The Role of Hepatic Nuclear Factor 1α and PDX-1 in Transcriptional Regulation of the pdx-1 Gene*

The PDX-1 homeodomain transcription factor regulates pancreatic development and adult islet β cell function. Expression of the pdx-1 gene is almost exclusively localized to β cells within the adult endocrine pancreas. Islet β cell-selective transcription is controlled by evolutionarily conserved subdomain sequences (termed Areas I (−2839 to −2520 base pairs (bp)), II (−2252 to −2023 bp), and III (−1939 to −1664 bp)) found within the 5′-flanking region of the pdx-1 gene. Areas I and II are independently capable of directing β cell-selective reporter gene activity in transfection assays, with Area I-mediated stimulation dependent upon binding of hepatic nuclear factor 3β (HNF3β), a key regulator of islet β cell function. To identify other transactivators of Area I, highly conserved sequence segments within this subdomain were mutagenized, and their effect on activation was determined. Several of the sensitive sites were found by transcription factor data base analysis to potentially bind endodermally expressed transcription factors, including HNF1α (−2758 to −2746 bp, Segment 2), HNF4 (−2742 to −2730 bp, Segment 4; −2683 to −2671 bp, Segment 7–8), and HNF6 (−2727 to −2715 bp, Segment 5). HNF1α, but not HNF4 and HNF6, binds specifically to Area I sequences in vitro. HNF1α was also shown to specifically activate Area I-driven transcription through Segment 2. In addition, PDX-1 itself was found to stimulate Area I activation. The chromatin immunoprecipitation assay performed with PDX-1 antisera also demonstrated that this factor bound to Area I within the endogenous pdx-1 gene in β cells. Our results indicate that regulatory factors binding to Area I conserved sequences contribute to the selective transcription pattern of the pdx-1 gene and that control is mediated by endodermal regulators like HNF1α, HNF3β, and PDX-1.

Several transcription factors that are enriched within pancreatic islet cells mediate differentiation during embryogenesis and the maintenance of specialized cellular functions in the adult (1)(2)(3). Some of these proteins are also dysfunctional in patients with diabetes, a disease caused in part by the failure of ␤ cells to produce sufficient insulin to meet the needs of the body (4 -6).
The sequences that control appropriate developmental and adult specific expression of the pdx-1 gene were demonstrated in transgenic studies to be located in the 5Ј-flanking region of the mouse (39,41) and rat (40) genes. Three nuclease hypersensitive sites, termed HSS 1 (Ϫ2560 to Ϫ1880 bp), 2 (Ϫ1330 to Ϫ880 bp), 3 (Ϫ260 to ϩ180 bp), were identified within this region of the endogenous mouse gene by DNase I and micrococcal nuclease analysis (39). However, only HSS1 sequences were capable of directing pancreatic ␤ cell-selective expression in transgenic and transfection studies (39). Sequence analysis of the mouse, chicken, and human pdx-1 genes revealed that the HSS1 region also represented the only area of significant identity within 4.5 kilobases of the transcription start site (41). Sequence conservation and function allowed the HSS1 region to be divided into the Area I (Ϫ2839 to Ϫ2520 bp), Area II (Ϫ2252 to Ϫ2023 bp), and Area III (Ϫ1939 to Ϫ1664 bp) subdomains (41). Areas I and II, but not Area III, were capable of independently directing ␤ cell-selective reporter gene activity in transfection assays, with Area I activation mediated by HNF3␤ (41). This forkhead transcription factor is also necessary for stimulation by Area II, although only in the context of both Area I and II sequences (39,41). The importance of HNF3␤ in pdx-1 transcription was also supported by results showing that pdx-1 mRNA levels were reduced in homozygous null HNF3␤ embryoid bodies (41).
Collectively, the data obtained from analysis of transgenic and transfected pdx-1-driven reporter constructs strongly suggested that the HSS1 region played an essential role in directing transcription during development and in the adult. We proposed that conserved subdomains of HSS1 (i.e. Areas I, II, and III) contained transcription factor binding sites that were critical in this process, like the HNF3␤ element (41). A comprehensive series of block mutants within the conserved sequences of Area I were prepared to test this hypothesis. Our results indicate that HNF1␣ and PDX-1 itself are key positive regulators of Area I activation in ␤ cells. These studies suggest that pdx-1 transcription is regulated by factors associated with controlling both the developmental and metabolic states of the ␤ cell and that an inactivating mutation in one of its critical regulators (e.g. PDX-1, HNF1␣, and HNF3␤) could affect its expression and contribute to ␤ cell dysfunction and disease in patients.
Chromatin Immunoprecipitation (ChIP) Assays-Monolayer cultures of mouse ␤TC-3 cells (ϳ0.5-1.0 ϫ 10 8 ) were exposed to 1% formaldehyde in Dulbecco's modified Eagle's medium for 5 min at 23°C; glycine was added to 0.125 M, and the cultures were incubated for 2 min. The cells were collected in cold phosphate-buffered saline, pelleted by centrifugation, and incubated for 10 min on ice in 0.6 ml of SDS lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris-HCl, pH 8.1, 1 mM phenylmethylsulfonyl fluoride). Lysed samples were transferred to prechilled microcentrifuge tubes containing 250 mg of glass beads (Յ106 m diameter, 106 m), and the chromatin was sonicated at power setting 4 with a Virsonic 100 sonicator (Virtis Company, Inc.) for twelve 10-s pulses at 4°C. The reactions were centrifuged for 10 min at 4°C to remove debris and stored at Ϫ70°C. A 100-l aliquot was diluted with 0.9 ml of buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris-HCl, pH 8.1, 167 mM NaCl, 1 mM phenylmethylsulfonyl fluoride, 0.1% protease inhibitor mixture for mammalian cells (Sigma)) and precleared with 60 l of bovine serum albumin-blocked protein A-Sepharose for 1 h at 4°C. After removal of the Sepharose beads by centrifugation, 1 l of anti-PDX-1 antisera, 10 g of normal rabbit IgG (Santa Cruz Biotechnology, Inc.), or no antibody was added to the supernatant, and the reaction was incubated for 1 h at 4°C. Antisera raised to the N-terminal (amino acids 1-75) and C-terminal (amino acids 271-283 (10)) regions of PDX-1 were used. Antibody-protein-DNA complexes were isolated by incubation with 60 l of blocked protein A-Sepharose for 3 h at 4°C. After extensive washing, bound DNA fragments were eluted and analyzed by PCR using Ready-to-Go PCR beads (Amersham Pharmacia Biotech), 15 pmol of each primer, and 10 l of immunoprecipitated DNA/reaction. Cycling parameters were: one cycle of 95°C for 2 min and 28 cycles of 95°C for 30 s, 61°C for 30 s, and 72°C for 30 s. The primers used for amplification of mouse Area I were 5Ј-CCACTAAGAAGGAAGGC-CAG-3Ј (Ϫ2785) and 5Ј-CTGAGGTTCTTTCTCTGCCTCTCTG-3 (Ϫ2435). Cycling parameters for amplification of mouse PEPCK were: one cycle of 95°C for 2 min and 28 cycles of 95°C for 30 s, 61°C for 30 s, and 72°C for 30 s. The primers used for amplification of mouse PEPCK were: 5Ј-GAGTGACACCTCACAGCTGTGG-3Ј (Ϫ434) and 5Ј-GGCAG-GCCTTTGGATCATAGCC-3Ј (Ϫ96). The PCR products were confirmed by sequencing. Amplified products were electrophoresed through a 1.4% agarose gel in Tris acetate/EDTA buffer and visualized by ethidium bromide staining.

RESULTS
Conserved Area I Sequences Are Essential for ␤ Cell Activation-Transfected pdx-1 driven reporter constructs spanning Area I (Ϫ2839 to Ϫ2520 bp) display ␤ cell-specific activation (41). To determine whether the conserved sequences between Ϫ2773 and Ϫ2643 bp are involved in regulation, each segment (S) of identity was independently mutated, and the effect on Area I activation was assayed in transfected HIT-T15, ␤TC-3, and MIN6 ␤ cell lines (Fig. 1A). Essentially equivalent results were obtained between these ␤ cell lines (Fig. 1B).
Mutations in S4, S5, S6, S7, and S8 all reduced Area I-driven activity (Fig. 1B), and their effect was comparable with mutat-ing the HNF3␤ site (41). The sequences found within S4 and S7-8 have a 9 of 13 nucleotide match to the consensus binding site for the HNF4 family of proteins (Refs. 59 and 60 and Fig.  2A). Gel mobility shift binding assays were performed with a probe corresponding to the HNF4 binding site in the apolipoprotein CIII gene (apoCIII) (47), in vitro translated HNF4␣, and S4, S7-8, and apoCIII competitors. In contrast to apoCIII, neither S4 nor S7-8 competed effectively for HNF4␣ binding (Fig. 2B). There was also no detectable binding between S4 or S7-8 probes and HNF4␣ (data not shown). These results demonstrated that HNF4␣ can not bind to S4 or S7-8, and as a consequence, neither HNF4␣ nor HNF4␥ appear to directly regulate Area I activation.
HNF1␣ Binds to S2-The HNF1␣ and HNF1␤ homeoproteins bind with identical specificity as homodimers or heterodimers (61). S2 has a 10 of 13 nucleotide homology with the consensus HNF1 binding site (Fig. 3A). Binding experiments were performed with the S2 probe and in vitro translated HNF1␣, and nuclear extracts were prepared from ␤ (HIT-T15, ␤TC3), islet ␣ (␣TC6), and pancreatic acinar (AR42J) cell lines and rat liver. The in vitro translated HNF1␣-bound complex co-migrated with one formed in liver, acinar, and ␤ extracts (Fig. 3B). The ␤ cell complex was supershifted with an antiserum raised to HNF1␣ but not to HNF1␤ (Fig. 3C). However, HNF1␤ was also found in the HNF1 complex from kidney extracts (data not shown). These results indicated that HNF1␣ and HNF1␤ could bind to S2. In addition, two other S2 complexes were identified, with the faster migrating ␤EF1 (␤enriched factor 1) complex only detected in ␤ cell extracts and the other in all (see ␤EF-1 and U (i.e. for ubiquitous) labeled complexes in Fig. 3B).
The binding properties of the S2 complexes formed in ␤ extracts were determined by competition analyses. All three appeared to be specific, because their formation was decreased by unlabeled wild type probe sequences and not by the S2 block mutant (Fig. 3D). In addition, mutating four base pairs within the HNF1␣/␤ binding core sequence also prevented competition (i.e. S2 7-11 MUT). In contrast, HNF1␣ binding was selectively eliminated using either the HNF1 binding site from the ␤-fi- brinogen (␤-FIB) gene or an S2 mutant that contained the HNF1 binding consensus sequences of the ␤-FIB competitor (S2 to ␤-FIB, 12 of 13 nucleotide homology; Fig. 3D).
HNF1␣ has also been shown to bind in vitro to a distinct control element of the rat I (28) and human (57) insulin genes. The homology between the human insulin and the consensus HNF1 binding site is poorer than to S2 (Fig. 3A), and competition analysis demonstrated that HNF1␣ has a 5-fold higher affinity for S2 than the human site. These results suggest that the diabetic condition found in mature onset diabetes of the young patients may not only be due to effects on insulin transcription but also to effects on pdx-1 expression.
HNF1␣ Potentiates Area I-driven Activation-To test whether HNF1␣ regulated Area I activation in ␤ cells, S2 sequences were changed to resemble the HNF1 binding site of the ␤-FIB gene. The S2 to ␤-FIB site serves as an effective competitor and probe for HNF1␣ binding, although it has little or no effect on the other S2 complexes (Fig. 3D; data not shown). This altered specificity mutant was made in Area I alone or Area I and Area II combined. The larger mouse pdx-1 gene promoter construct, termed Pst-Bst, directs ␤ cell-selective expression in transfected ␤ cell lines and transgenic animals (39).
The activity of the S2 to ␤-FIB mutant in Area I⅐pTk and Pst-Bst⅐pTk was compared with the wild type and S2 7-11 mutant. We believed that the S2 7-11 core mutant would be less active than the wild type, because this mutation prevents protein complex formation (Fig. 3). Each of these plasmids was introduced into HIT-T15, ␤TC-3, and MIN6 cells. As expected, the S2 7-11 mutant reduced both Area I and Pst-Bst-mediated activation (Fig. 4, A and B); the effect was more pronounced in Pst-Bst⅐pTk (Fig. 4B). The altered specificity mutant was consistently more active than the S2 7-11 mutant in Pst-Bst, although not in Area I alone. Similarly, the HNF3␤ binding site mutant in Area II only decreased ␤ cell stimulation within a Pst-Bst context, whereas both Area I and Pst-Bst activation were decreased by the HNF3␤ binding mutant in Area I (41).
To directly determine whether HNF1␣ activated Area I expression, an HNF1␣ expression plasmid was co-transfected with either the wild type or S2 block mutant Area I⅐pTk construct into NIH 3T3 cells, which lack endogenous HNF1␣. Overexpression of HNF1␣ activated Area I⅐pTk but had little effect on the S2 block mutant (Fig. 4C). Collectively, these results suggest that HNF1␣ regulates pdx-1 activation, but like HNF3␤ stimulation of Area II, this process involves functional interactions with a ␤ cell activator(s) residing in a neighboring control region.
HNF6 Binds Inefficiently to S5-Although S5 is highly homologous to the consensus HNF6 binding site, in vitro translated HNF6 bound very poorly when compared with the HNF6 site of the HNF3␤ promoter (Fig. 5, A and B). Similarly, S5 was only a slightly more effective HNF6 binding competitor than the S5 block mutant (Fig. 5C). In addition, HNF6 binding to S5 was not detected in ␤ or non-␤ (BHK and liver) nuclear ex- tracts, although binding was readily observed with the HNF3␤ site probe (Fig. 5D).
The prominent S5 binding complex was unique to ␤ cell lines and was termed ␤EF2 (Fig. 5D). Competition analysis demonstrated that ␤EF2 binding was reduced by the S5 wild type competitor but not the S5 block mutant or HNF3␤ promoter competitor (Fig. 5E). These results suggested that the protein(s) that forms the ␤EF2 complex is the S5 activator and not HNF6. The inability to see a change in PDX-1 protein levels in HNF6 null mice also indicates that this factor does not regulate pdx-1 transcription (35).
␤EF1 and ␤EF2 Contain PDX-1-The data from the preceding experiments indicated that a ␤ cell-enriched protein(s) binds to S2 and S5. Because these segments contain the characteristic homeodomain core binding sequence (i.e. TAATA), the islet- enriched PDX-1, Pax4, Pax6, Nkx2.2, Nkx6.1, and   FIG. 4. HNF1␣ stimulates Area I-driven expression in ␤ cells. A and B, the pdx-1 Area I (A) and Pst-Bst (B) region reporter plasmids are shown diagrammatically. Area I is shaded in the S2 core block mutant (S2 7-11) and hatched in the S2 altered specificity mutant (S2 to ␤-FIB). The wild type and mutant Area I⅐pTk (A) and Pst-Bst⅐pTk (B) plasmids were transfected into HIT-T15, ␤TC3, and MIN6 cell lines. The normalized activity Ϯ S.D. of each Area I mutant construct is presented as the percentage of activity of the corresponding wild type plasmid. C, NIH 3T3 cells were co-transfected with an HNF1␣ expression vector (pBJ5-HNF1␣ or vector alone (pBJ5) and the wild type or S2 block mutant Area I⅐pTk. Area I is shaded in the S2 block mutant (S2 BLOCK MUT). The normalized activity ϩ S.D. of each construct is presented as fold pBJ5-HNF1␣ activation relative to pBJ5.
Hlxb9 homeodomain proteins were analyzed for binding to S2 (Fig. 6A) and S5 (Fig. 6B). Binding was only observed with in vitro translated PDX-1, and the complex formed co-migrated with ␤EF1 and ␤EF2 (Fig. 6, A and B). Preincubating the ␤ extract with anti-PDX-1 antisera also specifically eliminated or supershifted the ␤EF-1 (Fig. 6C) and ␤EF-2 (Fig. 6D) complexes. As a final test for PDX-1 binding, a competition experiment was performed with the PDX-1 binding A3 element of the insulin gene. As expected, this competitor prevented formation of ␤EF-1 (Fig. 6C) and ␤EF-2 (Fig. 6D), although it did not affect HNF1␣ binding to S2 (Fig. 6C). S5 and A3 were found to bind with similar efficacy to PDX-1 in these assays and 5-fold higher than S2 (data not shown). These data demonstrated that PDX-1 can bind in vitro to two distinct cis-acting regulatory elements of Area I.
PDX-1 Stimulates Area I Activation-To determine whether PDX-1 binding to Area I mediated stimulation of Area I⅐pTk and Pst-Bst⅐pTk, S2 and S5 sequences were changed to more closely resemble the insulin A3 element. The S2 mutation specifically eliminated HNF1␣ binding (Fig. 6C and data not  shown). The activity of the A3 conversion mutants was compared in ␤ cells with the wild type and the core element block mutant.
There was little or no difference on S2 activation of Area I FIG. 5. S5 binds a ␤ cell-enriched factor but not HNF6. A, sequences of S5, the S5 core block mutant ( , and the HNF6 binding site from the HNF3␤ promoter are shown. The nucleotide similarity to the consensus HNF6 binding site is also shown, with nonmatching nucleotides in lowercase letters (57). D ϭ A, T, or G; H ϭ A, T, or C; W ϭ A or T; Y ϭ pyrimidine, T, or C). B, S5 and the HNF3␤ promoter probes were incubated with in vitro translated (IVT) HNF6. The binding reactions with S5 and HNF3␤ promoter probes contained 5 and 1 l of the HNF6 translation reaction, respectively. The HNF6 complex is indicated with an arrow. C, binding reactions were conducted with the HNF3␤ promoter probe using liver nuclear extracts. A 100-fold molar excess of unlabeled S5, S5 8 -11 MUT, and HNF3␤ promoter competitor to probe was used. D, nuclear extracts from liver, BHK, HIT-T15, ␤TC3, or MIN6 cells were analyzed for S5 (lanes 1-5) and HNF3␤ promoter (lanes 6 -10) binding. The principal complex formed with the S5 probe in liver extracts is nonspecific (NS) (data not shown). Gel shift assays performed with HNF6-specific antiserum confirmed the presence of HNF6 in the complex detected specifically with the HNF3␤ probe in liver and ␤ nuclear extracts (data not shown). E, binding reactions were conducted with the S5 probe and ␤TC-3 extracts either alone (Ϫ) or in the presence of a 100-fold molar excess of S5, S5 8 -11 mutant, or the HNF3␤ promoter competitor to probe. Note that only S5 competed effectively with the probe for ␤EF2 binding. alone in the A3 conversion mutant versus the 7-11 mutant in HIT-T15 and ␤TC3 cells, although stimulation was observed in MIN6 cells (Fig. 7A). In contrast, Pst-Bst activity was stimulated in all of the ␤ cell lines by the S2 to A3 mutation (Fig. 7B). The A3 conversion mutant in S5 more profoundly influenced activation, because both Area I (Fig. 7C) and Pst-Bst (Fig. 7D) activities were increased. Depending upon the cell line, Area I alone activation was potentiated 6 -13-fold over the core mutant and 1.5-2.5-fold over the wild type. The S5 to A3 conversion mutation in Pst-Bst was essentially the same as wild type Pst-Bst. These results suggested that binding of PDX-1 to S5 and/or S2 was involved in regulating pdx-1 transcription.
PDX-1 Binds to Area I in Vivo-The ChIP assay is a powerful tool for analyzing the occupancy of transcription factors on their cognate binding elements in vivo (62,63). To more fully address the possibility of PDX-1-mediated regulation of pdx-1 gene expression, we used the ChIP assay to determine whether PDX-1 binding to Area I could be observed within the context of the endogenous gene. Immunoprecipitation of formaldehyde cross-linked chromatin from ␤TC3 cells with antibodies specific to the N-or C-terminal region of PDX-1 precipitated Area I sequences, whereas rabbit IgG, CDX-4-specific antisera, or the no antibody controls did not (Fig. 8, top panel). In contrast, anti-PDX-1 antisera did not immunoprecipitate promoter sequences from the PEPCK gene, which is not transcribed in ␤TC3 cells (Fig. 8, bottom panel). Similar results were observed with MIN6 cells (data not shown). Together with the gel shift and transfection data described above, these results demonstrate that PDX-1 regulates its own transcription. DISCUSSION PDX-1 plays a critical role in both pancreatic development and adult islet ␤ cell function. The nuclease HSS1 region (Ϫ2560 to Ϫ1880 bp) of the pdx-1 gene contains crucial cisacting elements involved in expression. The basis for this proposal was supported by data demonstrating that: 1) only HSS1 was independently capable of directing islet ␤ cell selectively expression in transfection assays performed with pdx-1 driven constructs spanning the 4.5-kilobase promoter region (39); 2) the only significant sequence conservation within the promoter region was found within HSS1, as defined by the distinct functional subdomains (i.e. Areas I, II, and III) (41); and 3) a pdx-1-driven transgenic construct spanning Areas I and II was selectively expressed in islet ␤ cells in mice (39). Collectively, these results strongly suggested that HSS1 was regulated by transcription factors critical for ␤ cell formation and function. To localize regulatory elements shared among the mammalian pdx-1 genes, the segments of identity within Area I were specifically mutated, and the effect on activation in ␤ cells was determined in transfection assays. The analyses performed in HIT-T15, ␤TC-3, and MIN6 cells localized sequences that were involved in stimulation by this HSS1 subdomain (S2, S4, S5, S6, S7, and S8), as well as conserved segments of little or no regulatory importance (S1, S3, and S9). In addition, HNF-1␣ and PDX-1 itself were shown to mediate activation from Area I.
Strikingly, a transcription factor data base analysis of the mutational sensitive sequences in Area I suggested that proteins critical in endocrine cell development (HNF6, S5) and function (HNF1, S2; HNF4, S4 and S7-8) contribute to activation. However, only HNF1␣ was found to bind segment-specific (␤NE) are shown. The HNF1␣ and ␤EF binding complexes described in the text are labeled, as are the nonspecific complexes (NS). C and D, HIT-T15 nuclear extract binding to the S2 (C) and S5 (D) probes was conducted in absence (Ϫ) or presence (ϩ) of N-or C-terminal PDX-1 antisera and a 200-fold molar excess of the S2, S5, or insulin A3 competitor. probes in gel shift experiments performed with in vitro translated HNF1␣, HNF4␣, and HNF6 or nuclear extracts prepared from factor-producing cells. Our inability to detect HNF4␣ or HNF6 binding strongly suggests that these key regulatory proteins do not directly control Area I activation. The absence of a change in PDX-1 protein levels in HNF6 null mice also indicates that this factor does not play any role in regulating pdx-1 transcription (35). Endocrine islet cell development was severely impaired in these mice, apparently because of the requirement for HNF6 in neurogenin 3 expression (35), a transcription factor that is critical in determination of endocrine islet precursors (64).
To test whether HNF1␣ regulated Area I activation in ␤ cells, S2 sequences were changed to select for HNF1␣ binding (S2 to ␤-FIB) within the context of pdx-1 reporter constructs driven by human Area I alone or mouse Pst-Bst, which spans Areas I and II. The activity of the altered specificity mutants were compared with the wild type as well as the HNF1␣ binding defective mutant construct, S2 7-11 (Fig. 4). The S2 to ␤-FIB mutant only effectively activated the pdx-1 construct spanning Areas I and II, suggesting that stimulation by HNF1␣ involves functional interactions with an Area II activator, a situation that parallels the requirement of Area I for HNF3␤ activation of Area II (41). HNF1␣ was also shown to specifically activate S2-directed expression in co-transfection assays in NIH 3T3 cells. In contrast to S2, the HNF1␣ binding site in the human pdx-1 gene recently described may not be of general regulatory significance, because it is not conserved within the mouse or chicken genes (65).
The HNF-1 family of transcription factors bind DNA as homo-or heterodimers (61). Interestingly, HNF1␤ is co-expressed with pdx-1 in the ventral and dorsal walls of the primitive foregut during early pancreas development (66). In contrast, HNF1␣ is expressed at a later developmental stage and at much higher levels than HNF1␤ in adult ␤ cells (66). We were only able to detect HNF1␣ in the S2-specific HNF1 complex from HIT-T15, ␤TC-3, and MIN6 nuclear extracts, although HNF1␤ was independently shown to be capable of binding to S2 (data not shown). Thus, these results support the possibility that HNF1␣ and/or HNF1␤ function as activators of pdx-1 expression during pancreas specification and in the adult islet ␤ cell.
The decrease in insulin synthesis and secretion found in mice lacking HNF1␣ also provides support for a role in pdx-1 transcription (30,67), because decreased expression of PDX-1 would be expected to have a debilitating effect upon ␤ cell function through its actions on GLUT2 and insulin transcription. Two distinct lines of HNF1␣ null mice have been independently derived by Lee et al. (30) and Pontoglio et al. (68). The same HNF1␣ sequences were targeted for removal (i.e. amino acids 1-108), although each used a different strategy to eliminate them. Both lines result in decreased insulin mRNA expression. However, the absence of HNF1␣ expression had a distinct effect on islet pdx-1 expression as well as the overall physiology of these animals. The decrease in pdx-1 mRNA expression in HNF1␣ null mice from Pontoglio et al. was 2.4fold in the newborn pancreas islets and 2.9-fold in the adult (Fig. 1A of Ref. 69), whereas little or no effect was found in those from Lee et al. in islets from 2-week-old mice (Fig. 3 of Ref. 70). It is unclear which of these mouse models more closely resembles regulation in the human.
In addition to HNF1␣, PDX-1 was found to specifically bind to S2 in gel shift assays performed with ␤ cell nuclear extracts (Fig. 6, A and C) and was the principal S5 binding activity detected in these extracts (Fig. 6, B and D). Furthermore, Area I-and Pst-Bst-mediated activation were compromised in the S5 8 -11 mutant (Fig. 7, C and D), demonstrating the critical nature of this site in transcriptional regulation. PDX-1 activated the altered specificity mutants that replaced S2 or S5 with insulin A3. Importantly, PDX-1 was shown to specifically bind to Area I of the endogenous pdx-1 gene using the ChIP assay. In contrast, we were unable to demonstrate HNF1␣ binding to Area I in vivo (data not shown), which may either reflect occupancy of S2 by PDX-1 or simply a technical inability to recover HNF1␣-bound complexes in our ChIP assay.
These data strongly suggest that pdx-1 transcription is autoregulated. However, PDX-1 is not absolutely required for transcription because expression is detected during embryogenesis in pdx-1 Ϫ/Ϫ mice (9) and in the presence of a dominant negative PDX-1 mutant in ␤TC-3 cells (71). Autoregulation is utilized by other mammalian homeodomain-encoding genes, apparently to prevent extreme fluctuations in expression (72)(73)(74)(75)(76)(77). Because PDX-1 is the major S5 binding factor, it is likely to bind and regulate through this site. Support for this proposal is also provided by quantitative competition assays showing that PDX-1 binds with greater affinity to S5 than S2 in vitro and transfection data indicating that PDX-1 can transactivate through S5 (78). We are currently working toward developing model systems to address the importance of PDX-1 autoregulation in vivo.
Islet ␤ cell-specific expression clearly relies on the activity of many cell-enriched transcription factors, which function cooperatively to impart control (1,3). The data presented have defined conserved cis-acting regulatory elements that are necessary for pdx-1 expression in the ␤ cell. Furthermore, it is possible that the Area I sites that were insensitive to mutation in our assays reflect the limitations of our tumor cell lines rather than their importance in regulation in vivo. Identifying the regulatory factors that control the expression of genes such as pdx-1 may provide insight into the inherited defects that cause insulin deficiency and diabetes. For example, defects in ␤ cell function in mature onset diabetes of the young patients expressing a dysfunctional HNF1␣ or PDX-1 protein may in part result from defects in pdx-1 expression, because reduced PDX-1 levels would, in turn, likely have profound consequences on expression on many of the target genes involved in glucose sensing, including insulin, glucokinase, and GLUT2.