Interactions between Areas I and II Direct pdx-1 Expression Specifically to Islet Cell Types of the Mature and Developing Pancreas*

PDX-1 regulates transcription of genes involved in islet β cell function and pancreas development. Islet-specific expression is controlled by 5′-flanking sequences from base pair (bp) -2917 to -1918 in transgenic experiments, which encompasses both conserved (i.e. Area I (bp -2761/-2457), Area II (bp -2153/-1923)) and non-conserved pdx-1 sequences. However, only an Area II-driven transgene is independently active in vivo, albeit in only a fraction of islet PDX-1-producing cells. Our objective was to identify the sequences within the -2917/-1918-bp region that act in conjunction with Area II to allow comprehensive expression in islet PDX-1+ cells. In cell line-based transfection assays, only Area I effectively potentiated Area II activity. Both Area I and Area II functioned in an orientation-independent manner, whereas synergistic, enhancer-like activation was uniquely found with duplicated Area II. Chimeras of Area II and the generally active SV40 enhancer or the β cell-specific insulin enhancer suggested that islet cell-enriched activators were necessary for Area I activation, because Area II-mediated stimulation was reduced by the SV40 enhancer and activated by the insulin enhancer. Several conserved sites within Area I were important in Area I/Area II activation, with binding at bp -2614/-2609 specifically controlled by Nkx2.2, an insulin gene regulator that is required for terminal β cell differentiation. The ability of Area I to modulate Area II activation was also observed in vivo, as an Area I/Area II-driven transgene recapitulated the endogenous pdx-1 expression pattern in developing and adult islet cells. These results suggest that Area II is a central pdx-1 control region, whose islet cell activity is uniquely modified by Area I regulatory factors.

The pancreatic-duodenal homeobox-1 (PDX-1) 3 transcription factor is essential for pancreatic exocrine and endocrine islet cell development and the maintenance of adult islet ␤ cell function. Initially detected in dorsal gut endoderm by embryonic day (E) 8.5 in mice, PDX-1 later becomes widely expressed during development in the dorsal and ventral pancreatic buds, as well as the duodenal endoderm (1)(2)(3). Homozygous disruption of the pdx-1 gene in mice and humans results in pancreatic agenesis (1,2,4). The impact of PDX-1 on pancreatic development is more profound than any other characterized exocrine-or islet-enriched transcription factor, all of which act at later steps, with most affecting the formation of specific pancreatic cell types (5,6).
PDX-1 is principally expressed within adult pancreas islet ␤ cells and a small fraction of ␦ cells, where it regulates cell type-specific gene expression (e.g. insulin, glucose transporter type 2, ␤-glucokinase in ␤ cells; somatostatin in ␦ cells) (7). Strikingly, the conditional removal of pdx-1 from islet ␤ cells in mice results in a diabetic phenotype, presumably due (at least) to reduced insulin and glucose transporter type 2 expression (8). A PDX-1 heterozygous state causes glucose intolerance in mice, whereas humans develop maturity onset diabetes of the young (termed MODY4), a monogenic form of Type II diabetes (9,10).
Because PDX-1 is a critical transcriptional regulator of both pancreatic organogenesis and islet ␤ cell activity, efforts have focused on identifying the factors important for tissue-specific expression. Transgenic analysis has demonstrated that cell type-specific expression in the pancreas, stomach, and duodenum is mediated by sequences within 6.5 or 4.5 kilobase pairs upstream of the transcription start site in the rat and mouse genes, respectively (4,(11)(12)(13). This area contains several domains of significant identity between mouse, human, and chicken, termed Areas I (mouse bp Ϫ2761 to Ϫ2457), III (bp Ϫ1879 to Ϫ1600), and IV (bp Ϫ6529 to Ϫ6047), and one mammalian-specific identity domain, Area II (bp Ϫ2153 to Ϫ1923) (14,15). Areas I, II, and IV can independently direct ␤ cell type-specific reporter gene expression in cell line-based transfection assays (14 -16). Detailed analysis of conserved control elements within each of these regions has revealed that transcriptional control is mediated in part by factors involved in islet ␤ cell development, including FoxA2 (12,14,15,17,18), MafA (19), Pax6 (18), HNF1␣ (17,20), and PDX-1 (15,20) itself.
The Ϫ2917 to Ϫ1918-bp region of mouse pdx-1, a nuclease-hypersensitive domain that includes Areas I and II, is sufficient to direct transgene expression to islet ␤ cells in vivo (12,13). However, neither Area I nor contiguous Area III was capable of independently directing expression to islet cells in vivo (18). Furthermore, although an Area II transgene is expressed in a ␤ cell-specific manner, activity is only found in a fraction of the PDX-1 positive ␤ cell population (13,18). Collectively, these results suggested that factors binding to Area II and non-Area II sequences within bp Ϫ2917 to Ϫ1918 play a critical role in regulating pdx-1 transcription.
The primary objective of this study was to identify the sequences within bp Ϫ2917 to Ϫ1918 that cooperate with Area II to direct temporal and spatial transgenic expression in the islet. Cell line-based assays suggested that distinct properties of Area I were necessary for potentiating Area II-directed activity, in that neither non-conserved pdx-1 region sequences nor the ubiquitously active SV40 enhancer region were able to reproduce the precise ␤ cell-selective regulatory properties of bp Ϫ2917/Ϫ1918. In addition, the ␤ cell-specific enhancer of the rodent insulin gene was able to stimulate Area II activity, but differently than Area I. Activation of Area I/Area II was shown to be mediated by Nkx2.2, a homeodomain protein of the NK-2 class that is essential for proper islet ␤ cell formation and function (21). A transgene driven by Area I/Area II alone was first detected in vivo within developing insulin ϩ cells destined to populate the islet, and later to recapitulate the expression pattern of the endogenous gene in adult islets. The co-expression of Nkx2.2 in Area I/Area II transgene expressing cells is consistent with a role in both early developing and adult islet pdx-1 transcription. Collectively, the data indicate that islet cell-specific expression of pdx-1 is principally controlled by transcription factors controlling Area I/Area II-mediated activity.

MATERIALS AND METHODS
Transfection Constructs-The pdx-1 reporter constructs were made using mouse and human sequences, which were cloned directly upstream of the herpes simplex virus thymidine kinase (TK) minimal promoter in the chloramphenicol acetyltransferase (CAT) expression vector, pTK(An) CAT. The construction of the pdx-1:pTK constructs representing the Ϫ2917/Ϫ1918-bp region (mouse, Pst/Ϫ2917:Bst/ Ϫ1918 bp; human, Ϫ2839/Ϫ2023 bp), Area I (mouse, Ϫ2761/Ϫ2457 bp; human, Ϫ2839/Ϫ2521 bp), Area II (mouse, Ϫ2153/Ϫ1923 bp; human, Ϫ2252/Ϫ2023 bp), or Area IV (mouse, Ϫ6529/Ϫ6010 bp) were described previously (16). The Area I block transversion mutants and bp Ϫ2613,Ϫ2612 mutation in S6 were constructed in Ϫ2917/Ϫ1918:pTK and/or Area I:pTK using the QuikChange mutagenesis kit (Stratagene). The pdx-1 sequences were cloned in the forward orientation in each of the 2ϫArea I:pTK, 2ϫArea II:pTK, and 2ϫArea IV:pTK plasmids. The 2ϫ72 simian virus 40 (SV40) enhancer fragment was obtained by Hin-dIII digestion of p␤2ϫ72Ϫ (22,23) and subcloned with the same relative 5Ј to 3Ј orientation and nucleotide spacing just upstream of the Area I and Area II sequences in Area I:pTK or Area II:pTK, respectively. The Ϫ374 to Ϫ46-bp region of the rat insulin II gene was used to create InsII-AreaI:pTK, InsII-AreaII:pTK, and InsII:pTK (23,24). All construct sequences were confirmed by restriction enzyme digest and partial sequence analysis.
Cell Transfections-Monolayer cultures of pancreatic islet ␤ (␤TC-3 and MIN-6) and non-␤ (NIH-3T3) cell lines were maintained as described previously (24). Rous sarcoma virus enhancer-driven luciferase (pRSV-Luc) activity was used to normalize the CAT activity from co-transfected pdx-1:pTK. All cell lines were transfected using the Lipofectamine reagent (Invitrogen) with 1 g of pdx-1:pTK and 1 g of RSV-LUC (2 g total). Extracts were prepared 40 -48 h after transfection, and LUC and CAT enzymatic assays were performed as described (25,26). Each experiment was carried out at least three independent times.
Area I/Area II Transgene Construction and Generation of Transgenic Mice-Mouse Area I (Ϫ2759 to Ϫ2439 bp) was fused directly upstream of Area II (Ϫ2200 to Ϫ1923 bp) and ligated just 5Ј to the heat shock protein (hsp) minimal promoter in hsp681lacZpA, a bacterial ␤-galactosidase (lacZ) expression plasmid (12,13). DNA was NotI digested and the isolated Area I/Area II hsp-lacZ transgene injected within the Vanderbilt Transgenic Mouse/ES Cell Shared Resource into the pronuclei of one-cell embryos from B6D2 females, and embryos were transplanted into pseudopregnant ICR females. The F 0 founders were used to establish transgenic lines. Area I/Area II hsp-lacZ transgenics were identified by PCR analysis using primers to lacZ with DNA from neonatal leg or adult tail tissue prepared using the Puregene DNA Purification Kit (Gentra Systems, Minneapolis, MN).
X-Gal Staining-Neonatal digestive organs were fixed for 3 h in 4% paraformaldehyde at 4°C, then washed three times in permeabilization solution (0.2% Nonidet P-40 in phosphate-buffered saline). Activity from ␤-galactosidase was detected using X-gal as described (12). Macroscopic pictures were taken using a Stemi 2000-C microscope (Zeiss) and Nikon 5700 digital camera.

RESULTS
Area I/Area II Alone Controls the ␤ Cell Line-specific Activation Properties of the Ϫ2917/Ϫ1918-bp Region-A comprehensive transgenic analysis of the 5Ј-flanking region of the mouse pdx-1 gene revealed that the Ϫ2917 to Ϫ1918 bp region was uniquely capable of directing islet cell-selective expression (13). This region is found within a larger ␤ cell-specific nuclease-hypersensitive domain that spans blocks of conserved (i.e. Areas I (mouse bp Ϫ2761/Ϫ2457), II (bp Ϫ2153/Ϫ1923), III (bp Ϫ1879/Ϫ1600), and non-conserved sequences (12,14). Individually, only Areas I and II effectively directed ␤ cell-selective activation in transfection based assays, although their isolated activities were less effective than the Ϫ2917/Ϫ1918 bp-driven reporter (14,15). The activity of these conserved domains was also compromised relative to the Ϫ2917/Ϫ1918 bp-driven transgene in vivo. Thus, whereas Area II was able to direct transgenic expression to islet cells, activity was only found in a small fraction of the PDX-1 producing cells detected with the Ϫ2917/Ϫ1918 bp-driven transgene (18). Collectively, these results indicate that functional interactions between Area II and non-Area II sequences are required to obtain Ϫ2917/Ϫ1918 bp-like regulation.
To analyze the activation properties of the non-conserved sequences within the Ϫ2917/Ϫ1918-bp region, mouse pdx-1 sequences from Ϫ2917 to Ϫ2762 and Ϫ2456 to Ϫ2145 were subcloned directly upstream of the minimal TK promoter sequences in the CAT expression plasmid, pTK(An). Their activity was compared with Ϫ2917/ Ϫ1918 bp-, Area I-, and Area II:pTK constructs in transfected ␤ (␤TC-3, MIN6) and non-␤ (NIH-3T3) cell lines; the regulatory behavior of the various pdx-1:pTK constructs was found to be similar between ␤ cell lines. The ␤ cell line-specific activation properties of each construct is expressed as the ratio of pdx-1:pTK activity in ␤ cells to that of NIH3T3.
Only constructs containing Area I and/or Area II sequences were selectively activated in ␤ cells (Fig. 1), with non-conserved driven constructs non-functional (i.e. Ϫ2917/Ϫ2762:pTK, Ϫ2456/Ϫ2145:pTK). Significantly, a mouse pdx-1 construct driven by Area I/Area II alone had the same ␤ cell-selective stimulation properties as the expansive Ϫ2917/Ϫ1918:pTK. Reversing the order of Area I and II also did not change ␤ cell activity relative to Area I/Area II:pTK or Ϫ2917/Ϫ1918: pTK (see Area II/Area I in Fig. 1). In contrast, the non-conserved sequences in the Ϫ2917/Ϫ1918 region did not stimulate Area I or Area II activation (data not shown). The identity between mouse and human Area I and Area II is, respectively, 89 and 78% (14), presumably reflecting why cooperative activation was observed in chimeric Area I/Area II:pTK constructs (Fig. 2). These data suggest that conserved Area I and Area II control sequences are the principal mediators of the islet cellspecific activation properties of the Ϫ2917/Ϫ1918-bp region.
Area II Has Enhancer-like Properties Distinct from Other pdx-1 Control Domains-To further compare the functional properties of the conserved pdx-1 control domains, the effect of orientation and duplication (2ϫ) on Area I, II, and IV activity were compared. (Unfortunately, this analysis could not be extended to Area III, as this conserved region was inactive in transfection and transgenic experiments (13,14,18).) Each region was active in either the normal or reverse orientation (Fig. 3A), although orientation somewhat influenced the magnitude of stimulation.
The activity of the duplicated constructs were compared with both the single copy and the Area I/Area II construct, with control regions maintained in the normal orientation in each multimerized pdx-1 construct. The 2ϫArea II-driven reporter had unique functional properties when compared with 2ϫArea I:pTK or 2ϫArea IV:pTK in ␤ cells (Fig.  3B). Thus, the increase in activity of 2ϫ Area I was additive, whereas no effect was found upon duplicating Area IV. However, 2ϫArea II had cooperative enhancer-like stimulation characteristics, with an activity level equivalent to Area I/Area II. Multimerized human Area II also stimulated transcription in a synergistic manner (data not shown). Together, these data revealed the distinct enhancer-like properties of Area II and further supported a central role in pdx-1 expression in the pancreas.
The ␤ Cell-enriched Activators of Area I Appear Essential in Area II Activation-The nature of the factors involved in Area I/Area II control was investigated by fusing a ubiquitously expressed (SV40) or ␤ cellspecific (rat insulin II) enhancer upstream of Area I or Area II. The  activity of the heterologous enhancer-driven Area I and Area II constructs were compared with Area I/Area II:pTK, Area I:pTK, Area II:pTK, and enhancer alone:pTK (SV40:pTK, Ins II:pTK). The SV40 enhancer increased Area I and II activity to a much greater extent in non-␤ than ␤ cells (Fig. 4A). In contrast, the insulin enhancer had quite distinct effects on the activity of these control regions. Thus, there was no difference between InsII:pTK and InsII-Area I:pTK activity in ␤ cells, whereas Area II was activated by the insulin enhancer to a greater degree than by Area I (Fig. 4B). Notably, the results with InsII-Area II:pTK imply that Ϫ2917/Ϫ1918 bp-mediated activation involves communication between Area II and the ␤ cell-enriched control factors of Area I.

Areas I and II Direct Islet-specific Expression of pdx-1
Upon inspection of S4, S6, S7, and S8, a potential binding site for the islet-enriched Nkx2.2 transcription factor was found in S6 (Fig. 5C) (28,29). Nkx2.2 is a key regulator of islet ␤ cell development and function. Thus, the ␤ cell PDX-1 ϩ population is lost in Nkx2.2 Ϫ/Ϫ mice (21), whereas binding regulates rodent insulin transcription (28). The ability of Nkx2.2 to bind S6 was determined in gel mobility shift assays. A co-migrating S6 complex was found between the ␤TC-3 nuclear extract and in vitro translated Nkx2.2 that supershifted with ␣-Nkx2.2 antibody (Fig. 6). Competition assays also demonstrated that this complex was able to bind wild type S6, but not a 2-bp mutation that either blocks binding to the insulin II site or reduces S6-mediated activation (see the S6 mutant in Figs. 5C and 6). These data indicated that Nkx2.2 could specifically bind to the S6 control element of Area I.
The chromatin immunoprecipitation assay was used next to investigate if Nkx2.2 bound in vivo to Area I in ␤TC-3 cells. Area I sequences were selectively amplified by PCR from chromatin precipitated by ␣-Nkx2.2 antiserum, but not from chromatin treated with normal rabbit IgG or in the absence of antiserum (Fig. 7). In contrast, Nkx2.2 antiserum did not precipitate control sequences from the phosphoenolpyruvate carboxykinase gene, which is not expressed in ␤ cells. Collectively, these results strongly suggest that Nkx2.2 binding to S6 contrib-utes in the activation of Area I and Area I/Area II-mediated expression in ␤ cells.
Area I/Area II Drives Transgene Expression to All PDX-1 Producing Islet Cell Types in Vivo-Our in vitro data suggested that communication between Areas I and II was involved in regulating islet pdx-1 expression. To formally test this hypothesis in vivo, two independent lines of transgenic mice were derived that express an Area I/Area IIdriven lacZ transgene, termed Area I/Area II hsp-lacZ .
Expression from Area I/Area II hsp-lacZ was only seen in the pancreas, with the analysis of ␤-galactosidase expression revealing staining throughout the islet in tissues from neonates ( Fig. 8; 1-day-old, P1) and adults (data not shown). Quantitation of ␤-galactosidase co-staining in adult islet insulin (␤ cells), somatostatin (␦), glucagon (␣), and pancreatic polypeptide hormone-producing cells revealed that Area I/Area II hsp-lacZ expression was in 98% of insulin ϩ and 12% of somatostatin ϩ cells (TABLE ONE). In contrast, glucagon ϩ /␤-galactosidase ϩ or pancreatic polypeptide ϩ /␤-galactosidase ϩ cells were not observed in neonate or adult transgenic islets (data not shown). The Area I/Area II hsp-lacZ expression pattern is essentially identical to endogenous pdx-1 expression at these ages (3,12,13), clearly demonstrating the capacity of Areas I and II to selectively and comprehensively direct pdx-1 transcription in islet PDX-1 ϩ cells.
The activity of Area I/Area II hsp-lacZ was next examined at various stages of development to obtain insight into when control was implemented during pancreas formation. PDX-1 is essential at E8.5 for organogenesis (1), when it is widely expressed in proliferating exocrine and endocrine pancreatic precursors, and subsequently in forming islet ␤ and ␦ cells (3). Area I/Area II hsp-lacZ activity was first detected at E14.5, and restricted to most, but not all, insulin ϩ cells (Fig. 9). Because transgenic expression was found in most insulin ϩ PDX-1 ϩ cells at E14.5 and essentially all at E15.5, the absence of Area I/Area II hsp-lacZ expression at E13.5 indicates that Area I/Area II likely controls pdx-1 transcription in the secondary and principal wave of insulin ϩ cell production initiated at this time (30). Importantly, Nkx2.2 was also found in the Area I/Area II hsp-lacZ expressing cells (Fig. 9), consistent with a role in regulating pdx-1 transcription during development and in islet ␤ cells.

DISCUSSION
These studies have defined the sequences within the pdx-1 transcription unit that are capable of directing expression uniquely to pancreatic  islet cells. Through a systematic analysis of the nuclease-hypersensitive Ϫ2917 to Ϫ1918-bp region, we have demonstrated that Area I sequences augment Area II activity in both cell-based transfection and transgenic assays. An Area I/Area II-driven transgene targeted expression throughout the PDX-1 ϩ islet cell population in mice, a property distinct from Area II alone, which was only active in a fraction of producing cells. Mutational studies of Area I showed that Area I/Area II activation was regulated in part by binding of the islet-enriched Nkx2.2 transcription factor to conserved S6 at bp Ϫ2614/Ϫ2609. Our data further suggest that Area I and II act as the principal regulators of pdx-1 transcription during the start of the massive wave of islet ␤ cell differentiation during organogenesis.
Our deletional analysis of the Ϫ2917 to Ϫ1918 bp region demonstrated that non-conserved sequences play little, if any, role in regulation. Thus, Ϫ2917/Ϫ2762:pTK and Ϫ2456/Ϫ2145:pTK were not selectively activated in transfected ␤ cells (Fig. 1). Furthermore, Area I/Area II alone retained the regulatory properties of the larger transgene in in vitro and in vivo based pdx-1-driven reporter assays, consistent with non-conserved sequences having little, if any, influence on bp Ϫ2917/Ϫ1918 control. To begin to address the unique activation properties of Area I and II, we constructed pdx-1 control domain chimeric reporter constructs with either a general or ␤ cell-specific enhancer. In other studies, activation by individual ␤ cellenriched activators of the insulin gene was amplified by broadly expressed SV40 enhancer control factors (22,23,31). The SV40 enhancer also increased Area I-or Area II-directed stimulation in ␤ cells. However, the -fold increase in activity was much greater in non-␤ cells, resulting in an overall decrease in ␤ cell-specific activation (Fig. 4A). These results suggested that communication between ␤ cell-enriched regulatory factors was key to Area I/Area II activation. As a consequence, it was surprising that the rat insulin II enhancer was only able to stimulate Area II activity. The level of rat insulin-Area II activity was ϳ3-fold greater than Area I/Area II, whereas Area I chimeric activity was the same as the insulin enhancer alone (Fig. 4B). Although it is presently unclear why insulin enhancer regulatory factors only stimulated Area II activation, these results support the novel role of this pdx-1 control region implicated from the transgenic studies (18).

Area I/Area II hsp-LacZ is expressed in all islet ␤ cells and some ␦ cells
Adult islets were selected at random and the number of cells per islet producing insulin, somatostatin, and lacZ were determined by immunohistochemical staining. The percentage of cells coexpressing lacZ and insulin or somatostatin is shown.

␤ Cells
Insulin ϩ with ␤-galactosidase ϩ 98 % ␤-Galactosidase ϩ with insulin ϩ 98 % ␦ Cells Somatostatin ϩ with ␤-galactosidase ϩ 12 % ␤-Galactosidase ϩ with somatostatin ϩ 2 % Areas I and II Direct Islet-specific Expression of pdx-1 NOVEMBER 18, 2005 • VOLUME 280 • NUMBER 46 epithelium or adult islets also indicates that this factor does not play a defining role in cell-specific pdx-1 activation. Presumably, efforts focused on identifying the regulators of conserved S4, S7, and S8 will help clarify the nature of the factors that contribute to the ability of Area I to drive comprehensive expression of the Area II transgene throughout the endogenous islet PDX-1 ϩ cell population. Strikingly, Area I/Area II transgene activity was only first detected in insulin ϩ PDX-1 ϩ cells of E14.5 and E15.5, and not those at E13.5 (Fig. 9). In addition, expression from this transgene was not found in exocrine acinar cells or other islet hormone-producing cells, even though glucagon, amylase, somatostatin, and pancreatic polypeptide are first produced at around E10.5, E13.5, E15.5, and E18.5, respectively (30,40). The presence of Area I/Area II transgene activity in most insulin ϩ PDX-1 ϩ cells during development essentially recapitulates its endogenous penetrance within the mature islet ␤ cell population. A few insulin ϩ cells are detected as early as E10.5 in mice, but the majority of islet ␤ cells evolve from a distinct insulin-positive cell population that develops around E14 (41). As MafA is the only known islet-enriched transcription factor expressed this late in islet ␤ cell development or in such a cell restricted manner, pdx-1 expression in developing and adult ␤ cells may principally result from functional interactions between MafA and other islet-enriched Area I/Area II activators, which are expressed earlier and in a broader set of cell types (e.g. FoxA2 (42), BETA2 (35,43), and Pax6 (39)). The ability of MafA to coordinate the assembly of the insulin gene transcription complex by directly binding to BETA2 and PDX-1 provides support for such a hypothesis (44).
Collectively, our transfection and transgenic data strongly indicate that Areas I and II represent the islet cell-specific control domain of the pdx-1 transcription unit. However, the functional significance of Areas III and IV are less clear. Transfection analysis has shown that Area IV is selectively active in ␤ cell lines and capable of potentiating Ϫ2917/ Ϫ1918-driven reporter gene activity (16), suggesting an involvement in regulation. In contrast, Area III does not have ␤ cell-specific activity in cell line or transgenic assays (13,14). Importantly, recent gene manipulation experiments in mice provide support for a collective role for these conserved control domains in pdx-1 expression. Thus, when Areas I-III were deleted from the endogenous murine pdx-1 gene at the one-cell stage of embryogenesis using a Cre-LoxP strategy, homozygous mice displayed highly abrogated pancreas development with a profound reduction in islet cell numbers. 4 In contrast to the global pdx-1 knockout, duodenal and antral stomach development were relatively normal in Areas I-III-deficient mice, perhaps indicating a need for Area IV in expression of pdx-1 in non-pancreatic cell types. We believe that efforts focused on characterizing the regulators of pdx-1 transcription will provide valuable information on how islet expression is affected under normal and diabetic conditions.