Regulation of pdx-1 Gene Expression

The coding region of the pdx-1 gene comprises two exons. The first exon encodes for the NH₂-terminal region of PDX-1, and the second encodes for the homeodomain and COOH-terminal domain. The human, mouse, and rat genes are localized on chromosomes 13 (15,16), 5 (17), and 12 (18), respectively. Although the activation domain of PDX-1 is contained within the NH₂-terminal domain, its homeodomain is involved in DNA binding; both are involved in protein- protein interactions (19–25). The pdx-1 gene is TATA-less; thus, it utilizes three principal transcription initiation sites (17), followed by a short 5′ untranslated sequence of ∼100 nucleotides.

To understand the mechanisms that control the expression of PDX-1 during pancreas development and in the adult β-cell, the pdx-1 gene from different species was mapped. Regulatory regions lying upstream from the transcription start sites are under characterization in transgenic mice as well as in cultured cells in several laboratories. A genomic fragment containing ∼6.5 kb of the 5′-flanking rat pdx-1 sequence was sufficient to target β-galactosidase expression to pancreatic islets and duodenal cells in transgenic mice (17). A longer fragment containing the coding region and the 3′-flanking sequences of the gene restored the development of all pancreatic lineages and corrected glucose intolerance in pdx-1^−/− animals (24). In transiently transfected β-cells, the fragment extending from −6.2 to +68 linked to a reporter gene showed 20- to 100-fold higher activity than that in non-islet cells. Using deletion analyses, the β-cell-specific regulated expression of the rat sequence appeared to require a distal enhancer element located between the −6.2- and −5.67-kb region of the gene. This element was shown to bind the endodermal factors HNF-3β and Beta2, which act cooperatively to induce PDX-1 expression. Furthermore, glucocorticoids reduced pdx-1 gene expression by interfering with HNF-3β activity on the islet enhancer (26).

To characterize the regulatory elements and potential transcription factors necessary for the expression of human pdx-1 in β-cells, a series of 5′ and 3′ deletion fragments of a 7-kb sequence of the 5′-flanking region of the gene, fused to a reporter gene, was tested. By transient transfections in β-cells and non-β-cells, a β-cell- specific distal enhancer element located between −3.7 and −3.45 kb was delineated. This enhancer fragment strongly stimulated reporter gene activity in all β-cell lines tested and was much less active in non- β-cells, including glucagon-producing cells, and in acinar and hepatic cells. No sequence similarity was revealed between the enhancer sequence and the available mouse or rat pdx-1 genomic sequences. DNase I footprinting analysis revealed two protected regions: one binding the transcription factors SP1 and SP3 and another binding HNF-3β and HNF-1α. Similar to the rat distal enhancer, these factors act in concert to regulate the transcriptional activity of the pdx-1 gene (27).

In vivo characterization of the mouse pdx-1 sequences was independently initiated using a fragment of ∼4.5 kb upstream of the initiation start site (28–30). The transgene driven by these sequences was shown to approximately recapitulate the endogenous expression pattern. An extended fragment containing the coding region, 3 kb of the 3′-flanking region, and 6.2 kb of upstream sequences was able to completely rescue the apancreatic pdx-1 null phenotype and ultimately restore glucose homeostasis. Moreover, the sequences sufficient for appropriate developmental and islet-specific expression were located within ∼4.5 kb of 5′-flanking DNA (30). The β-galactosidase was detected in Brunner’s glands of the proximal duodenum and pyloric glands of the distal stomach and coincided with the expression of pdx-1 mRNA. Occasionally, ectopic activity was observed in exocrine tissue of the adult pancreas, submucosal layer of the duodenum, and even in the spleen (28).

Failure of the pancreas to develop in both humans and mice lacking PDX-1, as well as the dosage-dependent effect of PDX-1 on the expression of β-cell-specific genes (and on the maintenance of euglycemia), led to the assumption that sequences conserved between the two species could be essential for its transcriptional control. A striking divergence at the nucleotide level was observed between the two species with the exception of four regions that showed significant (94, 81, 73, and 78%) similarity. In addition to the conserved proximal promoter sequence (20), three short highly homologous regions were found between −2.81 and −1.67 kb of the human and between −2.7 and −1.8 kb of the mouse pdx-1 gene (Fig. 1). These regions were designated PH1, PH2, and PH3 for PDX-1 homologous regions 1–3 (22) or areas I, II, and III, as determined by Gerrish et al. (31). In transient transfection experiments, each of the conserved sequences was able to confer β-cell-specific activity on a heterologous promoter; however, it was done to different extents. PH1/areaI and PH2/areaII showed the highest preferential induction in β-cell versus non-β-cell activity (22,31). An interesting observation was the absence of the PH2/areaII domain in the chicken 5′-flanking region (31), suggesting that the regulation of pdx-1 expression in birds may differ from that in rodents and humans.

Attempting to identify factors that regulate the transcriptional activity of the conserved domains, DNase I footprinting analyses, gel electrophoretic mobility shift assays and mutational studies led to the identification of several transcription factors (Fig. 1). PH1/areaI and PH2/areaII sequences bind and are transactivated by HNF-3β. Although mutations in the HNF-3β binding site within the PH2/areaII sequence did not modify its transcriptional activity, in PH1/areaI, it had a profound effect. Interestingly, the PH1/areaI enhancer element was reported to bind the PDX-1 transcription factor itself both in vitro (22) and in vivo (32), suggesting a possible autoregulatory loop as a mechanism for PDX-1 to control its own expression. The involvement of HNF-1α in regulating PH1/areaI was also determined (32). It was further shown that the PDX-1 protein binds HNF-3β, and all three transcription factors appear to act cooperatively to regulate transcription. Identification of factors binding and regulating conserved sequence PH3 is the focus of ongoing studies.

Thus, from the studies on the binding and the cooperativity between the different factors acting in concert to control the transcriptional activity of pdx-1 regulatory elements, it emerges that at least some aspects of the expression of the gene rely on the transcription factor HNF-3β. Indeed, its absence in mouse embryonic stem cells had a profound effect on pdx-1 gene expression (29). HNF-3β is a member of the forkhead/winged helix family of transcription factors and is essential for endodermal cell lineages (33,34). It is structurally related to histone H5, which can alter the nucleosomal structure and thus prime target genes for expression by opening the chromatin structure and providing promoter access to other transcription factors (35,36). Because HNF-3β is not restricted to β-cells, the selective transcription of PDX-1 is likely to rely on the combination of additional factors, among them HNF-1α, SP1, PDX-1 itself, and possibly other unidentified factors.

In addition, the distal human enhancer element and the PH1/areaI domain bind the HNF-1 members of transcription factors. HNF-1α and HNF-1β homeoproteins are capable of binding the HNF-1 site as homodimers or heterodimers (37,38). Transient transfection experiments in fibroblasts demonstrated that both HNF-3β and HNF-1α independently activate the distal human enhancer and, when cotransfected, act in a synergistic manner. However, HNF-1β did not affect enhancer-driven transcription separately or in combination with HNF-3β. Although high levels of HNF-1β mRNA are observed at 6–7.5 days of gestation (39), HNF-1α is expressed at a later developmental stage (40) and appears to be the predominant form present in adult β-cells (27). We found that the relative abundance of HNF-1α and HNF-β proteins differ in various pancreatic cell lines, suggesting that differences in HNF-1 subtype ratios may be one of the factors contributing to tissue-specific expression of the pdx-1 gene. The relative abundance of the two major HNF-1 species was recently suggested to be a mechanism for expression pattern determination for the glut-2 gene (41). The in vivo studies on HNF-1α knockout mice did not give an unequivocal answer regarding the importance of this transcription factor for pdx-1 gene expression. Shih et al. (42) reported that in HNF-1α null animals’ pdx-1 mRNA levels were significantly decreased (42); however, little or no effect on its expression was obtained by Parrizas et al. (43). Furthermore, no reduction in PDX-1 protein was observed in transgenic mice with selective expression of a dominant-negative form of HNF-1α in pancreatic β-cells (44).

Mutations in the HNF-1α gene represent the most frequent form of MODY (45). The observed diabetic phenotype was first suggested to be the result of impaired binding of HNF-1α to the insulin promoter, thus causing decreased transcriptional activity of the gene. However, binding of HNF-1α to the flanking AT sequences (FLAT)-F element of the insulin gene appears to be unique to the rat insulin I gene because this AT-rich motif is not conserved among insulin promoter sequences. In fact, HNF-1α is a weak transactivator of the human insulin gene; we therefore favor the explanation that the impact on insulin gene expression is indirect, via its effect on pdx-1 gene transcription. Impaired pancreatic function in MODY3 is also the result of defective transcription of genes involved in β-cell glucose sensing and glucose metabolism (42).

The in vivo importance of the conserved domains was addressed in transgenic mice (28–30). Fragment XbaI–XhoI, spanning the sequence between −4.3 and −1.88 kb of the mouse pdx-1 gene, which includes both PH1/areaI and PH2/areaII domains, directed the transgene expres‘sion to pancreatic islets but not to any other cell population in which PDX-1 is normally expressed. However, a smaller region (PstI–BstEII) that still contained PH1/areaI and PH2/areaII sequences was active in the islets as well as in the pyloric sphincter and the common bile duct. Furthermore, within the pancreas, expression of the reporter gene driven by the PstI–BstEII fragment was found in the majority of insulin-, glucagon-, and somatostatin-producing cells; therefore, this fragment was considered to be an endocrine-specific enhancer. Furthermore, the fragment (XhoI–BglII) containing PH3/areaIII drove the reporter expression to clusters of insulin-producing cells, whereas it was almost inactive in glucagon- and somatostatin-positive cells of the neonatal pancreas. However, this β-cell-specific activity was transient and lost in adult pancreases (30). Although the reason for the silencing of this transgene is not clear, it may be proposed that the XhoI–BglII region plays a specific role in immature β-cells or provides critical developmental cues for the initiation of the mature β-cell lineage. Altogether, these data suggest that the conserved domains confer islet-specific, and to a certain extent β-cell- specific, transcription both in vivo and in vitro. However, separately or in combination, these elements cannot faithfully recapitulate the expression of endogenous pdx-1, thus implicating that additional regulatory elements must be involved in this process. Mutations or deletions of important regulatory sequences in the context of the endogenous pdx-1 gene will help assess their critical importance in the regulated expression of the gene.

As mentioned earlier, the pdx-1 gene is TATA-less, and although the sequences corresponding to the transcription start sites among the rat, mouse, and human genes are highly conserved (17) (Fig. 2), great heterogeneity is revealed further upstream. The β-cell-specific transcriptional regulation of the rat pdx-1 gene was reported to rely in part on a proximal promoter sequence containing an E-box motif located at −104. Pursuing our search for functional regulatory elements in the human pdx-1 5′-flanking region, deletion analysis of the proximal promoter was performed, and a β-cell-specific regulatory sequence between −160 and −100 bp was identified. Comparison of this region between the promoters of insulin genes from different origins showed a relatively high degree of homology (78%), with greater heterogeneity further upstream. DNAse I footprinting, using the conserved proximal promoter of the human gene and β-cell extracts, revealed a specific protected region around the conserved E-box motif (CACGTG) (17). This site predominantly binds a complex containing the transcription factor USF (17). Mutations abolishing its binding impaired the activity of the pdx-1 promoter. Expression of a dominant-negative form of USF-2 in β-cells reduced both the pdx-1 promoter activity as well as PDX-1 mRNA and protein levels. This led to a reduction of PDX-1 binding to the insulin promoter and consequently a dramatic decrease in insulin gene expression (46). Thus, USF1 and USF2 interactions with the E-box sequence in the pdx-1 promoter appear to contribute to the preferential expression of the gene in β-cells.

Taken together, these results suggest that the transcriptional stimulation of pdx-1 in β-cells is mediated by a unique combination of protein-protein interactions and that separate modules in the gene can be active at a given stage by binding a specific set of transcription factors. Indeed, the transcription factors HNF-3β, HNF-1α, HNF-1β, SP1/3, USF1/2, and PDX-1 itself regulate the expression of the pdx-1 gene. Most of these factors have been previously shown, mainly by knockout experiments in mice, to be important developmental regulators. PDX-1 is a key factor with multiple functions both during development and in the differentiated β-cell. In the adult, its role in regulating islet-specific genes and, most importantly, in mediating the glucose effect on insulin gene transcription emphasizes its particular role in normal and diabetic states.

The study was supported by the Juvenile Diabetes Research Foundation International (1-2001-325).

The authors wish to express their gratitude to Tamara Gurevich, Michal Shoshkes, and Etti BenShushan for their contributions.