Subcellular Localization of Glycogen Synthase Kinase 3β Controls Embryonic Stem Cell Self-Renewal

Abstract

Phosphatidylinositol 3-kinase (PI3K), protein kinase B (AKT1), and c-myc have well-established roles in promoting the maintenance of murine embryonic stem cells (mESCs). In contrast, the activity of glycogen synthase kinase 3β (GSK3β), a negatively regulated target of AKT1 signaling, antagonizes self-renewal. Here, we show that PI3K/AKT1 signaling promotes self-renewal by suppressing GSK3β activity and restricting its access to nuclear substrates such as c-myc. GSK3β shuttles between the cytoplasm and nucleus in mESCs but accumulates in the cytoplasm in an inactive form due to AKT1-dependent nuclear export and inhibitory phosphorylation. When PI3K/AKT1 signaling declines following leukemia inhibitory factor withdrawal, active GSK3β accumulates in the nucleus, where it targets c-myc through phosphorylation on threonine 58 (T58), promoting its degradation. Ectopic nuclear localization of active GSK3β promotes differentiation, but this process is blocked by a mutant form of c-myc (T58A) that evades phosphorylation by GSK3β. This novel mechanism explains how AKT1 promotes self-renewal by regulating the activity and localization of GSK3β. This pathway converges on c-myc, a key regulator of mESC self-renewal.

The self-renewal of murine embryonic stem cells (mESCs) is controlled through the activity of leukemia inhibitory factor (LIF) by a mechanism requiring the activation of STAT3 (12, 14). Also critical for self-renewal is the canonical phosphatidylinositol 3-kinase (PI3K) pathway, which signals through AKT1 (19, 28). Interrupting the activity of these signaling pathways results in loss of the capacity for long-term self-renewal (16). Under self-renewing conditions, elevated PI3K/AKT1 activity levels are proposed to directly inhibit glycogen synthase kinase 3β (GSK3β) through the phosphorylation of serine 9 (S9) (4, 19, 23). Then, as mESCs differentiate following LIF withdrawal, PI3K/AKT1 activity declines (19, 23) and the catalytic activity of GSK3β becomes elevated due to the hypophosphorylation of S9 (4). The importance of GSK3β in mESC self-renewal is highlighted by several key observations. First, inhibition of GSK3β activity is required to maintain mESCs in a pluripotent, self-renewing state (20, 32). Second, genetic inactivation of GSK3α/β severely compromises the ability of mESCs to differentiate (8). Finally, inhibition of GSK3β significantly enhances the derivation frequencies of mESC lines from blastocyst stage embryos (27).

The proto-oncogene c-MYC is a direct transcriptional target of LIF/STAT3 signaling and is essential for the maintenance of mESC self-renewal (4). Inactivation of Myc activity leads to differentiation, and its ectopic expression relieves the requirement for LIF/STAT3 signaling. Consistent with this, Myc levels are elevated in pluripotent mESCs but decline markedly during the initial stages of differentiation. Together, these findings establish Myc as a key regulator of mESC self-renewal. The importance of Myc activity in establishing and maintaining the pluripotent state is underscored by recent experiments showing that c-myc is important for the efficient generation of induced pluripotent stem (iPS) cells, in conjunction with other factors such as Oct4, Sox2, and Klf4 (17, 25, 29).

The transcriptional downregulation of the c-MYC gene represents one of the first responses to the cessation of LIF signaling and the inactivation of STAT3 in mESCs. Besides being controlled by transcription, c-myc levels are primarily regulated by changes in protein stability. Although c-myc is an unstable protein in nontransformed cell lines, with a half-life (t_1/2) of ∼10 to 5 min (31), it exhibits unusual stability in mESCs (t_1/2 of ∼105 min) (4). This is comparable to the stability of oncogenic forms of Myc expressed by transforming viruses (v-myc) or in tumors such as Burkitt's lymphoma (31). Following the decline in PI3K/AKT1 activity that accompanies the loss of LIF signaling, c-myc turnover is accelerated by a mechanism requiring its phosphorylation on threonine 58 (T58) (4). Failure to phosphorylate c-myc on T58 following LIF withdrawal maintains it in a stable state, sustains elevated levels of c-myc protein, and supports LIF-independent self-renewal (4). Phosphorylation of c-myc on T58 and its accelerated degradation is therefore a key event required for transition from the self-renewing state to commitment to differentiation.

Several major questions remain regarding the respective roles of PI3K/AKT1, GSK3β, and Myc in the determination of the fate of mESC. While PI3K/AKT1 is recognized to be critical for self-renewal, its mechanism of action has not been defined. Moreover, although suppression of GSK3β activity is a requirement for self-renewal, how it is regulated in mESCs and how it antagonizes self-renewal is unclear. Finally, although the collapse of Myc activity is a defining event in early cell fate commitment, a mechanism connecting it with PI3K/AKT1 and GSK3β has not been previously defined. Together, these questions represent significant shortfalls in our understanding of self-renewal, pluripotency, and early cell fate decisions made by mESCs. This report addresses the above-mentioned issues and defines a mechanism linking PI3K/AKT1, GSK3β, and Myc. These data establish a mechanism for how AKT1 and GSK3β perform opposing roles in the control of mESC cell fate decisions, involving a pathway that converges on Myc transcription factors.

MATERIALS AND METHODS

Cell culture and reagents.

mESCs were cultured in the absence of feeders on tissue culture grade plastic precoated with 0.2% gelatin-phosphate buffered saline, as described previously (22). Stable cell lines were generated by the transfection of supercoiled constructs with Lipofectamine 2000 (catalog no. 11668-027; Invitrogen) according to the manufacturer's instructions. After a 24-h recovery period, transfected cells were selected in puromycin (1 μg/ml) or neomycin (200 μg/ml) for between 7 and 10 days and then clonally expanded in the presence of drug selection.

The following materials were used in this study: (2′Z,3′E)-6-bromoindirubin-3′-oxime (BIO) (BIO/GSK3 inhibitor IX, catalog no. 361552; Calbiochem), MeBIO (catalog no. 361556; Calbiochem), lithium chloride (catalog no. 7447-41-8; Fisher), GSK3 inhibitor XV (Calbiochem), CHIR 99021/CT 99021 (Stemgent), AKT inhibitor V (AKTV) (catalog no. 124012; Calbiochem), LY294002 (catalog no. ST-420; Biomol), PI-103 (Calbiochem), leptomycin B (LB) (catalog no. 431050; Calbiochem), and 4-hydroxytamoxifen (catalog no. H6278; Sigma).

Plasmid constructs and antibodies.

c-myc^T58A-ER and myr.AKT1-ER expression constructs have been described previously (4, 28). Rat GSK3β expression constructs were generated by insertion into the EcoRI site of pCAGipuro. All site-directed mutagenesis was performed by using a QuikChange kit (Stratagene) and was confirmed by sequencing on both DNA strands. Hemagglutinin (HA) epitope-tagged constructs were generated by inserting a triple HA tag by PCR, immediately before the STOP codon. Constructs with engineered triple HA tags and two tandem copies of the simian virus 40 (SV40) nuclear localization signal were also generated by insertion immediately before the STOP codon (details available on request). Antibodies used in this study were as follows: GSK3β (catalog no. 610202; BD Biosciences), GSK3β phospho-S9 (catalog no. 9336; Cell Signaling), β-actin (catalog no. A2066; Sigma), Oct4 (catalog no. SC-8628; Santa Cruz), c-myc (catalog no. 9402; Cell Signaling), c-myc phospho-T58 (catalog no. 9401; Cell Signaling), AKT1 (catalog no. 4685; Cell Signaling), AKT1 phospho-S473 (catalog no. 9271; Cell Signaling), Nanog (catalog no. RCBAB0002FF; Cosmo Bio), Cdk2 (catalog no. SC-6248; Santa Cruz), and fibrillarin (a gift from M. Terns, University of Georgia).

Immunoblotting, subcellular fractionation, and transcript analysis.

Immunoblot analysis was performed as described previously (22). Nuclear and cytoplasmic subcellular fractions were prepared by using a Nxtract CelLytic NuCLEAR extraction kit (Sigma). Fractions were prepared from ∼5 × 10⁶ cells, and for the determination of subcellular distribution, equal proportions of nuclear and cytoplasmic extracts were probed following immunoblotting and sodium dodecyl sulfate-polyacrylamide gel electrophoresis. Quantitative reverse transcriptase PCR assays were performed, using TaqMan gene expression assays (Applied Biosystems).

Immunostaining, confocal microscopy, and Leishman staining.

Cells were plated on gelatin-coated LabTec slides. Prior to being mounted, cells were stained with a To-Pro-3-phosphate-buffered saline solution (1:300) (catalog no. T3605; Invitrogen) for 5 min. All images were acquired with a Zeiss LSM 510 META confocal microscope (40× oil objective; 1.3 numerical aperture). Staining of cells with Leishman reagent was done as described previously (21). Colonies were scored as being positive only if >90% of the cells in a colony stained with Leishman reagent. All counts were performed in triplicate on at least 100 individual colonies.

RESULTS

Activation and nuclear localization of GSK3β during early mESC differentiation.

While evaluating the regulation of GSK3β in mESCs, we observed that within 24 h of LIF withdrawal its enzymatic activity increased over 15-fold (4), it became hypophosphorylated on S9, and it relocalized from the cytoplasm to the nucleus (Fig. 1A and B). This occurred in adherent cultures and when cells were grown in suspension as embryoid bodies (EBs) (Fig. 1A and B), indicating that this is a general phenomenon associated with commitment to early differentiation.

FIG. 1.

GSK3β relocalizes to the nucleus during early mESC differentiation. (A) R1 mESCs (plus LIF), day 2 (d2) EBs, R1 mESCs grown without LIF for 2 days (d2 −LIF), and mESCs treated with LB (110 nM) for 6 h were probed with antibodies as indicated. To-Pro-3 (blue) was used to visualize DNA. All staining was visualized by confocal microscopy. Scale bar, 50 μm. (B) Nuclear (N) and cytoplasmic (C) extracts from R1 mESCs and mESCs grown in the absence of LIF for 1 day (d1) or 2 days (d2) were subjected to immunoblot analysis as indicated. Oct4 and β-actin were used as nuclear and cytoplasmic markers, respectively. Right panel: analysis of nuclear and cytoplasmic fractions from mESCs treated with LB for 6 h. Equal proportions of nuclear and cytoplasmic fractions were loaded for direct comparison. (C) Whole-cell lysates (20 μg) from R1 mESCs or mESCs grown in the absence of LIF for up to 3 days or as EBs over 4 days were immunoblotted as indicated. (D) mESCs (plus LIF) or mESCs cultured for 1 day in the absence of LIF were immunostained for c-myc^pT58 and visualized by confocal microscopy.

Nuclear accumulation of GSK3β coincided with the decline of c-myc protein levels, increased c-myc T58 phosphorylation, and loss of AKT1 activity, as indicated by decreased phosphorylation on serine 473 (S473) (Fig. 1C and D). In contrast to changes in levels of pluripotency markers, these events coincided with the loss of Nanog but preceded the decline in Oct4 levels (Fig. 1C; see also Fig. S1 in the supplemental material). The addition of GSK3 inhibitors such as BIO, lithium chloride, GSK3 inhibitor XV, and CHIR 99021/CT 99021 to EBs blocked phosphorylation of c-myc on T58 (see Fig. S2 in the supplemental material), indicating that GSK3α/β is required for the accelerated turnover of c-myc during early mESC differentiation. The inactive BIO analogue, MeBIO, had no effect on T58 phosphorylation under these conditions. These observations are consistent with our previous model (4) that activation of GSK3β promotes mESC differentiation by promoting the degradation of c-myc through a T58-dependent mechanism. Although the accumulation of GSK3β in the nucleus is specifically associated with its hypophosphorylation on S9, minor amounts of S9-phosphorylated GSK3β were detected in differentiating cells, but only in the cytoplasm (Fig. 1A). The relocalization and activation of GSK3β following LIF withdrawal was observed in all mESC lines tested, including R1, D3, E14tg2a, and CGR8 (Fig. 1; see also Fig. S3 in the supplemental material; also data not shown), indicating that these events are a general feature of early differentiation commitment.

One previous report showed that in L cells and 10T1/2 cells, GSK3β shuttles between the cytoplasm and the nucleus, even though it has a predominantly cytoplasmic localization (9). However, it is unclear how general a phenomenon this is, and moreover, the biological significance of GSK3β shuttling and its mechanism of regulation have not been elucidated. To establish if GSK3β nuclear shuttling occurs in mESCs, they were treated with LB, an inhibitor of Crm1-mediated nuclear export. This treatment resulted in the accumulation of S9-phosphorylated GSK3β in the nucleus (Fig. 1A and B), indicating that GSK3β shuttles between the nucleus and the cytoplasm in self-renewing mESCs. The simplest interpretation of these results is that although GSK3β shuttles, it is located primarily in the cytoplasm, because the rate of nuclear export exceeds that of import. Similar experiments were performed with two other pluripotent cell types of embryonic origin, epiblast stem cells (EpiSCs) (26) and induced pluripotent stem cells (iPS cells) (see Fig. S4 in the supplemental material). In both cases, LB treatment resulted in the redistribution of GSK3β from the cytoplasm to the nucleus within 6 h (see Fig. S4 in the supplemental material), indicating that GSK3β shuttling is a general attribute of pluripotent cells.

GSK3β accumulates in the nucleus of mESCs when PI3K is inhibited.

The accumulation of active GSK3β in the nucleus coincides with decreased AKT1 activity (Fig. 1C) and is possibly linked to the loss of PI3K signaling following LIF withdrawal (19). A role for PI3K/AKT1 in the control of GSK3β subcellular localization has not been proposed previously, however, although it has well-defined roles in promoting self-renewal (28). To further evaluate the connection between PI3K/AKT1 activity and GSK3β subcellular localization, we added inhibitors of PI3K (LY294002 and PI-103) and AKT1 (AKTV) to mESCs cultured in the presence of LIF and asked if this was sufficient to relocalize GSK3β to the nucleus. Treatment of mESCs with LY294002, PI-103, or AKTV promoted the nuclear accumulation of GSK3β (Fig. 2A; see also Fig. S5 and S6 in the supplemental material), which coincided with decreased c-myc levels and increased T58 phosphorylation (Fig. 2B and C). T58-phosphorylated c-myc colocalized with fibrillarin in nucleoli (Fig. 2B), the site where c-myc is targeted for ubiquitination, a key step required for its subsequent proteolysis (31). Nanog responded to LY294002 treatment in a manner similar to that of c-myc, although Oct4 levels remained stable (Fig. 2C; see also Fig. S7 in the supplemental material). Inhibition of PI3K/AKT1 therefore reproduces events associated with early differentiation with regard to GSK3β activity status and subcellular localization.

FIG. 2.

PI3K/AKT activity controls the cytoplasmic-nuclear distribution of GSK3β in mESCs. (A) R1 mESCs (plus LIF) were treated with vector alone or with LY294002 (40 μM, 24 h). Cells were then probed as indicated and visualized by confocal microscopy (scale bar, 50 μm). (B) mESCs treated with LY294002 (40 μM) or LY294002 plus BIO (2 μM) were probed with antibodies for c-myc^pT58 or the nucleolar marker fibrillarin. (C) Immunoblot analysis of untreated mESCs or mESCs treated for 24 h with LY294002 (40 μM) and, where indicated, with MeBIO (2 μM) or BIO (2 μM). Whole-cell lysates (20 μg) were probed with antibodies as indicated.

As GSK3β is directly regulated by AKT1, its nuclear localization following decreased PI3K/AKT1 signaling could be accounted for by the loss of S9 phosphorylation and by its catalytic activation. S9 regulation of GSK3β and its enzymatic activation do not appear to be critical for shuttling or nuclear localization, however, because the GSK3β^S9A mutant shuttles normally (see also Fig. S8 in the supplemental material), and inhibition of GSK3β with BIO/GSK3 inhibitor IX has no effect on its subcellular localization or ability to shuttle (data not shown). The subcellular localization of GSK3β is therefore controlled by AKT1 but independently of regulation through S9 (see Discussion).

The addition of the GSK3 inhibitor BIO, but not its inactive analogue MeBIO, blocked downregulation of c-myc levels and T58 phosphorylation following the addition of LY294002 (Fig. 2B and C) but had no impact on LY294002-dependent relocalization to the nucleus (data not shown). This finding indicates that although the loss of PI3K/AKT1 signaling is required for GSK3β nuclear accumulation, GSK3β activity is not required. These results show that PI3K/AKT1 is required to maintain GSK3β in a cytoplasmic, inactive state in self-renewing mESCs and implies a mechanism for how c-myc levels are regulated in mESCs and during early differentiation.

AKT1 controls nuclear export of GSK3β in mESCs.

Our results so far indicate that the loss of PI3K/AKT1 activity is sufficient to promote the nuclear accumulation of active GSK3β, elevate c-myc phosphorylation on T58, and accelerate the turnover of c-myc protein (4). Collectively, these events are sufficient to trigger mESC differentiation, and their close temporal coordination suggests a mechanism where PI3K/AKT1, GSK3β, and c-myc activities are linked by a common pathway. To further establish that AKT1 activity blocks nuclear accumulation of active GSK3β, we employed a cell line expressing a constitutively active, myristoylated version of AKT1 fused to the steroid binding domain of the estrogen receptor (ER; myr.AKT1-ER). The addition of 4-hydroxytomoxifen (4OHT) to this cell line then allowed us to investigate the role of AKT1 in GSK3β regulation. Previously, Watanabe and colleagues (28) showed that when this cell line was treated with 4OHT, the differentiation of mESCs was severely delayed following LIF withdrawal, but the mechanism underpinning the ability of sustained AKT1 activity to maintain mESC self-renewal was not addressed.

To investigate the link between AKT1 activity, GSK3β subcellular localization, and c-myc regulation, we evaluated the status of GSK3β and c-myc in the absence of LIF in the presence or absence of 4OHT (Fig. 3A). In the absence of 4OHT (myr.AKT1-ER off), cells grown in the absence of LIF downregulated c-myc protein and displayed a corresponding increase in T58 phosphorylation (Fig. 3B). Furthermore, GSK3β accumulated in the nucleus in an S9-hypophosphorylated state, T58-phosphorylated c-myc accumulated in nucleolar speckles, and Nanog levels declined (Fig. 3C through F). All of these are signatures of early mESC differentiation, as observed previously in R1 mESCs (Fig. 1). In the presence of 4OHT (myr.AKT1-ER on), however, cells failed to downregulate c-myc, T58 phosphorylation levels remained low, and GSK3β S9-phosphorylation levels remained elevated (Fig. 3B). At the cellular level, GSK3β remained in the cytosol in a S9-phosphorylated state and Nanog remained elevated in the nucleus, but T58 staining of nucleolar speckles was absent (Fig. 3B through F). Throughout these experiments, enhanced green fluorescent protein (eGFP) fluorescence was used to confirm that cells were carrying the myr.AKT1-ER construct, as previously described (28). These results demonstrate that in the absence of LIF, mESCs fail to differentiate normally when AKT1 activity is maintained and retain a dome-shaped colony morphology, elevated Nanog and c-myc levels, and a low level of c-myc T58 phosphorylation (Fig. 3B through F). Significantly, sustained AKT1 activity in the absence of LIF blocked the nuclear accumulation of active GSK3β. Consequently, these cells maintain a large pool of stable Myc protein in the nucleus to support self-renewal.

FIG. 3.

Ectopic AKT1 activity blocks the activation and nuclear accumulation of GSK3β and c-myc T58 phosphorylation following LIF withdrawal. (A) Experimental scheme for panels B through F, involving a mESC line expressing a myristoylated form of AKT1 fused to the steroid binding domain of the ER (myr.AKT1-ER). The construct introduced expresses eGFP from an internal ribosome entry site (IRES) linked to the myr.AKT-ER cassette (28). (B) Cell lysates from mESCs (plus LIF) or cells cultured in the absence of LIF with (+) or without (−) 4OHT (1 μM) for 1 to 4 days (see panel A) were resolved by polyacrylamide gel electrophoresis and probed as indicated. (C through F) Levels of myr.AKT1-ER and its activity status were determined with antibodies for pan-AKT1 and AKT1^pS473, respectively. mESCs (plus LIF) and cells cultured in the absence of LIF with or without 4OHT were probed with antibodies as indicated. Expression of myr.AKT1-ER was established by eGFP fluorescence from a linked IRES.

One caveat to the interpretation of these results is that GSK3β remained inactive in the cytoplasm because mESCs failed to differentiate due to maintained AKT1 activity. This would be consistent with AKT1 maintaining self-renewal but regulating GSK3β localization indirectly. To address the question of whether AKT1 activity or differentiation per se was responsible for GSK3β localization, a separate approach was taken. In the following experiments, mESCs were allowed to differentiate for 3 days following the withdrawal of LIF. Then AKT1 was reactivated (myr.AKT1-ER on) by the addition of 4OHT, and GSK3β localization was assayed after a further 24 h (Fig. 4A). As expected, GSK3β accumulated in the nucleus in an active, S9-hypophosphorylated state following LIF withdrawal, with elevated levels of T58-phosphorylated c-myc (Fig. 4B through D). When myr.AKT1-ER was reactivated by the addition of 4OHT at day 3, GSK3β relocalized to the cytoplasm in a S9-phosphorylated, inactive state (Fig. 4B and C). The residual phosphorylation of c-myc on T58 was extinguished upon the reestablishment of elevated AKT1 activity and reduced GSK3β activity (Fig. 4D). To establish that cytoplasmic accumulation of GSK3β under these conditions was due to nuclear export and not due to the accumulation of newly synthesized protein in the cytosol, LB was added to block the nuclear export pathway. This analysis unequivocally showed that LB treatment blocked the accumulation of GSK3β in the cytoplasm, indicating that the reestablishment of AKT1 activity in differentiating mESCs was sufficient for GSK3β inactivation and for its nuclear export (Fig. 4E). AKT1 is therefore required for the nuclear export of GSK3β in mESCs and appears to promote self-renewal by regulating GSK3β at multiple levels.

FIG. 4.

Nuclear GSK3β can be redirected from the nucleus to the cytoplasm of differentiating mESCs by reactivation of AKT1. (A) Experimental scheme for panels B through E, where myr.AKT1-ER was activated in E14Tg2a mESCs by the addition of 4OHT (1 μM) for 24 h following 3 days of differentiation (−LIF) and relocalization of GSK3β to the nucleus. (B through D) After 4 days (4d), cells grown in the absence of LIF with or without 4OHT were probed as indicated. eGFP fluorescence (GFP) marks the positions of myr.AKT1-ER-expressing cells. (E) mESCs expressing myr.AKT1-ER were analyzed as for panel B except that LB was added 6 h before fixation.

GSK3β regulates T58 phosphorylation of c-myc following loss of AKT1 activity.

Next, we set out to formally establish if GSK3β was required for the degradation and phosphorylation of c-myc following the collapse of PI3K/AKT1 activity. This follows work presented earlier in this report in which the inhibition of PI3K activity with LY294002 was shown to promote the accumulation of active, nuclear GSK3β and to elevate levels of c-myc phosphorylation on T58 (Fig. 2). Regulation of T58 phosphorylation was a focus, since we previously showed that collapse of c-myc protein levels during early differentiation occurs through a T58-dependent mechanism and that a T58A mutant form of c-myc can maintain self-renewal in the absence of LIF (4). This question was addressed for wild-type (WT) and GSK3α^−/− GSK3β^−/− (double knockout [DKO]) E14 mESCs (8) by evaluating the T58 phosphorylation status of c-myc under conditions of high and low levels of PI3K/AKT activity. In unperturbed, WT E14 mESCs, GSK3β is inactive in the cytoplasm and c-myc is hypophosphorylated on T58 (Fig. 5A). Upon treatment with LY294002, GSK3β localizes to the nucleus, coinciding with the elevation of c-myc T58 phosphorylation. When BIO is added in combination with LY294002, however, GSK3β still accumulates in the nucleus, but c-myc T58 phosphorylation remains low (Fig. 5A). These data are consistent with earlier results obtained in R1 mESCs that linked the loss of AKT1 activity to GSK3β nuclear localization and c-myc T58 phosphorylation (Fig. 1; see also Fig. S2 in the supplemental material). When this experiment was repeated in an isogenic DKO E14 mESC line, c-myc was maintained in an unphosphorylated state, even in the presence of LY294002 (Fig. 5B). Transient expression of GSK3β in DKO cells restored cytoplasmic staining, and when treated with LY294002, GSK3β localized to the nucleus (Fig. 5C). Increased T58 phosphorylation of c-myc accompanied the nuclear accumulation of GSK3β in these experiments.

FIG. 5.

GSK3β is required for phosphorylation of c-myc on T58, following its entry into the nucleus and collapse of AKT1 activity. (A) Untreated (−) E14 mESCs (plus LIF) were treated with LY294002 (40 μM) or LY294002 (40 μM) plus BIO (2 μM) for 24 h and were then subjected to immunostaining as indicated. Images were captured by confocal microscopy. Scale bar, 50 μm. (B) GSK3α^−/− GSK3β^−/− E14 mESCs were treated as for panel A with LY294002 and were probed with antibodies as indicated. (C) GSK3α^−/− GSK3β^−/− E14k mESCs (as for panel B) were transiently transfected with a GSK3β_HA expression plasmid. Where indicated, cells were then treated with LY294002 (40 μM; 2 days posttransfection). Three days posttransfection, cells were probed with antibodies for HA and c-myc^pT58, and staining was visualized by confocal microscopy. Scale bar, 50 μm. (D) WT E14k mESCs, E14k GSK3α^−/− GSK3β^−/− DKO mESCs, and DKO mESCs transfected with a GSK3β_HA expression plasmid (as for panel C) were treated as indicated: untreated (−), treated with 40 μM LY294002 (+LY), or treated with 2 μM BIO (+BIO). Inhibitors were added 2 days posttransfection. Cell lysates were analyzed by immunoblot analysis following polyacrylamide gel electrophoresis and probed with antibodies as indicated.

Immunoblot analysis was used to corroborate these observations (Fig. 5D). As expected, inhibition of PI3K in WT E14 mESCs decreased S9 phosphorylation of GSK3β but increased phosphorylation of c-myc on T58, the latter of which was blocked by the addition of BIO. In the untreated E14 DKO line, GSK3β levels were absent and c-myc T58 phosphorylation was low. Restoration of GSK3β activity following transfection reestablished the sensitivity of S9 (GSK3β) and T58 (c-myc) phosphorylation to LY294002, as seen in WT E14 mESCs. Elevated levels of c-myc T58 phosphorylation in the presence of LY294002 was once again blocked by BIO. Together, these results demonstrate that GSK3β is responsible for c-myc phosphorylation in mESCs when PI3K/AKT signaling declines. Nuclear localization of GSK3β and its ability to target Myc is therefore likely to be a critical determinant of differentiation commitment.

GSK3β interferes with mESC maintenance by targeting c-myc.

Although nuclear accumulation of GSK3β requires the downregulation of PI3K/AKT1 activity following LIF withdrawal and results in increased phosphorylation of c-myc on T58, the biological consequences were not defined by these experiments. This question was addressed by evaluating the effects of a nucleus-localized, constitutively active mutant form of GSK3β on self-renewal. This was approached by generating an R1 mESC cell line expressing an HA-tagged, constitutively active GSK3β mutant (serine 9 to alanine substitution [S9A]) with two SV40 nuclear localization signals (GSK3β^S9A.NLS) under the control of Cre recombinase (Fig. 6A). Cre recombinase expression, following transient transfection, was monitored by the loss of eGFP-positive (eGFP+) immunostaining after 3 days. Nanog expression was used as the readout for mESC maintenance in this assay. In eGFP⁺ Nanog⁺ cells (−Cre cells), GSK3β^S9A.NLS was not expressed, as indicated by the absence of HA immunostaining (Fig. 6A). Cre-mediated excision, however, monitored by the loss of eGFP fluorescence, established conditions where ∼85% of eGFP⁻ cells (+Cre cells) expressed nuclear GSK3β^S9A.NLS. Only ∼10% of eGFP⁻ HA⁺ cells retained Nanog expression, compared to 90% in eGFP⁺ HA⁻ cells (Fig. 6A and B). Under similar conditions, cells expressing vector alone (Fig. 6A and B) retained levels of Nanog comparable to that in untransfected cells. Using Nanog as a readout, these data indicate that enforced localization of constitutively active GSK3β has a clear effect on mESC maintenance.

FIG. 6.

Enforced nuclear localization of active GSK3β disrupts self-renewal and promotes differentiation of mESCs. (A) R1 mESCS were transfected to generate a stable cell line carrying a cassette flanked by loxP sites (denoted by solid triangles) that included the eGFP gene driven by the CAG1 promoter and an IRES linked to a puromycin resistance gene (puro^R). Upon Cre expression in this line, the eGFP/puro^R cassette was excised, allowing for the expression of a constitutively active mutant (S9A) form of GSK3β linked to two concatenated SV40 nuclear localization signals and three concatenated HA tags (GSK3β^S9A.NLS). The GSK3β^S9A.NLS line was transiently transfected with an empty vector (−Cre) or the same vector carrying the gene for Cre recombinase (+Cre). After 3 days, cells were evaluated by eGFP fluorescence and immunostaining for Nanog and HA. (B) Cell lines generated by transfection with pCAGipuro (vector; top panel) or GSK3β^S9A.NLS (lower panel) were transfected with an empty expression construct (−) or the same construct expressing Cre recombinase (+) as described for panel A. The percentages of Nanog- and HA-positive (+ve) cells were scored, and the data were graphed to establish the effects of conditional GSK3β^S9A.NLS expression. At least 500 cells were scored under each of the experimental conditions. (C) R1 mESCs were transiently transfected with vector alone (pCAGipuro) or the equivalent vector expressing GSK3β^S9A.NLS in the presence or absence of 2 μM BIO. Twenty-four hours posttransfection, cells were selected with 1 μg/ml puromycin for 4 days. Colonies were then stained with Leishman reagent to identify undifferentiated mESC colonies (left panels). In parallel, colonies were analyzed by bright field microscopy (right panels). Scale bar, 50 μm. Cells (1 × 10⁴/cm²) were plated onto each dish, in each case generating equivalent numbers of colonies (±10%). (D) R1 mESCs were selected with puromycin (as described for panel C) following transfection with vector (pCAGipuro), WT GSK3β (WT), a GSK3β^S9A mutant (S9A), GSK3β^S9A.NLS (S9A.NLS), or GSK3β^S9A.NLS in the presence of 2 μM BIO (S9A.NLS + BIO). More than 100 colonies were scored under each condition. For a colony to be designated “stained” in the Leishman assay, >90% of cells within the colony were required to stain strongly. Colonies scored as “unstained” in this assay consisted of <10% of the stained cells. (E) Quantitative reverse transcriptase PCR analysis of Oct4 transcript levels in R1 mESCs transfected with vector only or GSK3β^S9A.NLS (as for panels C and D) following 4 days of puromycin selection. Transcript levels were normalized to a glyceraldehyde-3-phosphate dehydrogenase (GAPDH) control and were determined in three biologically independent experiments. Error bars represent ± the standard errors of the means.

The previous observations suggested that nuclear GSK3β interferes with mESC self-renewal and promotes differentiation. To obtain further evidence for this, R1 mESCs were transfected with vector alone, GSK3β, or GSK3β^S9A or GSK3β^S9A.NLS constructs. Transfectants were then selected with puromycin over 4 days, followed by staining with Leishman reagent (21) (Fig. 6C and D). Cells transfected with vector alone, GSK3β, or GSK3β^S9A or GSK3β^S9A.NLS constructs generated cultures where >80% of the colonies stained strongly in the Leishman assay and maintained a typical three-dimensional dome-shaped colony morphology (Fig. 6C and D). The remaining 20% of colonies only partially stained with Leishman reagent and exhibited an intermediate (flat/domed) colony morphology. However, only ∼3% of colonies generated by transfection with GSK3β^S9A.NLS stained in this assay. This corresponded to a general disruption of colony architecture, with most cells exhibiting a flattened morphology reminiscent of differentiating primitive endoderm cells that form following LIF withdrawal (15) and loss of Myc function (4). Most of the flattened colonies consisted of less than 5% Leishman-stained cells. When BIO was added to GSK3β^S9A.NLS-transfected cultures at 24 h posttransfection, the percentage of Leishman-stained cells was restored to almost 90% and a typical dome-shaped colony morphology was retained (Fig. 6C and D). Expression of GSK3β^S9A.NLS also caused a major decrease in Oct4 transcript levels (Fig. 6E) and increased T and Cdx2 transcript levels, and ∼50% of cells stained positive for the endoderm marker FoxA2 (see Fig. S9 in the supplemental material).Together, these data demonstrate that a biological consequence of active, nucleus-localized GSK3β is the promotion of mESC differentiation. A mechanism for this was suggested by an increase in c-myc T58 phosphorylation and turnover at this time (this report and reference 4).

To establish if nuclear GSK3β triggers mESC differentiation by targeting c-myc, we repeated the previous experiment in a cell line expressing a stable, mutant form of c-myc (T58A) fused to the hormone binding domain of the ER (Fig. 7A). Previously, we showed that this stable form of Myc (c-myc^T58A-ER) under the control of 4OHT can promote self-renewal in the absence of LIF (4). Critical to this was the ability of c-myc to avoid phosphorylation on T58, thereby retaining its comparatively long t_1/2 (∼105 min) (4). In this report, we have formally established that in mESCs, the regulation of c-myc through T58 is controlled by GSK3β. To address the possibility that GSK3β antagonizes self-renewal by targeting Myc, we asked if ectopic c-myc activity could rescue the differentiation-promoting effects of GSK3β^S9A.NLS. If c-myc is the main target for active, nuclear GSK3β following LIF withdrawal or loss of PI3K/AKT signaling, ectopic expression of c-myc^T58A-ER should circumvent the effects of GSK3β and cells should retain the capacity to self-renew, even over short periods of time. Transfection of vector alone, with or without 4OHT, was unable to block differentiation caused by GSK3β^S9A.NLS, as determined by staining with Leishman reagent (<5% stained colonies) and colony morphology (Fig. 7B and C). As previously noted, in the presence of activated nuclear GSK3β, cells adopt a morphology that is typical of a primitive endoderm (Fig. 7C), similar to that seen following Myc inactivation or LIF withdrawal in mESCs (4). However, when the c-myc^T58A-ER fusion was activated by 4OHT, a typical dome-shaped colony morphology was maintained and cells retained uniform staining with Leishman reagent in the presence of GSK3β^S9A.NLS (Fig. 7B and C). These data show that if Myc activity is maintained in mESCs, the ability of GSK3β to trigger differentiation is severely compromised. In conclusion, c-myc is a major target of GSK3β when it enters the nucleus following loss of PI3K/AKT1 signaling. The ability of c-myc to be targeted by GSK3β following its accumulation in the nucleus is shown to be part of a pathway that impacts on mESC fate determination following the loss of PI3K/AKT1 signaling. A model summarizing our findings is shown in Fig. 7D.

FIG. 7.

Effects of GSK3β on mESC self-renewal can be blocked by enforced expression of a c-myc^T58A mutant. (A) The c-myc^T58A-ER expression plasmid or its corresponding empty vector (pCAGiNeo) was used for the generation of two cell lines. A second construct in pCAGipuro that expresses GSK3β^S9A.NLS was engineered into these lines. neo^R, neomycin resistant; puro^R, puromycin resistant. (B) A stable neomycin-resistant R1 mESC cell line was generated by transfection with a c-myc^T58A-ER expression construct (see legend for panel A). This cell line was transiently transfected with a GSK3β^S9A.NLS expression construct (see legend for panel A) and selected for 4 days in the presence of 1 μg/ml puromycin in the presence (+) or absence (−) of 5 nM 4OHT. Colonies were then scored for Leishman staining: colonies were scored as “undifferentiated” if they were composed of >90% stained cells. Error bars, ± standard errors of the means. (C) Morphology of pCAGipuro (vector) or c-myc^T58A-ER colonies following transfection with a GSK3β^S9A.NLS expression construct and puromycin selection (1 μg/ml, 4 days) in the presence or absence of 4OHT (as described for panel B). Scale bar, 50 μm. (D) Model depicting the central findings of this report, describing roles for PI3K/AKT in the regulation of GSK3β activity and subcellular localization in mESCs and how GSK3β promotes differentiation when it localizes to the nucleus by targeting c-myc.

DISCUSSION

Prior to this report, there were major gaps in our knowledge relating to the mechanisms of mESC self-renewal and early cell fate commitment. Although PI3K/AKT1 and GSK3β were previously known to be critical determinants of these cell fate decisions, their mechanisms of action in mESCs was poorly understood. Our work defines novel roles for PI3K/AKT signaling through GSK3β and shows that this pathway converges on key nuclear targets such as Myc.

GSK3β shuttling and its regulation by AKT1.

Although GSK3β is generally considered to be a cytoplasmic protein, there are numerous examples where it has been shown to partially or transiently accumulate in the nucleus, for example, during apoptosis (3), replicative senescence (33), and the S phase of the cell cycle (7). Once in the nucleus and depending on cell type, GSK3β can phosphorylate substrates such as cyclin D (7), NFAT (2), and c-myc (31). The GSK3β import mechanism appears to be dependent on a bipartite nuclear localization signal (NLS) (13), while export under some circumstances requires FRAT (9). Only one report (9) has described the nuclear-cytoplasmic shuttling of GSK3β, and so the generality of this observation was previously unclear. We have investigated this in 14 primary and transformed cell lines of murine and human origin. While the relative distribution of GSK3β between the nucleus and cytoplasm varies in different cell lines, its localization is dynamic in all cases, involving continual shuttling between the nucleus and cytoplasm (M. Bechard and S. Dalton, unpublished data). Nuclear-cytoplasmic shuttling of GSK3β is therefore a general phenomenon that has major biological implications in a broad range of cell types.

Our results show that AKT1 activity is the principal determinant of GSK3β localization and activity in pluripotent cells. Under self-renewing conditions, when PI3K signaling is active nuclear GSK3β is rapidly exported back into the cytoplasm by an AKT1-dependent mechanism. This finding is supported by several lines of evidence. First, LB-dependent trapping of GSK3β in the nucleus is reproduced by the inhibition of PI3K and AKT1. Second, under conditions where GSK3β accumulates in the nucleus during differentiation, reestablishment of elevated AKT1 activity restores the GSK3β nuclear export pathway. Although AKT1 directly regulates GSK3β through S9, GSK3β activity does not seem to be directly linked to the import-export mechanism, since shuttling continues in the presence of BIO and a GSK3β^S9A mutant shuttles normally.

Previous work indicated that PI3K/AKT1 signaling is central to the maintenance of murine, monkey, and human ESCs (19, 24, 28), but the mechanism for this has not been previously defined in any detail. Our work addresses this by showing that AKT1 regulates GSK3β activity by at least two nonredundant mechanisms: first, by S9-dependent regulation of catalytic activity, and second, by regulation of nuclear export. Since GSK3β has no definable nuclear export signal, it is likely that chaperone proteins are involved and are themselves subject to regulation by AKT1. Candidate chaperone molecules include FRAT (9, 10), 14-3-3 (1), and Axin (6, 10, 30). These proteins are known to form complexes with GSK3β and, in the cases of 14-3-3 and Axin, are known to shuttle in and out of the nucleus. Whether the shuttling of these proteins is dependent on PI3K/AKT1 is not known and will be the subject of further investigation. The involvement of PI3K/AKT1/GSK3β signaling in a wide range of biological processes suggests that these findings have major implications for development, cell cycle control, apoptosis, replicative senescence, tissue homeostasis, and cancer.

GSK3β disrupts mESC maintenance by targeting Myc.

The accumulation of active GSK3β in the nucleus following the loss of PI3K/AKT1 activity disrupts mESC maintenance (this report). We show that a major target for nuclear GSK3β is c-myc, a transcription factor that plays major roles in maintaining and establishing the pluripotent state (17, 25, 29). Nmyc, a functionally redundant member of the Myc family, is also known to be targeted by GSK3β in other cell types (31) and is likely to be subject to similar regulation in mESCs. The connection between GSK3β and c-myc was first suggested by us in 2005 (4), following observations that the c-myc^T58A mutant, which evades GSK3β, sustains mESCs in the absence of LIF. Additional reports have highlighted the importance of suppressing GSK3β activity in order to maintain ESC self-renewal (20, 32) and its requirement for normal differentiation (8).

How GSK3β antagonizes self-renewal was not previously understood, but in this report we demonstrate that one of its key functions is to target c-myc. Understanding the exact role of Myc transcription factors in ESC maintenance will be guided by the identification of target genes. Recently, a large cohort of Myc target genes have been identified in mESCs, many of which seem to be involved in cell cycle control, metabolic activity, and epigenetic regulation (5, 11). Another potential target of GSK3β regulation could be Nanog. Storm and coworkers (23) have shown that the inhibition of PI3K activity reduces Nanog protein levels, suggesting that it could be subject to regulation by GSK3β once it enters the nucleus (Fig. 7).

The roles of PI3K/AKT1 and GSK3β in regulating mESC maintenance.

In a broad context, our results provide a mechanistic rationale for several key reports documenting the effects of GSK3β on human and mouse ESC self-renewal. For example, it was previously unclear why suppression of GSK3β activity is critical for self-renewal (20, 32) and improved mESC derivation (27) and why GSK3β activation is an important biochemical event required for early mESC fate commitment (8). This report answers these questions by showing that upon LIF withdrawal, GSK3β accumulates in the nucleus, where it gains access to substrates such as c-myc. Our model (Fig. 7D) indicates that GSK3β needs to be suppressed in mESCs so that key nuclear factors such as c-myc can be maintained at levels compatible with their roles in self-renewal. These results explain the observations of Sato et al. (20) and Ying et al. (32), who showed that suppression of GSK3β activity is a prerequisite for ESC maintenance. The inability of GSK3β to access nuclear substrates is also consistent with our previous observations that c-myc has an unusually long t_1/2 in mESCs (∼105 min) (4). Our results also address the observations of Doble and coworkers (8), who showed that genetic inactivation of GSK3α/β severely compromises the ability of ESCs to differentiate. It is unclear, however, if GSK3 inactivation is sufficient to maintain self-renewal in the absence of LIF. There is evidence that in short-term experiments, inhibition of GSK3 can maintain many mESC characteristics (20). There is reason to suspect, however, that the loss of GSK3 activity per se will not substitute for the loss of LIF or PI3K/AKT1 signaling in mESCs. For example, c-myc-dependent self-renewal requires multiple signaling inputs. LIF/STAT3 signaling is required for transcriptional regulation of the c-MYC gene, and PI3K/AKT1 is required to suppress GSK3 activity, which results in stabilization of Myc protein. Together, these multiple inputs are critical for self-renewal but are not sufficient by themselves for self-renewal (4, 29).

Several reports have documented a role for PI3K in the rapid proliferation, survival, and maintenance of mESCs (24), but it is unclear if and how the AKT1/GSK3/c-myc signaling axis impacts these properties. We have not observed any dramatic differences in proliferation rates between WT mESCs or those expressing c-myc^T58A or myr-AKT1 or in cells deficient for GSK3α/β. This can be explained, however, by the already elevated levels of AKT1 and c-myc in mESCs and by low steady-state levels of GSK3 activity. It will be important to establish if this signaling axis is important in the maintenance of other pluripotent cell types such as EpiSCs, iPSCs, and human ESCs.

In summary, we have established a mechanism to explain why PI3K/AKT1 is important for mESC self-renewal, why and how GSK3β activity is suppressed in mESCs, and why GSK3β activation is a critical event during early differentiation. Finally, we have defined Myc as a nuclear target of this pathway and provided an explanation for how Myc is regulated in self-renewal and early differentiation. In total, this work describes a pathway required for early differentiation commitment. We believe that control of GSK3β localization and enzymatic activity in mESCs, each controlled independently by PI3K/AKT1, is likely to reflect a general mechanism applicable to the regulation of GSK3β and Myc in a wide range of cell types. This has profound implications for development, cancer, and the regulation of other stem cell compartments in vivo where Myc has well-defined roles (18).