Glycogen Synthase Kinase 3 (GSK3) is a multifunctional kinase involved in diverse cellular activities such as metabolism, differentiation, and morphogenesis. Recent studies showed that GSK3 in Dictyostelium affects chemotaxis via TorC2 pathway and Daydreamer. Now we report that GSK3 affects PI3K membrane localization, of which the mechanism has remained to be fully understood in Dictyostelium. The membrane localization domain (LD) of Phosphatidylinositol-3-kinase 1 (PI3K1) is phosphorylated on serine residues in a GSK3 dependent mechanism and PI3K1-LD exhibited biased membrane localization in gsk3− cells compared to the wild type cells. Furthermore, multiple GSK3-phosphorylation consensus sites exist in PI3K1-LD, of which phosphomimetic substitutions restored cAMP induced transient membrane localization of PI3K1-LD in gsk3− cells. Serine to alanine substitution mutants of PI3K1-LD, in contrast, displayed constitutive membrane localization in wild type cells. Biochemical analysis revealed that GSK3 dependent serine phosphorylation of PI3K1-LD is constitutive during the course of cAMP stimulation. Together, these data suggest that GSK3 dependent serine phosphorylation is a prerequisite for chemoattractant cAMP induced PI3K membrane localization.
Understanding how eukaryotic cells manage their locomotive behaviors has been one of the major goals of current cell biology. Years of investigation uncovered that chemotaxis is synergistically controlled through multiple pathways that include PI3K, TORC2, soluble Guanylyl Cyclase (sGC), and PLA2 in Dictyostelium (Swaney et al. 2010; Kortholt et al. 2011). Recent studies using Dictyostelium cells demonstrated that GSK3 affects phosphatidylinositol-3,4,5-triphosphate (PIP3) metabolism, TORC2 signaling, and F-Actin remodeling via Daydreamer (Teo et al. 2010; Kolsch et al. 2013). It remains to be determined, however, how GSK3 affects these pathways at the molecular level. Furthermore, there exist contrasting phenotypes manifested by cells lacking GSK3: Teo et al. (2010)reported lack of increase in PIP3 level in response to cAMP, no TorC2 activation, and decreased Adenylyl Cyclase expression in gsk3− cells, whereas Kolsch et al. (2012) showed different results. The results reported here were obtained from gsk3− cells with JH10 background, which are largely consistent with those of Teo et al. (2010).
Dictyostelium GSK3 affected prestimulus control of PIP3 level through regulating PI3K membrane localization. Previous studies showed Dictyostelium PI3K is regulated at the following two levels: one is its localization control between the cytosol and the plasma membrane, and the other is regulating lipid kinase activity (Funamoto et al. 2002). Activation of PI3K is mediated through Ras, but the mechanism of PI3K membrane recruitment is not well understood. We noticed that the membrane localization domains of PI3K1 and PI3K2, the two main PI3Ks responsible for the majority of PIP3 generation in Dictyostelium cells (Huang et al. 2003), contain the consensus sequence for GSK3 phosphorylation (SxxxSxxxS, where ‘x’ indicates any amino acid) and that the PI3K1 membrane localization domain (LD) is phosphorylated on serine residues in wild type cells but significantly underphosphorylated in gsk3− cells. This serine phosphorylation of PI3K1-LD seems to be a requirement for optimal transient membrane localization of PI3K in response to cAMP stimulation. Biochemical as well as imaging analyses of the effects of GSK3 on the PI3K are described below.
Materials and methods
Dictyostelium culture, chemotaxis and motility assays, and Fluorescence microscopy
Dictyostelium wild-type strain (JH10) and gsk3− cells were grown as described previously (Kim et al. 1999). Exponentially growing cells were differentiated with 50 nmol/L pulses of cAMP for 4 h and assayed for either random motility or challenged with a cAMP-filled capillary needle (Eppendorf Femtotip) at a density of 6 × 104 cells/cm2. Images were recorded using a CoolSNAP digital camera with OpenLab software. Chemotactic indices and speed of movements were computed as described before (Veeranki et al. 2008). All fluorescent images were obtained using a 100× oil lens on a Leica DM IRB inverted epifluorescence microscope equipped with a CoolSNAP digital camera and relative fluorescence profiles were obtained using Openlab image software.
GSK3 kinase assay
The GSK3 peptide kinase assay was performed as previously reported (Ryves et al. 1998; Kim et al. 2002). The primed peptide substrate sequences are as follows. GSM peptide –RRRPASVPPSPSLSRHSSPHQRR, P1 peptide – RRRMNSIESS7SNDS11NRR, P2 peptide – RRRNDSNCSS109GSSS113-PGRR, and P3 peptide – RRRGSSSGSS177SGGS181-PDRR. Control samples were assayed in the presence of 50 mmol/L LiCl, a GSK3 inhibitor, and specific activities were calculated by subtracting these values from the total counts. Specific activity is expressed as picomole of phosphate transferred/assay. Each experiment was repeated three times and error bars represent standard deviation.
Ras binding assay
Active Ras proteins were assayed as before (Veeranki et al. 2008). Whole cell lysates prepared after 4 h of cAMP pulsing were mixed with 5 μg of purified GST-Raf1-RBD or GST-Byr2-RBD on Glutathione-sepharose beads at 4°C for 2 h, and the pellets were washed three times with cell lysis buffer (20 mmol/L TrisCl [pH7.7], 150 mmol/L NaCl, 1% Triton X-100, 5% glycerol, 1 mmol/L EDTA, 0.1% beta-mercaptoethanol, and 1× Roche Protease Inhibitor mix). The active Ras proteins bound with GST-RBD were visualized by western blotting with the anti-Pan-Ras antibody.
Antibodies and λ-phosphatase
Anti-GFP antibodies were from Covance for western blot analysis. Anti-GST antibody was purchased from Santa Cruz Biotech and anti-Pan-Ras antibodies were from Calbiochem (Ab-3). Rabbit anti-phosphoserine antibody was purchased from Invitrogen (Catalog No. 61-8100). For each western, 1:250 dilution was made to achieve a final concentration of 1 mg/mL. λ-phosphatase was purchased from Biolabs, the reaction was conducted according to the provider's suggestions (30 min at 30°C in a total volume of 50 μL).
GST-PI3K1-LD expression vector construction
pEXP4(+)-PI3K1LD-GFP (Funamoto et al. 2002) was used as a template for generating GST-PI3K1-LD. The full length PI3K1LD cDNA (the intron within the PI3K1LD was removed by site-directed mutagenesis) is 1476 bp long and contains a ClaI site at +1457. pEXP4(+)-PI3K1LD was obtained by removing GFP with KpnI digestion. By using a primer set (5'-GAAGATCTATGTCCCCTATACTAGGTTATTGG-3’ and 5’-GAAGATCTGAATTCCGGGGATCCACG-3’), a cDNA encoding GST, a thrombin site, a BamHI, an EcoRI site was generated by polymerase chain reaction (PCR) from PGEX-4T-1 vector. The GST gene was inserted into the upstream of the region of PI3K1LD in pEXP4(+)-PI3K1LD. All constructs above were confirmed by sequencing.
Mutations were generated using the QuikChange Site Directed Mutagenesis Kit (Stratagene) and confirmed by sequencing, and expressed in pEXP4. The serine to alanine mutations were generated by PCR with mutants primers as indicated: PI3K1-LD (S7A, S11A) – primers SDM11F and SDM11R, PI3K1-LD (S109A, S113A) – primers SDM12F and SDM12R, PI3K1-LD (S177A, S181A) – primers SDM13F and SDM13R, PI3K1-LD (S370A, S374A) – primers SDM14F and SDM14R. The sextuple alanine to serine mutant (SAS-PI3K1-LD) was generated by PCR from PI3K1-LD (S7A, S11A, S177A, S181A) using mutant primers SDM12F and SDM12R. The Sextuple Phosphomimetic mutant (SPM-PI3K1-LD) was generated through the following steps: PI3K1-LD(S7E, S11E) was generated by using the mutant primers SDM21F and SDM21R, PI3K1-LD(S109E, S113E, S177E, S181E) was generated with SDM22F and SDM22R mutant primers using PI3K1-LD (S177E, S181E) as the template, and SPM-PI3K1-LD (S7E, S11E, S109E, S113E, S177E, S181E) was generated by substituting the SpeI-SpeI fragment of the PI3K1-LD (S109E, S113E, S177E, S181E), which contains the four mutations, into the PI3K1-LD (S7E, S11E). Primers used in the mutants generation are as follows:
Subcellular fractionation of wild type and mutants GFP-PI3K1-LD proteins
Cell fractionation assay were performed as described previously with minor modifications (Han et al. 2006). Twenty five million aggregation competent cells were spun down and washed once with ice-cold PBS (137 mmol/L NaCl, 2.7 mmol/L KCl, 10 mmol/L Na2HPO4, 1.8 mmol/L KH2PO4) and the cell pellets were treated with 400 μL of 0.02% TritonX-100 and incubated 10 min at 4°C with agitation. All solutions contained protease inhibitors (Roche, Complete Mini). Mixtures were centrifuged at 12 000 g for 5 min at 4°C. The supernatants were then mixed with 4 × SDS protein loading dye and the pellet were mixed with 100 μL of 1 × SDS protein loading dye. Forty microliters of soluble (cytosolic) fractions and 1 μL of pellet (membranous) fractions were loaded onto sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) and analyzed by western blotting using anti-GFP antibody. Ras proteins were used as a marker enriched at the pellet fraction.
gsk3− cells display aberrant membrane localization of GFP-PHcrac and GFP-PI3K1-LD proteins
Vegetative wild type Dictyostelium cells expressing GFP-PH proteins as a PIP3 marker (generous gift from Dr. Devreotes Lab) display largely cytoplasmic localization of GFP-PHcrac (Parent et al. 1998; Kortholt et al. 2011). In contrast, vegetative gsk3− cells expressing GFP-PHcrac proteins displayed significantly elevated level of plasma membrane localization of GFP-PHcrac proteins compared to wild type cells (Fig. 1A). Similarly, wild type cells pulsed with cAMP for 4 h exhibited low basal level of GFP-PHcrac proteins at the plasma membrane. In contrast, cAMP pulsed gsk3− cells displayed relatively high basal level of GFP-PHcrac proteins at the plasma membrane (Fig. 1B, time 0). In response to cAMP stimulation, GFP-PHcrac proteins translocalized to the plasma membrane in a transient manner in wild type cells, but not in gsk3− cells (Fig. 1B, Movies S1 and S2). Our results using gsk3− cells with JH10 background are consistent with the results of Teo et al. (2010) using gsk3− cells of Ax2 background.
One of the causes that gsk3− cells were unable to further recruit GFP-PHcrac proteins to the plasma membrane in response to chemoattractant could be that PI3K proteins are not properly regulated in gsk3− cells. PI3K1 and PI3K2, two of the major proteins responsible for cAMP induced PIP3 generation, are regulated at the level of lipid kinase activation by small G protein Ras as well as at the plasma membrane recruitment step through their membrane localization domain (PI3K-LD) (Huang et al. 2003). A previous study showed that constitutive localization of PI3K to the plasma membrane by myristoylation tagging led to higher prestimulus level of PIP3 (Huang et al. 2003), underscoring the importance of proper prestimulus membrane localization of PI3K.
Given that both elevated production of PIP3 through PI3K misregulation and sustained maintenance of PIP3 via downregulation of PTEN, a PIP3 phosphatse, could result in high basal PIP3 level, cAMP mediated regulation of both PI3K and PTEN were analyzed. Wild type cells expressing either GFP-PI3K1-LD or GFP-PTEN exhibited transient subcellular relocalization of each protein in response to cAMP as expected (Movies S3 and S4). gsk3− cells, however, displayed aberrantly high prestimulus level of GFP-PI3K1-LD at the plasma membrane at both vegetative and pulsed stages (Fig. 2A,C). No clear difference of prestimulus localization of GFP-PTEN was observed from wild type and gsk3− cells (Fig. 2A,D). gsk3− cells were unable to further mobilize PI3K1-LD to the plasma membrane in response to cAMP (Fig. 2C and Movie S5), yet displayed a transient membrane dislocalization of GFP-PTEN proteins upon cAMP stimulation similarly to wild type cells (Fig. 2D and Movie S6). The levels of GFP-PI3K1-LD proteins were comparable in both wild type and gsk3− cells (Fig. 2B). These data showed that aberrant prestimulus PI3K membrane localization was observed in gsk3− cells, which is insensitive to cAMP stimulation.
To determine the possibility that whether an upstream activator of PI3K, Ras, is misregulated in gsk3− cells or not, the basal and cAMP induced Ras activities were measured using two different active Ras binding domains (RBD) from Raf1 and Byr2 proteins. Raf1-RBD preferentially binds to active RasB, RasD, and RasG and Byr2-RBD strongly to RasC but weakly to RasG (Kae et al. 2004; Sasaki et al. 2004). gsk3− cells displayed normal prestimulus Ras activities, but Ras activity persisted longer in gsk3− cells than in wild type cells when assayed by Raf1-RBD (Fig. 3A). In addition, cAMP induced Byr2-RBD binding Ras activities were exaggerated in gsk3− cells, but clearly responsive to cAMP stimulation. These data suggest that the lack of detectible change of cAMP induced PIP3 level in gsk3− cells when assayed using GFP-PHcrac protein as a PIP3 marker is not likely due to the misregulation of Ras activation. We prefer the view that the defect of cAMP-induced PIP3 level change in gsk3− cells are likely due to the limited level of PI(4,5)P2 subpopulation accessible to PI3K, which was previously proposed by Teo et al. (2010).
GSK3 dependent serine phosphorylation of PI3K1-LD is critical for its cAMP induced membrane shuttling
Given that PI3K1-LD proteins are not properly regulated in cells lacking GSK3, GSK3 may phosphorylate PI3K1-LD and thus regulate its membrane localization. Furthermore, the membrane localization domains of PI3K1 and 2 contain multiple GSK3 phosphorylation consensus sequences (Fig. 4A). Systematic serine to aspartic acid as well as alanine sextuple substitution mutants of PI3K1-LD-GFP were generated and expressed in wild type and gsk3− cells and biochemically analyzed (Fig. 4B). Wild type PI3K1-LD-GFP proteins displayed contrasting subcellular localization in wild type and gsk3− cells; PI3K1-LD displayed approximately 50% enrichment in the pellet fraction of gsk3− cells compared to the wild type (Fig. 4B middle panel). Substitutions of potential GSK3 phosphorylation sites with aspartic acid (Sextuple Phosphomimetic Mutant [SPM]) restored cytosolic enrichment of PI3K in gsk3− cells, whereas serine to alanine mutant (Sextuple Alanine Substitution [SAS]) proteins were enriched in the pellet fraction of wild type cells. The total amount of PI3K1-LD proteins in Pellet and Supernatant fractions were comparable as shown in Figure 4B (lower panel) and the relative distributions of various PI3K1-LD proteins in each fraction were shown in Figure 4C.
The same sets of PI3K1-LD-GFP proteins were then analyzed microscopically. The SPM-PI3K1-LD-GFP protein displayed transient membrane translocalization in response to cAMP stimulation in both wild type and gsk3− cells (Fig. 5 and Movies S7 and S8), suggesting that phosphorylation of these sites is a prerequisite for chemoattractant-induced regulation of PI3K1 subcellular localization. In other words, dephosphorylation of these sites would not be necessary for PI3K to relocalize back to the cytosol after the peak activation of PI3K. This point is further supported by biochemical analysis of PI3K1-LD phosphorylation described later (Fig. 6). In contrast to phosphomimetic mutant, the SAS-PI3K1-LD proteins in wild type cells displayed the high basal level of membrane localization, which was refractory to further activation by cAMP, resembling the situation of PI3K1 in gsk3− cells (Figs 5B and 2B, and Movie S9). Relative intensities of GFP signals from each experiment were shown in Figure 5C.
PI3K1-LD proteins are constitutively phosphorylated on serine residues in a GSK3 dependent manner
To further dissect the role of the phosphorylation of PI3K1-LD, we generated and expressed GST-PI3K1-LD in wild type and gsk3− cells. Western blot analysis using anti-phospho-serine specific antibody disclosed differential phosphorylation of PI3K1-LD proteins in wild type and gsk3− cells, which is sensitive to phosphatase treatment (Fig. 6A). We also noticed a slight, but consistent mobility difference between the GST-PI3K1-LD protein bands purified from wild type and gsk3− cells, which is sensitive to phosphatase treatment as well (Fig. 6A). As mentioned previously, the serine phosphorylation of PI3K1-LD protein is likely a prerequisite or permissive signal for chemoattractant-induced regulation of PI3K. Consistently, GST-PI3K1-LD proteins displayed no significant change in its serine phosphorylation level in response to cAMP stimulation (Fig. 6B). We also determined that synthetic and primed peptides encoding the three potential GSK3 phosphorylation sequences were phosphorylated by whole cell extract in a lithium sensitive manner (Fig. 6C). In contrast, only minimal level of lithium sensitive phosphorylation was observed from gsk3− cell lysates as expected. The simplest interpretation of these data is that GSK3 directly phosphorylates these sites upon priming by other kinases. Although, it is possible that GSK3 regulates PI3K phosphorylation indirectly through other kinase.
Previous studies on the role of GSK3 in the context of cell motility regulation uncovered that GSK3 affects TorC2 (Teo et al. 2010) and Daydreamer (Kolsch et al. 2012) mediated signaling. This study details that GSK3 is essential for proper membrane targeting of PI3K. Misregulation of PI3K in gsk3− cells led to an aberrantly high level of prestimulus PIP3, which resembles the situation of cells lacking PTEN, and inefficient polarization in response to cAMP gradient. Furthermore, that gsk3− cells were refractory to cAMP mediated transient PIP3 production is likely due to a limited pool of PIP2 accessible for PI3K as suggested by Teo et al. (2010), given that a set of Ras can still be transiently activated in gsk3− cells by cAMP. Both GFP imaging experiments and biochemical fractionation supported the view that GSK3 affects serine phosphorylation of PI3K1 membrane localization domain, which is likely a permissive signal for chemoattractant-induced plasma membrane localization of PI3K. Considering that Sextuple Phosphomimetic PI3K1-LD mutant was able to display transient membrane localization in response to cAMP and the level of serine phosphorylation of GST-PI3K1-LD did not change during the course of cAMP stimulation, GSK3 dependent phosphorylation of PI3K1-LD is a prerequisite for proper chemotactic responses. It is thus likely that chemoattractant signals control PI3K localization by modulating PI3K binding targets either in the cytosol or on the plasma membrane as diagramed in Fig. 7.
In addition to the aberrant PI3K membrane recruitment, it is clear that gsk3− cells suffer from other defects. As reported by Teo et al. (2010) TorC2 activation and PKB and PKBR activations were severely compromised. Consistently, alleviating PIP3 level by treating gsk3− cells with PI3K inhibitor LY294002 did not improve motility properties of gsk3− cells (data not shown). Another possibility is that gsk3− cells either underexpress or overexpress signaling component critical for motility regulation. One such target is SodC, which is significantly underexpressed in gsk3− cells (Strmecki et al. 2007). sodC− cells display aberrant Ras activation with a concomitant increase in the prestimulus PIP3 level (Veeranki et al. 2008). Overexpression of SodC in gsk3− cells, however, failed to rescue any of the compromised motility parameters (data not shown). Wild type cells expressing SodC displayed modest inhibition of the speed of movement, indicating that a fine-tuning of the superoxide level by this enzyme is necessary for optimal motility behavior of cells. Finally, it would be worthy to mention that the proper regulation of Raf1-RBD binding Ras species, likely RasD, may constitute a critical part of executing cell fate decisions mediated by GSK3.
This work was supported by Grant-In-Aid from American Heart Association (0555264B) and NIH SC1 (CA143958) from the National Cancer Institute.