Regulation of autophagic cell death by glycogen synthase kinase-3β in adult hippocampal neural stem cells following insulin withdrawal

In our previous reports, we demonstrated that HCN cells undergo a genuine ACD without signs of apoptosis upon insulin withdrawal [4,6]. Of note, HCN cells are subject to apoptosis in response to prototypical apoptosis inducers, such as staurosporine (STS). These findings indicate that insulin-deprived HCN cells adopt ACD as the primary mode of cell death despite their intact apoptotic capability. To confirm the non-apoptotic nature of HCN cell death induced by insulin withdrawal, insulin-deprived HCN cells were treated with a pan-caspase inhibitor Z-VAD.fmk. The insulin-withdrawn and insulin-containing conditions are denoted as I(−) and I(+), respectively, throughout this report. Caspase activation was not observed in HCN cells cultured in either condition. Consistent with this observation, Z-VAD.fmk failed to protect HCN cells from insulin withdrawal (Figure 1A). In sharp contrast to this ineffectiveness against insulin withdrawal, Z-VAD.fmk efficiently blocked apoptotic cell death induced by etoposide and STS in a dose-dependent manner (Figure 1A). Inhibition of caspase-3 activation by Z-VAD.fmk was confirmed by Western blotting analysis in apoptotic cells treated with STS (Figure 1B). The normal induction of apoptosis by prototype apoptotic inducers verified the intact apoptotic capability of HCN cells.

GSK-3β expression levels are particularly high in the brain [19], and broadly implicated in neuronal cell death and neurodegeneration [27-29]. Additionally, GSK-3β activation promotes apoptosis in various stress conditions, and over-expression of GSK-3β leads to neuronal apoptosis [29]. On the contrary, GSK-3β is inhibited by prosurival growth factors including insulin. Therefore, the pro-survival and proliferative effects of insulin signaling seem, at least in part, to be mediated through suppression of GSK-3β. Based on these findings, we hypothesized that GSK-3β may positively regulate the induction of autophagy and subsequent cell death in HCN cells following insulin withdrawal. Thus, an increase in GSK-3β activity should result in higher levels of autophagic flux and ACD. On the other hand, the opposite should occur when GSK-3β is decreased by pharmacological or genetic means. Herein, we addressed our hypotheisis by modulating GSK-3β activity in insulin-deprived HCN cells.

First, to examine whether GSK-3β was activated following insulin withdrawal, we monitored the level of inhibitory phosphorylation at the GSK-3β serine 9 residue by Western blotting analysis. Stimulation of cells with insulin causes inactivation of GSK-3 via a phosphatidylinositol 3 kinase (PI3K)-dependent mechanism. PI3K-induced activation of protein kinase B (PKB; also known as AKT) phosphorylates serine 9 on GSK-3β and serine 21 on GSK-3α [26,30]. Phosphorylation of these residues inhibits GSK-3 activity. As expected, AKT was inactivated and phosphorylation levels of GSK-3β (Ser9) decreased in insulin-deprived HCN cells, indicating activation of GSK-3β (Figure 2A, B). As an additional indicator of GSK-3β activity, we also monitored the protein level of β-catenin, which is phosphorylated by GSK-3β and subsequently degraded through the ubiquitin-proteasome system [31]. As such, the β-catenin level was substantially decreased in insulin-deprived HCN cells (Figure 2A, B). Consistent with our previous reports, insulin withdrawal caused a significant increase in the level of the type II of microtubule-associated protein light chain 3 (LC3-II), the most reliable biochemical marker of autophagy (Figure 2A, B) [32]. In combination with our earlier work demonstrating the prevention of cell death by Atg7 knockdown in HCN cells, these data revealed that ACD was the primary cell death mechanism in insulin-deprived HCN cells, which manifested increased levels of autophagy and GSK-3β activation.

Given the role of GSK-3β in neuronal apoptosis, autophagy, and insulin signaling, we postulated that GSK-3β may regulate autophagy induction and cell death in HCN cells following insulin withdrawal. To test this, we used 6-bromoindirubin-3'-oxime (BIO), a specific GSK-3 inhibitor [33] which effectively prevented insulin withdrawal-induced cell death at a concentration of 0.25 μM as assessed by PI/Hoechst staining (Figure 3A). BIO strongly suppressed autophagy induction, as the increased level of LC3-II was lowered to the level comparable with the insulin-containing condition (Figure 3B, C) and the high number of the GFP-LC3 puncta observed in I(−) condition was significantly decreased (Figure 3D, E). Inhibition of GSK-3β activity was confirmed by the significant accumulation of β-catenin (Figure 3B, C). BIO itself had no effect on cell viability under the insulin-containing condition (data not shown).

The pharmacologic inhibition of GSK-3 should affect both GSK-3α and -3β equally. To distinguish the relative contribution of the two isoforms to HCN cell death and demonstrate the essential role of GSK-3β in the regulation of ACD, we knocked down each isoform using specific small interfering RNAs (siRNAs; Figure 4A). In accordance with the pharmacological inhibition of GSK-3, GSK-3β siRNA attenuated insulin withdrawal-induced cell death (Figure 4B) and autophagy induction, as shown by a decrease in LC3-II level (Figure 4C, D). Of note, the GSK-3α-targeting siRNA yielded a marginal level of protection (Figure 4B), and double knockdown of both isoforms of GSK-3 led to a similar degree of the protective effect as observed for the GSK-3β single knockdown (Figure 4). These data supported our hypothesis regarding the critical role of GSK-3β in the regulation of ACD in HCN cells following insulin withdrawal.

To further establish the role of GSK-3β in the regulation of ACD in insulin-deprived HCN cells, we transfected HCN cells with the GSK-3β constructs encoding the WT and CA mutant forms using Lipofectamine method (Figure 5A). The CA form of GSK-3β was generated by mutation of the serine 9 residue to an alanine residue. This serine 9 residue is a key regulatory site for the inactivation of GSK-3β, and mutation of this residue to alanine prevents the inhibitory phosphorylation of the serine 9 residue, thereby avoiding enzyme inactivation [34,35]. Ectopic expression of the constructs was confirmed by Western blotting analysis with an antibody against the hemagglutinin (HA) epitope tag (Figure 5C). Over-expression of the GSK-3β WT form significantly increased cell death in insulin-deprived HCN cells (Figure 5B). In particular, the over-expression of GSK-3β CA mutant increased cell death to a greater extent than the WT form. Both the GSK-3β WT and CA forms elevated the level of LC3-II autophagy marker with more profound effects caused by the GSK-3β CA mutant form (Figure 5C, D). However, in spite of this marked increase in cell death, caspase-3 activation was not detected (Figure 5C). Furthermore, when apoptosis was monitored by Annexin V dye, which binds to the phosphatidylserine exposed on the outer leaflet of apoptotic cells, the low percentage of Annexin V-positive cells remained unaltered in the cells over-expressing the GSK-3β CA form, as well as the WT form, under insulin withdrawal condition for 24 h (Figure 5E). A high percentage of cells were positively stained by Annexin V after STS treatment for 6 h, verifying the accuracy of the assay. To confirm the autophagic nature of the GSK-3β-induced cell death, Atg7 was knocked down by infection of Atg7 shRNA-expressing lentivirus. Knockdown of Atg7 significantly attenuated cell death induced by over-expresion of GSK-3β CA (Figure 5F, G). Therefore, a lack of apoptosis indices despite a robust increase in LC3-II levels and cell death rates by GSK-3β provided further support for our hypothesis that GSK-3β was a critical upstream regulator of ACD in insulin-deprived HCN cells.

In contrast to the absence of apoptosis, the autophagic flux rate was significantly boosted by high levels of GSK-3β activity. The furtherance of autophagic flux by the GSK-3β CA form was examined by analyses of the LC3-II protein levels and the GFP-LC3 puncta numbers in combination with the autophagic flux blocker, Bafilomycin A1 (BafA). An impairment of autophagy flux, as well as the induction of autophagy, which are opposite incidences in terms of autophagosome production and clearance, can lead to seemingly similar accumulations of LC3-II protein or LC3 puncta. Therefore, an increase in LC3-II levels or LC3 puncta numbers does not necessarily correlate with increased autophagic flux, and caution is required for the interpretation of results. A comparison between the steady-state levels of autophagosomes (before treatment with autophagic flux blockers, such as autophagosome-lysosome fusion inhibitors or intralysosomal protease inhibitors) and the accumulated levels of autophagosomes (after inhibitor treatment) has been recommended as a result. A high rate of autophagic flux would result in significantly different levels of autophagy markers between the steady state and accumulated autophagosome conditions, while marginal differences may denote an impaired autophagic flux [36]. An increase in the rate of autophagosome formation and clearance in insulin-deprived HCN cells was previously shown, demonstrating an increase of autophagic flux [4,6]. To measure the alteration in autophagic flux following over-expression of the GSK-3β CA form in insulin-deprived HCN cells, we used BafA which blocks the fusion between the autophagosome and lysosomes, preventing autophagic flux [37]. Ectopic expression of the GSK-3β CA form in I(−) led to greater increases in LC3-II level than insulin withdrawal alone (Figures 5C and 6B), and BafA treatment further boosted the increase in the LC3-II level induced by the GSK-3β CA form (Figure 6B). These results indicated that the GSK-3β CA mutant form accelerated autophagic flux more in insulin-deprived HCN cells, which already exhibited high levels of autophagic flux. An additional measure for assessing autophagic flux is the number of GFP-LC3 puncta. Autophagy activation induces a change in the GFP-tagged LC3 pattern from the diffused signal in the cytosol to a puncta/dot pattern following enrichment of LC3 in autophagosome [38]. Consistent with the notion of elevated autophagic activity, an increase in the number of GFP-LC3 puncta in the GSK-3β CA form-expressing HCN cells was facilitated to a greater extent by BafA treatment under the insulin-deficient condition (Figure 6C-D). Collectively, these data provided important evidence suggesting that GSK-3β may be the key regulator of ACD following insulin withdrawal in HCN cells.