The cellular regulators PTEN and BMI1 help mediate NEUROGENIN-3–induced cell cycle arrest

Neurogenin-3 (NEUROG3) is a helix-loop-helix (HLH) transcription factor involved in the production of endocrine cells in the intestine and pancreas of humans and mice. However, the human NEUROG3 loss-of-function phenotype differs subtly from that in mice, but the reason for this difference remains poorly understood. Because NEUROG3 expression precedes exit of the cell cycle and the expression of endocrine cell markers during differentiation, we investigated the effect of lentivirus-mediated overexpression of the human NEUROG3 gene on the cell cycle of BON4 cells and various human nonendocrine cell lines. NEUROG3 overexpression induced a reversible cell cycle exit, whereas expression of a neuronal lineage homolog, NEUROG1, had no such effect. In endocrine lineage cells, the cellular quiescence induced by short-term NEUROG3 expression required cyclin-dependent kinase inhibitor 1A (CDKN1A)/p21CIP1 expression. Expression of endocrine differentiation markers required sustained NEUROG3 expression in the quiescent, but not in the senescent, state. Inhibition of the phosphatase and tensin homolog (PTEN) pathway reversed quiescence by inducing cyclin-dependent kinase 2 (CDK2) and reducing p21CIP1 and NEUROG3 protein levels in BON4 cells and human enteroids. We discovered that NEUROG3 expression stimulates expression of CDKN2a/p16INK4a and BMI1 proto-oncogene polycomb ring finger (BMI1), with the latter limiting expression of the former, delaying the onset of CDKN2a/p16INK4a-driven cellular senescence. Furthermore, NEUROG3 bound to the promoters of both CDKN1a/p21CIP1 and BMI1 genes, and BMI1 attenuated NEUROG3 binding to the CDKN1a/p21CIP1 promoter. Our findings reveal how human NEUROG3 integrates inputs from multiple signaling pathways and thereby mediates cell cycle exit at the onset of differentiation.


Neurogenin-3 (NEUROG3) is a helix-loop-helix (HLH) transcription factor involved in the production of endocrine cells in the intestine and pancreas of humans and mice.
However, the human NEUROG3 loss-of-function phenotype differs subtly from that in mice, but the reason for this difference remains poorly understood. Because NEUROG3 expression precedes exit of the cell cycle and the expression of endocrine cell markers during differentiation, we investigated the effect of lentivirus-mediated overexpression of the human NEUROG3 gene on the cell cycle of BON4 cells and various human nonendocrine cell lines. NEUROG3 overexpression induced a reversible cell cycle exit, whereas expression of a neuronal lineage homolog, NEUROG1, had no such effect. In endocrine lineage cells, the cellular quiescence induced by short-term NEUROG3 expression required cyclin-dependent kinase inhibitor 1A (CDKN1A)/p21 CIP1 expression. Expression of endocrine differentiation markers required sustained NEUROG3 expression in the quiescent, but not in the senescent, state. Inhibition of the phosphatase and tensin homolog (PTEN) pathway reversed quiescence by inducing cyclin-dependent kinase 2 (CDK2) and reducing p21 CIP1 and NEUROG3 protein levels in BON4 cells and human enteroids. We discovered that NEUROG3 expression stimulates expression of CDKN2a/p16 INK4a and BMI1 proto-oncogene polycomb ring finger (BMI1), with the latter limiting expression of the former, delaying the onset of CDKN2a/ p16 INK4a -driven cellular senescence. Furthermore, NEUROG3 bound to the promoters of both CDKN1a/p21 CIP1 and BMI1 genes, and BMI1 attenuated NEUROG3 binding to the CDKN1a/p21CIP1 promoter. Our findings reveal how human NEUROG3 integrates inputs from multiple signaling path-ways and thereby mediates cell cycle exit at the onset of differentiation.
NEUROGENIN-3 (NEUROG3) is a basic helix-loop-helix (HLH) 2 transcription factor involved in gastrointestinal endocrine (GIE) cell lineage determination (1). In the mouse intestine, some NEUROG3 ϩ cells are in cycle, but the majority are postmitotic and beginning to express early GIE cell markers (2). Similarly, in the developing mouse pancreas, mitotically active progenitor cells exit the cell cycle after the initiation of Neurog3 expression, in a process involving the cyclin-dependent kinase inhibitor 1a (Cdkn1a/p21 CIP1 ) (3). Furthermore, ectopic expression of Neurog3 in early postnatal ␤-cells attenuates their proliferation, and results in a reduction of total ␤-cell mass leading to diabetes mellitus. Neurog3 expression seems to drive mouse pre-endocrine lineage cells from the cell cycle, initially inducing cellular quiescence. Cellular senescence sets in as the exit from cell cycle becomes irreversible.
The polycomb gene Bmi1 (4) is also involved in cell cycle exit during mouse endocrine cell lineage development. Proliferating ␤-cells from young mice express high levels of Bmi1, and low levels of Cdkn2a/p16 INK4 (5-7). As mice age, BMI1 is displaced from the p16 INK4 locus, resulting in increased p16 INK4 expression and the consequent withdrawal of proliferating ␤-cells from the cell cycle. Similarly, ␤-cells in Bmi1-null mice This work was supported by National Institutes of Health NIDDK Grants DK083762 and DK41301, and NIAID Grant U01DK085535. The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This article contains Figs. S1-S3 and Tables S1-S3. 1 To whom correspondence should be addressed. cro ARTICLE become quiescent prematurely, reducing ␤-cell numbers with consequent failure of glucose homeostasis (5)(6)(7).
The role of BMI1 as a mediator of senescence and quiescence is further highlighted by the recent observation that the population of Bmi1 ϩ cells in the mouse intestinal crypt is enriched with quiescent early GIE lineage cells (8), likely related to Neu-rog3-expressing cells known to be concentrated in the same crypt domain (2). Since Neurog3 expression is weak in the Bmi1 ϩ population (8), Bmi1 expression is likely downstream of Neurog3. Similar quiescent cells, identified by mTert expression, respond to post-fasting refeeding by repressing Phosphatase and tensin homolog (Pten) expression, and re-entering the cell cycle (9). These observations suggest that in mice, Neurog3's cell cycle effects occur in early GIE lineage cells and are mediated downstream by Bmi1 and Pten.
The phenotype of loss of NEUROG3 function is similar in humans and mice, however, there are clinically significant differences. Notably, whereas loss of NEUROG3 function in both humans and mice abolishes production of intestinal endocrine cells, ␤-cell development is less impacted in humans than in mice at birth, but results in severe diabetes in early childhood (10,11). Therefore, it is of interest to establish whether NEUROG3 expression in human pre-endocrine cells similarly induces exit from the cell cycle, and if so identify the mechanisms involved.
Here, we introduce the use of the BON4 cell line as a model for investigating the response of the human endocrine cell lineage to NEUROG3. We report that NEUROG3 initially induces quiescence in a p21 CIP1 -dependent manner. Furthermore, if NEUROG3 expression is prolonged, then the initially reversible cycle arrest (cellular quiescence) gradually becomes permanent (cellular senescence). The early NEUROG3-induced cellular quiescence is reversible by inhibition of PTEN, due to reduction in steady-state NEUROG3 levels in BON4 cells and human enteroids. We also report that the onset of NEUROG3-induced senescence is slowed due to concomitant induction of BMI1 expression, which acts by attenuating p16 INK4 expression.

NEUROG3 inhibits the cell cycle in an endocrine cell line, but NEUROG1 does not
BON cells are a well-characterized human endocrine lineage cell line (12), from which a homogenous appearing BON4 subline was isolated; it expresses very low levels of NEUROG3 and NEUROG1 (Fig. S1, A and B). BON4 cells are readily transduced with constitutively active lentiviruses-encoding NEUROG3 (cNEUROG3), NEUROG1 (cNEUROG1), or Control sequence (cControl); expression is detectable within 1 day of transduction (Fig. S1, A and B). NEUROG1 is a close proneural homolog of NEUROG3, not involved in GIE specification (13). In an unrelated study, 3 to be published separately, we confirmed that the cNEUROG3 lentivirus drives levels of NEUROG3 expression similar to that seen in untransduced primary cultures of human intestinal epithelium undergoing active endocrine cell production.
Cumate (Cum)-inducible cell lines, denoted BON4 Cum3NEUROG3 , BON4 Cum3NEUROG1 , and BON4 Cum3Control , were generated by transducing BON4 cells with Cum-inducible lentiviruses encoding NEUROG3, NEUROG1, or a Control sequence, respectively. Expression and protein were detectable within 1 day of adding Cum to the culture medium (Fig. S1, B and E).
We were unable to establish similar cell lines from BON4 cells successfully transduced with cNEUROG3 because the cells quickly ceased proliferating, as assessed by the MTT assay (Fig. 1A), and loss of expression of the active cell cycle marker Ki67 (Fig. 1B, Fig. S1C). Analysis of the DNA content of the transduced cells by flow cytometry confirmed that the majority were in G 0 -phase of the cell cycle, confirming that the cells exited the cell cycle (Fig. 1C). In contrast, control cells transduced with cNEUROG1 proliferated actively and were indistinguishable from those transduced with control lentivirus (cControl) (Fig. 1, A-C).
The nuclear localization signal was deleted from the cNEUROG3 construct (cNEUROG3:⌬NLS), to test the effect of reducing NEUROG3 translocation to the nucleus. MTT assays show that cNEUROG3:⌬NLS has a largely impaired ability to induce cell cycle exit (Fig. S1D). These results demonstrate that NEUROG3, but not NEUROG1, induces human endocrine lineage cells to exit the cell cycle, via mechanisms requiring translocation to the nucleus.

NEUROG3 inhibits the cell cycle in most nonendocrine cell lines, but NEUROG1 does not
We also found that transduction with cNEUROG3 induces cell cycle exit in cell lines derived from mesodermal tissues (macrophages, fibroblasts, and kidney) or endodermal tissues (stomach, colon, liver, and insulinoma), but transduction with cNEUROG1 had no effect (Fig. 1D).
NEUROG1 was previously reported to induce cell cycle arrest in P19 cells, a mouse teratoma cell line with a wellcharacterized ability to differentiate into neural cells (14). We confirm that P19 cells exit the cell cycle when transduced with cNEUROG1, whereas cNEUROG3 and cControl transduced cells were unaffected. Furthermore, a Cum-inducible P19 Cum3 NEUROG1 cell line was produced and responded to Cum induction by exiting the cell cycle, whereas a P19 Cum3 NEUROG3 cell line did not (Fig. 1, E and F).

NEUROG3 induces the expression of various mature GIE cell lineage markers in BON4 cells
Transduction of BON4 cells with cNEUROG3 induces both protein and transcript expression of common GIE cell markers including chromogranin A (CHGA), proprotein convertase subtilisin/kexin type 1 (PCSK1), gastrin (GAST), and somatostatin (SST) (Fig. 2, A and B). Thus, NEUROG3 not only induces cell cycle exit, but it also drives GIE cell differentiation.

PTEN/BMI1 and NEUROG3 cellular arrest
we used to assess whether p21 CIP1 is required for NEUROG3induced cell cycle arrest. Dox induction of the shRNA decreased the p21 CIP1 transcript level by 90% (Fig. 3, B, inset, and C). A control cell line, BON4 Cum3 NEUROG3 Dox3shRNA(Control) , transduced with a Dox-inducible control shRNA was also generated.
The BON4 Cum3 NEUROG3 Dox3shRNA(p21CIP1) and BON4 Cum3 NEUROG3 cells proliferated normally in the absence of Cum or Dox. However, proliferation in both lines was strongly repressed upon addition of Cum to the culture medium, due to the induction of NEUROG3 expression (Fig. 3B). In the absence of Cum, proliferation of BON4 Cum3 NEUROG3 Dox3shRNA(p21CIP1) cells were not impacted by depleting p21 CIP1 by adding Dox to the medium (Fig. 3B). Importantly, blocking p21 CIP1 expression by adding Dox to the medium 3 days before inducing NEUROG3, strongly attenuated the cell cycle arrest and expression of endocrine differentiation markers induced by NEUROG3 (Fig. 3, B and C).
The rate of NEUROG3 degradation is increased if it is phosphorylated by cyclin-dependent kinases (CDKs) (15,16). Depletion of p21 CIP1 , a CDK inhibitor, could thereby lead to reduced NEUROG3 protein levels. Indeed, treatment of Cuminduced BON4 Cum3 NEUROG3 Dox3shRNA(p21CIP1) cells with Dox results in a dramatic reduction in steady-state levels of NEUROG3 (Fig.  3D). However, depletion of p21 CIP1 expression in control BON4 Cum3 NEUROG1 Dox3shRNA(p21CIP1) cells had no effect on proliferation and NEUROG1 levels (Fig. 3D, data not shown).
Finally, we confirmed that NEUROG3, but not NEUROG1, directly interacts with E box elements within the immediate promoter region of p21 CIP1 . Direct functional interaction of NEUROG3 with the human p21 CIP1 promoter was demonstrated using a luciferase expression reporter assay (Fig.  S1F), and a chromatin immunoprecipitation (ChIP) assay of BON4 cells expressing FLAG-tagged NEUROG3 (Fig. S1G). These finding demonstrate that NEUROG3-induced cell cycle exit is mediated by direct interaction with the p21 CIP1 promoter.

NEUROD1 expression is not required for NEUROG3-induced quiescence
The bHLH transcription factor NEUROD1 is a major downstream NEUROG3 target (17), so we assessed its role in NEUROG3-induced quiescence. NEUROD1 protein is readily detectable in BON4 cells, and its expression increased following transduction with cNEUROG3, but not following cNEUROG1 transduction (Fig. S2A). We transduced BON4 Cum3 NEUROG3 cells with a Dox-inducible NEUROD1 shRNA lentivirus, establishing a BON4 Cum3 NEUROG3 Dox3shRNA(NEUROD1) doubly inducible cell line. Dox induction of shRNA expression significantly decreased NEUROD1 mRNA and protein levels (Fig. S2B). Importantly, the NEUROG3-induced cell cycle exit was not rescued by blocking the NEUROG3-induced up-regulation of NEUROD1 (Fig. S2C); therefore, NEUROD1 up-regulation is not required for the NEUROG3-induced cell cycle exit.

PTEN/BMI1 and NEUROG3 cellular arrest
observation indicates that baseline NEUROD1 expression promotes BON4 cell proliferation.

NEUROG3-induced cell cycle exit is initially reversible but becomes irreversible with time
BON4Cum3 NEUROG3 cells undergoing various durations of Cum-induced cell cycle exit differ in the extent of their recovery 2 days after cessation of Cum treatment, by which time NEUROG3 levels are significantly reduced (Fig. S1E, Fig.  4B). The cell cycle exit induced by short-term treatment (Ͻ6 days) was reversible, whereas long-term treatment (10 days) was not (Fig. 4A). Similarly, the induction of expression of the endocrine differentiation marker CHGA declined 2 days after a short-term (Ͻ6 days), but not after long-term treatment, suggesting that short-term exposure to NEUROG3 is insufficient to "lock in" an endocrine cell fate (Fig. 4B).

PTEN inhibition reverses NEUROG3-induced quiescence, by decreasing NEUROG3 protein
The phosphatase PTEN has been implicated in the process of cell cycle re-entry by quiescent mouse intestinal cells (9). PTEN is ubiquitously expressed in human cells (18), prompting us to test if PTEN function is involved in NEUROG3-induced quiescence in human BON4 cells.
A NEUROG3-induced quiescent state was invoked by adding Cum to the culture medium of BON4 Cum3 NEUROG3 . On day 3 of Cum treatment, bisperoxovanadium-pic (bpV(pic), a specific PTEN inhibitor (19); 100 nM final concentration), or an equal volume of vehicle in matched controls, was added to the culture medium. Control BON4 Cum3 NEUROG1 and BON4 Cum3 Control cells underwent similar Cum and bpV(pic) treatments. Cell proliferation and NEUROG3 protein levels were measured 1 day later. cells were treated with Dox, and processed every other day for 13 days and a MTT assay was performed at different time points. Inset, p21 mRNA levels with or without Dox at day 3. C, BON4 Cum3 NEUROG3 Dox3shRNA(p21CIP1) cells treated with Cum with or without Dox, and Western blot analysis performed at various time points using anti-CHGA, p21, and tubulin. D, Western blotting of BON4 Cum3 NEUROG1 Dox3shRNA(p21CIP1) and BON4 Cum3 NEUROG3 Dox3shRNA(p21CIP1) cells treated with Cum with or without Dox, and protein isolate was probed with anti-FLAG and tubulin Abs. Statistics: B, two-way ANOVA and two-way Tukey's multiple comparisons with adjusted p values.

Prolonged NEUROG3 expression leads to cellular senescence in a p21 CIP1 -dependent manner
Prolonged exposure of BON4 Cum3 NEUROG3 cells to Cum gradually induces strong p16 INK4a expression (Fig. 3A). Such cells fail to re-enter the cell cycle following Cum withdrawal (Fig. 4A), indicating that they have become senescent. To confirm their senescence, we stained a time series of such cells for senescence-associated ␤-Gal (SA-␤-Gal), a marker of cellular senescence (20). SA-␤-Gal activity was undetectable in BON4 Cum3 NEUROG3 cells during the first 5 days of Cum exposure (Fig. 5A) despite the detection of low and slowly increasing levels of p16 INK41 expression at that time (Fig. 3A). However, SA-␤-Gal activity was discernible after 6 days of Cum exposure, and by 8 days most cells were SA-␤-Gal positive (Fig. 5A), confirming their senescence. Control BON4 Cum3 Control and BON4 Cum3 NEUROG1 cells treated with Cum continued proliferating (Fig. 1A) and remained negative for SA-␤-Gal (data not shown).
Robust expression of p21 CIP1 is necessary to institute NEUROG3-induced quiescence (Fig. 3, A and B), but it remains unclear whether prolonged NEUROG3 expression, in the absence of up-regulation of p21 CIP1 , is sufficient on its own to eventually drive cells into quiescence and senescence. We tested this by the simultaneous exposure of BON4 Cum3 NEUROG3 Dox3shRNA(p21CIP1) cells to Cum and Dox, which prevented quiescence and senescence for as long as we tested (15 days). In contrast, BON4 Cum3 NEUROG3 Dox3shRNA(p21CIP1) Cum alone rapidly became quiescent and eventually senescent. This demonstrates that NEUROG3-induced quiescence and senescence requires p21 CIP1 (Fig. 5, B and C). It should be noted, however, that in the absence of p21 CIP1 , the resultant Cum-induced NEUROG3 protein level is diminished (Fig.  3D), which may have weakened its impact.
For comparison, we examined the cell cycle state induced by treating BON4 Cum3 NEUROG3 Dox3shRNA(NEUROD1) cells with Dox, which represses baseline NEUROD1 expression, greatly attenuating proliferation (Fig. S2, A-C). These cells remained in a slowly cycling quiescent state for a prolonged period without becoming senescent, as indicated by their minimal growth and lack of SA-␤-Gal staining even after 15 days (Fig. S2D). Thus, prolonged NEUROG3 exposure in the absence of p21 CIP1 -induced quiescence, or prolonged quiescence in the absence of NEUROG3-induced p21 CIP1 up-regulation, are both insufficient to induce senescence in BON4 cells.

NEUROG3 induces BMI1 expression, which attenuates the induction of p16 INK4a , delaying the onset of senescence
The polycomb group protein BMI1 is involved in regulating pancreatic ␤-cell senescence by repressing p16 INK4a expression (see Introduction). This led us to suspect that BMI1 may similarly be involved in limiting p16 INK4a expression in the early quiescent phase of NEUROG3 expression in BON4 cells (Fig.  3A), thereby delaying the onset of senescence. Indeed, Cum treatment of BON4 Cum3 NEUROG3 cells led to increased BMI1 mRNA and protein levels within hours, whereas BMI1 did not change in Cum-treated control BON4 Cum3 NEUROG1 cells ( Fig.  6A; Fig. S3A; data not shown). Direct functional interaction of NEUROG3 with the human Bmi1 promoter was demonstrated using a luciferase expression reporter assay (Fig. 6B), and a ChIP assay of BON4 cells transfected with FLAG-tagged NEUROG3 (Fig. 6C) (21).
We assessed whether Bmil impairs NEUROG3 binding to the E box in the p21 CIP1 promoter (22). We examined the loss of BMI1 function with specific Bmil inhibitor PTC-209 (1.25 M). The addition of Bmil inhibitor significantly enhanced NEUROG3 induction of the human p21 CIP1 promoter using a luciferase expression reporter assay (Fig. 6D), and by ChIP assay in BON4 cells (Fig. 6E).
The impact of loss of BMI1 in the context of NEUROG3induced quiescence was determined by simultaneous treatment of BON4 Cum3 NEUROG3 cells with Cum and PTC-209. The cells exhibited a dramatic increase in expression of p16 INK4a , p19 INK4d , and CHGA expression (Fig. 6F), accompanied by an accelerated onset of cellular senescence as assessed by SA-␤-Gal (Fig. 6G). In contrast, inhibition of BMI1 in cells not treated with Cum (and hence not up-regulating NEUROG3) did not undergo senescence. Taken together, these data indicate that within this in vitro system, short-term NEUROG3 induces p21 CIP1 and BMI1 expression, the latter acting to attenuate the up-regulation of p16 INK4a and p19 INK4d levels, thereby delaying the onset of NEUROG3-induced cellular senescence.

Discussion
We have found that the human endocrine-derived BON4 cell line responds to the pro-endocrine transcription factor NEUROG3 by rapidly exiting the cell cycle and expressing markers of endocrine differentiation as outline in Fig. 6H. A, MTT assay of BON4 Cum3 NEUROG3 cells treated with either vehicle or Cum for 5 or 14 days and processed, whereas others were treated for a similar amount of time, and then Cum was removed for 2 days before processing. B, BON4 Cum3 NEUROG3 cells were treated with either vehicle or Cum for various times, whereas others were treated for a similar amount of time, and then Cum was removed for 2 days before processing. Cell extracts were assessed by Western blotting with anti-NEUROG3, CHGA, or tubulin. C, MTT assay of BON4 Cum3 NEUROG3 , BON4 Cum3 NEUROG1 , and BON4 Cum3 Control cells treated with Cum for 5 days with or without the PTEN inhibitor, bpV(pic), at a concentration of 100 nM on the last day. D, Western blotting of BON4 cells transiently transfected with a NEUROG3 expressing vector 5 days earlier with or without bpV(pic). E, Western blotting of BON4 Cum3 NEUROG3 , and BON4 Cum3 NEUROG1 cells treated with Cum for 3 days, and then with or without the PTEN inhibitor for 3 additional days. Blot was probed with anti-FLAG (to detect the C-terminal FLAG tag of NEUROG3/NEUROG1), anti-p21 CIP1 , or anti-tubulin antibody. F, Western blotting of human enteroids treated with and without the PTEN inhibitor for 3 additional days. Statistics: C, two-way ANOVA and two-way Tukey's multiple comparisons with adjusted p values.

PTEN/BMI1 and NEUROG3 cellular arrest
Interestingly, we also found that NEUROG3 drives a similar cell cycle exit in a wide variety of cell lines, but that its homolog NEUROG1 does not. P19 cells is a singular exception. We found that NEUROG3 expression has little effect on P19 cells, whereas NEUROG1 stimulates p27 KIP1 expression and induces cell cycle exit (Fig. 1, E and F, data not shown), consistent with the cell line's neurogenic response to NEUROG1 (14). The failure of NEUROG3 to act in P19 cells, despite P19 cell's well-characterized pluripotency, warrants future investigation.
We also found that if p21 CIP1 expression is repressed, then NEUROG3 fails to induce BON4 cell quiescence and endocrine differentiation (Fig. 3, B and C), as others have shown in mouse pancreas (3,23). This dependence of NEUROG3 on p21 CIP1 is due to the direct action of NEUROG3 at the p21 CIP1 promoter (Fig. S1, F and G), explaining why deleting the nuclear localization signal impairs NEUROG3's ability to induce cell cycle arrest (Fig. S1D).
Phosphorylation of NEUROG3 has recently been shown to increase its degradation (15,16). This explains our observation that NEUROG3 protein levels were diminished after p21 CIP1 repression (Fig. 3D) or PTEN inhibition (Fig. 4, E and F), both of which regulate protein phosphorylation (24 -26).
In mouse pancreas, newly differentiating endocrine cells require a short burst of high levels of NEUROG3 exposure to induce differentiation, whereas adult mouse ␤-cells require tonic low levels of NEUROG3 expression to sustain hormone production (27). We have similarly found that transient NEUROG3 expression in BON4 cells induces cellular quiescence and commencement of endocrine differentiation, both of which are reversible if NEUROG3 is reduced (either by removing Cum-induction, or via increased NEUROG3 phosphorylation and consequent degradation following p21 CIP1 or PTEN inhibition). Prolonged expression of NEUROG3 results in senescence and sustained endocrine differentiation.
In BON4 cells, NEUROG3 also drives BMI1 expression, which in turn acts to delay senescence by constraining NEUROG3-induced expression of p21 CIP1 and p16 INK4a (Fig. 6,  A-F and H). Chemical inhibition of BMI1 function accelerates the NEUROG3-induced expression of p16 INK4a , p19 ARF , endocrine differentiation, and senescence (Fig. 6F). This may explain the pancreatic endocrine maturation strategy, intense short bursts induce cell cycle exit, initiates differentiation, and increases BMI1 expression, which acts to restrain premature p16 expression, facilitating cell cycle re-entry during the postnatal islet growth, and remodeling. As ␤-cells mature, lowgrade expression of NEUROG3 reinforces the endocrine program, but the resultant low BMI1 and high p16 levels render ␤-cells more refractory to cell cycle re-entry.
It is important to note that whereas the BON4 cell line provides a tractable model in which to explore the consequences of a NEUROG3-induced endocrine state in human cells, its use has caveats. The parental BON cell line was generated from a metastatic human carcinoid tumor of pancreas (12), and hence displays abnormalities including continuous proliferation and high basal levels of hair/enhancer of split-1 (HES1) and NEUROD1 expression (Fig. S2A, data not shown). Hes1 is a direct target of Notch and is a potent transcriptional inhibitor

PTEN/BMI1 and NEUROG3 cellular arrest
of bHLH transcription factors, as well as CDK inhibitors (28,29). Therefore, it is plausible that the high basal levels of HES1 in BON4 cells represses endogenous expression of NEUROG3 and the CIP/KIP genes. Robust HES1 expression in BON4 cells would be expected to inhibit NEUROD1 expression; but in fact, NEUROD1 is surprisingly abundant in these cells, and as we have shown here, is required to drive their baseline prolifera-tion. In contrast, NEUROD1 is known to induce cell cycle arrest when expressed in various cell lines or in secretin-expressing GIE cells (30). Interestingly, we also found that BON4 cells express high levels of RFX6 mRNA, a transcription factor that functions both up-and downstream of NEUROG3 (data not shown). Although not assessed here, RFX6 may have a role in driving NEUROD1 expression in the absence of NEUROG3 (31).  BMI1 expression, which attenuates p16, p19, and the GIE phenotype. A, Western blotting of BON4 Cum3 NEUROG3 cells stimulated with Cum for 5 days in the presence or absence of BMI1 inhibitors for up to 2 days and assessed for BMI1 and tubulin. B, cells transiently transfected with human BMI1 promoter (Ϫ100/ϩ1) luciferase construct, and with various amounts of either NEUROG3 or NEUROG1 pGL3-basic expressing vector, and luciferase were measured and normalized with ␤-Gal after 3 days. C, ChIP assay for NEUROG3-FLAG binding to E-box elements in the BMI1 promoter. D, cells transiently transfected with human p21 promoter (Ϫ592/ϩ41) and various amounts of NEUROG3 pGL3-basic expressing vector with and without BMI1 inhibitor for the last 24 h and luciferase were measured and normalized with ␤-Gal after 3 days. E, ChIP assay for NEUROG3-FLAG binding to E-box elements in the p21 promoter of cells transfected with NEUROG3 and treated with and without Bmi1inh. F, quantitative PCR assessments of CHGA, NEUROG3, p16, p19, and p21 were performed from total RNA of BON4 Cum3 NEUROG3 cells treated with or without Cum for 3 days and BMI1 inhibitor for 3 days. G, SA-␤ Gal staining assay of BON4 Cum3 NEUROG3 cells treated with or without Cum in the presence or absence of BMI1 inhibitor. All experiments were performed 3-4 times in triplicate and representative data are presented. H, proposed model of PTEN and BMI1 modulation NEUROG3 effects on the cell cycle. Statistics: A-D, one-way ANOVA and Dunnett's multiple comparison and adjusted p values; E and F, two-way ANOVA and two-way Tukey's multiple comparisons with adjusted p values.

Statistics
The data throughout the manuscript are presented as the mean Ϯ S.E. Multiple t tests were performed using GraphPad Prism software, and significant p values (*, Ͻ0.01; **, Ͻ0.001; ***, Ͻ0.0001) are indicated in all the figures.

Antibodies, oligonucleotides, and vectors
The various antibodies, primers used for quantitative PCR (qPCR), vectors and viruses used are listed in Tables S1-S3. Oligonucleotides used for ChIP, and site-directed mutagenesis will be made available upon request.

Cell lines
Most established cell lines where obtained from ATCC and propagated and maintained as suggested. The BON4 cell line used in this study was generated from the BON cell line, obtained from Dr. Townsend (University of Texas Medical Branch, Galveston, TX). The parental BON that had a more heterogeneous appearance and growth characteristics were maintained in Dulbecco's modified Eagle's medium/F-12 media supplemented with 10% FBS (12).

Lentiviral development and transduction
The single exon NEUROG1 was PCR amplified from human genomic DNA with a C-terminal FLAG tag oligonucleotide. Sequence encoding a C-terminal FLAG tag was also added to a previously described human NEUROG3 cDNA and along with NEUROG1 the constructs were cloned into a lentivirus-based vector to generate constitutively active lentivirus (10). To generate the NEUROG1-and NEUROG3-inducible lentiviruses, both were subcloned into SparQ-IRES-GFP lentivector. Attenuation of various transcripts was performed using lentiviral shRNA that were purchased from a commercial vendor and are displayed in Table S3. Cells' containing an inducible system were supplemented with either Cum (30 g/ml) or Dox (1 g/ml) and replaced every other day.

Cell proliferation analysis by MTT and FACS
Cell proliferation was assessed by standard MTT assay (32). By FACS analysis, cells were harvested and fixed with 4% PFA. After fixation, cells were resuspended in PBS containing propidium iodine and RNase A, and analyzed for DNA content by flow cytometry using ModFitLt V3 (pMAC). Data are representative of at least 3 separate experiments performed in triplicate.

Luciferase reporter assay
Luciferase assay was performed as previously described (10). Various constructs were transfected using Lipofectamine, and lysates were used to measure luciferase that was normalized by Renilla luciferase.

Western blot analysis
Cells were lysed and 10 g of protein was loaded onto a 10% polyacrylamide gel and subjected to SDS-PAGE before blotting onto an immune blot from Amersham Biosciences, Hybond. Membranes were blotted using standard procedures using the antibodies shown in Table S1. Western blots were visualized using the ECL plus Western blotting kit and scanned on an imager. qPCR qPCR with cDNA synthesized from 5 g of total RNA was performed either with TaqMan assay, or using Perfecta SYBR Green SuperMix, Low ROX TM in a real-time Thermocycler. The results obtained for each individual gene were normalized to ␤-tubulin expression in the sample. The primers used were designed with Primer3, or as suggested by TaqMan, and are shown in Table S2.

SA-␤-Gal staining
SA-␤-Gal staining was used to determine replicative senescence (33). Cells were washed with PBS and fixed with 4% PFA, and SA-␤-Gal staining solution was incubated, and observed every 4 h for 12 h, and then every 12 h. The optimal incubation period was determined based on the visibility of stained cells in the test sample, but not in the control sample.

Chromatin immunoprecipitation (ChIP)
ChIP experiments were carried out using modifications to the micro-ChIP protocol (34). Briefly, BON4 cells were treated with 1% PFA to cross-link the DNA with bound proteins. After washing, the cells were lysed in SDS-based lysis buffer with protease inhibitors, sonicated to shear chromatin to an average size of 500 bp, and cellular debris was removed by centrifugation. Anti-FLAG-M2 antibody was bound to protein G-coated magnetic beads, and diluted chromatin lysate was incubated overnight with antibody-bound magnetic beads and washed in RIPA buffer and TE. Cross-links were reversed and the DNA was purified and quantified by qPCR using SYBR Green PCR. DNA enrichment data were estimated as the percentage bound/input ratio, presented relative to the un-induced control.

Immunofluorescent staining
BON4 cells transduced with one of several lentiviruses and were grown on glass coverslips and subsequently fixed in 4% PFA and washed with PBS before treating with blocking buffer and incubating with primary and secondary antibodies listed in Tables S1-S3.

Statistics
For all other multiple comparisons, one-, or two-way ANOVA was followed by Tukey's or Dunnett's multiple comparisons test. For comparisons involving one independent variable with two groups, we performed Student's t tests. Statistical significance was set at: *, p Ͻ 0.05; **, 0.01; and ***, 0.005, with respect to final (adjusted) p values. Variation is reported as S.E.