GSK-3 inactivation or depletion promotes β-cell replication via downregulation of the CDK inhibitor, p27 (Kip1)

Abstract

Diabetes (T1DM and T2DM) is characterized by a deficit in β-cell mass. A broader understanding of human β-cell replication mechanism is thus important to increase β-cell proliferation for future therapeutic interventions. Here, we show that p27 (Kip1), a CDK inhibitor, is expressed abundantly in isolated adult human islets and interacts with various positive cell cycle regulatory proteins including D-type cyclins (D1, D2 and D3) and their kinase partners, CDK4 and CDK6. Also, we see interaction of cyclin E and its kinase partner, CDK2, with p27 suggesting a critical role of p27 as a negative cell cycle regulator in human islets. Our data demonstrate interaction of p27 with GSK-3 in β-cells and show, employing rodent β-cells (INS-1), isolated human islets and purified β-cells derived from human islets, that siRNA-mediated depletion of GSK-3 or p27 or 1-AKP/BIO-mediated GSK-3 inhibition results in increased β-cell proliferation. We also see reduction of p27 levels following GSK-3 inactivation or depletion. Our data show that serum induction of quiescent INS-1 cells leads to sequential phosphorylation of p27 on its S10 and T187 residues with faster kinetics for S10 corresponding with the decreased levels of p27. Altogether our findings indicate that p27 levels in β-cells are stabilized by GSK-3 and thus p27 downregulation following GSK-3 depletion/inactivation plays a critical role in promoting β-cell replication.

Introduction

Both type 1 and 2 diabetes mellitus (T1 and T2 DM) are characterized by reductions in β-cell mass, nearly completely in T1DM and partially in T2DM.^1,2 In T1DM, β-cell mass is reduced due to autoimmune-mediated β-cell destruction and accumulating data from clinical studies and animal models suggest increased β-cell formation in T1DM.^1,3 T2DM, the most common form, is characterized by relative reductions in β-cell mass, β-cell dysfunction and peripheral insulin resistance.⁴ Several studies have demonstrated that replication of pre-existing differentiated β-cells plays an important role in regulating expansion and maintenance of postnatal/adult β-cell mass.^5–8 Studies using β-cell ablation models in rodents have shown that on cessation of β-cell damage, β-cell mass is increased because of enhanced proliferation of surviving mature β-cells.^9,10 Also, all β-cells can contribute equally to islet growth and maintenance and may not involve specialized progenitors.^11,12

The molecular basis of pancreatic β-cell replication and its regulation particularly in human islets is not well understood. A recent study on human pancreatic β-cell G₁/S proteome has demonstrated the presence of both positive and negative cell cycle regulatory proteins¹³ suggesting their functional importance in controlling human β-cell replication. Eukaryotic cell cycle progression, particularly the G₁ to S-phase transition, is controlled by a series of protein complexes composed of cyclins (e.g., cyclin D1, D2, D3 and E) and cyclin-dependent kinases (CDKs) (e.g., CDK4, CDK6 and CDK2) the activity of which is in turn controlled by a group of CDK inhibitors (CKIs).¹⁴ Among these CKIs, p27 (also called Kip1 and encoded by CDKN1B) plays a pivotal role in controlling cell proliferation.^15,16 While overexpression of p27 causes cell cycle arrest in G₁ phase,^15,16 inhibition of p27 expression prevents cell cycle arrest in response to mitogen depletion.¹⁷ p27 knockout (^-/-) mice show larger than normal body size and exhibit multiple organ hyperplasia.¹⁴ The abundance of p27 protein is primarily regulated at the posttranslational level by various mechanisms, particularly degradation/proteolysis, nuclear export and import and stabilization, which are critical for the control of p27 function and are modulated by phosphorylations, such as, T187 and S10 phosphorylations and protein-protein interactions.^18–24 In addition, a stable interaction with a cyclin-CDK complex, particularly cyclin E-CDK2, is critical for p27 degradation.²⁵ These studies illustrate that p27 is an important cell cycle regulator, particularly for G₁/S transition.

Glycogen synthase kinase-3 (GSK-3), a multifunctional serine-threonine kinase, is ubiquitously expressed and constitutively active in resting cells and is primarily regulated through inhibition of its activity by various mitogenic and hormonal signaling pathways.^26,27 GSK-3 has two mammalian isoforms, encoded by distinct genes, GSK-3α (51 kDa protein) and GSK-3β (47 kDa protein), which have high homology within their kinase domains (98% identity). GSK-3 is a key regulator of various signaling pathways including receptor tyrosine kinases, Wnt and G-protein-coupled receptors and is implicated in a variety of important cellular processes including cell cycle regulation and proliferation.^26,27 Several studies indicated that GSK-3, also localized in the nucleus besides cytoplasm,²⁸ has significant potential to act as a negative regulator of proliferation since it facilitates degradation of the two important positive cell cycle regulators, D1 and E cyclins,^29,30 and stabilizes p27,³¹ a key negative cell cycle regulator.

Here, we show using isolated adult human islets, βpurified from such islets and rodent β-cells, INS-1, that reduction in p27 or GSK-3 levels via short interfering RNA (siRNA) or small molecule-mediated GSK-3 inactivation enhances β-cell replication.

Results

We have measured, employing isolated adult human islets, the protein levels of various cell cycle regulators including the negative regulators, such as, p27 (Fig. 1A a and 1B) and p21 (Fig. 1C d and 1D), and the positive regulators, such as, D-type cyclins, D1, D2 and D3 (Fig. 1A b–d and 1B) and their kinase partners, cyclin-dependent kinase-4 (CDK4) and CDK6 (Fig. 1C a and b, and 1D), and cyclin E (Fig. 1A e and 1B) and its kinase partner, CDK2 (Fig. 1C c and 1D), and also, the amounts of retinoblastoma protein (Rb) (Fig. 1A f and 1B).¹⁴ Our WB analysis followed by scanning of protein band intensities (see Materials and Methods for details) suggest that p27 is a predominant protein among the several cell cycle modulators that we have examined using adult human islets (Fig. 1A–D) indicating its potentially critical role in human islets. It's worth mentioning that employing those antibodies used widely by various independent groups (see “Materials” for details), we find higher levels of CD3 compared to CD1 and CD2 (Fig. 1B), increased amounts of CDK2 relative to CDK4 and CDK6 (Fig. 1D) and also, mostly unphosphorylated form of Rb (Fig. 1A f) in adult human islets. We have measured the levels of GSK-3 and phospho-GSK-3 and have found high amounts of both the proteins with both of their isoforms, α and β, in adult human islets (Figs. 1A g and h, and B). We have analyzed the important role of GSK-3 in regulating p27 levels in the current study as described below. We have examined the levels p27 and GSK-3 employing two independent batches of either low BMI (26–27) (HI-1 and HI-2) or high BMI (45–50) (HI-3 and HI-4) adult islets and have found following immunoblotting and scanning of band intensities (see Materials and Methods) that p27 and GSK-3 levels are almost 2-fold higher in low BMI islets relative to high BMI (Fig. 1E) suggesting that the presence of high levels of these two proteins, p27 (a negative cell cycle regulator),¹⁴ and GSK-3 (a multifunctional serine-threonine kinase),^26,27 in adult islets (low BMI) likely plays a critical role in maintaining adult β-cell quiescence. Surprisingly, we have found 3- to 5-fold higher levels of CD3 (a positive cell cycle regulator)¹⁴ in low BMI islets compared to high BMI (Fig. 1E) indicating that βin adult islets (low BMI) have the potential to enter into the cell cycle if required.

Figure 1.

p27 is a predominant cell cycle protein in adult human islets. Isolated adult human islets from several independent batches (n = 8) having 70–90% purity (determined by dithizone staining) were cultured and total lysates were prepared (see “Methods” for details) and immunoblotted as shown in the representative figures [human islets (HI) I–III]. INS-1 cell extracts (CE) were used for comparison. (A) a–i and (B) a–e, equal amounts (60 µg) of human islet lysates or INS-1 CE were fractionated in SDS-PAGE and the transferred peptides were incubated with specific primary antibodies as shown (see “Materials and Methods” for details) for O/N (∼20 h) at 4°C followed by secondary antibody incubation for 1 hour at RT. The exposure time for all the blots following ECL was the same except for GSK-3, which was five times shorter and for phospho GSK-3 (pGSK-3) and tubulin that were three times lower. Tubulin was used as loading controls. The immunoreactive band intensities were measured using Bio-Rad Fluorescent Imager FX Pro Plus and the data were analyzed by Quantity One software. Individual protein band intensities were graphically represented in (B) and (D). To measure individual band intensity, background was subtracted for each protein band followed by normalization against the control protein (tubulin). (E) a–e, isolated adult human islets from four independent batches (two [HI-1 and HI-2] from low BMI [26–27] and two [HI-3 and HI-4] from high BMI [40–50] cadaveric donors) with high purity (80–90% for low BMI islets) and low purity (35–40% for high BMI islets) were cultured and equal amounts (50 µg) of human islet lysates or INS-1 CE were fractionated in SDS-PAGE and the transferred peptides were incubated with specific primary antibodies as shown. *p < 0.05; NS, not significant.

To understand the biological implication of p27, we have examined the ability of p27 to interact with various cyclins and CDKs in adult human islets since p27 possesses specific cyclin/CDK binding domains.^15,16,24 Our IP + WB studies using isolated adult human islet extracts show that p27 is able to interact not only with various D-type cyclins and their kinase partners, CDK4 and CDK6, it also binds robustly with GSK-3 (Fig. 2A a and b). While CDK6 appears to interact more with hyperphosphorylated form of p27, GSK-3, all of D cyclins and CDK4 tend to bind either unphosphorylated or hypophosphorylated form of p27 (Fig. 2A a). We see increased binding of GSK-3 with D-type cyclins compared to their kinase partners (Fig. 2B a and b). Also, p27, D cyclins and their kinase partners interact mostly with GSK-3β isoform than α (Fig. 2B a). We have analyzed, employing adult human islets, the interactions of p27 with cyclin E or CDK2 and also, have examined if antibodies against phosphorylated forms of p27, p-p27 (S10) and p-p27 (T187),^21–24 can pull down any detectable levels of p27 protein. While we see robust binding of p27 with cyclin E or CDK2 following IP + WB analysis, antibodies against p-p27(S10) and p-p27(T187) were unable to pull down any detectable amounts of p27 (Fig. 2C a and b). We have found similar results using INS-1 cell extracts (Fig. 2C, c and d) suggesting a critical importance of such robust interaction of p27 with cyclin E and CDK2 in adult human islets. Also, the data suggest that the antibodies against the two phosphorylated forms of p27, p-p27(S10) and p-p27(T187), have either undetectable or very low affinity for p27. We then examined the interactions of cyclin E with either CDK2 or p-p27(S10) or p-p27(T187) in adult human islets and have found, following IP + WB assays that while CDK2 has robust binding ability with cyclin E, neither p-p27 (S10) nor p-p27(T187) has any detectable interaction with cyclin E (Fig. 2D, a and b). We have found similar results using INS-1 cell extracts (Fig. 2D, c and d). Also, we see that anti-p27 antibody has the ability to pull down considerable amounts of cyclin E using both human islet and INS-1 extracts (Fig. 2D a–d) suggesting that p27 via interaction with cyclin E and CDK2 can form trimeric complexes in adult human islets. We have examined the subcellular distribution of p27, p-p27 (S10) and p-p27 (T187) in purified adult human β-cells (FACS-sorted β-cells following Newport Green staining) (see Materials and Methods) and have found that while p27 is present in both the nucleus and cytoplasm, p-p27 (T187) is localized primarily in nucleus and p-p27 (S10) is distributed mostly in cytoplasm (Sup. Fig. 1A and B). We have analyzed the ratio of β-(insulin-positive) and α- (glucagon-positive) cells in isolated human islets (see Materials and Methods) and have found that β-cells are much more abundant than α-cells (Sup. Fig. 1C). Also, we have demonstrated expression of the important marker genes in isolated human islets and purified human β-cells by RT-PCR (Sup. Fig. 2A and B) indicating maintenance of their functional integrity in our culture conditions. The findings altogether suggest a critical role of p27 in negatively regulating cell cycle of β-cells in adult human islets via protein-protein interactions with D-cyclins and E cyclin and their kinase partners as well as through GSK-3 binding.

Figure 2A–D.

p27 interacts with various cyclins, CDKs and also, GSK-3 in adult human islets. (A–D) (Aa, Ba, Ca, Cc, Da and Dc), equal amounts (50 Ég) of islet or INS-1 extracts for each lane were subjected to IP using various antibodies as shown (see “Materials and Methods” for details) and the immunoprecipitates were subjected to western blotting (WB) using antibodies against p27 or GSK-3 or cyclin E as shown. Antibodies used against CD1 and cyclin E were from two different sources, Santa Cruz Biotechnology (SC) and BD Biosciences (BD). Individual band intensities for IP+WB or WB were graphically represented in (Ab, Bb, Cb, Cd, Db and Dd) following densitometric scanning using Bio-Rad Fluorescent Imager FX Pro Plus and analysis by Quantity One software as described in Figure 1 legend. Quantitative data are presented as the mean ± SD (n = 2).

To further explore the function of p27 on β-cell replication, we have employed INS-1 cells (a well differentiated insulin secreting rat β-cell line) (see Materials and Methods). We have induced quiescent (serum deprived) INS-1 cells with serum (see Materials and Methods) for various time points and show that while the p27 level decreases gradually and is reduced 3.5-fold after O/N (∼22 h) serum stimulation (Fig. 3A, a, 0 h vs. O/N and B, 0 h vs. O/N), the cyclin D2 level increases gradually and is enhanced 1.4-fold following O/N serum induction (Fig. 3A b, 0 h vs. O/N and B, 0 h vs. O/N). The robust decrease of p27 (a negative growth regulator) and the concomitant moderate increase of D2 cyclin (a positive growth regulator) in response to mitogens (i.e., serum) following starvation indicate that these rodent β-cells restore cell cycle regulatory mechanism for replication. We checked the levels of cyclin D3 following serum starvation and stimulation and didn't find any appreciable changes (data not shown). We have examined the phosphorylation kinetics of p27 on its two important residues, S10 and T187, after serum starvation and stimulation of INS-1 cells since nuclear degradation of p27 and its translocation from nucleus to cytoplasm and cytosolic degradation are regulated by T187 and S10 phosphorylation respectively.²³ We have found that S10 phosphorylation peaks much faster than T187 phosphorylation (Fig. 3C b, lanes, 15 min-1 h vs. c, lanes, 5 h-O/N and D) ensuring p27 degradation (Fig. 3C a, 0 h vs. O/N and D) as also demonstrated using non-β-cells.¹⁸ The results suggest that INS-1 retains its growth responsive function via regulating the G₁-S-phase transition. Unfortunately a similarly robust human β-cell model still doesn't exist for such studies.

Figure 3.

Serum deprivation and stimulation of rodent b-cells (INS-1) results in p27 degradation and cyclin D2 induction and influences phosphorylation kinetics of p-p27 (S10) and p-p27 (T187). Total extracts of INS-1 cells after serum starvation (∼44 h culture in 0.5% FBS-containing medium) and induction (in the presence of 10% FBS) for various time points [0 h, 15 min, 1, 3, 5 h and O/N (∼22 h)] were prepared (see Materials and Methods for details) (n = 3 or more) and equal amounts (50 µg) of cell extracts for each time point were fractionated in SDS-PAGE. Peptides, following membrane transfer, were incubated with specific primary antibodies as shown. Hsp90 and tubulin were used as loading controls for (A and C) respectively. The immunoreactive band intensities for (A and C) were measured (see Fig. 1 legend and Materials and Methods) and graphically represented in (B and D) respectively. *p < 0.05; for p27 and CD2 in (B), 1 h-O/N vs. 0 h and 5 h-O/N vs. 0 h respectively. *p < 0.05; for p27, p-p27(S10) and p-p27(T187) in (D) 1 h-O/N vs. 0 h and 15 min-O/N vs. 0 h and 15 min-O/N vs. 0 h respectively.

We, therefore, have used INS-1 cells to examine the effects of siRNA-mediated p27 knock down on replication. We have transiently transfected various concentrations of individual siRNAs (control/scrambled or experimental including VIII66 and II47) and have shown that particularly, 100 pmol (pm) of II47 and both 50 and 100 pm of VIII66 are effective in significantly reducing p27 levels (1.6-fold reduction by 100 pm of II47 and 1.9–2.3-fold reduction by VIII66) [Fig. 4A b (100 pm of II47 siRNA) and c (50–100 pm of VIII66 siRNA) vs. a (20–100 pm of control siRNA) and B]. We were unable to detect any appreciable changes in the levels of other important cell cycle regulatory proteins including cyclin E and CDK2 (Fig. 4C b and c and lanes, VIII66 vs. control and D) in spite of significant reduction of p27 levels in VIII66-transfected cells (Fig. 4C a, lanes, VIII66 vs. control and D) suggesting the specificity of siRNA-mediated p27 knock down. We then have measured the effects of p27 depletion on β-cell replication by various assays including ³H-thymidine incorporation, EdU incorporation and cell counting. We show using low (1%) or high (10%) serum-containing medium that relative to control/scrambled siRNA, VIII66 promotes DNA synthesis, 1.6–1.8-fold, (Fig. 4E and VIII66 vs. control) and enhances cell numbers, 2.4–2.6-fold, (Fig. 4F and VIII66 vs. control). We have examined the effects of VIII66 on the proportion of S-phase cells by Edu incorporation and show a moderate increase in the S-phase population [Fig. 4G and c (VIII66) vs. b (control)]. We have treated VIII66-transfected INS-1 cells without (DMSO) or with roscovitine (an inhibitor of cyclin E-CDK2) in the presence of 10% serum and have found roscovitine-mediated significant blockage (1.7-fold) of tritiated thymidine incorporation relative to VIII66 alone (Fig. 4H and VIII66 + roscovitine vs. VIII66), which is comparable to control siRNA (Fig. 4H and VIII66 + roscovitine vs. control) indicating that p27 depletion by VIII66 siRNA likely leads to increased cyclin E-CDK2 activity and results in enhanced DNA synthesis. We also found that INS-1 cell numbers, increased following VIII66 transfection, were markedly reduced when treated in the presence of roscovitine (data not shown). The results indicate that cyclin E-CDK2 activity is inversely proportional to the p27 level in β-cells and important for p27 degradation.²⁴ We have analyzed the effects of siRNA-mediated p27 knock down on adult human β-cell replication employing isolated human islets and have found that DNA synthesis is increased particularly in the presence of II47 or V91 siRNA relative to controls (1.3- to 1.45-fold) (Sup. Fig. 3, II47 or V91 vs. control) without any marked changes in the presence of VIII66 (not shown) because of significant lack of sequence homology with human. The results taken together indicate that depletion of p27 levels via RNAi promotes β-cell replication.

Figure 4.

siRNA-mediated p27 knock down promotes β-cell proliferation. INS-1 cells were transiently transfected, for (A) with each of control or experimental siRNAs at various concentrations as shown and for (C) with control or VIII66 siRNA. Cells after transfection were cultured, for (A) in 10% serum-containing medium and for (C) in medium containing either 1 or 10% serum. Equal amounts of cell extracts were fractionated [for (A) 70 µg/sample and for (C) 40 µg/sample] and the transferred peptide bands were immunoblotted as shown (n = 3 or more). Tubulin was used as loading controls. The immunoreactive band intensities for (A and C) following densitometric scanning (see Fig. 1 legend and also, Materials and Methods) were graphically plotted in (B and D) respectively. INS-1 cells after transfection of either control or experimental (VIII66) siRNA were cultured in medium containing 1 or 10% serum and then assayed for incorporated radioactivity (E) and cell numbers (F) (n = 4 or more) and also for incorporated EdU (G) (n = 2). (H) INS-1 cells after control or VIII66-siRNA transfection were treated without (DMSO) or with roscovitine (cyclinE-CDK2 inhibitor) (5–10 µM) and incubated in the presence of ³H-thymidine followed by assessment of incorporated radioactivity (n = 3). *p < 0.05; for (B) 100 pm of II47 and 20–100 pm of VIII66 vs. control and for (D) VIII66 of 1 or 10% serum vs. control. **p < 0.01; for (E and F) VIII66 vs. control. **p <0.01; for (H) VIII66 vs. control. *p <0.05; for H, VIII66 + roscovitine vs. control.

We then explored the roles of glycogen synthase kinase-3 (GSK-3),²⁷ in controlling p27 stability since we have identified robust protein-protein interaction between p27 and GSK-3 in our IP+WB studies (see Fig. 2). We have transiently transfected INS-1 cells with each of control/scrambled or GSK-3α or GSK-3β siRNAs and have found that while GSK-3α siRNA specifically depletes the α-isoform (50–60% reduction in the presence of 50–100 pm of GSK-3α siRNA compared to the controls) without significantly interfering with β-isoform levels (Fig. 5A, b vs. a), GSK-3β does the opposite without significantly affecting the α-isoform levels (Fig. 5A c vs. a). We have analyzed the levels of two key phosphorylation sites of GSK-3, S21 for α and S9 for β,^26,27 after individual siRNA transfection (control/scrambled or experimental) and have found that while GSK-3α siRNA (particularly in the presence of 50–100 pm), relative to controls, specifically lowers S21 phosphorylation (>70%) (Fig. 5B, b vs. a), S9 levels are particularly downregulated (>60%) following GSK-3β siRNA transfection (particularly in the presence of 50–100 pm) (Fig. 5B, c vs. a). The next question we have asked is whether GSK-3α or β depletion in INS-1 cells can affect the p27 levels and we have found about 60% reduction in the p27 levels after GSK-3α or β siRNA transfection (specifically in the presence of 100 pm) compared to controls (Fig. 5C, c and b vs. a). The data indicate that depletion of GSK-3 leads to p27 downregulation in β-cells.

Figure 5.

siRNA-mediated GSK-3 depletion leads to p27 downregulation in β-cells. (A–D) INS-1 cells were transiently transfected with each of control/scrambled or experimental GSK-3 siRNAs (to knock down either α or β isoform of GSK-3) at various concentrations (20, 50 and 100 pm) (n = 2). Cells after transfection were cultured in 10% serum-containing medium. Equal amounts (60 µg/sample) of cell extracts following siRNA (control or experimental) transfection were fractionated in SDS-PAGE and the transferred peptides were immunoblotted with various antibodies [(A) anti-GSK-3, (B) anti-p-GSK-3, (C) anti-p27 and (D) anti-Hsp90]. Hsp90 was used as loading controls. The immunoreactive band intensities were measured (see Fig. 1 legend and Materials and Methods).

We then analyzed the effects of, following transient transfection, either individual GSK-3 siRNAs (α or β isoform) or α + β together or α or β in combination with p27 siRNA (VIII66) on INS-1 cell proliferation. We show that while individual α or β siRNA promotes tritiated thymidine incorporation 2.8–3.2-fold relative to controls, α and β together (co-transfection) result in 2.6-fold increase (Fig. 6A) suggesting lack of any additive effects. We have examined the effects of co-transfection of p27 with either GSK-3α or β and show enhanced DNA synthesis (3-fold) comparable to individual GSK-3α or β (Fig. 6A). It's worth mentioning that p27 siRNA (VIII66) transfection alone increased DNA synthesis that was comparable to GSK-3α + β co-transfection (data not shown). We have found, after transfection of individual GSK-3α or β or α + β together, comparable levels of increase in β-cell numbers (1.69–1.76-fold) relative to controls (Fig. 6B). Altogether, the data (Figs. 5 and 6) suggest that p27 levels in β-cells are stabilized by GSK-3 and depletion of GSK-3 results in p27 downregulation and in turn, increases β-cell replication.

Figure 6.

GSK-3 depletion promotes β-cell proliferation. INS-1 cells were transiently transfected with either control or experimental siRNAs, individually or in combination, as shown (n = 3). Cells following transfection were cultured in 10% serum-containing medium and subjected to two different assays, tritiated thymidine incorporation (A) and cell counting (B) as described in Materials and Methods. **p < 0.01; for (A) GSK-3α or β vs. control. *p < 0.05; for (A) GSK-3α+β or p27 + GSK-3α or p27 + GSK-3β vs. control and for (B) GSK-3α or GSK-3β or GSK-3α+β vs. control.

We next asked whether GSK-3 inactivation by small molecule inhibitors (1-AKP or BIO) can result in p27 degradation in β-cells. We have induced quiescent INS-1 cells with 10% serum in the presence of DMSO and have found marked degradation (2.6-fold) of the p27 levels after O/N (∼22 h) serum stimulation relative to 0 h (Fig. 7A a, lanes, O/N vs. 0 h) as expected (see Fig. 3). We have induced INS-1 cells with 10% serum in the presence of roscovitine, an inhibitor of cyclinE-CDK2, and have found that the levels of p27 at 5 h are comparable to 0 h (Fig. 7B a, lanes, 5 h vs. 0 h) and are slightly reduced, 1.3-fold, after O/N ttreatments (Fig. 7B a, lanes, O/N vs. 0 h) suggesting cyclin E-CDK2 activity is critical for p27 degradation.²⁴ We have analyzed the effects of small molecule GSK-3 inhibitors (e.g., 1-AKP) on the p27 levels and have demonstrated marked p27 reduction after 5 h and O/N treatments (3–3.5-fold) (Fig. 7C a, lanes, 5 h and O/N vs. 0 h). We have seen decreased effects of BIO on p27 degradation (about 2-fold) (Sup. Fig. 4). However, when INS-1 cells were treated with 1-AKP in the presence of roscovitine, p27 levels were restored (Fig. 7D a, lanes, 5 h and O/N vs. 0 h) indicating cyclin E-CDK2 activity is required for p27 degradation in β-cells. We didn't find any significant changes in the amounts of total GSK-3 or Hsp90 (Fig. 7A–D). We also have asked if 1-AKP or BIO can promote human β-cell proliferation and thus have incubated isolated adult human islets in the absence (DMSO) or presence of either 1-AKP or BIO for 3–4 days (see Materials and Methods). We have found enhanced incorporation of tritiated thymidine in the presence of BIO or 1-AKP (1.5 to 1.7-fold) relative to DMSO-treated controls (Fig. 7E, BIO or 1-AKP vs. DMSO). We also have shown employing Newport Green-stained and FACS-sorted single human β-cells (see Materials and Methods) that treatments with either 1-AKP or BIO result in mostly cytosolic localization of p27 relative to untreated (DMSO) controls where p27 is localized in both nuclei and cytosol (Sup. Fig. 5) indicating nuclear degradation and/or cytosolic translocation of p27 following GSK-3 inactivation. The data altogether suggest that GSK-3 inactivation via small molecule inhibitors can play a critical role in reducing p27 levels in β-cells as well as promote β-cell replication.

Figure 7.

GSK-3 inactivation leads to p27 instability and enhances β-cell replication. INS-1 cells after ∼44 h serum starvation were induced with 10% FBS in the presence of DMSO (A), roscovitine, an inhibitor of cyclinE-CDK2 (B), 1-AKP, a small molecule GSK-3 inhibitor (C) and 1-AKP + roscovitine (D) for various time points [0 h, 5 h and O/N (∼22 h)] (n = 2) and equal amounts (50 µg/lane) of total cell lysates were fractionated in SDS-PAGE and immunoblotted with specific antibodies (anti-p27, anti-GSK-3 and anti-Hsp90) as shown. (E) isolated human islets were cultured in the absence (DMSO) or presence of BIO (a small molecule GSK-3 inhibitor) (2 µM) or 1-AKP (5 µM) followed by incubation in the presence of tritiated thymidine (n = 4) to assess the effects of individual GSK-3 inhibitor treatment on β-cell proliferation. **p < 0.01; BIO or 1-AKP vs. control (DMSO).

Discussion

We have shown that p27, a negative cell cycle regulator, is a predominant protein among various important cell cycle regulators including D-type cyclins and their kinase partners, CDK4/6, E cyclin and its kinase partner, CDK2, p21 and retinoblastoma protein (Rb) that are present in isolated adult human islets (Fig. 1). Studies from several laboratories using adult mice^6,7,32 and also, adult rat and human islets/β-cells^13,33 have demonstrated the importance of positive cell cycle regulators, particularly D2 and D1 cyclins and their kinase partners, CDK4/6, in controlling β-cell mass/growth via regulation of β-cell proliferation. However, the regulatory effects of protein-protein interactions of cyclins (D-type and E) and their kinase partners with p27 to control cyclin D-CDK4/6 and/or cyclin E-CDK2 function, particularly in human β-cells/islets, are not well understood. It is known that p27 has cyclin and CDK binding domains.^15,24 We have noted robust interaction of p27 with E cyclin and its kinase partner, CDK2, as well as with D-cyclins and their CDK partner, CDK4, but only at moderate levels with CDK6 (Fig. 2). Also, we have found robust interaction of p27 with GSK-3 (Fig. 2). We have shown that serum starvation followed by serum induction of rodent β-cells (INS-1) result in significant downregulation of p27 and moderate upregulation of cyclin D2 and simultaneously, our data on the kinetics of p27 phosphorylation suggest that faster and delayed phosphorylation of S10 and T187 respectively likely expedite nuclear export as well as degradation of p27 to ensure the G₁ to S-phase progression (Fig. 3).^19–23 We have demonstrated that blocking cyclin E-CDK2 function by roscovitine can lead to robust inhibition of β-cell proliferation (Fig. 4). Recently a group has shown using mouse model that CDK4 regulates β-cell mass by recruiting quiescent β-cells to enter the cell cycle.³⁴ Another recent study has revealed that CDK6 is capable of driving human β-cell proliferation in vitro and also, it is able to increase human islet engrafment/proliferation in vivo.³⁵ The findings altogether indicate that inhibition of cyclin-dependent kinase (CDK) function, particularly of CDK4/6 and CDK2, via protein-protein interaction with p27 may exert a critical effect in maintaining normal adult human β-cell quiescence and thus downregulation of p27 levels in β-cells may play a key role in promoting adult human β-cell replication in vitro and in vivo. We have shown using INS-1 cells and isolated adult human islets that RNAi-mediated p27 depletion promotes β-cell proliferation (Fig. 4 and Sup. Fig. 3). A recent study on the G₁/S proteome using adult human islets has demonstrated the presence of various important cell cycle proteins including the positive and negative regulators¹³ indicating the potential of adult human β-cells to replicate and expand upon receiving appropriate stimuli/signals. Also, several studies using animal models of diabetes showed the importance of downregulation or loss of p27 in β-cells for accelerating replication of β-cells and thereby expansion of β-cell mass for correcting diabetes.^36–38

We have found robust interaction of p27 with GSK-3 in human islets and rodent β-cells (Fig. 2). Furthermore, p27 has multiple consensus binding sites for GSK-3 (S/T-X-X-X-S/T-P),^26,27 (more in human than rodent p27). We have shown that siRNA-mediated GSK-3 downregulation, either α or β isoform, results in p27 degradation in β-cells (INS-1) and our transient transfection and co-transfection studies demonstrate that GSK-3α or β siRNA, either individually or in combination with p27 siRNA, can promote β-cell proliferation (Figs. 5 and 6). Furthermore, our data show that small molecule-mediated (1-AKP or BIO) GSK-3 inactivation can lead to p27 downregulation and enhancement of β-cell proliferation (Fig. 7). Studies using in-vitro and in-vivo models have demonstrated that GSK-3 inactivation/deficiency results in p27 instability/degradation.^31,37 Also, increased proliferation of both primary and transformed rodent β-cells following GSK-3 inactivation by small molecule inhibitors including BIO or 1-AKP have been reported.³⁹ A study demonstrated increased proliferation of highly differentiated primary rodent cardiomyocytes after BIO-mediated GSK-3 inhibition.⁴⁰ GSK-3 is localized in both cytosol and nuclei^26–28 and we also have seen nuclear and cytoplasmic localization of GSK-3 in purified single adult human α-cells (Sup. Fig. 5). We thus propose that GSK-3, present in high amounts in adult human islets (Fig. 1), plays a critical role in preventing p27 degradation and thereby block normal adult β-cells from exiting quiescence. The ability of GSK-3 to act as an important negative regulator of β-cell cycle is highly likely because the roles of GSK-3 in facilitating degradation of the positive cell cycle regulators, cyclin D1 and cyclin E, have been reported.^29,30 It is not clear at this point if GSK-3 augments the stability of other CKIs, particularly INK4 (inhibitors of CDK4/6) group of proteins (p15, p16, p18 and p19),¹⁴ in β-cells, specifically adult human β-cells. Our findings thus underscore the importance of either direct decrease of p27 or downregulation of p27 via GSK-3 inactivation/depletion to promote β-cell proliferation. A recent study shows that conditional ablation of GSK-3β in islet β-cells in mice results in increased β-cell mass accompanied by increased proliferation and decreased apoptosis.⁴¹ The regenerative potential of human β-cells via proliferation following exposure to growth factor/s has been demonstrated.⁴² An unanswered question is if Wnt pathway activation plays any major role in promoting human β-cell proliferation since the importance of Wnt signaling in stimulating primary rodent β-cell proliferation has been reported.⁴³ Recently a study has demonstrated increased nuclear translocation of β-catenin in human islets/β-cells following GSK-3 inhibition via LiCl or 1-AKP, which is rapamycin sensitive and the same study has also shown the importance of synergistic effects of glucose and GSK-3 inhibition in enhancing human β-cell proliferation.⁴⁴ GSK-3 inhibitors are known to mimic some actions of insulin and GSK-3 is thus considered as a potential drug target for T2DM.²⁷ Furthermore, GSK-3 inhibitors, due to their proliferation-promoting abilities, may be suitable in combination with immunosuppressive agents for treatment of T1DM patients because studies suggest that in most cases, even those with long-standing T1DM, some β-cells persist and are continually destroyed and also because of enhanced β-cell formation in T1DM.^1,3,45

Materials and Methods

All the cell and tissue culture media and reagents as well as Click-iT EdU (Alexa Fluor 488) flow cytometry assay kit, Newport Green (NG)-PDX, Pluronic F127 and Superscript First Strand RT-PCR kit were purchased from Invitrogen. The following antibodies that we used in our studies were purchased from the companies mentioned below—monoclonal antibodies (mAbs) against p27^Kip1, cyclin-dependent kinase (CDK) 4, cyclin D1 (CD1), cyclin D3 (CD3), cyclin E and CDK2 were bought from BD Biosciences; mAbs against CD1, p21^Cip1, PCNA, Rb and GSK-3 α/β and polyclonal antibodies (pAbs) against cyclin E, CDK2, p-p27 (Thr187), p-p27 (Ser10), Skp2, β-catenin and Hsp-90 and mAb against cyclin E were from Santa Cruz Biotechnology, Inc.; mAb against CDK6 and pAbs against p-GSK-3α/β (Ser21/9) and β-catenin and the Rb antibody sampler kit (#9969) were from Cell Signaling Technology; mAb against cyclin D2 (CD2) was from Lab Vision Corporation/Neomarkers; guineapig anti-human insulin serum was from Linco Research; monoclonal mouse anti-human Ki-67 and monoclonal anti-α-tubulin were from Dako Cytomation and Sigma-Aldrich respectively. The secondary antibodies including anti-guineapig, anti-mouse and anti-rabbit, conjugated with FITC, Texas Red and Cy3 respectively, were purchased from Jackson Immuno Research Laboratories Inc. The small molecule GSK-3 inhibitors, 1-Azakenpaullone (1-AKP) and BIO (6-bromoindirubin-3′-oxime) and the CDK2 inhibitor, roscovitine (has no detectable effect on either CDK4 or CDK6 activity at low micromolar concentrations⁴⁶), were purchased from Calbiochem.

Culture of INS-1 cells and human pancreatic islets.

INS-1 cells, differentiated insulin secreting pancreatic β-cells derived from rat insulinoma,⁴⁷ were provided by Lou Philipson, University of Chicago, and cultured as described in reference ⁴⁷. For serum starvation and stimulation experiments, INS-1 cells were plated in medium containing 10% FBS the day before starvation and deprived of serum by culturing in 0.5% serum-containing medium for ∼44 h. Adult human islets (isolated from male and female cadaveric donors of varying ages, 50 ± 15 years) were obtained through the NIH- and JDRF-supported Islet Cell Resource Consortium (ICRC) and Integrated Islet Distribution Program (IIDP). The Institutional Review Board, University of Chicago, approved the use of human islets for our studies. Human islets, immediately after receiving, were washed with PBS (Ca⁺⁺ and Mg⁺⁺-free) and cultured in CMRL 1066 medium (usually for ∼48 h for recovery) as described in reference ⁴⁸. Islet purity and viability assessed by dithizone staining and trypan blue exclusion respectively varied from 65–90% among batches.

Purification of human β-cells.

β-cells from isolated adult human islets were purified following published protocols.^48,49 Briefly, islets were trypsinized and pipetted gently to make single cell suspensions and after filtering through 70 µm sterile cell strainer (BD Falcon), cells were resuspended in PBS containing BSA (0.5%) and glucose (2.8 mM). Cells were then incubated with NG-PDX (permeant ester form) (2 µM) in the presence of Pluronic F127 at 37°C for 30 min. Following incubation, cells were resuspended in the same buffer containing BSA and glucose and subjected to FACS sorting using a Beckman-Coulter MoFlo High Throughput Sorter (excitation-488 nm and emission-530/30 band pass filter). The FACS-sorted β-cells were cultured in the same medium as used for human islet culture (see above).

Immunofluorescence.

Single β-cell suspensions (human and rodent) after plating on poly-l-lysine (Sigma) coated coverslips were cultured and, following PBS wash, cells were first fixed in 2–4% paraformaldehyde (PFA) for 15 min (for insulin) and then in acetone for 10 min at 4°C. After fixation, cells were permeabilized in PBS containing 0.2% Triton X-100 for 5 min at room temp (RT) followed by blocking in PBS with 3% BSA + 50 mM ammonium chloride + 10 mM glycine for 45 min at RT. Cells were then incubated with specific primary antibodies diluted in PBS containing 3% BSA and 0.2% Tween-20 for O/N at 4°C followed by incubation with appropriate secondary antibodies in the same buffer for 1 h at RT. Cells were finally mounted in slow fade gold containing DAPI (Invitrogen) and examined by fluorescence microscopy (Nikon E800).

Preparation of cell/islet lysates, WB and IP.

INS-1 cells and human islets were brought into ice-cold immunoprecipitation (IP) buffer containing freshly added protease inhibitor cocktails⁵⁰ and phosphatase inhibitor cocktails A and B (Santa Cruz Biotechnology Inc.). Cells/islets were sonicated and supernatants were collected after microcentrifugation. Protein concentrations were measured in replicates for western blotting (WB) or IP + WB by Bio-Rad protein assay and equal amounts of freshly prepared extracts were fractionated in 10% SDS-PAGE either for direct WB studies or subjected to IP followed by WB analysis. The immunoreactive bands after transfer onto Hybond-P membranes (GE Healthcare) were detected by ECL-plus. Band intensities were measured using Bio-Rad Fluorescent Imager FX Pro Plus and the data were analyzed by Quantity One software. Individual band intensity was calculated after subtracting each background followed by normalization against the control protein (tubulin and/or Hsp90).

RT-PCR.

Total RNA was extracted from cells (INS-1 and purified human β-cells) as well as human islets as described in reference ⁵¹. Equal amounts of total RNA were used to make cDNAs and then equal amounts of cDNAs for each sample were subjected to PCR amplifications (30 cycles) using specific forward and reverse primers for rat and human respectively.

Synthesis of short interfering RNAs (siRNAs) and transient transfection.

To knockdown p27 and GSK-3, we designed several siRNAs with the help of siRNA prediction tool of the Whitehead Institute (jura.wi.mit.edu/bioc/siRNAext) and as described previously in reference ⁵¹. The target sequences designed to make the duplex siRNAs are the following: control/scrambled (for p27), TAC GCG CAT AAG ATT AGG G and experimental, II47-GCA AGT GGA ATT TCG ACT T; V91-AAT GTT TCA GAC GGT TCC C; VIII66-GAT GTA GCA TTG CGC AAT T. For GSK-3, control/scrambled, TTC TCC GAA CGT GTC ACG T, GSK-3α, GGG TGT AAA TAG ATT GTT A and GSK-3β, CGA TTA CAC GTC TAG TAT A. All siRNAs were chemically synthesized by Dharmacon. Each siRNA was transiently transfected (for cells 1–2 days and for islets 3 days) using Lipofectamine 2000 (Invitrogen) according to the protocols from the company and a published report in reference ⁵².

³H-thymidine incorporation.

Cells (INS-1-0.2-0.3 × 106; purified human β-cells-0.05-0.1 × 10⁶) and human islets (200) were plated in replicates. Cells and islets after siRNA transfection (control or experimental) or treatments without (DMSO) or with GSK-3 inhibitor (3–4 days) were incubated with ³H-thymidine (Specific Activity-6.7 Ci/mmol; Perkin Elmer) (5–10 µCi/ml of medium) for either 6–8 h (cells) or O/N (islets). Following incubation, cells and islets were washed with cold PBS and radioactivity precipitated by trichloroacetic acid was measured in a liquid scintillation counter (Packard).⁵³

Cell counting.

INS-1 cells were seeded in replicates and following O/N culture, were transiently transfected by siRNA (control or experimental) (see above). After transfection, cells were cultured in medium containing low (1%) or high (10%) serum and following PBS wash, were trypsinized and single cells were counted using hemocytometer.

EdU incorporation.

Equal numbers of cells following transient transfection of either control or experimental siRNA were cultured for 24–48 h and finally incubated without (DMSO) or with EdU (5 µM) for several hours. Detection of incorporated EdU (with Alexa Fluor 488 azide) and analysis of cell cycle were performed using Becton Dickinson LSRII following the company's (Molecular Probes, Invitrogen) protocol and the data were analyzed using FlowJo (Treestar) software.

Staining of human islets for confocal imaging and 3D reconstruction.

Isolated adult human islets were immunostained for insulin (β-cells) and glucagon (α-cells) and confocal imagings and reconstruction analysis of stacks of images in 3D were performed as described in reference ⁵⁴.

Statistical analysis.

Quantitative data are presented as the mean ± the standard error of the mean (SEM) (n = 3 or more), unless indicated. The differences considered statistically significant (Student's t test) were at p values less than 0.05.

Financial Support

This work was supported by National Institutes of Health Grants RO1-DK-013914 to D.F.S. and by the University of Chicago Diabetes Research and Training Center Grants DK-020595.