thr³²⁸) is crucial for the recognition of MST3 by anti-MST3(275–393) mAbs.

We next examined the relative mobilities of FLAG–MST3(FL) containing mutations in or close to the activation loop by SDS/PAGE. Mutation of the activation loop Thr¹⁷⁸ phosphorylation site [FLAG–MST3(T178A)] prevented the appearance of any 58 kDa FLAG–MST3 band following exposure of myocytes to calyculin A (Figure 3A, left-hand panel) and the anti-MST3(275–393) mAb epitope was also preserved (Figure 3A, right panel). Phos-tag™/SDS gels (Figure 3B) confirmed the result shown in the left-hand panel of Figure 3(A). Thus phosphorylation of MST3(Thr³²⁸) requires MST3 kinase activity and the data suggest that this autophosphorylation occurs in cis rather than in trans (since endogenous MST3 would have been activated by calyculin A in these experiments). GCKs reportedly possess a ‘secondary’ phosphorylation site lying in the conserved GTP(C,F,Y)WMAPE sequence in the P+1 loop (Thr¹⁸² in MST3), phosphorylation of which may be required for activation [2]. Phosphorylation of this residue has been detected experimentally by proteomics approaches [44,45], but care must be taken in annotation because of sequence similarities between GCKIII members in this region. Compared with FLAG–MST3(FL) (Figure 3A), there was only a partial shift of FLAG–MST3(T182A) to a lower mobility species in myocytes exposed to calyculin A (Figure 3C), possibly indicating a requirement of MST3(Thr¹⁸²) phosphorylation for MST3 activation as assessed by MST3(Thr³²⁸) cis-autophosphorylation. However, the analogous threonine residue in the catalytic subunit of PKA is important in the kinase–protein substrate interaction [46], and this is also true for other protein serine/threonine kinases. Thus it is equally likely that the FLAG–MST3(T182A) mutation reduces the affinity of MST3 for its substrates, in this case, FLAG–MST3(Thr³²⁸). Gordon et al. [15] recently obtained similar results to those of the present study using FLAG–MST(T178A) and FLAG–MST3(T182A). They also studied the FLAG–MST3(T172A) mutation since MST3(phosphoThr¹⁷²) has also been detected by proteomics approaches [44,47,48]. The mobility of FLAG–MST3(T172A) was still reduced by okadaic acid [15], and the conclusion was that phosphorylation of this site is unimportant in the appearance of 58 kDa FLAG–MST3(FL) [15].

The phosphomimetic mutation, FLAG–MST3(T178D), resulted in the constitutive appearance of 58 kDa FLAG–MST3(T178D), suggesting stimulation of cis-autokinase activity, though 54 kDa FLAG–MST3(T178D) was never completely lost (Figure 3C). Calyculin A did not cause a complete shift of FLAG–MST3(T178D) to the 58 kDa band (Figure 3C), although there may have been an increase in the relative abundance of 58 kDa FLAG–MST3(T178D) species following exposure of myocytes to calyculin A compared with untreated myocytes (Figure 3C), which presumably results from increased preservation of FLAG–MST3(Thr³²⁸) phosphorylation. Mutation of the essential Lys⁵³ in the MST3 ATP-binding site to an arginine residue [FLAG–MST3(K53R)] prevented the appearance of any 58 kDa FLAG–MST3 on exposure of myocytes to calyculin A and the MST3(275–393) mAb epitope was preserved (Figure 3C).

In vitro immunokinase activities of the various MST3 mutants were assessed by phosphorylation of MBP. Myocytes were infected with AdV–FLAG–MST3 constructs titrated to produce essentially equal expression of FLAG–MST3 species (Figure 3D). MST3 is reportedly more active in such assays with Mn²⁺ rather than with Mg²⁺ as the metal ion cofactor, although the relative magnitude varies with metal ion concentration [6,19,49]. In our assays, FLAG–MST3(FL) was only ~30% as active with 20 mM Mn²⁺ in untreated or calyculin A-treated myocytes as with 20 mM Mg²⁺ (results not shown), thus Mg²⁺ was routinely used. Activities of FLAG–MST3(FL), FLAG–MST3(Δ314–431) and FLAG-MST3(T328A) were stimulated to a similar overall level by calyculin A (Figure 3D), suggesting that FLAG–MST3(Thr³²⁸) phosphorylation does not play any direct role in the regulation of MST3 activity. Furthermore, the fact that activity of FLAG–MST3(1–313) was indistinguishable from FLAG–MST3(FL) (Figure 3D) confirms that phosphorylation of MST3(Thr¹⁷⁸) in the catalytic domain did not contribute significantly to the reduced migration of FLAG–MST3(FL). Because FLAG–MST3(Δ314–431) was not constitutively active, the concept that caspase cleavage of MST3 results in constitutive activation [24,28] could not be confirmed in our system. FLAG–MST3(Δ341–431) and FLAG–MST3(Δ377–431) were activated similarly to FLAG–MST3(Δ314–431) and FLAG–MST3(FL) (results not shown). As expected, FLAG–MST3(T178A) exhibited a very low level of relative activation by calyculin A (Figure 3D) and FLAG–MST3(K53R) was completely inactive (results not shown), consistent with the findings of Lu et al. [19]. FLAG–MST3(T178D) may be slightly more active in the absence of calyculin A exposure than FLAG-MST3(FL) and, as expected, was only very slightly activated by calyculin A. Our general conclusion is that the activation of FLAG–MST3(FL) and the appearance of the FLAG–MST3 [i.e. FLAG–MST3(phosphoThr¹⁷⁸/phosphoThr³²⁸)] band at 58 kDa are different aspects of the same process.

Although we have concluded that the phosphorylation of MST3(Thr³²⁸) represents a cis-autophosphorylation, the kinase(s) responsible for phosphorylation of MST3(Thr¹⁷⁸) is unclear. Previous work suggests that the MST3(Thr¹⁷⁸) phosphorylation is also an autophosphorylation [6,19]. FLAG–MST3(T178A) expresses a very low, but detectable, MBP kinase activity (Figure 3D), whereas in contrast FLAG–MST3(K53R) is completely inactive (results not shown). Hence autophosphorylation of MST3(Thr¹⁷⁸) could lead to rapid autocatalytic activation following PP1/PP2A inhibition. In addition, we examined the rat MST3 primary sequence (Supplementary Figure S1) with Scansite (http://scansite.mit.edu). At medium stringency, Scansite identified PKA as a putative MST3(Thr¹⁷⁸) kinase. However, isoprenaline did not cause any reduction in mobility in FLAG–MST3(FL) on SDS/PAGE that should result from MST3(Thr¹⁷⁸) phosphorylation and subsequent MST3(Thr³²⁸) cis-autophosphorylation (results not shown). At high stringency, MST3(Ser³⁰⁵) was identified as a putative casein kinase 2 site. Although it is possible that this site is phosphorylated, its phosphorylation does not contribute to the reduction in mobility seen with calyculin A (Supplementary Figure S3B).

Lu et al. [19] identified two tryptic phosphopeptides in autophosphorylated recombinant MST3(FL), one of which contained phosphoThr¹⁷⁸ that was responsible for autoactivation. The second tryptic phosphopeptide was not identified and could contain MST3(phosphoThr³²⁸). We did not detect ‘constitutive’ autophosphorylation (Figure 1B, lower panel) or autoactivation (Figure 3D) of FLAG–MT3(FL) in myocytes in the absence of calyculin A. However, immunoblots in the study by Lu et al. [19] of immunoprecipitated HA–MST3(FL) expressed in COS-1 cells [with antibodies against residues 310–432 (sic)] detected two bands. These were identified as ‘unphosphorylated’ HA–MST3(FL) and autophosphorylated HA–MST3(FL). but, from the extent of the reduction in mobility, the latter may represent HA–MST3(phosphoThr¹⁷⁸/phosphoThr³²⁸) rather than HA–MST3(phosphoThr¹⁷⁸).

Others have shown that MST3 is cleaved immediately C-terminal to Asp³¹³ by caspases during apoptosis, generating a species that translocates to the nucleus [24,28]. The anthracycline daunomycin stimulates apoptosis and caspase activation in cardiac myocytes [50–52], thus we examined its effects on cleavage of FLAG–MST3(FL). As expected, daunomycin increased the appearance of the active 17 kDa fragment of caspase 3 in uninfected cardiac myocytes or myocytes expressing FLAG–MST3(FL), confirming stimulation of apoptosis (Figure 4A). There was detectable cleavage of FLAG–MST3(FL) to a form that co-migrated with FLAG–MST3(Δ314–431) (Figure 4B), indicating that induction of apoptosis resulted in MST3 cleavage. However, there was no detectable loss of the endogenous immunosignal detected by MST3(275–393) mAbs [which do not recognize FLAG–MST3(Δ314–431); Figure 2B] in uninfected myocytes (Figure 4C). We also examined the subcellular localization of FLAG–MST3(FL) following activation by calyculin A and of FLAG–MST3(Δ314–431). In unstimulated cells, FLAG–MST3(FL) was mainly cytoplasmic [Figure 4D(i)], whereas FLAG–MST3(Δ314–431) constitutively accumulated in the nucleus [Figure 4D(ii)] even though it cannot be enzymically active (Figure 3D). No significant accumulation of FLAG–MST3(FL) in the nucleus was detected following exposure to calyculin A [Figure 4D(iii)]. We could not detect any nuclear translocation of cleaved FLAG–MST3 into the nucleus in myocytes expressing FLAG–MST3(FL) following exposure to daunomycin [Figure 4D(iv)], presumably because it represents only a comparatively minor species (Figure 4B). Thus aspects of the canonical pathway for activation of MST3 operate in cardiac myocytes, in that MST3 is cleaved during apoptosis (Figure 4B) and the inactive cleaved form can potentially migrate to the nucleus [Figure 4D(ii)].

Since the principal mechanism of activation of MST3 by calyculin A does not appear to involve caspase cleavage, we took a computational and biochemical approach to investigate the structure of MST3. The X-ray structure of the catalytic domain (residues 1–303) of MST3 has been determined [53,54], but little is known about the regulatory domain. We first assessed the degree of disorder of rat MST3 using DISOclust [32]. The initial ~20 amino acids at the N-terminus constitute a disordered region that precedes a highly ordered structure (~270 residues) associated with the (conserved) kinase catalytic domain (Figure 5A). In contrast, the regulatory domain [MST3(314–431)] is highly disordered apart from a relatively low-probability-ordered region (residues 397–418) towards the C-terminus. Residues 281–431 were examined using PSIPRED [33]. This detected a region of low-probability strand structure [W³²⁵IFTIR³³⁰; note that this contains MST3(Thr³²⁸)] and a helical region between residues 389 and 421 (Figure 5B) corresponding to the dip in disorder probability between residues 397 and 418 in Figure 5(A). A representation of the three-dimensional model for MST3(FL) using the IntFOLD-TS method [34,35] shows the disordered N-terminal domain, the ordered catalytic domain and the largely disordered regulatory domain (Figure 5C). The strand W³²⁵IFTIR³³⁰ and the C-terminal predicted helical region sequences are marked.

Stokes radii (a) of the various FLAG–MST3 species were determined by gel filtration. The Stokes radius of FLAG–MST3(FL) by Sephacryl S200HR (Figures 6A and 6B) or S300HR (Figures 7A, 7C and 7E) chromatography was 47.3–49.6 Å (where 1 Å=0.1 nm; Table 1), considerably greater than the value of 29.2 Å calculated for a globular FLAG–MST3 monomer of ~49 kDa molecular mass. Either FLAG–MST3(FL) is highly asymmetric (axial ratio ~20 for an unhydrated prolate ellipsoid as calculated from the frictional coefficient [55]) which could result from the highly disordered structure of its regulatory domain (Figure 5) or it is complexed with other proteins (including possibly itself). Using myocytes infected variously with combinations of FLAG–MST3(FL), HA–MST3(FL) and GST–MST3(FL) followed by immunoprecipitation or glutathione–Sepharose precipitation then immunoblotting, we were unable to find any evidence for homomultimer formation (results not shown). The Stokes radius of FLAG–MST3(Δ377–431) was significantly less than that of FLAG–MST3(FL) and, although still highly asymmetric, was closer to the Stokes radius calculated for a globular protein (Figures 6A and 6B, and Table 1), suggesting that the C-terminal MST3(377–431) sequence may be involved in a heterologous interaction with another protein. As would be predicted, FLAG–MST3(Δ341–431) and FLAG–MST3(Δ314–431) also eluted at volumes closer to their calculated Stokes radii for corresponding globular proteins (Figure 6A and 6B, and Table 1).

Exposure of myocytes expressing FLAG–MST3(FL) to calyculin A caused a dramatic reduction in the Stokes radius, suggesting dissociation of a macromolecular complex involving FLAG–MST3(FL) or a dramatic shape change (Figures 7A–7C and 7E, and Table 1). Mixing of FLAG–MST3(FL) extracts from myocytes exposed or not exposed to calyculin A clearly showed the separation of inactive FLAG–MST3(FL) (Figure 7C, lower band) from activated FLAG-MST3(FL) (Figure 7C, upper band). Furthermore, the Stokes radius of FLAG–MST3(T328A) is significantly less (P<0.001) than FLAG–MST3(FL) (Figures 7D and 7E, and Table 1) and its elution volume was not significantly altered by exposure of myocytes to calyculin A (Figure 7D), suggesting that it does not interact with other proteins in the same manner as FLAG–MST3(FL).

YSK1/SOK1 (another member of the GCKIII subgroup) interacts with the Golgi protein GOLGA2/gm130 [56–59] with residues 20–302 of YSK1/SOK1 (this represents the entire catalytic domain) being important in this interaction [56]. Fidalgo et al. [58] have interpreted the data from [56] as indicating that residues 270–302 are critical in the interaction of YSK1/SOK1 with GOLGA2/gm130. However, an interaction between MST3 and GOLGA2/gm130 could not be identified [56]. This is surprising, given that the residues in rat MST3 {MST3(274–306), which also contain the bipartite NLS [28]} are homologous to YSK1/SOK1(270–302) (24/33 identity with 4/33 conservative replacements). We examined the interaction between FLAG–MST3(FL) and GOLGA2/gm130 in myocytes with and without calyculin A exposure. GOLGA2/gm130 was detected in immunoprecipitates of FLAG–MST3(FL) in unstimulated myocytes indicating an association (Figure 8A). Incubation of myocytes with calyculin A resulted in a decrease in this association (Figures 8A and 8B). Interactions (not necessarily direct and sometimes mediated by GOLGA2/gm130 [58]) between GCKIII members and CCM3/PDCD10, one of the proteins important in cerebral cavernous malformations [11], have been demonstrated [12,15,58,60,61]. We therefore also attempted to detect interactions between FLAG–MST3 and CCM3/PDCD10, but were unsuccessful (results not shown). However, further investigation of interactions between MST3 and heterologous proteins is warranted especially in view of recent work on striatins. Striatins (most abundant in the brain [62,63] although splice variants are expressed in the heart [64]) are PP2A B-type regulatory subunit scaffolding proteins which form complexes [STRIPAKs (striatin-interacting phosphatase and kinases)] with PP2A and additionally contain GCKIII members, CCM3/PDCD10 and phocein/Mob3 (which is present in the Golgi and other membranous structures) [15,60,65,66].

MO25 (of which there are two isoforms, MO25α and MO25β) is an evolutionarily conserved protein first identified by its expression during early murine embryogenesis [67,68]. It is best characterized as a scaffolding protein in a heterotrimer with the STRAD pseudokinases and the LKB1 (liver kinase B1) tumour suppressor kinase [3]. Binding of STRADα to MO25 involves four interaction sites (sites A–D) in STRADα: Glu¹⁰⁵ (site A), Ile¹³⁸ (site B), Tyr¹⁸⁵ (site C) and the C-terminal Trp⁴²⁹–(Glu⁴³⁰)–Phe⁴³¹ [W(E)F] motif (site D) [69,70]. A recent analysis of human Ste20-related kinases revealed that all four sites are conserved in the GCKIII and GCKVI subgroups, with sites A–C lying in the catalytic domain and site D in the C-terminal regulatory domain [4]. In rat MST3 (Supplementary Figure S1), at Glu⁵⁸, Leu⁹⁰ and Tyr¹³⁴ constitute sites A–C respectively, and the STRADα site DW(E)F motif is replaced by a WIF motif at residues 325–327, immediately N-terminal to MST3(Thr³²⁸). Surface plasmon resonance showed that MO25 isoforms (MO25α and MO25β) bind to the two GCKVI members [OXSR1 (oxidative-stress responsive 1)/OSR1/KIAA1101 and SPAK (SPS1-related proline-alanine-rich protein kinase)/PASK/Stk39 (serine threonine kinase 39) and the GCKIII member MST4/MASK with K_d1 values in the 3–10 μM range (as opposed to about 10–50 nM with STRADα) [3,4]. MST3 and YSK1/SOK1 were not studied. The activities of the wild-type GCKIII kinases are stimulated 2–5-fold by MO25α [4]. Stimulation of the activities of the GCKVI members is greater (70–100-fold), but the threonine residue of their activation loops had been mutated phosphomimetically to glutamic acid, so the situation is not necessarily comparable.

Because the sites involved in interaction with MO25 are conserved in all five GCKIII/GCKVI members, we examined the interaction between MO25 and FLAG–MST3(FL). Neither MO25α nor MO25β co-immunoprecipitated with FLAG–MST3(FL) in myocyte extracts in the absence of exposure to calyculin A (Figure 8C, top and middle panels). However, following exposure of myocytes to calyculin A and associated phosphorylation of FLAG–MST3(Thr³²⁸) (Figure 8C, bottom panel)], MO25α and MO25β each co-immunoprecipitated with FLAG–MST3(FL) (Figure 8C, top and middle panels). In an analogous experiment with myocytes expressing FLAG–MST3(FL) or FLAG–MST3(T328A) at approximately equal levels (Figure 8D, upper panel), MO25α did not co-immunoprecipitate with FLAG–MST3(T328A) to any significant extent following exposure of myocytes to calyculin A (Figure 8D, lower panel), indicative of an absolute requirement for phosphorylation of this site for association. Thus phosphorylation of MST3(Thr³²⁸) leads to decreased association with GOLGA2/gm130, together with increased association with MO25.

Filippi et al. [4] were unable to demonstrate co-immunoprecipitation of MO25 with any GCKIII or GCKVI kinase and attributed this to the high K_d values for the interactions. Our co-immunoprecipitation experiments show that calyculin A-activated FLAG–MST3(FL) does interact with MO25α (and MO25β) (Figures 8C and 8D), an interaction that does not occur with calyculin A-activated FLAG–MST3(T328A) (Figure 8D). Interestingly, the W³²⁵IFTIR³³⁰ sequence of MST3 corresponding to a low-probability strand region (Figures 5B and 5C) is conserved in both YSK1/SOK1 (W³¹⁹TFPPTIR³²⁶) and MST4/MASK (W³²⁴SFTTVR³³⁰) (Table 2). In the GCKVI members, where the WIF motif is replaced by WEW motif, the residue in the +1 position relative to the C-terminal tryptophan is a serine which lies in a highly acidic region in these kinases (Table 2). Although phosphorylation of a serine in the regulatory domain of SPAK/PASK/Stk39 or OXSR1/OSR1/KIAA1101 has been identified, the relevant residue lies in the −14 position relative to the serine residue of the WEWS motif [71–73]. Given that all of the GCKIII and GCKVI members except YSK1/SOK1 contain a serine/threonine residue +1 to the W(I/S/E)(F/W) motif (and, even in YSK1/SOK1, there is a threonine residue at +3), we suggest that the scheme of activation that we are proposing for MST3 might be more widely applicable to other GCKIII/VI members. Furthermore, we can speculate that the relatively low affinity of GCKIII/GCKVI members for MO25 [4] might be attributable to the serine/threonine residue immediately C-terminal to the STRADα-like site D being unphosphorylated.

From our experiments, we have developed the following model (Figure 9). MST3 is bound to the Golgi apparatus protein GOLGA2/gm130 [59] in the basal state, probably through the MST3(377–431) region. GOLGA2/gm130 might retain MST3 in the cytoplasm which may be why FLAG–MST3(Δ314–431) (which gel-filtration experiments suggest is unable to interact with GOLGA2/gm130) constitutively localizes to the nucleus [Figure 4D(ii)]. Calyculin A activates MST3 by promoting phosphorylation of MST3(Thr¹⁷⁸), and this leads to cis-autophosphorylation of MST3(Thr³²⁸) and dissociation of the MST3–GOLGA2/gm130 complex. The MST3(Thr³²⁸) cis-autophosphorylation requires the ‘docking domain’ encompassed by MST3 residues 341–376 and does not directly affect MST3 activity. MST3(Thr³²⁸) lies within a conserved WIFTIR sequence that may represent a structured region in the regulatory domain which is, overall, highly disordered. Cis-autophosphorylation of MST3(Thr³²⁸) simultaneously allows MO25 to bind to MST3. The function of the MST3–MO25 interaction is unknown as yet. In the MO25–STRAD–LKB1 complex, the GCKVI homologue STRAD is involved in nucleocytoplasmic shuttling of LKB1 and the interaction between LKB1 and STRAD (stabilized by MO25) localizes LKB1 to the cytoplasm [69,74]. It is possible that the various complexes that MST3 forms are related to the determination of subcellular localization.

AdV

adenovirus

CCM3

cerebral cavernous malformation 3

ECL

enhanced chemiluminescence

GCK

germinal centre kinase

GOLGA2

golgin A2

gm130

Golgi matrix protein 130

GST

glutathione transferase

HA

haemagglutinin

LKB1

liver kinase B1

mAb

monoclonal antibody

MASK

MST3 and SOK1-related kinase

MBP

myelin basic protein

MST

mammalian Sterile20-related kinase

NLS

nuclear localization signal

OXSR1

oxidative-stress responsive 1

PAK

p21-activated kinase

PASK

proline-alanine-rich Sterile20-related kinase

PDCD10

programmed cell death 10

PKA

cAMP-dependent protein kinase

PP

protein serine/threonine phosphatase

SOK1

Sterile20/oxidant stress response kinase1

SPAK

Sterile20/Sps1-related proline/alanine-rich kinase

Ste20

Sterile20

STRAD

Ste20-related adapter

TBST

Tris-buffered saline plus Tween 20

YSK1

yeast Sps1/Ste20-related kinase 1

Stephen Fuller performed the majority of the experimental work including the preparation of the many AdV constructs. Liam McGuffin was responsible for the modelling of the MST3 structure. Andrew Marshall performed some of the immunoblots. Alejandro Giraldo performed the immunofluorescence staining. Sampsa Pikkarainen made the initial observation shown in Figure 1(A) and assisted with the molecular cloning of MST3. Angela Clerk prepared the cardiac myocytes. Peter Sugden performed the hydrodynamic characterization of MST3. The experiments were conceived by Angela Clerk and Peter Sugden in conjunction with Stephen Fuller. The paper was written primarily by Peter Sugden with the assistance of Angela Clerk, and Stephen Fuller and Liam McGuffin also contributed.

We also thank Dario Alessi and Paola de los Heros (MRC Protein Phosphorylation Unit, University of Dundee, Dundee, U.K.) for providing us with the MO25β antibody.

This work was supported by the British Heart Foundation [grant number PG/07/060/23276] and Fondation Leducq [grant number CV05-02].