• Open Access

PTEN controls β-cell regeneration in aged mice by regulating cell cycle inhibitor p16ink4a

Authors


Summary

Tissue regeneration diminishes with age, concurrent with declining hormone levels including growth factors such as insulin-like growth factor-1 (IGF-1). We investigated the molecular basis for such decline in pancreatic β-cells where loss of proliferation occurs early in age and is proposed to contribute to the pathogenesis of diabetes. We studied the regeneration capacity of β-cells in mouse model where PI3K/AKT pathway downstream of insulin/IGF-1 signaling is upregulated by genetic deletion of Pten (phosphatase and tensin homologue deleted on chromosome 10) specifically in insulin-producing cells. In this model, PTEN loss prevents the decline in proliferation capacity in aged β-cells and restores the ability of aged β-cells to respond to injury-induced regeneration. Using several animal and cell models where we can manipulate PTEN expression, we found that PTEN blocks cell cycle re-entry through a novel pathway leading to an increase in p16ink4a, a cell cycle inhibitor characterized for its role in cellular senescence/aging. A downregulation in p16ink4a occurs when PTEN is lost as a result of cyclin D1 induction and the activation of E2F transcription factors. The activation of E2F transcriptional factors leads to methylation of p16ink4a promoter, an event that is mediated by the upregulation of polycomb protein, Ezh2. These analyses establish a novel PTEN/cyclin D1/E2F/Ezh2/p16ink4a signaling network responsible for the aging process and provide specific evidence for a molecular paradigm that explain how decline in growth factor signals such as IGF-1 (through PTEN/PI3K signaling) may control regeneration and the lack thereof in aging cells.

Introduction

Loss of proliferation capacity and the inability of cells to respond to mitogenic stimulation are the major hallmarks of aging cells in vivo and in vitro (Sharpless & DePinho, 2007). The expression of genes encoded by the Ink/Arf locus, particularly p16ink4a, increases remarkably with age and may control age-related cellular changes (Serrano, 1997; Krishnamurthy et al., 2004). In rodent models, transgenic expression of p16ink4a leads to early loss of regeneration capacity in several tissue types including pancreatic islets (Krishnamurthy et al., 2006). The inability of pancreatic islet β-cells to efficiently regenerate and compensate for hyperglycemic conditions as individuals age is proposed to underlie diabetes pathogenesis (Buchanan, 2001). In this study, we investigated the role of the PI3K/AKT pathway downstream of mitogenic signaling in cellular aging using β-cell proliferation/regeneration as a model.

In pancreatic β-cells, several mitogens were found to regulate β-cell proliferation. Mice lacking hepatic growth factor (HGF) receptor suffer from smaller islet mass and enhanced apoptosis (Dai et al., 2005; Roccisana et al., 2005; Mellado-Gil et al., 2011), whereas overexpression of HGF leads to enhanced islet function with increased proliferation (Garcia-Ocana et al., 2000, 2001). Similarly, insulin-like growth factor (IGF-1) treatment prevents apoptosis and is thought to be a mitogen for β-cell growth (George et al., 2002; Yu et al., 2003; Agudo et al., 2008; Robertson et al., 2008). These studies suggest that the PI3K/AKT pathway plays a mitogenic role in β-cell growth as IGF-1, insulin, and HGF signal through the PI3K/AKT signaling pathway (Holst et al., 1998; Tuttle et al., 2001; Liu et al., 2002; Bernal-Mizrachi et al., 2004). Consistently, mice overexpressing a constitutive active form of AKT displayed larger islets and reduced apoptosis rate in response to streptozotocin (STZ) treatment (Bernal-Mizrachi et al., 2001), whereas mice with overexpression of GSK-3β, the substrate inhibited by AKT activity, led to reduced β-cell mass (Liu et al., 2008).

In cultured primary cells, PI3K inhibition leads to cell cycle arrest and premature aging (Collado et al., 2000). Perturbation of PI3K/AKT signaling in C. elegans and Drosophila also significantly affects their lifespan (Burgering & Kops, 2002). In pancreatic β-cells, we and others showed that deletion of Pten leads to the upregulation of PI3K/AKT pathway and induces β-cell proliferation capacity (Nguyen et al., 2006; Stiles et al., 2006), particularly in adult animals. This model enables us to determine the molecular consequence of PTEN loss/PI3K activation on signals that regulate cellular aging in vivo. Our study reveals that Pten null β-cells exhibit a marked increase in mitotic activity as they age. The downregulation of p16ink4a is associated with this regeneration phenotype observed in the aged Pten null β-cells. We further establish a novel cyclin D1/E2F/Ezh2 signal node that mediates PTEN-regulated p16ink4a expression.

Results

PTEN loss blocks the decline in proliferation and restores the proliferation potential in adult murine β-cells

The proliferation rate of pancreatic β-cells starts to decline early in life (Teta et al., 2005). We observed a 1.4% static mitotic index (percent of β-cells positive for Ki67 staining) in 1.5-month-old mouse β-cells, which reduces to 0.2% when the mice are 3 months old (Fig. 1A). A further decline in mitotic activity is observed in older mice. We investigated the effect of PTEN loss on this proliferation decline by comparing the percentage of β-cells positive for Ki67 in control (PtenloxP/loxP; Rip-Cre; RosaLacZ or Ptenwt/wt; Rip-Cre+; RosaLacZ, Con) and Pten null (PtenloxP/loxP; Rip-Cre+; RosaLacZ) mice (Fig. 1A, top panel). The percentage of Ki67 positive cells was dramatically increased in the Pten null mice that are 3 months old and older (Fig. 1A, bottom right panel). Particularly, in 11- to 13-month-old mice, the proliferation index is about 0.5% in Pten null mice as compared to < 0.05% in the Cons. Interestingly, in 1.5-month-old mice where the Con β-cells retain relatively high mitotic activity and responsiveness to physiological and developmental regulations, PTEN loss had little effect on β-cell mitotic activity (Fig. 1A, bottom left panel). This divalent effect of PTEN on proliferation in young vs. old mice suggests that PTEN may selectively control β-cell proliferation in adult animals. Consistent with this hypothesis, we observed that the increase in islet mass in the Pten null mice is age dependent (Fig. 1B). In Con mice, the islet/pancreas ratio remained unchanged from 1 to 13 months of age (Fig. 1B, left panel). The islet mass in 1- to 3-month-old Pten null mice is twice that of the control's, presumably resulting at least partially from the effects of PTEN loss on β-cell size (Fig. S1). Starting from 3 months of age when higher mitotic activity is observed in the Pten null vs. Con mice (Fig. 1A), the islet/pancreas ratio increased exponentially (Fig. 1B). Additionally, Pten loss has a more profound effect on fasting plasma glucose and insulin levels in old vs. young mice (Fig. 1C,D). As reported previously, the islet morphology and hormone expression are normal (Fig. S2). Together, these results indicate that PTEN negatively regulates β-cell proliferation in aged animals.

Figure 1.

PTEN loss slows age-induced decline in β-cell proliferation and enhances islet mass in aged mice. (A) Mitotic activity of control and Pten null mice measured as Ki67 staining. Upper panels, representative images of Ki67 (green) and insulin (red) staining in young (1.5 months) vs. old (11–13 months) pancreas. Blue, DAPI for nuclei. Arrows indicate Ki67-positive cells within islets. Lower panels, quantitation of all stained sections showing the ratio of Ki67 positive to total β-cells. *Different from the age-matched controls, < 0.05. **Different from 3-month-old controls, < 0.05. n ≥ 5. Scale Bar, 50 μm. (B) Islet/pancreas ratio (bottom) from H&E-stained sections shown as representatives on top. *Different from age-matched controls, P < 0.05. **Different from age-matched controls and from 1.5-month-old Pten null mice, < 0.05. Scale Bar, 500 μm. n ≥ 5. n = 4 for 11- to 13-month-old Pten null mice. (C) Fasting plasma glucose levels. n ≥ 6, *< 0.05. Young: 1.5 months old. Old: 8–9 months old. (D) Fasting plasma insulin levels in mice where Pten deletion is induced at 3 months (young) and 9 months (old) of age. Insulin levels are measured 1 months after the last dose of tamoxifen. *< 0.05. n ≥ 4. Control, PtenloxP/loxP; Rip-CreER+; RosaLacZ injected with corn oil; Pten null, PtenloxP/loxP; Rip-CreER+; RosaLacZ injected with tamoxifen.

This initial observation suggested that PTEN may regulate the age-related regeneration potential of β-cells. We treated 9- to 15-month-old Pten+/− (PtenloxP/wt; Rip-Cre+) mice with STZ to induce injury and examined whether reduced PTEN level also rescues the decline in regeneration potential of β-cells following injury in the aged animals (Fig. 2). Islets induced by STZ have been shown to undergo spontaneous recovery in young but not old mice (Hartmann et al., 1989). We show here that 2 months after STZ treatment in 3-month-old mice, islet clusters are found in Pten null pancreas as well as surviving control mice (Fig. 2A) confirming previous observations. In older mice (9–15 months of age), proliferating cells (Ki67 staining) are observed 5 days following initial STZ treatment in Pten+/− islets but not control ones (Fig. 2B, 0.22 ± 0.06% vs. 0.04 ± 0.02%). As AKT has an important role in cell survival and Pten deletion also protects the islets from STZ-induced apoptosis (Stiles et al., 2006), a higher dose of STZ is used to induce more severe β-cell destruction. The high-dose STZ led to more cell death in Pten+/− mice vs. controls compared with low-dose STZ (Fig. S3). In both the low- and high-dose STZ-treated mice, PTEN loss led to higher mitotic activity (Figs 2B and S3), suggesting that the regenerative capacity in the old islets is induced when PTEN is inhibited. Two months after treatment with low-dose STZ, islet atrophic phenotype remains in the Con mice (Fig. 2C, bottom panel), whereas islet morphology is fully recovered in the Pten+/− mice (Fig. 2C, right panels) with islet/pancreas ratio being almost ten-fold higher in Pten+/− mice (1.3 ± 0.5%) vs. controls (0.15 ± 0.05%). This recovery is accompanied by the presence of Ki67-positive β-cells (Fig. 2C, bottom panel), suggesting that inhibition of PTEN reversed the inability of β-cells to regenerate in these aged mice. Together with the age-dependent effect of Pten deletion on β-cell proliferation (Fig. 1), these results support a role for PTEN and PTEN-regulated pathways in the age-dependent loss of regeneration potential of β-cells.

Figure 2.

PTEN loss restores the proliferation potential of aged β-cells. (A) Pancreas morphology in 3-month-old mice 2 months after streptozotocin (STZ) treatment. Inset, high magnification images. Representative image of pancreas sections from 20 animals. Scale Bar, 500 μm. (B) Mitotic index (Ki67 staining, green) in pancreas of 13-month-old control (Con) and Pten+/− (PtenloxP/wt; Rip-Cre+) mice 5 days after treatment with STZ. Left two panels, representative image of pancreas sections from three animals. Green, Ki67; red, insulin; blue, DAPI. Arrows indicate Ki67-positive β-cells. Right panel, quantitation of Ki67 ratio. n = 4; *P < 0.05. Scale Bar, 50 μm. (C) Two months after STZ treatment in 9- to 13-month-old control and Pten +/− mice. Top, H&E; bottom, Ki67 (green) and insulin (red). Scale Bar, 50 μm. Representative of four mice.

PTEN controls p16ink4a-regulated aging signals

A primary biochemical function of PTEN is to inhibit the action of PI3K. We show that phosphorylation of the PI3K effector, serine/threonine kinase AKT, is dramatically induced in the Pten null islets compared with the Con ones, whereas the amount of p-ERK is minimally altered (Fig. 3A). Thus, PI3K/AKT signaling is likely the major pathway through which PTEN loss induces the β-cell proliferation phenotype.

Figure 3.

Pten deletion leads to upregulation of cyclin D1 and downregulation of p16ink4a and p27. (A) Protein profiles in islets isolated from 3- to 6-month-old control (Con) and Pten null mice. (B) Left, protein profiles in islets isolated from 9-month-old mice. Right, quantitation of p16ink4a, cyclin D (cynD1), and p27 in 9-month-old isolated islets. n = 3; *different from that of control islets, < 0.05. Bottom, p16ink4a (green) and insulin (red) staining of pancreas sections from 9-month-old mice. Representative image of five mice. Scale bar, 50 μm. (C) Mitotic activity of Ink4a/Arf null β-cells compared with that from Pten null and control β-cells reported in Fig 1. n ≥ 3. *different from that of control islets, < 0.05. (D) Double-nucleotide labeling with CldU (red) and IdU (green) in 3-month-old mice. Blue, insulin. Solid white arrows point to nuclei with both CldU and IdU labels. Dashed arrow points to nuclei with single IdU or CldU labels. Inset: higher magnification images of labeled cells pointed by white arrows. Scale bar, 50 μm. (E) Percentage of double-positive β-cells among all labeled (either CldU or IdU) β-cells. n = 10; *different from that of control islets, < 0.05.

To further explore how PTEN-regulated signals control β-cell proliferation in older islets, we evaluated the levels of the cell cycle proteins, cyclin Ds, p27, and p16ink4a, in 3- to 6-month-old and 9-month-old islets (Figs 3 and S4). PTEN loss led to a reduction in p27 (32%) and p16ink4a (58%) expression and an increase in cyclin D1 (12%) expression in 3- to 6-month-old mice (Fig. 3A). However, neither cyclin D2 nor cyclin D3 amounts were altered. The downregulation of p16ink4a observed upon PTEN loss is more profound in older mice (> 9 months) where p16ink4a levels are higher and proliferation rate is lower (Figs 3B and S5). Interestingly, cyclin D1 expression is also significantly induced in 9-month-old islets when PTEN is lost, whereas no change in p27 upon PTEN loss is observed (Fig. 3B). These results suggest that the increased level of cyclin D1 and the decreased level of p16ink4a may play an important role in the ability of β-cells to proliferate in older mice that lack PTEN. To explore the importance of p16Ink4a in the age-dependent decline in pancreatic β-cell renewal capacity, we evaluated the proliferation rate of β-cells in young and old mice lacking p16ink4a (Figs 3C and S6). Similar to what was observed with Pten deletion, p16ink4a loss has little or no effect on the rate of proliferation measured by Ki67 in 1.5-month-old mice. In older mice (3 and 5–7 months old), Ki67 index in p16ink4a islets is comparable to that observed in Pten null islets. Both of which are significantly higher than that in the controls. This observation suggests that p16ink4a plays a major role in mediating the pro-mitotic and anti-aging effects of PTEN pathway.

Consistent with these observations, we found that PTEN regulates the cell cycle re-entry of β-cells. Using double-nucleotide labeling technique (Teta et al., 2007; Salpeter et al., 2010), we determined whether β-cells undergo one or multiple cell divisions within a defined experimental window. We found that majority of the β-cells in 3-month-old control mice only incorporate the nucleotides once, indicating that they only undergo one cell division (Fig. 3D, left panel). Deletion of Pten led to a doubling of cells capable of entering two cell divisions in the short experimental labeling time (Fig. 3D, right panel and E), suggesting that PTEN may function as a block to delay cell cycle re-entry (Fig. S7).

To investigate how PTEN regulates p16ink4a expression, we used immortalized mouse embryonic fibroblasts (mEFs) in which PTEN expression can be manipulated. In mEFs that lack PTEN, p-AKT, and D cyclins are unequivocally induced. Similar to pancreatic islets, the expression of both p16ink4a protein and mRNA, but not p53, is low in immortalized mEFs lacking PTEN (Pten null) and high in mEFs with intact PTEN (Con; Fig. 4A). As these cells recapitulated the expression profiles that we observed in mouse islets, it provided us the opportunity to directly evaluate the relationship between PTEN and p16ink4a. In Con mEFs, we showed that PTEN knockdown resulted in reduction in p16ink4a in both mRNA and protein levels (Fig. 4B). Induction of PTEN expression in PTEN-deficient U87 cells led to an increase in p16ink4a expression, concomitant with a decrease in AKT phosphorylation and cyclin D1 expression (Fig. 4C).

Figure 4.

PTEN regulates expression of p16ink4a. (A) Expression of cell cycle proteins in mouse embryonic fibroblasts (mEFs) with (control) or without PTEN (Pten null). Quantitative analyses of p16ink4a protein and mRNA are shown in bottom left two panes. Lower right, quantitation of p53 protein levels. Tubulin or GAPDH was used as loading control. n = 3; *different from that of control mEF,< 0.05. (B) Expression of shPTEN in control mEFs led to upregulation of cyclin D1, phosphorylation of AKT, and decreases in p16ink4a. Actin is used as loading control. Bottom two panels, quantitative analysis of PTEN and p16ink4a mRNA levels in mEFs treated with scramble RNA (scRNA) or shPTEN. n = 3; *different from that of scRNA-transfected control cells, < 0.05. (C) Induced expression of PTEN in a human glioma cell line lacking PTEN (U87) led to increases in p16ink4a levels. (D) Immunohistostaining of p16ink4a (green) and insulin (red) in PtenloxP/loxP; RipCreER+; RosaLacZ mice (9 months old) treated with tamoxifen (+Tm, bottom) to induce PTEN loss. Top, vehicle (−Tm)-treated mice with intact PTEN. Representative image of six mice. Scale bar, 10 μm. (E) Western blot analysis of p16ink4a proteins in PtenloxP/loxP; RipCreER+; RosaLacZ islets with (+Tm) or without (−Tm) tamoxifen. Actin is detected as loading control. Representative results of three mice.

Furthermore, we induced Pten deletion in β-cells postnatally by injecting tamoxifen in PtenloxP/loxP; Rip-CreER+; RosaLacZ mice. Induced deletion of Pten in 9-month-old β-cells led to reduced p16ink4a expression and increased proliferation compared with age-matched vehicle-treated controls (Figs 4D,E and S8). Taken together, our in vitro and in vivo data demonstrate that PTEN regulates p16ink4a expression.

PTEN regulates p16ink4a expression through cyclin D1

To elaborate the mechanism by which PTEN regulates p16ink4a, we analyzed the histone methylation pattern of the p16ink4a promoter and gene, a major mechanism for the regulation of p16ink4a transcription (Agherbi et al., 2009). Using mEFs, we found that deletion of Pten led to enhanced trimethylation at histone 3 lysine 27 (H3K27) on the Arf/Ink locus where the p16ink4a gene is located. Both the promoter- and the gene-coding regions, particularly exon 1, are methylated when PTEN is lost (Fig. 5A). As polycomb (PRC) proteins regulate methylation of histone 3 of the Arf/Ink locus (Agherbi et al., 2009), we investigated the expression of two polycomb proteins, Ezh2 and Bmi-1. PTEN loss led to the induction of both Ezh2 and Bmi-1 (Fig. 5B,C). Downregulation of Ezh2 with shRNA, but not Bmi-1, led to increased expression of p16ink4a (Figs 5D and S9). These results indicate that Ezh2 regulates PTEN-mediated p16ink4a expression. To confirm that this signaling pathway is also induced in the islets when PTEN is lost, we determined the levels of Ezh2. We show here that Ezh2 expression is undetectable in control islets and is induced in islets lacking PTEN (Fig. 5E). These results indicate that Ezh2 may regulate p16ink4a in islets, similar to its role in mEFs, consistent with the idea that the PRC complex is required for the PTEN-mediated regulation of p16ink4a.

Figure 5.

PTEN regulates methylation of the Arf/Ink4 locus through modulating the expression levels of Ezh2. (A) ChIP assay of ink/Arf promoter and coding region occupied by methylated H3K27 vs. total histone 3 is higher in Pten null mEFs vs. controls. n = 3; *different from the controls at the same site. < 0.05. (B,C) Two polycomb group genes, Ezh2 (B) and Bmi-1 (C), are upregulated in Pten null mEFs. Top panels, western blotting images. Bottom panels, quantitation of protein and mRNA levels of Ezh2 and Bmi-1 in control (Con) and Pten null mEFs. n = 3; *different from that of the controls, < 0.05. (D) Knocking down of Ezh2 (left two panels), but not Bmi-1 (right panel), enhances the expression of p16ink4a in Pten null mEFs. Left two panels, p16ink4a and Ezh2 mRNA expression after introduction of Ezh2 shRNA. Right panel, p16ink4a and Bmi-1 mRNA expression after introduction of Bmi-1 shRNA. n = 3; *different from that of scramble RNA-transfected cells, < 0.05. (E) Expression of Ezh2 in control (Con) or Pten null mouse islets. Representative results of two mice 8–10 months old.

The cellular amounts of polycomb proteins are reported to depend on the cell cycle (Bracken et al., 2003). In agreement with this idea, our data show that serum withdrawal resulted in reduced expression of cyclin D1 and Ezh2 (and to a lesser extent Bmi-1; Fig. 6A). In the Pten null mEFs and islets, increased cyclin D1 expression correlated with downregulation of p16ink4a. Consistent with a role for cyclin D1 in regulating p16ink4a expression, we found that expression of a nuclear persistent and stabilized cyclin D1, cyclin D1-T286A, led to downregulation of p16ink4a, whereas knockdown of cyclin D1 expression resulted in its upregulation (Fig. 6B). In contrast, stabilization (expression of cyclin D2-T280A) or knockdown of cyclin D2 has less effect on p16ink4a expression (Fig. 6C). Together, these data indicate that cyclin D1 modulates the PTEN-mediated p16ink4a expression.

Figure 6.

PTEN regulates p16ink4a through cyclin D1/E2F/Ezh2. A. Serum starvation reduces levels of cyclin D1, Bmi-1, and Ezh2 in Pten null mEFs. *different from that of control cells, P < 0.05; **different from that of both control and Pten null cells, P < 0.05. n = 3. (B) Left panels, introducing a stable cyclin D1 mutant (T286A) led to downregulation of p16ink4a. Right panels, introducing shRNA for cyclin D1 resulted in robust induction of p16ink4a. *different from that of vector or scrambled RNA-transfected cells, < 0.05. n = 3. Bottom panels, protein expressions of cyclin D1. Vec, vector-transfected samples. (C) mRNA levels of cyclin D2 (top left) and p16ink4a (top right) in control cells transfected with vector or stable cyclin D2-T280A mutant; in scramble or cyclin D2 shRNA-transfected cells (bottom). n = 3. (D) Left panel, mRNA levels of p16ink4a decreases with expression of E2F1. Right panel, p16ink4a expression increases in E2F1 shRNA-transfected Pten null cells. *different from that of vector or scrambled RNA-transfected cells, < 0.05. n = 3. (E) Control mEFs were co-transfected with the Ezh2 promoter-driven luciferase construct and indicated transcription factors (E2F1-4). n = 3, *different from vector-transfected cells, < 0.05. (F) Levels of cyclin Ds, p16ink4a, PTEN, p-RB, Ezh2, and p-AKT in young (1–2 months) vs. old (12–15 months) mice. Tubulin, actin, and GAPDH are detected as loading controls. One to three mice were used for each time point.

Elevated cyclin D1 leads to the activation of E2F through phosphorylation of retinoblastoma (RB) proteins. We found that overexpression of E2F1 led to downregulation of p16ink4a expression, whereas knockdown of E2F1 robustly induced expression of p16ink4a (Fig. 6D). pRB-E2F signaling was shown to be an upstream regulatory node for Ezh2 expression (Bracken et al., 2003). We found that expression of E2Fs1-4, especially E2Fs1-3, resulted in the dramatic induction of an Ezh2 promoter activity (Fig. 6E). This suggests that E2Fs likely serve as the transcription factors downstream of cyclin D1 that promote Ezh2 expression. Together, these data establish a PTEN/cyclin D1/E2F/Ezh2/p16ink4a signaling axis for the regulation of p16ink4a expression. Collectively, data from our study suggest that this newly discovered signaling network controls the ability of aged β-cells to self-renew. Consistent with this idea, we observe a decrease in cyclin D1, and to a lesser extend cyclins D2 but not D3, in old vs. young islets, concurrent with downregulation of pRB and Ezh2 (Fig. 6F). These results correlate with the observed differences in p16ink4a and mitotic activity in the old and young mice. When PTEN is lost, restoration of β-cell growth and regeneration occurs as a result of removal of this p16ink4a-regulated block by upregulation of cyclin D1. In young mice where p16ink4a is not robustly expressed, and not playing a significant role in overall β-cell growth, PTEN loss has a less significant effect on proliferation.

Discussion

We have demonstrated a molecular and phenotypic connection between PTEN-regulated PI3K pathway and p16ink4a, the signal implicated in cellular aging of pancreatic β-cells. This represents a novel link because neither the relationship between PTEN and p16ink4a nor that of PI3K with p16ink4a had been established previously. In tumor cells, PTEN loss was shown to correlate with allelic loss of p16ink4a (You et al., 2002). Our study reveals that loss of PTEN in nontumor forming β-cells leads to downregulation of p16ink4a expression and partial block of the aging-related loss of proliferation capacity. Further analysis of the molecular mechanism establishes a novel PTEN/cyclin D1/E2F/Ezh2/p16ink4a signaling node. These results are likely relevant to the physiological function of aging as PI3K/AKT signal has been implicated in the aging process. Localized muscle expression of IGF-1, the ligand for PI3K/AKT signal, was recently found to prevent age-related muscle atrophy and potentiate regeneration (Musaro et al., 2001). Thus, our study provides a novel molecular mechanism for altered regeneration of aged cells.

The primary and characterized function of PTEN is its role as a negative regulator of PI3K/AKT signaling pathway. The PI3K pathway modulates growth factor signaling, including IGF-1, HGF, and PDGF among other mitogenic factors. Thus, through this function, PTEN loss is expected to have a broad spectrum of function on the molecular level. In this study, we demonstrate that the signals controlled by PTEN are linked with the molecular processes (cyclin D1/p16ink4a) that govern the ability of aged cells to proliferate. p16ink4a, of which the expression is reduced when PTEN is lost, has been used as a biomarker for physiological aging (Krishnamurthy et al., 2004). Expression of p16ink4a increases remarkably with age in a variety of human and rodent tissues (Beausejour & Campisi, 2006). Overexpression of p16ink4a leads to premature aging (Krishnamurthy et al., 2004, 2006). Thus, our results demonstrating that p16ink4a is downregulated when PTEN is lost provided a potential signaling mechanism that may govern the aging process through regulating p16ink4a. Indeed, expression of PTEN and p16ink4a is positively correlated with each other in healthy human islets (R2 = 0.84, n = 4 and 0.65, n = 3 in two independent experiments).

A potential process of cellular aging regulated by p16ink4a is cellular senescence. PTEN (and PI3K/AKT) has been shown to regulate the cellular senescence process in experimental models (Veldhuis et al., 2004; Chen et al., 2005; Alimonti et al., 2010; Kennedy et al., 2011). Although inconsistent data are present for how PTEN may induce senescence and the involvement of p16ink4, likely due to the different experimental systems used, PTEN downregulation and AKT induction have been linked to inhibition of p16ink4a (Veldhuis et al., 2004; Kennedy et al., 2011), supporting our observations.

p16ink4a binds to CDK4/6 and cyclin D complexes and inhibits the downstream phosphorylation of the retinoblastoma (RB) protein (Serrano, 1997). Genetic studies showed that the CDK4/6 signals are important for maintaining β-cell mass (Cozar-Castellano et al., 2004). Deletion of cdk4 (Kushner, 2006) or D-type cyclins (Kushner et al., 2005), the primary partners for CDK4/6, both leads to significantly smaller islet mass. Concurrent loss of RB and another pocket protein, p130, also allows β-cells to overcome the G1/S checkpoint (Harb et al., 2009). These studies have established that checkpoint entry into the S phase needs to be overcome for β-cells to regenerate. Our finding that PTEN regulates p16ink4a is a novel discovery demonstrating how the PTEN/PI3K signal interacting with this checkpoint machinery. Consistent with this idea, our results reveal that PTEN prevents β-cells from re-entering another cell cycle after they finish one cell division. Previous studies have suggested that cyclin D2 may be involved in the regulation of this refractory period in mice challenged with glucose (Salpeter et al., 2010, 2011). Whether the age-dependent change in refractory period depends on p16ink4a or other mechanism downstream of PTEN remains to be elucidated.

Accumulation of D-type cyclins is necessary for cells to progress from G1 into the S phase. The presence of p16ink4a blocks the cell cycle progression by binding and inhibiting the CDK/cyclin complex. In tumors and other cases of pathologically excessive proliferation, D-type cyclins and p16ink4a expression are often negatively correlated. Thus, accumulation of D-type cyclins and degradation of p16ink4a are thought to be both necessary for cells to progress through the G1/S transition. In addition, expression of Ezh2, which negatively regulates the p16ink4a promoter, increases significantly in dividing cells where cyclin D is high (Chen et al., 2011). This suggests that p16ink4a, Ezh2, and cyclin D expression may be co-regulated by the same signals. Our observation in mEFs and islets suggests that cyclin D1 is the major cyclin that controls Ezh2 and p16ink4a expression. These results are also supported by the recent report that PDGF regulates islet mass by inducing cyclin D1, but not cyclin D2 (Chen et al., 2011) even though genetic studies suggest that cyclin D2 plays a role in islet size (Georgia & Bhushan, 2004; Kushner et al., 2005). In younger mice, PTEN loss indeed led to increases in cyclin D2 expression (data not shown). This alteration is not observed in older mice where p16ink4a expression is high. Particularly, levels of cyclin D2 in islets remained fairly consistent with age. Thus, cyclin D1 appears to drive the age-dependent decline in proliferation, whereas cyclin D2 may have an effect on the development and induced response of the islets to mitogens when p16ink4a expression is relatively low.

Activation of the E2F transcription factors as a result of D-type cyclin accumulation is required for the transition to S phase. E2Fs were reported to bind the promoter of Ezh2, a negative regulator of p16ink4a expression in human diploid fibroblasts (Bracken et al., 2003). We find that E2Fs1-3 robustly induces the expression of Ezh2. Thus, the induction of cyclin D1 resulting from PTEN loss may lead to inhibition of p16ink4a by inducing E2Fs activity and subsequently Ezh2 expression, although we do not exclude the possibility that these factors may act independent of each other. Ezh2, a PRC2 member, is a methyltransferase responsible for methylation of H3K27 at the Ink/Arf locus where p16ink4a is located (Agherbi et al., 2009). Our data demonstrate that PTEN loss leads to increased methylation of the Ink/Arf locus. We show that inhibiting Ezh2 can restore p16ink4a expression, which is downregulated in Pten null mEFs. This observation is consistent with the requirement of Ezh2 to methylate H3K27 and establishes for the first time that PTEN and PTEN-regulated pathways control the expression of Ezh2. In islets, loss of either Bmi-1 or Ezh2 reduces the proliferation capacity of β-cells (Chen et al., 2009; Dhawan et al., 2009).

As a negative regulator of PI3K signaling, PTEN regulates growth factor signals, including insulin, IGF-1, HGF, and PDGF. Previous studies have reported that mice lacking various components of the IGF-1/insulin signaling pathway in β-cells have impaired β-cell functions (Kulkarni et al., 1999, 2002; Xuan et al., 2002). Mice lacking either insulin or IGF-1 receptor are unable to maintain the islet mass postnatally. Deletion of both receptors leads to more severe loss of islet mass, suggesting that both receptors support the growth of islet β-cells (Ueki et al., 2006). Consistently, mice lacking insulin receptor substrate 2 develop β-cell failure due to decreased proliferation/increased apoptosis (Withers et al., 1998). Similarly, manipulating PDGF and HGF signals also results in islet mass changes (Dai et al., 2005; Roccisana et al., 2005; Chen et al., 2011; Mellado-Gil et al., 2011). Genetic manipulation of PI3K, AKT, and GSK-3β, three downstream kinases, as well as forkhead transcriptional factor FOXO1, a downstream transcriptional factor, also leads to altered β-cell proliferation and function (Tuttle et al., 2001; Eto et al., 2002; Buteau & Accili, 2007; Tanabe et al., 2008). Collectively, these signals converge on G1/S transition machineries, which have the potential to regulate p16ink4a expression in aged β-cells. Recent GWAS studies suggest that polymorphisms of cell cycle regulatory genes are important for β-cell functions and are correlated with the development of type II diabetes (Voight et al., 2010). Thus, the PTEN/cyclin D1/E2F/Ezh2/p16ink4a signaling network that we discovered may be relevant to the physiological aging of β-cells and pathogenesis of diabetes.

Experimental procedures

Animals

Except where it indicates, PtenloxP/loxP; Rip-Cre+; RosaLacZ mice (Pten null) are used as experimental mice; and Ptenwt/wt; Rip-Cre+; RosaLacZ and PtenloxP/loxP; Rip-Cre; RosaLacZ mice were used as controls (Con). Consistent with previous observations (Stiles et al., 2006), the Pten null mice displayed an increase in islet mass, lower fasting glucose levels, and higher mitotic activity in the β-cells at 3 months of age (Fig. S2). β-cell injury is achieved by i.p. injection of (STZ; Sigma, St. Louis, MO, USA) (50 mg kg−1 BW daily for 5 days in low dose or 1 × 200 mg kg−1 BW high dose) in PtenloxP/+; Rip-Cre+ mice with the PtenloxP/+; Rip-Cre mice as controls. For the specific effect of Pten deletion in aged mice, 9-month-old PtenloxP/loxP; Rip-CreER+; RosaLacZ mice were used for inducing deletion of Pten in adult mice by five injections of tamoxifen (a total dose of 30 mg) as reported (Dor et al., 2004). Mice were analyzed 1 month after the last dose of tamoxifen. Corn oil (vehicle)-injected mice were used as controls. Animals were housed in a temperature-, humidity-, light-controlled room (12-h light/dark cycle), allowing free access to food and water. All experiments were conducted according to the USC IACUC research guidelines.

Cell culture

Control (Pten+/+) and Pten null (Pten−/−) mEFs were cultured as previously reported (Zeng et al., 2011). U87 cells were engineered to express PTEN under the control of a doxycycline inducible promoter (Woiwode et al., 2008). Cells were grown in DMEM supplemented with 10% Tet-free FBS, G418 at 1 mg mL−1, and blasticidin at 10 μg mL−1. Cells were treated with 1 μg mL−1 doxycycline for 0, 6, 24, and 48 h to induce PTEN expression.

Human islets

Human islets were isolated from brain dead cadaveric donor pancreata. Islet isolation was carried out using the modification of the previously published work (Ricordi et al., 1988) and in accordance with City of Hope Institutional Review Board (IRB) approved protocols.

Mouse islet isolation

The pancreas was perfused with collagenase P solution (0.5–0.8 mg mL−1) and digested at 37 °C for 13–17 min. Islets were then purified using Ficoll gradients with densities of 1.108, 1.096, 1.069, and 1.037 (Cellgro; Stiles et al., 2006). Dithizone was used to verify the purity of the islets.

Immunohistochemistry

Zn-formalin-fixed and paraffin-embedded sections were stained as reported (Stiles et al., 2006). Antibodies used are Ki67 (Thermo Scientific, Waltham, MA, USA), PTEN (Cell Signaling Tech., Danvers, MA, USA), cyclin D1 and p16ink4a, p27 (Santa Cruz, Dallas, TX, USA), and insulin (Invitrogen, Carlsbad, CA, USA). Antibodies for other pancreatic hormones were provided by Zymed (now Invitrogen). BrdU staining was performed as previously described (Stiles et al., 2006). TUNEL was performed using a kit from Roche.

CldU and IdU labeling

Ten-week-old mice were fed with drinking water containing 1 mg mL−1 CldU for 8 days, followed by 2 days of regular drinking water with no label, followed by another 8 days of 1 mg mL−1 IdU-containing water. CldU and IdU staining were performed as previously reported (Teta et al., 2007).

Plasmids/shRNA

pcDNA cyclin D1 T286A was obtained from Addgene (Newman et al., 2004). Cyclin D2 was amplified from mouse cDNA library and cloned into pcDNA plasmid (Invitrogen). Cyclin D2 T280A was generated using a mutagenic kit (Invitrogen). Ezh2-1283 to +60 bp promoter was cloned from mouse genomic DNA and inserted into pGL2 luciferase reporter vector (Promega, Madison, WI, USA). Gene knockdown was performed by introducing pSilencer-U6 neo or puro plasmids containing sequence-specific shRNA using the BioT transfection reagent (Bioland Scientific LLC, Paramount, CA, USA) followed by selection.

Western blot

Cell or tissue lysates were subjected to SDS-PAGE. Antibodies used are as follows: p16ink4a, AKT, p27, cyclin D1, and GAPDH from Santa Cruz; PTEN, pAKT, Cyclin D2 and D3, p53, p-ERK, and Ezh2 from Cell Signaling Tech.; β-actin (Sigma); Bmi-1 (Millipore, Billerica, MA, USA); and Tubulin (Abcam, Cambridge, MA, USA).

RNA isolation and quantitative RT–PCR analysis

Total RNA was isolated using TRizol reagent (Invitrogen) and was reverse-transcribed using Promega RT system. Real-time PCRs were performed on the ABI 7900HT Fast Real-Time PCR System. Primers for real-time PCR are listed in Table S1 and S2.

Chromatin immunoprecipitation assay

mEFs were treated with 1% formaldehyde upon confluence, followed by sonication to shear genomic DNA. The supernatant was then immunoprecipitated with anti-H3K27 m3 or anti-H3 antibodies (Santa Cruz). After reverse crosslink, the precipitated DNA fragments were used as templates for PCR analysis. Primers used for qPCR are listed in Table S3.

Luciferase assay

Ezh2 promoter luciferase reporter construct was co-transfected into mEFs with pRL-TK, which serves as internal control (Promega). Luciferase activity was assayed 48 h. post-transfection using the Dual-Luciferase Reporter Assay System (Promega).

Statistical analysis

Data are presented as means ± the standard error of the mean and analyzed by Student's t-test, with two-tailed P-values of < 0.05 considered statistically significant.

Acknowledgments

We thank Dr. Bryan Stiles for his editorial input. We thank Mr. James Gubbins, Ms. Rima Deshpande, Anketse Kassa, Melissa Kim, and Zhenrong Shi for their technical assistance. Wealso thank Drs. Jake Kushner, Norman Sharpless, and Bruce R. Zetter for helpful suggestions and for providing reagents.

Author contributions

N.Z. and B.L.S. involved in the study design; N.Z., K.Y., J.B., L.H., D.A., B.M.P., R.A., J.W. S., X.H., and V.M. conducted the experiments; N.Z., I.A., F.K., D.L.J., and B.L.S. analyzed the data; N.Z., D.L.J., and B.L.S. prepared the manuscript.

Financial disclosure

This work is funded by Zumberg Individual Award, NIDDK R21 DK075928-02 (B.L.S.) and R01 DK084241-01A1 (B.L.S.). N.Z. is partially supported by the USC CBM training fellowship.

Conflict of interest

There are no conflict of interests to disclose for any of the authors.

Ancillary