Nkx6.1 Controls a Gene Regulatory Network Required for Establishing and Maintaining Pancreatic Beta Cell Identity

All pancreatic endocrine cell types arise from a common endocrine precursor cell population, yet the molecular mechanisms that establish and maintain the unique gene expression programs of each endocrine cell lineage have remained largely elusive. Such knowledge would improve our ability to correctly program or reprogram cells to adopt specific endocrine fates. Here, we show that the transcription factor Nkx6.1 is both necessary and sufficient to specify insulin-producing beta cells. Heritable expression of Nkx6.1 in endocrine precursors of mice is sufficient to respecify non-beta endocrine precursors towards the beta cell lineage, while endocrine precursor- or beta cell-specific inactivation of Nkx6.1 converts beta cells to alternative endocrine lineages. Remaining insulin+ cells in conditional Nkx6.1 mutants fail to express the beta cell transcription factors Pdx1 and MafA and ectopically express genes found in non-beta endocrine cells. By showing that Nkx6.1 binds to and represses the alpha cell determinant Arx, we identify Arx as a direct target of Nkx6.1. Moreover, we demonstrate that Nkx6.1 and the Arx activator Isl1 regulate Arx transcription antagonistically, thus establishing competition between Isl1 and Nkx6.1 as a critical mechanism for determining alpha versus beta cell identity. Our findings establish Nkx6.1 as a beta cell programming factor and demonstrate that repression of alternative lineage programs is a fundamental principle by which beta cells are specified and maintained. Given the lack of Nkx6.1 expression and aberrant activation of non-beta endocrine hormones in human embryonic stem cell (hESC)–derived insulin+ cells, our study has significant implications for developing cell replacement therapies.


Introduction
Innovative strategies for diabetes therapy aim to replace lost insulin-producing beta cells by reprogramming other cell types or by deriving beta cells from pluripotent cells. Ectopic expression of the transcription factors Pdx1, Neurogenin 3 (encoded by the Neurog3 gene; Ngn3), and MafA has been shown to reprogram pancreatic exocrine acinar cells into beta-like cells [1]. Similarly, some success in reprogramming of liver cells into beta cells has been reported after misexpression of Pdx1, Ngn3, MafA, NeuroD, or Nkx6.1 [2][3][4][5][6]. Moreover, recent studies have demonstrated that pancreatic endocrine alpha cells can spontaneously convert into beta cells after near complete ablation of beta cells in adult mice [7]. Conversely, loss of beta cell identity and partial conversion of beta cells into other endocrine cell types has recently been identified as an early event marking beta cell failure in diabetes [8]. Thus, substantial plasticity exists between pancreatic cell types, and this plasticity could potentially be exploited to halt diabetes progression or to replenish beta cells in diabetic individuals. However, little is still known about the factors that control this plasticity.
During embryonic development, all endocrine cell types are derived from a common endocrine precursor population marked by the transcription factor Ngn3 [9,10]. Ngn3 activity is required for the specification of all endocrine cells [11] and the expression of Arx and Pax4, two transcription factors that control endocrine subtype choices downstream of Ngn3. Arx-deficient mice display a loss of alpha cells and concomitant increase in beta and delta cells, while Pax4-deficiency results in the opposite phenotype of reduced beta and delta cells but increased alpha cells [12,13]. Strikingly, forced expression of Pax4 in endocrine precursors and their differentiated progeny imparts a beta-like cell identity to differentiating precursors, resulting in hyperplastic islets with an excess of beta-like cells at the expense of the other endocrine cell types [14]. However, despite their increased beta cell mass, mice misexpressing Pax4 eventually become diabetic and succumb prematurely, suggesting that sustained expression of Pax4 is not compatible with normal beta cell function. Since Pax4 is normally absent from beta cells and only transiently expressed in endocrine precursors during embryogenesis [15], it is possible that proper beta cell development and maturation requires Pax4 downregulation. Similar to Pax4, misexpression of Pdx1 in endocrine precursors has also been shown to favor a beta cell fate choice over other endocrine cell types [16]. Unlike ectopic Pax4 expression, forced expression of Pdx1 did not reduce the numbers of delta and PP cells, but selectively affected the ratio between beta and alpha cells. Therefore, Pdx1 activity appears to primarily control the alpha versus beta cell fate decision, which is consistent with its expression in both beta and delta cells [13]. Nkx2.2 has recently been identified as a beta cell maintenance factor and stabilizes beta cell fate by repressing the alpha cell fate determinant Arx [17]. While these studies have provided insight into the factors involved in endocrine cell type specification and maintenance, still little is known about how these factors interact to establish and maintain gene expression programs characteristic of each endocrine cell type. In particular, it is unclear which molecular mechanisms operate in beta cell precursors to ensure that alternative endocrine lineage programs are repressed, while beta cell-specific programs are activated. Given the simultaneous initiation of multiple endocrine subtype programs in one cell with current human embryonic stem cell (hESC) differentiation protocols [18,19], such knowledge is critical for refining these protocols to support the differentiation of mature and functional beta cells in vitro.
In addition to MafA and Mnx1 (also called Hb9) [20][21][22], in the adult pancreas Nkx6.1 is among the few transcription factors exclusively detected in beta cells. During development, Nkx6.1 is first expressed in multipotent pancreatic progenitors, where it specifies an endocrine identity by repressing the pre-acinar transcription factor Ptf1a [23]. At later developmental stages, Nkx6.1 expression persists in common progenitor cells for the ductal and endocrine cell lineages before becoming eventually restricted to the beta cell lineage [24]. Whether or not Nkx6.1 plays a role in beta cell specification and maintenance remains unknown, largely due to the lack of appropriate genetic models to study this question. Excessive early acinar cell specification and reduced numbers of Ngn3 + cells in Nkx6.1 null mutant mice preclude their utility for such studies.
To determine the function of Nkx6.1 in endocrine cell type specification and beta cell maintenance, we generated novel genetic mouse models to conditionally inactivate or misexpress Nkx6.1 after endocrine precursors have been specified. Our studies reveal that Nkx6.1 is both necessary and sufficient to specify the beta cell lineage. Nkx6.1 prevents alpha cell specification in cooperation with Pdx1 by directly repressing Arx through competition with the Arx gene activator Isl1. Furthermore, inactivation of Nkx6.1 in beta cells causes loss of beta cell identity and conversion into delta cells. Our findings identify Nkx6.1 as a beta cell programming factor and uncover a transcriptional network that initiates and maintains beta cell-specific gene expression programs, while repressing programs of alternative endocrine lineages.

Results
Heritable expression of Nkx6.1 in endocrine precursors favors a beta cell fate choice at the expense of other islet cell types During pancreatic development, Nkx6.1 is expressed in a subset of Ngn3 + cells and is then exclusively maintained in beta cells [25], suggesting that the specification of non-beta endocrine cell types might require Nkx6.1 downregulation. To explore whether expression of Nkx6.1 in all, or at least the majority of, Ngn3 + cells is sufficient to allocate precursors to the beta cell lineage, we heritably expressed Nkx6.1 in Ngn3 + cells, utilizing a mouse line that allows for conditional overexpression of Nkx6.1 after expression of Cre recombinase (Nkx6.1 OE mice). In Nkx6.1 OE mice, concomitant expression of Nkx6.1 and enhanced green fluorescent protein (eGFP) is induced by Cre recombinase-mediated excision of a lacZ expression cassette flanked by loxP sites. The Nkx6.1 OE transgene design is analogous to the Z/EG transgene, in which Cre recombinase induces expression of GFP by recombining loxP sites flanking a lacZ cassette. Therefore, Z/EG mice were used as controls for the Nkx6.1 OE strain ( Figure 1A, 1B). We induced transgene recombination with Ngn3-Cre and compared the relative contribution of recombined GFP + cells to each of the five endocrine cell types in Ngn3-Cre;Nkx6.1 OE and Ngn3-Cre;Z/EG control mice. Notably, due to mosaic expression of the transgenes, not all hormone + cells expressed GFP. At postnatal day (P) 2, 55.161.7% of the GFP + cells expressed insulin in Ngn3-Cre;Z/EG control mice, while 86.162.5% of the GFP + cells were insulin + in Ngn3-Cre;Nkx6.1 OE mice ( Figure 1C, 1H, 1M; P,0.001), suggesting that Nkx6.1 favors a beta cell fate choice. Consistent with this notion, Nkx6.1-expressing endocrine precursor cells displayed a significantly decreased propensity to differentiate into glucagon + , somatostatin + , pancreatic polypeptide (PP) + , or ghrelin + cells compared to endocrine precursor cells from control mice ( Figure 1D-1G, 1I-1M). Since Nkx6.1 expression did not affect cell replication, as shown by analysis of the proliferation marker Ki67 ( Figure 1N), or survival [23], these data indicate that Nkx6.1 promotes beta cell differentiation from endocrine precursors at the expense of alternative endocrine fates.
Similar to Nkx6.1, conditional expression of Pax4 in mouse endocrine precursors results in beta cell specification at the expense of all other endocrine cell types [14]. The expression of Pax4 leads to oversized islets and is eventually accompanied by

Author Summary
Diabetes is a disease caused by the loss or dysfunction of insulin-producing beta cells in the pancreas. Recent studies suggest that modification of the beta cells' differentiation state is among the earliest events marking the progressive failure of beta cells in diabetes. Currently, very little is known about the factors that instruct cells to adopt beta cell characteristics and maintain the differentiated state of beta cells. We have discovered that a single transcription factor can instruct precursor cells of other endocrine cell types to change their identity and differentiate into beta cells. Conversely, inactivation of the transcription factor in endocrine precursors prevents their differentiation into beta cells and results in excess production of other endocrine cell types. When the factor is specifically inactivated in beta cells, beta cells lose their identity and adopt characteristics of other endocrine cell types, similar to what is seen in animal models of diabetes. Thus, we have identified a single factor that is both sufficient to program beta cells and necessary for maintaining their differentiated state. This factor could be an important target for diabetes therapy and could help reprogram other cell types into beta cells.
beta cell dysfunction and diabetes. To determine whether transgenic Nkx6.1 expression similarly causes islet and beta cell hyperplasia, we compared islet size in Ngn3-Cre;Nkx6.1 OE and control mice at P2 and at 5 months of age. Consistent with our observation that Nkx6.1 overexpression in adult beta cells does not stimulate beta cell expansion [26], Ngn3-Cre;Nkx6.1 OE mice displayed normal islet cell mass ( Figure S1A, S1B). Furthermore, 5-month-old Ngn3-Cre;Nkx6.1 OE mice exhibited normal glucose tolerance ( Figure S1C), showing that sustained expression of the Nkx6.1 transgene does not perturb glucose homeostasis.
To next explore the extent of endocrine precursor reprograming and to assess the maturity of beta cells in Ngn3-Cre;Nkx6.1 OE mice, we analyzed insulin + progeny of targeted endocrine precursors for the expression of critical beta cell markers and possible ectopic expression of non-beta endocrine cell markers. As expected, insulin + GFP + cells in Ngn3-Cre;Nkx6.1 OE mice expressed the beta cell marker Pdx1, MafA, and Pax6 at P2, showing that transgenic Nkx6.1 expression in endocrine precursors and their progeny does not impair beta cell maturation ( Figure S1D-S1I). Next, to determine whether Nkx6.1 is sufficient to fully repress alternative endocrine lineage programs during endocrine cell differentiation, we analyzed insulin + cells in Ngn3-Cre;Nkx6.1 OE mice for expression of Arx, Brn4, glucagon, and somatostatin. At P2, targeted GFP + cells in Ngn3-Cre;Nkx6.1 OE mice were largely indistinguishable from their counterparts in control mice and rarely displayed coexpression of insulin with any of these non-beta endocrine markers (Figure S1J-S1Q; arrowheads), suggesting that Nkx6.1 is effective in fully establishing a beta cell expression program.
Forced Nkx6.1 expression is not sufficient to reprogram differentiated non-beta endocrine cells into beta cells We next sought to determine whether Nkx6.1 acts by inducing cell fate conversion during differentiation of Ngn3 + endocrine precursors or by converting already differentiated non-beta endocrine cells. Since we observed residual glucagon + , somatostatin + , pancreatic polypeptide (PP) + , and ghrelin + cells misexpressing Nkx6.1 at P2 ( Figure 1M), we first examined whether these cells convert into beta cells during postnatal life, as observed after Pdx1 expression in Ngn3 + cells [16]. Different from Ngn3-Cre;Pdx1 OE mice, we found that targeted non-beta endocrine cells persisted in 5-month-old Ngn3-Cre;Nkx6.1 OE mice ( Figure S2A-S2H). Notably, the endocrine cell type ratios observed in Ngn3-Cre;Nkx6.1 OE mice at P2 are largely maintained at 5 months of age ( Figure S2I), suggesting that additional non-beta-to-beta cell fate conversion does not occur postnatally. To directly test whether forced expression of Nkx6.1 in differentiated alpha cells triggers their conversion into beta cells, we induced recombination of the Nkx6.1 OE transgene with Glucagon-Cre (Glc-Cre). Consistent with the persistence of targeted non-beta endocrine cells in adult Ngn3-Cre;Nkx6.1 OE mice, we failed to observe insulin + cells expressing GFP ( Figure S3). These findings pinpoint Nkx6.1 beta cell programming activity to a period between the Ngn3 + state and activation of hormone gene expression.
Nkx6.1 f/2 ;Ngn3-Cre;Z/EG mice were born at the expected Mendelian frequency, but died within the first few days after birth from dehydration and hyperglycemia; a phenotype indicative of a beta cell defect. To determine whether loss of Nkx6.1 affects the cell fate choice of endocrine precursors, we analyzed the fate of Ngn3 + cells in Ngn3-Cre;Z/EG and Nkx6.1 f/2 ;Ngn3-Cre;Z/EG mice. Based on our previous finding that Nkx6.1 prevents acinar cell fate specification [23], we first examined whether loss of Nkx6.1 in Ngn3 + cells allocates endocrine precursors to the acinar lineage. The ability of Ngn3 + cells to undergo endocrine-to-acinar cell fate conversion has been previously demonstrated in conditions of reduced Ngn3 gene dosage or impaired Notch signaling activity [30,31]. At P2, we found GFP + cells to be exclusively restricted to endocrine islets in both Ngn3-Cre;Z/EG and Nkx6.1 f/2 ;Ngn3-Cre;Z/ EG mice ( Figure 2E, 2F), revealing that Nkx6.1 deletion in Ngn3 + cells does not cause endocrine-to-acinar fate conversion. Thus, unlike multipotent pancreatic progenitors, which adopt an acinar cell identity in the absence of Nkx6.1 activity [23], Ngn3 + endocrine precursors are no longer competent to activate acinar gene expression programs after deletion of Nkx6.1.
Because Nkx6.1-deficient endocrine precursors differentiate into endocrine cells ( Figure 2F), we next sought to determine whether loss of Nkx6.1 affects the relative proportion of the different endocrine cell types arising from Ngn3 + cells. To examine the endocrine cell fate choice of Ngn3 + cells, we quantified how many of the recombined Ngn3-expressing cells were allocated to each endocrine cell lineage by co-staining for GFP as a lineage marker of Ngn3-cell progeny together with each of the five hormones individually. At P2, 55.161.7% of recombined cells were insulin + in Ngn3-Cre;Z/EG control mice, while only 16.665.4% of recombined cells were insulin + in Nkx6.1 f/2 ;Ngn3-Cre;Z/EG mice ( Figure 2G, 2L, 2Q; P,0.01), suggesting that Nkx6.1-deficient precursors have a lower propensity to differentiate into insulin + cells. To test whether Nkx6.1-deficient Ngn3 + cells instead adopt non-beta endocrine cell identities, we compared the percentage of Ngn3 + cells that contributed to each non-beta endocrine cell lineage in Ngn3-Cre;Z/EG and Nkx6.1 f/2 ;Ngn3-Cre;Z/EG mice.

Nkx6.1-deficient insulin + cells are polyhormonal and ectopically express alpha cell markers
To investigate whether Nkx6.1 mediates beta cell specification downstream of Ngn3 by regulating transcription factors necessary for beta cell development, we analyzed expression of the beta cell progenitor markers Pax4, MafB, and Pdx1 in Ngn3-Cre;Z/EG and Nkx6.1 f/2 ;Ngn3-Cre;Z/EG embryos during the peak period of beta cell differentiation at e15.5. Confirming our previous findings in Nkx6.1 null mutant embryos [24], Pax4 expression was not affected in Nkx6. Although reduced in numbers, we still observed targeted GFP + cells expressing insulin in neonatal Nkx6.1 f/2 ;Ngn3-Cre;Z/EG mice, raising the question of whether these Nkx6.1-deficient insulin + cells properly differentiate into beta cells. To investigate how lack of Nkx6.1 affects beta cell gene expression programs, we analyzed  [20,24]. We conclude that insulin expression can still be initiated in the absence of Nkx6.1, but that these insulin + cells lack key features of normal beta cells.
The lack of beta cell-specific markers in Nkx6.1-deficient insulin + cells raised the question of whether these insulin + cells also carry features of alternative endocrine lineages. To determine whether loss of Nkx6.1 in endocrine precursors results in the activation of mixed endocrine gene expression programs, we analyzed insulin + cells in Nkx6.  arrowheads in L; 0% of insulin + GFP + cells expressed somatostatin in control mice vs. 19.6% in Nkx6.1 f/2 ;Ngn3-Cre;Z/EG mice). These findings demonstrate that Nkx6.1 is critical for repressing alternative endocrine lineage programs and that beta cell-specific programs can only be induced to a limited extent when Nkx6.1 is lost.
Previous studies have shown that the number of glucagon + cells is increased in Pdx1 heterozygous mutant mice [32,33]. Furthermore, beta cells lose Nkx6.1 expression upon Pdx1 deletion in beta cells [34]. Combined with our observation that Nkx6.1 maintains Pdx1 expression during beta cell differentiation, these findings raise the possibility that Pdx1 and Nkx6.1 cooperate through a positive feedback loop to establish and maintain beta cell identity and to repress non-beta endocrine lineage programs. To test this idea, we analyzed wild type, Nkx6.1 f/+ ;Ngn3-Cre, Pdx1 +/2 , and compound heterozygous Nkx6.1 f/+ ;Ngn3-Cre;Pdx1 +/2 mice for the ectopic expression of non-beta endocrine hormones in insulin + cells at P2. While we saw no coexpression of somatostatin or PP with insulin in any of the four genotypes ( Figure 5A-5D; data not shown), glucagon and insulin co-positive cells were occasionally detected in all genotypes, including wild type mice ( Figure 5A-5D, 5I). Quantification of the percentage of insulin + cells also expressing glucagon revealed significantly more dual hormone-positive cells in compound heterozygous Ngn3-Cre;Nkx6.1 f/+ ;Pdx1 +/2 mice than in either single heterozygous mutant or in wild type mice ( Figure 5A-5D, 5I). As previously reported [32,33], Pdx1 +/2 mice displayed an increase in the glucagon to insulin cell ratio that was also seen in compound heterozygous Ngn3-Cre;Nkx6.1 f/+ ;Pdx1 +/2 but not in Nkx6.1 f/+ ;Ngn3-Cre mice ( Figure 5J). Thus, haploinsufficiency for Pdx1 but not Nkx6.1 increases alpha cell numbers, which may reflect a non-cell autonomous effect on alpha cell proliferation, as recently shown in a mouse model of conditional deletion of Pdx1 in beta cells [35]. To determine whether the mixed lineage identity of insulin + glucagon + cells is associated with the expression of Arx, we co-stained pancreatic sections of mice at P2 from all genotypes for insulin, glucagon, and Arx. The majority of insulin + glucagon + cells expressed Arx in all genotypes, although occasional insulin + glucagon + Arx 2 cells were also detected ( Figure 5E-5H; arrowheads). The observed increase in cells exhibiting mixed alpha/beta cell identity and Arx expression in Nkx6.1/Pdx1 compound heterozygous mice supports the notion that Pdx1 and Nkx6.1 cooperate in beta cell fate specification by preventing activation of alpha cell-specific gene expression programs during endocrine cell differentiation.
Nkx6.1 is a direct repressor of Arx and competes with the Arx activator Isl1 at an Arx intronic enhancer To further explore how Nkx6.1 prevents endocrine precursors from adopting alpha cell identity, we examined the relationship of the Nkx6.1 and Arx expression domains in progeny of Ngn3-expressing cells during development. Consistent with the Nkx6.1 Specifies and Maintains Beta Cells dependence of Arx expression on Ngn3 [12], Arx was confined to a domain that marks descendants of Ngn3-expressing cells ( Figure 6A90, 6B90; note, occasional Arx + GFP 2 cells in A and B can be explained by mosaic expression of the Z/EG and/or Ngn3-Cre transgenes). At e14.5, preceding the onset of the major wave of beta cell differentiation, the majority of Arx + cells also expressed Nkx6.1 and Arx + Nkx6.1 2 cells were rare ( Figure 6A; arrowheads). By e16.5, however, when large numbers of beta cells arise [36], GFP + cells seldom coexpressed Nkx6.1 and Arx ( Figure 6B; arrowheads). Thus, endocrine precursors initially activate Nkx6.1 and Arx concomitantly, but their expression domains become mutually exclusive during beta cell differentiation. Together with our finding that Nkx6.1 regulates Arx during this time window, the observed expression pattern of Arx and Nkx6.1 raised the possibility that Nkx6.1 functions as a transcriptional repressor of Arx.
To explore this hypothesis, we tested whether Nkx6.1 occupies Arx regulatory sequences and is capable of repressing Arx transcription. We identified two conserved Nkx6.1 binding motifs within 5 kb of the 59 end flanking region from the Arx transcriptional start site ( Figure 6C; site A and B) as well as 12 potential Nkx6.1 binding sites located in two previously characterized enhancers within the third intron of the Arx genomic sequence ( Figure 6C; site C, Re1) and the 39 flanking region ( Figure 6C; site D, Re2) [37]. Recently, the Arx Re1 and Re2 enhancers have been shown to be required for Isl1-mediated activation of Arx in alpha cells, suggesting that these two enhancers are critical for Arx transcription [37]. Chromatin immunoprecipitation (ChIP) analyses for Nkx6.1 revealed that Nkx6.1 directly and specifically associates with the Arx Re1 enhancer (site C) in the Min6 beta cell line ( Figure 6D) and in embryonic endocrine precursors isolated by fluorescence-activated cell sorting (FACS) from Ngn3 eGFP/+ embryonic pancreata at e15.5 ( Figure 6E). Discordant with a previous report, which reported binding of Nkx6.1 to site B in a beta cell line [17], no association was observed with the other sites (sites A, B, or D) containing Nkx6.1 motifs ( Figure 6D). Transfection of the aTC1-6 alpha cell line with an expression plasmid for Nkx6.1 and a luciferase reporter construct containing the Arx Re1 enhancer sequence revealed that Nkx6.1 significantly reduced reporter gene activity ( Figure 6F). Confirming previous findings [37], transfection of an expression plasmid for Isl1 activated the Re1 enhancer ( Figure 6G). Cotransfection of CMV-Nkx6.1 abolished the ability of Isl1 to activate the Re1 enhancer in a dosage-dependent manner, showing that Nkx6.1 and Isl1 regulate Arx antagonistically through competition at the Re1 enhancer. However, in the presence of Nkx6.1, addition of CMV-Isl1 was not sufficient to revert Nkx6.1-mediated repression ( Figure 6G), indicating dominance of Nkx6.1 repressive over Isl1 activator activity. While our experiments show that Nkx6.1 is able to repress the Arx Re1 enhancer in alpha cell lines, Nkx6.1 cannot evoke an alpha-to-beta cell fate change when misexpressed in differentiated alpha cells ( Figure S3). Thus, Nkx6.1-dependent Arx repression through the Re1 enhancer appears to be functionally most relevant during endocrine cell type specification, when Nkx6.1 prevents initiation of Arx expression ( Figure 4E, 4F).

Nkx6.1 is necessary for maintaining beta cell identity
It has recently been shown that Nkx2.2 is an obligatory repressor of Arx in differentiated beta cells and that the absence of Nkx2.2 repressor activity causes beta-to-alpha cell conversion in mice [17]. Since Nkx2.2 is expressed in both beta and alpha cells, it has been speculated that beta cell-specific repression of Arx might depend on Nkx6.1 [17]. To directly test this hypothesis, we deleted Nkx6.1 selectively in beta cells, using the rat insulin promoter II (RIP)-Cre transgene to recombine the Nkx6.1 f allele and the Rosa26 (R26)-YFP reporter allele. As expected, the YFP lineage label was largely confined to beta cells in control Nkx6.1 f/+ ;RIP-Cre;R26-YFP mice at 6 weeks of age ( Figure 7A-7E, 7O). In striking contrast, only a few Nkx6.1-deficient, YFP-labeled cells expressed insulin in Nkx6.1 f/2 ;RIP-Cre;R26-YFP mice ( Figure 7F, 7G, 7O), suggesting that cells that once activated the insulin promoter no longer expressed insulin. Analysis of YFP expression in conjunction with glucagon, somatostatin, and PP revealed that Nkx6.1-deficient beta cells adopted delta cell identity, but did not convert into alpha or PP cells ( Figure 7H-7J, 7O). Thus, loss of Nkx6.1 in differentiated beta cells no longer causes activation of glucagon and PP, as observed after Nkx6.1 inactivation in Ngn3 + endocrine precursors. These findings suggest that Nkx6.1 is necessary to repress delta cellspecific genes in beta cells, but that expression of alpha and PP cell-specific genes are inhibited through an Nkx6.1-independent mechanism once beta cells have formed.
To further test whether Nkx6.1 deficiency in beta cells could lead to partial activation of an alpha cell gene expression program, we examined YFP + cells for Arx expression. YFP + cells rarely expressed Arx in both control and Nkx6.1 f/2 ;RIP-Cre;R26-YFP mice ( Figure 7K, 7M), demonstrating that Nkx6.1 is no longer necessary for Arx repression after beta cells have differentiated. Likewise, Pdx1 expression in adult islet cells was Nkx6.1independent, as somatostatin + cells that arose from Nkx6.1deficient insulin-expressing cells were Pdx1 + ( Figure 7L, 7N). Our data reveal that gene regulation by Nkx6.1 is highly contextdependent. While Nkx6.1 is necessary for Arx repression and Pdx1 activation in beta cell precursors, both genes are regulated by Nkx6.1-independent mechanisms in mature beta cells.

Discussion
The specification of pancreatic endocrine cell types is governed by the transcription factors Pdx1, Pax4, and Arx [12,13,16]. However, as none of these transcription factors control the development of solely one endocrine cell type, the molecular mechanisms that confer lineage specificity have remained largely elusive. Here, we demonstrate that the beta cell-specific transcription factor Nkx6.1 is both necessary and sufficient to specify the beta cell lineage ( Figure 8A). We show that Nkx6.1 and Isl1 antagonistically regulate Arx expression in endocrine precursors through a direct transcriptional mechanism ( Figure 8B). Thus, our study identifies Nkx6.1 as a critical beta cell programming factor that promotes the beta cell fate choice by simultaneously inducing beta cell genes and repressing non-beta endocrine genes.

Nkx6.1: a master regulator of the beta cell fate choice
We found that forced expression of Nkx6.1 allocates endocrine precursors to the beta cell lineage. While a similar activity has been described for Pdx1 [16], Nkx6.1 and Pdx1 display different kinetics of beta cell programming. As evidenced by cells coexpressing insulin, glucagon, and Arx at birth [16], forced expression of Pdx1 in endocrine precursor cells initially produces cells with mixed alpha and beta cell identity. In contrast to Ngn3-Cre;Pdx1 OE mice, we rarely observed cells exhibiting both alpha and beta cell features in neonatal Ngn3-Cre;Nkx6.1 OE mice, suggesting that Nkx6.1 plays a critical role early during cell fate specification of endocrine precursors. This notion is consistent with our finding that Nkx6.1 acts as a direct transcriptional repressor of the Arx gene. While Pax4 has also been shown to act as a direct repressor of Arx [38], our observation that Pax4 expression is not affected by Nkx6.1 ablation suggests that Pax4 alone is not sufficient to repress Arx in beta cell precursors. Given  [15], it is possible that the observed derepression of Arx and glucagon in Pax4 mutant mice [12,38] is a consequence of Nkx6.1-deficiency.
In addition to repressing alpha cell-specific genes through the regulation of Arx, Nkx6.1 also reallocated delta, PP, and epsilon cell precursors to the beta cell lineage. Likewise, inactivation of Nkx6.1 in endocrine precursors resulted in increased production of cells of all non-beta endocrine cell types and ectopic expression of non-beta cell hormones in insulin + cells. The phenotype observed upon conditional activation or deletion of Nkx6.1 in endocrine precursors identifies Nkx6.1 as a potent general repressor of non-beta endocrine gene expression programs. While our study establishes Nkx6.1 as a direct repressor of the alpha cell fate determinant Arx, Nkx6.1 likely represses additional cell fate determinants critical for the specification of delta, PP, and epsilon cells. As little is known about the transcription factors mediating non-beta endocrine cell fate choices, identification of additional direct target genes for Nkx6.1 in endocrine cell fate specification will have to await future studies.
Similar to Nkx6.1, forced expression of Pax4 also conferred beta cell identity to precursors of all endocrine cell lineages [14]. However, in contrast to Pax4 misexpressing mice [14], we did not distance from the transcriptional start site. Nkx6.1 binds to site C (Re1 element) in the Arx locus in chromatin from Min6 cells (D) and FACS-sorted GFP + cells (E) from e15.5 pancreata of Neurog3 eGFP embryos analyzed by ChIP with antibodies against Nkx6.1 or control immunoglobulin G (IgG). Mouse glucagon promoter and intergenic primers were used as positive (+) and negative (2) controls, respectively. (F) Co-transfection of aTC1-6 cells with the Arx Re1 enhancer-luciferase construct, the CMV-Renilla expression construct, and with or without the CMV-Nkx6.1 expression construct. Lane one (M) represents basal luciferase expression of the minimal promoter. Luciferase activity was quantified relative to the expression of the minimal promoter. Activity of the Re1 enhancer is repressed by Nkx6.1. (G) Co-transfection of aTC1-6 cells with the Arx Re1 enhancer-luciferase construct, CMV-Renilla, and with different concentration of CMV-Nkx6.1 and CMV-Isl1, as indicated. Nkx6.1 prevents activation of the Arx Re1 enhancer by Isl1 in a dose-dependent manner (lanes 2-7). Luciferase activity was quantified relative to the expression of the Re1 enhancer. Increasing concentrations of Isl1 are not sufficient to restore baseline activity of the Re1 enhancer in the presence of 200 ng of CMV- Nkx6.1 (lanes 8-12). Scale bar = 50 mm. Error bars represent S.E.M; *p,0.05, **p,0.01, ***p,0.001. doi:10.1371/journal.pgen.1003274.g006 observe oversized islets or diabetes in adult mice misexpressing Nkx6.1. Our observations in Ngn3-Cre;Nkx6.1 OE mice are consistent with our previous study, showing that transgenic overexpression of Nkx6.1 in beta cells does not stimulate beta cell proliferation or perturb beta cell function [26]. Thus, despite their shared property as a direct repressor of Arx [38], Nkx6.1 and Pax4 must also have distinct targets in endocrine cells. The lack of adverse effects on beta cell function makes Nkx6.1 an excellent candidate for beta cell programming strategies.
The transcriptional network that specifies beta cells Unlike Pax4 expression, which we found to be independent of Nkx6.1, maintenance of Pdx1 expression downstream of Ngn3 required Nkx6.1 activity. Previous genetic studies in mice have shown that the expression of Nkx6.1 in beta cells also depends on Pdx1 activity [34], suggesting that these two ''pro-beta'' transcription factors reinforce each other's expression. Since both Pdx1 and Nkx6.1 control the expression of critical genes for beta cell function, such as MafA and Glut2 ( [20,25,34]; this study), the positive regulatory loop between Pdx1 and Nkx6.1 might be critical for initiating and stabilizing beta cell-specific gene expression programs during endocrine cell differentiation. The mutual reinforcement in gene expression between these two transcription factors also explains why the combined deletion of one copy of Pdx1 and Nkx6.1 was sufficient to destabilize the beta cell fate choice and to cause ectopic expression of glucagon in insulin + cells. Nkx6.1 could regulate Pdx1 either directly or indirectly by repressing an inhibitor of Pdx1 expression. Since Nkx6.1 directly represses Arx and Arx has previously been shown to repress both Pdx1 and Nkx6.1 [39], loss of Pdx1 in Nkx6.1-deficient precursors is most likely the consequence of Arx derepression ( Figure 8B).

Competence window of beta cell programming
We found that expression of Nkx6.1 did not force all Ngn3 + cells to differentiate into beta cells and that small numbers of alpha, delta, PP, and epsilon cells expressing Nkx6.1 persisted during adulthood. This implies that the competence of precursors to adopt a beta cell fate upon Nkx6.1 activation is limited to a short period during development when cells are still plastic and lineagespecific gene expression programs have not been fully established. Our observation that misexpression of Nkx6.1 in alpha cells failed to induce conversion of alpha into beta cells supports the notion that cells quickly lose the competence to respond to Nkx6.1 repressive cues as they undergo endocrine differentiation.
Our study shows that Nkx6.1 occupies the Arx enhancer in beta cells ( Figure 6D), but unlike its inactivation in endocrine precursors, Nkx6.1 inactivation in beta cells does not activate Arx. This finding suggests that Nkx6.1 is only capable of repressing Arx during endocrine cell differentiation. One potential mechanism that could account for Nkx6.1-independent Arx repression in beta cells is DNA methylation, which has recently been shown to occur at the Arx locus [40]. Interestingly, deletion of the DNA methyltransferase Dnmt3a or Dnmt1 in beta cells results in Arx derepression and spontaneous conversion of beta into alpha cells [17,40], suggesting that DNA methylation alone is sufficient to keep Arx repressed in beta cells. Dnmt3a is recruited to Arx by Nkx2.2 and loss of Nkx2.2 repressor activity causes spontaneous conversion of beta into alpha cells [17], demonstrating that Arx expression in beta cells can be readily induced when Nkx2.2 repressor function is removed. Whereas Nkx2.2 forms a complex with Dnmt3a in the 59 regulatory region of Arx [17], we show that Nkx6.1 occupies an intronic Arx enhancer, where it competes with Isl1 and prevents Isl1 from activating Arx. Thus, it appears that DNA methylation-dependent repression of Arx is particularly important for keeping Arx repressed in beta cells, whereas Nkx6.1 prevents Arx activation in differentiating beta cell precursors. Overall, the role of DNA methylation in restraining cell plasticity during development is still poorly understood. Knowledge of how islet cell type-specific genes are epigenetically modified as cells differentiate and how this process can be reversed will prove important for devising effective cell programming and reprogramming strategies.  [39] and activation of Nkx6.1 by Pdx1 [34], our data support a model whereby cross-repression between Arx and Nkx6.1 confers alpha versus beta cell precursor identity. In beta cell precursors, Nkx6.1 expression is reinforced by Pdx1, which is repressed by Arx in alpha cell precursors. doi:10.1371/journal.pgen.1003274.g008 Our study highlights the importance of Nkx6.1 as a beta cell programming factor during endocrine cell differentiation and shows that insulin expression can be initiated independent of Nkx6.1 and Pdx1. Strikingly, the insulin + Nkx6.1 2 Pdx1 2 cells observed after conditional inactivation of Nkx6.1 in Ngn3 + cells display a similar molecular profile as insulin + cells generated in vitro with current hESC differentiation protocols. Like Nkx6.1-deficient insulin + cells in Nkx6.1 f/2 ;Ngn3-Cre mice, hESC-derived insulin + cells are polyhormonal and fail to express Nkx6.1, Pdx1, MafA, and critical glucose transporters ( [18,19]; Sander laboratory, unpublished data), which suggests that Nkx6.1 is required for complete beta cell programming in mice and humans. Our studies now pave the way for exploring the effectiveness of Nkx6.1 in (re)programming strategies to generate functional beta cells for diabetes therapy.
Midday on the day of vaginal plug appearance was considered e0.5. Glucose tolerance tests were performed as previously described [26].

Immunohistochemistry, morphometry, and cell quantification
Tissue preparation, immunofluorescence and TUNEL staining were performed as previously described [49]. For detection of nuclear antigens, antigen retrieval was performed in pH 6.0 citrate buffer and sections were permeabilized in 0.15% Triton X-100 in PBS. The following primary antibodies were used at the given dilutions: When necessary, nuclei were counterstained with Hoechst 33342 (Invitrogen) at 10 mg/ml. Primary antibodies were detected with donkey-raised secondary antibodies conjugated to Cy3, Cy5, DyLight488 (Jackson ImmunoResearch) or Alexa488 (Molecular Probes) at 1:1500 dilution (1:500 for Cy5). ApoTome images were captured on a Zeiss Axio Observer Z1 microscope with Zeiss AxioVision 4.8 and figures prepared with Adobe Photoshop/Illustrator CS4. Where necessary, the Cy5 channel was pseudo-colored white. Images were processed in accordance with the Journal of Cell Biology figure manipulation guidelines.
For all morphometric analyses and cell quantifications, a total of 10 sections per mouse from at least three mice per genotype were analyzed. For TUNEL, proliferation, and cell lineage analyses, the number of GFP + Hoechst + marker + cells was manually counted, divided by the total number of GFP + Hoechst + (Ngn3-Cre-mediated lineage tracing) or marker + Hoechst + (RIP-Cre-mediated lineage tracing) cells, and multiplied by 100. For the analysis of Pdx1 and Nkx6.1 single and compound heterozygous mice, sections were stained for insulin, glucagon, and somatatostatin and all marker + cells (on average 2100 cells per pancreas) were manually counted. Insulin + cells were analyzed for the expression of glucagon or somatostatin. For quantification of GFP + marker + cells, on average 500 GFP + cells from at least five different sections per mouse were counted.
For islet cell mass measurements, images covering an entire pancreas section were tiled using a Zeiss Axio Oberver Z1 microscope with the Zeiss ApoTome module. The hormone + area and total pancreas area were measured using ImagePro Plus 5.0.1 software (Media Cybernetics) and islet cell mass was calculated as follows: (hormone + area/total pancreatic area) multiplied by pancreatic weight.
Based on previously identified motifs [50,51], a custom positional weight matrix was used to identify putative Nkx6.1 binding sites. ChIP assays were performed as described [52], using Nkx6.1 (P. Serup, 1:250) or rabbit IgG antisera. Immunoprecipitations were performed on Min6 cells (ATCC) or on 1.5610 6 GFP + cells isolated from 300 pancreata of Neurog3 eGFP/+ embryos at e15.5 by fluorescence activated cell sorting. Each ChIP assay was quantified in triplicate by qPCR. The following primer sequences were used: The Isl1 Re1 enhancer and CMV-Isl1 constructs as well as procedures for transient transfections and luciferase assays have been described previously [53]. Unless indicated otherwise, 0.1 mg of CMV-Nkx6.1 was transfected. Luciferase and Renilla expression were measured 48 hours post transfection. For each data point relative luciferase activity was quantified as the total luciferase units divided by the total Renilla units. All reporter gene analyses were performed in triplicate.