Early pancreatic development requires the vertebrate Suppressor of Hairless (RBPJ) in the PTF1 bHLH complex

The RBPJL form of the PTF1 complex binds and activates the Rbpjl promoter

To determine whether PTF1a might indeed control the transcription of Rbpjl, we searched the 10-Kb 5′ flanking region and intronic sequences of the Rbpjl gene for potential PTF1-binding sites comprising an E-box and a TC-box spaced one or two helical DNA turns apart (Cockell et al. 1989; Rose and MacDonald 1997; Beres et al. 2006). Two potential binding sites with sequences that conform to the consensus were detected; both are within 1.4 Kb of the transcriptional start site (Fig. 2A). The proximal PTF1-binding site near the transcriptional start (at −91) is composed of an RBP-binding sequence with tandem E-boxes one and two helical turns away, and is retained from birds to primates. The distal site (at −1341) has an E-box two DNA turns from an RBP-binding sequence and is conserved well only among mammals.

Both PTF1 sites can bind the authentic RBPJL form of the PTF1 complex isolated from adult pancreas, and the complex can be supershifted by antibody against PTF1a (Fig. 2B). Moreover, the two sites also bind PTF1 complexes reconstituted from the individual subunits synthesized by cell-free translation. Binding requires both an E-box and the TC-box with proper spacing (see Fig. 6, below; data not shown).

A 1.4-Kb Rbpjl promoter region containing the transcriptional start and both potential PTF1-binding sites increased the activity of the promoterless reporter plasmid pGL3b at least 140-fold in two acinar cell lines tested by transfection (Fig. 3A,B). Mutational inactivation of the TC-box (Fig. 3B) or the E-boxes (Fig. 3C) of the proximal site eliminated the acinar activity, whereas similar mutations of the distal site had no affect. The mutations had only modest effects on the low activity of the 1.4-Kb promoter in the nonpancreatic 293 kidney cell line. A 367-base-pair (bp) promoter region that retains the proximal but not the distal site also retained high acinar-specific activity (Fig. 3A). Therefore, in this assay, the proximal PTF1-binding site is critical for activity of the promoter in acinar cells, but appears to be irrelevant in nonacinar cells. It is noteworthy that this is the first PTF1-binding site recognized that has two E-boxes, which we later show are redundant (see Fig. 6, below).

Not only was the proximal PTF1-binding site necessary to drive acinar cell transcription, but it was also sufficient when multimerized and linked to a passive minimal promoter (Fig. 2C). The acinar activity for the multimerized PTF1-binding site from the Rbpjl promoter could be reconstituted in 293 cells by the ectopic expression of the PTF1 subunits (see Fig. 7A, below). Thus, the proximal PTF1-binding site is an acinar-specific promoter element that is highly responsive to components of the PTF1 complex.

To determine whether either PTF1-binding site is active in vivo, we tested the normal and mutated 1.4-Kb regions in founder transgenic embryos (Fig. 4). Of the 22 transgenic embryos obtained from fertilized eggs microinjected with the normal transgene, eight (36%) stained intensely for the transgenic β-galactosidase reporter in the pancreas, but nowhere else in the viscera. A mutation of the proximal site sufficient to eliminate PTF1 binding abolished expression of the transgene, whereas the equivalent mutation of the distal PTF1 site had little if any effect (25% expressed). Thus, the proximal site is also critical for expression in the pancreas of mouse embryos, and the relative importance of the two sites was paralleled in vivo and in vitro.

The PTF1 complex can be formed with either RBPJ or RBPJL (see Fig. 2D; Beres et al. 2006). In the adult pancreas, the RBPJL form of the complex is bound to the promoters of the digestive enzyme genes, including Ela1 (Beres et al. 2006). The results from immunoprecipitation of chromatin from adult mouse pancreas demonstrated that both PTF1a and RBPJL were present on the proximal Rbpjl promoter region containing the PTF1-binding site, whereas RBPJ was not (Fig. 5A). Consecutive immunoprecipitations of pancreatic chromatin, first with the antibody for PTF1a, and then the antibody for RBPJL (and vice versa), enriched the Rbpjl promoter equivalent to the arithmetical product of the individual enrichments (Fig. 5B), which is consistent with coresidence of the two proteins at the same allele (Geisberg and Struhl 2004). These results indicate that RBPJL acts within the PTF1 complex to autoactivate the transcription of its gene in adult acinar cells.

The unusual nature of the proximal PTF1-binding site

The trimeric PTF1 complex forms in solution and subsequent binding to DNA requires an E-box and a TC-box separated by one or two helical turns of DNA (Beres et al. 2006). These properties are diagnostic of the concerted binding of the PTF1 complex and distinguish the binding of an intact complex from the independent binding of individual components or subcomplexes. The pairing of two E-boxes with a TC-box in the proximal site was unprecedented among known PTF1-binding sites, and we tested whether both E-boxes are necessary for complex binding and transcriptional activation. The results of EMSA analyses showed that either E-box was effective in combination with the TC-box, because disruption of both E-boxes was required to abolish binding of a reconstituted trimeric PTF1 complex containing either RBPJ or RBPJL (Fig. 6A, cf. lanes 3,7,11 and 15, and cf. lanes 4,8,12 and 16). Similarly, the disruption of both E-boxes was required to eliminate the activity of the −389 proximal promoter fragment in transfected acinar cells (Fig. 6B).

Mutation of both E-boxes also eliminated the binding of the PTF1a:E12 heterodimer (Fig. 6A, cf. lanes 2,6,10 and 14). To confirm that the critical requirement for an E-box is the binding of a trimeric PTF1 complex and not only the PTF1a–E12 heterodimer, we tested the effects of changing the spacing between an E-box and a TC-box. Inserting or removing a half DNA helical turn reduced binding of both the J and JL forms of the trimeric PTF1 (Fig. 6C) without affecting the independent binding of the PTF1a:E12 dimer (Fig. 6C, cf. lanes 2,6,10) or RBPJ (Fig. 6C, cf. the monomer band in lanes 3,7,11). The half-turn changes also decreased the activity of the Rbpjl promoter 90%–97% in transfected acinar cells (Fig. 6D). These results verify that the activity of the PTF1-binding site in the Rbpjl promoter requires the concerted binding of a trimeric complex.

To examine the regulatory properties of the isolated proximal PTF1-binding site, we compared the activity of a three-copy repeat of the site with that of three copies of the well-characterized PTF1 site from the Ela1 promoter (Rose and MacDonald 1997; Rose et al. 2001; Beres et al. 2006). Each repeat was linked to the minimal promoter of the Ela1 gene driving a luciferase reporter and assayed by transfection into the 266-6 acinar cell line or the 293 kidney cell line. The PTF1 site from the Rbpjl promoter was highly active in 266-6 acinar cells, compared with the site from the Ela1 promoter (see Fig. 2C). The acinar activity of the Rbpjl site could be reconstituted in 293 kidney cells by the ectopic expression of the PTF1 subunits (Fig. 7A). Although the Rbpjl site was much more active than the Ela1 site with either PTF1 form, its greater effectiveness with the J form (90-fold) was particularly striking compared with the Ela1 site (sixfold) (Fig. 7B). The Rbpjl version is favored because its TC-box is identical to the optimal RBP-binding sequence (Supplementary Fig. S1). We reasoned that the enhanced ability to be activated by the J form of the PTF1 complex might be important for the acinar-specific induction of Rbpjl during pancreatic development.

PTF1 switches from an RBPJ to an RBPJL form during pancreatic development

To examine whether a PTF1 complex containing RBPJ may be the initial activator of Rbpjl transcription, we monitored the presence of PTF1a, RBPJL, and RBPJ on the proximal Rbpjl promoter just prior to (E12.5), during (E15.5), and after (E17.5) the secondary transition (Fig. 8A). PTF1a and RBPJ were present early at E12.5, when acinar cell development is beginning and Rbpjl is activated. As development progressed, the ratio of RBPJL to RBPJ on the promoter increased seven- to eightfold. Thus, the PTF1 complex bound to the Rbpjl promoter changes from one containing RBPJ, when the promoter is being activated, to one with RBPJL, as the RBPJL protein accumulates. This switch does not occur for all PTF1 targets, however; only the RBPJL form of PTF1 is detected on the CPA1 promoter in the developing pancreas and is retained in adult acinar cells (Fig. 8B). It appears that the production of RBPJL is an early event in acinar development that then contributes to the activation of the digestive enzyme genes.

The early developmental function of PTF1a requires its interaction with RBPJ

Previously (Beres et al. 2006), we showed that the substitution of alanine for tryptophan at residue 298 in the PTF1a-binding site for RBPJ extinguished the ability of PTF1a to interact with RBPJ, yet retained its ability to interact with RBPJL (see Fig. 2D). This PTF1a mutant provides the means to investigate the role of the J form of the trimeric PTF1 complex independent of the JL form. To determine whether the J form of the PTF1 complex is required for the early growth and branching morphogenesis phase of pancreatic development, we generated a line of mice bearing the W298A mutation (Ptf1a^WA) using recombination-mediated cassette exchange (Long et al. 2004; J. Burlison, Q. Long, C.V.E. Wright, and M.A. Magnuson, in prep.). As for mice heterozygous for the Ptf1a^cre-null allele (Kawaguchi et al. 2002), pancreatic development and function for heterozygous Ptf1a^WA/+ mice were normal. However, the pancreatic development of embryos that expressed only the W298A form of PTF1a was severely affected and mirrored that of homozygous Ptf1a-null embryos (see below).

We used the activation of the Rosa26R locus and β-galactosidase staining of Ptf1a^WA/cre embryos to follow the fate of the PTF1a-deficient cells initially specified for pancreas and the growth of the mutant epithelium (Fig. 9). The cre allele has the entire coding region of Ptf1a replaced by that of cre, and therefore is null for PTF1a function (Kawaguchi et al. 2002). The developmental defects of Ptf1a^WA/cre embryos were indistinguishable from those of Ptf1a^WA/WA embryos (Supplementary Fig. S2). At E10.5, the dorsal bud of the Ptf1a^WA/WA pancreas appeared normal, including the size of the epithelium and the presence of PDX1 (see Fig. 10, top). A Ptf1a expression domain was present in the duodenal epithelium at the normal position of the ventral bud, but budding did not occur (data not shown). Indeed, as for Ptf1a^cre/cre-null embryos (Kawaguchi et al. 2002), the ventral pancreas never formed. Rather, pancreatic precursor cells that activated the Ptf1a locus became marked by β-galactosidase activity and reverted to an intestinal fate within the duodenal epithelium (Fig. 9C′, inset; data not shown). Developmental defects of the dorsal epithelial bud in Ptf1a^WA/cre embryos became evident at E11.5. As for Ptf1a^cre/cre embryos (Fig. 9A), the growth of the dorsal epithelium was retarded (Fig. 9A′); by E12.5, the epithelium was markedly smaller than normal (Fig. 9B′). Because the PTF1a^WA protein was present (see Fig. 10B′), developmental defects were not due to an inability to accumulate the mutant protein. At 1 d prior to birth, the dorsal pancreatic epithelium was a truncated, poorly branched, duct-like structure (Fig. 9C′) with columnar epithelial cells, akin to the ductal vestiges of Ptf1a-deficient embryos (Krapp et al. 1998; Kawaguchi et al. 2002). Because the W298A mutant so closely phenocopies the null, it appears that the first crucial function of PTF1a during pancreatic development requires its interaction with RBPJ.

The inability of PTF1a to bind RBPJ prevents the secondary transition

Normal growth and morphogenesis of the early pancreatic epithelia establish an extensive precursor cell population that fuels the genesis of islet and acinar cells during the secondary transition. Exhausting this precursor population by forcing precocious differentiation (Apelqvist et al. 1999; Hald et al. 2003; Murtaugh et al. 2003) or preventing its formation by inactivation of Pdx1 (Ahlgren et al. 1996; Hale et al. 2005) or Ptf1a (Krapp et al. 1998; Kawaguchi et al. 2002) precludes this developmental transition. Because the growth and branching of the early Ptf1a^WA/WA pancreas are also greatly attenuated, we examined developmental markers to determine whether the secondary transition occurs (Fig. 10). The appearance of PTF1a in proacini emerging during the secondary transition (Fig. 10F) did not occur in the mutant embryos (Fig. 10F′). Pdx1 and Sox9 expression continued in the mutant (Fig. 10G′; data not shown), although at the low levels characteristic of the precursor epithelium prior to the secondary transition. Acinar structures did not form, and acinar differentiation markers such as CPA1 did not appear (Fig. 10I′). The characteristic surge of endocrine cells also did not occur. The induction of high PDX1 levels that normally occurs in precursor cells directed to the β-cell lineage (Fujitani et al. 2006a) did not occur in the mutant and the insulin-producing islet β-cells that normally form in large numbers at this stage are far fewer in the mutant (see the insets of Fig. 10L, L′). The endocrine cell clusters present in the mutant pancreas appeared early, prior to the secondary transition, and were composed of glucagon⁺ cells that also contained prohormone convertase 1/3 (PC1/3) (Fig. 10J–L, J′–L′) rather than PC2. The early appearance and the combination of markers indicate that these cell clusters are of the early endocrine lineage (Kim and MacDonald 2002; Wilson et al. 2002) rather than the islet lineage. Thus, neither islet nor acinar developmental programs begin in the mutant pancreas.

Notch-dependent and Notch-independent functions of RBPJ

Compelling evidence for Notch signaling through RBPJ during pancreatic development derives from the effects of genetic alterations of Notch pathway components. Germline mutations that inactivate genes encoding components of the canonical Notch pathway (i.e., Dll1, Hes1, Rbpj, or Notch1) disrupt early pancreatic growth and branching, deplete epithelial precursor cells by overcommitment to the early endocrine lineage, and eliminate or delay the secondary transition and consequently the formation of islets and acini as well (Apelqvist et al. 1999; Jensen et al. 2000; Fujikura et al. 2006). Similar developmental defects occur by the premature overexpression of the proendocrine Ngn3 gene, which overrides Notch control by antagonizing Hes1 (Jensen et al. 2000; Schwitzgebel et al. 2000; Lee et al. 2001). Conversely, misexpression of activated Notch locks the progenitor cells of the pancreatic epithelium into an undifferentiated state (Hald et al. 2003; Murtaugh et al. 2003). Thus, Notch signaling is required for proper pancreatic development, and RBPJ is the transcriptional mediator. However, these genetic experiments do not distinguish Notch-dependent from Notch-independent roles for pathway components such as RBPJ.

Several lines of evidence had suggested that such a role for RBPJ was likely. (1) Using RBPJ as bait, Obata et al. (2001) caught PTF1a in a yeast two-hybrid screen with a cDNA library from embryonic neural tissue and showed that coexpression of PTF1a and RBPJ could activate a reporter gene driven by the PTF1-binding site from the Ctrb promoter in transfected cells. (2) RBPJ can form a trimeric complex with PTF1a and an E-protein that can bind a known PTF1-L target site and activate transcription through that site in transfected cells (Beres et al. 2006). (3) C-terminal truncations of the human PTF1a that eliminate binding of either RBP (Beres et al. 2006) disrupt the development of the cerebellum (Sellick et al. 2004) where RBPJ is present but RBPJL is not. (4) Orthologs of PTF1a, E2A, and RBPJ (but not RBPJL) are conserved to insects and are able to form an equivalent trimeric complex with identical DNA-binding properties and transactivation potential in mammalian cells (Beres et al. 2006). Thus, PTF1-J appears to be the ancestral complex.

Because PTF1a is required for pancreatic development prior to the appearance of RBPJL, we reasoned that RBPJ might have a Notch-independent function as an obligate partner for PTF1a early, and that the Rbpjl gene might be a target of the complex. We found two potential PTF1-binding sites in the Rbpjl promoter. The proximal site could bind and be activated by either the PTF1-J or PTF1-L, and mutations in this site disrupted promoter activity in transfected acinar cells and the pancreas of transgenic embryos. Moreover, PTF1-J was present on the Rbpjl promoter at the onset of acinar cell differentiation when transcriptional activation of Rbpjl begins, and was gradually replaced by the JL form as differentiation progressed and mature acinar cells accumulated.

To test the importance of the J form of PTF1 directly, we derived a line of mice that express a mutant PTF1a with a single amino acid change (W298A) that disrupts its interaction with RBPJ, but not RBPJL. Pancreatic development of homozygous Ptf1a^WA/WA embryos was arrested in the same manner as Ptf1a-null embryos, indicating that the first critical function of PTF1a requires the partnership of RBPJ within PTF1-J, a role for RBPJ distinct from that in Notch signaling.

Although effects on pancreatic development of Rbpj^−/− and Ptf1a^WA/WA embryos are similar, important differences indicate that RBPJ is required as a partner for Notch as well as for PTF1a, with roles that overlap temporally but not functionally. In the absence of RBPJ, pancreatic epithelial growth is slowed by the premature and enhanced formation of differentiated endocrine cells (Apelqvist et al. 1999; Fujikura et al. 2006) due to the unrestricted differentiation of the early endocrine lineage. The Rbpj^−/− defects are identical to those of null mutations for the other Notch pathway components Delta-like1 and Hes1 (Apelqvist et al. 1999; Jensen et al. 2000) and are consistent with the expected consequence of lost Notch signaling, which is premature differentiation due to an inability to sustain a population of progenitor cells. In mutant embryos with the version of PTF1a that is unable to bind RBPJ, epithelial growth and branching are also attenuated, and the early endocrine cells are present. These similar developmental abnormalities indicate that the principle cause of the pancreatic defects for the W298A version of PTF1a is its inability to interact with RBPJ, although a formal possibility remains that an interaction of PTF1a with an as yet unidentified partner is also disrupted. Because Notch signaling remains intact for Ptf1a^WA/WA embryos, the early pancreatic endocrine population is not expanded as it is in the Rbpj-deficient embryos. This important difference indicates that RBPJ serves two separate functions in pancreatic development, one for Notch control of the progenitor state and another for PTF1 control of the early developmental program. Rbpj mutants are deficient for both, whereas the inability to recruit RBPJ into the PTF1 complex only affects the latter.

The tryptophan to alanine mutation is near the C terminus in a heptapeptide sequence that is conserved between vertebrates and flies. The PTF1a motif (WTPEDPR) is similar to the critical sequence in the NotchICD (WTPEGFK) for binding to RBPJ (Kovall and Hendrickson 2004). Indeed, a synthetic peptide containing the Notch motif inhibits the recruitment of RBPJ by PTF1a, and overexpression of either the NotchICD or PTF1a in cell transfection experiments interferes with the activity of the other (Beres et al. 2006). It would be instructive to know whether this mutual antagonism has a regulatory function during development. The transition to RBPJL in acini may serve to unlink the RBP subunit from Notch signaling. Alternatively, the selective use of the RBPJL in the PTF1 complex of adult acini may ensure the availability of RBPJ for its Notch pathway function.

The developmental abnormalities caused by the W298A form of PTF1a prove our model (Beres et al. 2006) for the defects in human PTF1a shown to be associated with permanent neonatal diabetes mellitus (PNDM) (Sellick et al. 2004). Two human mutations delete short regions of the C terminus of PTF1a containing an evolutionarily conserved hexapeptide sequence that includes W298. Neither mutant protein is able to recruit RBPJ into a trimeric PTF1 complex (Beres et al. 2006). Our results show that a much subtler mutation with the same biochemical consequences causes the same defects of pancreatic and cerebellar development as the PNDM newborns (this study; H. Kei, Y. Nakada, S.M. Glasgow, T. Masui, M. Henke, T.M. Beres, M.A. Magnuson, R.J. MacDonald, and J.E. Johnson, in prep.).

Autoregulation of Rbpjl

The proximal PTF1-binding element of the Rbpjl promoter is required for the activation of Rbpjl transcription in nascent acinar cells. The presence of PTF1a and RBPJ on the proximal promoter region at this time indicates that PTF1-J initiates this first stage of Rbpjl expression. The presence of RBPJ in an organ-specific transcription factor complex is novel. As acinar differentiation progresses, RBPJL gradually replaces RBPJ on the promoter. Because transcription is begun by PTF1-J and continued by PTF1-L, the switch must occur in the same cell, either by an exchange of RBP subunits or of complete PTF1 complexes.

Several lines of experimental evidence indicate that a complete trimeric PTF1 complex binds and activates the Rbpjl promoter rather than individual subunits that bind and act independently. PTF1a and RBPJL coreside within the Rbpjl promoter region containing a consensus binding site for the trimeric PTF1 complex essential for the acinar activity of the promoter; the vast majority of the RBPJL DNA-binding activity in nuclear extracts from adult acinar cells is in PTF1-L (Beres et al. 2006); the formation of a stable trimeric complex with highly cooperative binding properties occurs prior to binding to DNA; alterations of the spacing between the E-box and TC-box that prevent binding of either PTF1 complex, but allow independent binding of a PTF1a–HEB heterodimer to the E-box and RBPJL (or RBPJ) to the TC-box, eliminate the transcriptional activity of the PTF1-binding site. The sum of this evidence indicates that the trimeric PTF1 complex mediates the cell-specific activity of the Rbpjl promoter in mature acinar cells.

The autoactivation of Rbpjl by PTF1-L may provide a stable autoregulatory loop that ensures the maintenance of the acinar cell phenotype. PTF1-L is also present on PTF1-binding sites necessary for acinar transcription of all the pancreatic digestive enzyme genes examined (Beres et al. 2006). Thus, the stable maintenance of Rbpjl expression can directly control the continued transcription of genes that define the acinar cell phenotype. Preliminary observations of a Rbpjl knockout line of mice (T. Masui, M.E. Magnuson, and R.J. MacDonald, unpubl.) indicate that the mutant pancreas at birth is consistently one-third less than normal and acinar tissue is less differentiated, as judged by 70% lower levels of the mRNAs for two of the five digestive enzymes so far examined. Thus, RBPJL is important for a maximal acinar phenotype, though not critical for formation of acinar cells. It will be interesting to determine whether RBPJL is required for an optimal response to dietary changes or for acinar regeneration after injury.

The PTF1 complex has an ancient function

An orthologous trimeric PTF1 complex is conserved to insects (Beres et al. 2006). Drosophila melanogaster has a PTF1a ortholog (FER1) that can a form a complex with the Drosophila E2A equivalent (Daughterless; DA) and Su(H). The Drosophila complex can bind the PTF1 site from the mouse Ela1 promoter and has transactivation potential in transfected mammalian cells intermediate to the mammalian PTF1-J and PTF1-L. Remarkably, the conservation is so great that mixed complexes of Drosophila and mammalian subunits are functional.

Insects do not have a digestive gland homologous to the vertebrate pancreas. The pancreatic endocrine function is served by cells in the brain that deliver insulin-like peptides to the heart tube and gut and neuroendocrine cells of the ring gland that release peptides with glucagon-like functions into the hemolymph (Rulifson et al. 2002; Kim and Rulifson 2004). Although digestive serine proteases are produced in the ventriculus and gastric caecae (Davis et al. 1985), Fer1 is not expressed there (http://flybase.bio.indiana.edu; J. Zoloty, R.J. MacDonald, and D.M. McKearin, unpubl.), and most digestive hydrolases are made in the midgut mucosa. Consequently, Fer1 does not have a regulatory role equivalent to that of Ptf1a in pancreatic development and acinar function. Rather, the extensive expression of both Fer1 (http://flybase.bio.indiana.edu) and Ptf1a (Obata et al. 2001; Sellick et al. 2004; Glasgow et al. 2005) in the embryonic central nervous system suggest that the common, and therefore, ancestral function of Ptf1a/Fer1 is for the development of the CNS. In vertebrates, and likely in flies, this function involves a Notch-independent collaboration with Rbpj/Su(H) (H. Kei, Y. Nakada, S.M. Glasgow, T. Masui, M. Henke, T.M. Beres, M.A. Magnuson, R.J. MacDonald, and J.E. Johnson, in prep.). It would be of interest to determine whether RBPJ has Notch-independent roles in other programs of organogenesis with partners other than PTF1a.

Recombinant DNA and cell transfection

Recombinant plasmids were constructed using standard molecular biological techniques. The promoter of mouse Rbpjl (−1380 to +103) was amplified by PCR from mouse BL6 genomic DNA and subcloned into pGL3b in front of the luciferase reporter gene (RbpLp.luc). The minimal promoter construct (EIp.luc) has the Ela1 basal promoter linked to the 5′ end of the luciferase gene of PGL3basic (Promega). Mutagenesis of PTF1-binding sites was performed with QuickChange (Stratagene); the E-boxes were modified from CANNTG to TCNNTA and the central cores of the TC-boxes from TTCCCA to TTCAAT. A repeat of three 32-bp PTF1-binding sites of Rbpjl promoter or three 26-bp PTF1 sites of the Ela1 promoter was placed upstream of EIp.luc. The cDNA expression plasmids for HEB (TCF12), mouse PTF1a (pcDNA3/ptf1a), mouse RBPJL (pcDNA3/Rbpjl), and myc-tagged human RBPJ have been described (Beres et al. 2006). Transfections of 293 human embryonic kidney cells (American Type Culture Collection [ATCC] CRL-1573), AR4-2J rat acinar cells (ATCC CRL-1492), and 266-6 mouse acinar cells (ATCC CRL-2151) were performed as described previously (Liu et al. 2001).

Quantitative RT–PCR

Complementary DNA was synthesized using SuperScript II reverse transcriptase (Invitrogen,) and oligo(dT) with RNA isolated from embryonic and adult mouse pancreas by the guanidine thiocyanate technique (MacDonald et al. 1987). Samples of cDNA derived from the equivalent of 15 ng of total cellular RNA template were amplified in 25-μL reactions with the following gene-specific primers: Ptf1a, 5′-CTTGCAGGGCACT CTCTTTC-3′ and 5′-CGATGTGAGCTGTCTCAGGA-3′; Rbpjl, 5′-CAGAGCATGCCATCATCCTA-3′ and 5′-AGTCCCATG TAACCGCAGAC-3′; Rbpj, 5′-GGTCCCAGACATTTCTGCAT-3′ and 5′-GGAGTTGGCTCTGAGAATCG-3′; Ela1, 5′-AGCAGA ACCTGAGCCAGAAT-3′ and 5′-TTGTTAGCCAGGATGGT TCC-3′; and β-actin, 5′-AGCCATGTACGTAGCCATCC-3′ and 5′-CTCTCAGCTGTGGTGGTGAA-3′. The amount of cDNA was quantified with SYBR Green MasterMix (Applied Biosystems) using the ABI prism 7700 (Applied Biosystems) and normalized to the level of β-actin.

Hybridization in situ

Hybridization in situ was performed with digoxygenin-labeled cRNA probes on frozen sections of paraformaldehyde-fixed mouse pancreas as described by others (Birren et al. 1993). Thirty-micron cryosections of mouse embryos were incubated with 1–2 μg/mL digoxygenin-labeled cRNA probes overnight at 65°C. The sections were then incubated with alkaline-phosphatase-conjugated anti-digoxygenin antibody and visualized with a 5-Bromo-4-Chloro-3′-Indolylphosphate p-Toluidine Salt and Nitro-Blue Tetrazolium Chloride solution. The mouse cRNA probes for Ptf1a, Rbpjl, and Ngn3 were synthesized with T7 or SP6 polymerase from PCR-amplified fragments of cDNA inserts cloned in pBSII SK⁺.

Electrophoretic mobility shift assays (EMSA)

The subunits of PTF1 were synthesized by in vitro transcription and translation using the TnT Reticulocyte Lysate System (Promega). Nuclear extract was prepared from rat adult pancreas as described previously (Rose et al. 2001). The products of IVT were quantified by [³⁵S]methionine incorporation; relative molar ratios of the PTF1 subunits were calculated from the number of methionine residues in each. The double-stranded oligonucleotide probes for the PTF1-binding sites from the Rbpjl and the Ela1 promoter were labeled with ³²P by polynucleotide kinase and EMSAs were performed as described (Rose et al. 1994).

Mouse genetic engineering

Antibodies and immunofluorescence microscopy

For EMSA supershift experiments, the rabbit RBPJL antiserum and the affinity-purified rabbit anti-PTF1a have been described (Rose et al. 2001; Beres et al. 2006); anti-RBPJ was purchased from the Institute of Immunology Co., Ltd. For chromatin immunoprecipitation (ChIP), anti-RBPJ (SC-8213) and anti-cMYC were purchased from Santa Cruz Biotechnology. The anti-RBPJL did not cross-detect RBPJ, and the anti-RBPJ did not cross-detect RBPJL in EMSA and Western blotting experiments (Beres et al. 2006). For immunofluorescence, anti-CPA1 (used at 1:1000) and anti-PC1/3 (1:200) were from Chemicon, anti-insulin (1:2000) and anti-glucagon (1:8000) were from Linco, and anti-E-cadherin (1:1000) was from Sigma Aldrich. Affinity-purified Guinea pig anti-PTF1a (1:2000) for immunofluorescence was the kind gift of Drs. Kei Hori and Jane Johnson (University of Texas Southwestern Medical Center, Dallas, TX), and affinity-purified rabbit anti-PDX1 (1:1000) was a gracious gift from Michael Ray and Christopher Wright (Vanderbilt University School of Medicine, Nashville, TN). Immunofluorescent localization was performed with 10-μm frozen tissue sections derived from paraformaldehyde-fixed embryonic mouse pancreas. Staining with hematoxylin and Periodic Acid-Schiff was performed as described (Offield et al. 1996).

ChIP

Chromatin from adult mouse pancreas and liver was prepared from formaldehyde cross-linked nuclei as described for rat liver chromatin (Chaya and Zaret 2003), and sheared further to an average of 500-bp by sonication. Chromatin was prepared similarly from pooled embryonic pancreases, and the chromatin equivalent to that of one embryonic pancreas in 1 mL of ChIP dilution buffer (Upstate Biotechnology) was immunoprecipitated with 3 μg of antibody and blocked protein G-Sepharose beads (Upstate Biotechnology). After the beads were washed and cross-linking was reversed, the immunoprecipitated DNA was analyzed by quantitative real-time PCR. Sequential ChIP (Geisberg and Struhl 2004) was performed as described previously (Beres et al. 2006). The primer pairs to detect specific gene promoter regions were as follows: Rbpjl, 5′-TGCTGGGTCTG GCTTCTACT-3′ and 5′-CCGATCCTCACACTGGATTT-3′; Cpa1, 5′-CATGGTCAAGGGTGAAAGC-3′ and 5′-CTGAG GTCTGAGGCCTTTTT-3′; and R28S, 5′-CTGGGACATAGT GGGTGCTT-3′ and 5′-GAGCCTAGAGATGGGCTGTG-3′.

The amounts of immunoprecipitated DNA for the promoter regions were quantified with SYBR Green Mastermix (Applied Biosystems) using the ABI prism 7700 (Applied Biosystems). The level of ChIP enrichment was calculated from the amount of target promoter DNA relative to the amount of the 28S region of the large ribosomal RNA gene in the immunoprecipitated chromatin versus the starting chromatin.