Transcriptional and Translational Regulation of β-Cell Differentiation Factor Nkx6.1*

In the mature pancreas, the homeodomain transcription factor Nkx6.1 is uniquely restricted to β-cells. Nkx6.1 also is expressed in developing β-cells and plays an essential role in their differentiation. Among cell lines, both β- and α-cell lines express nkx6.1 mRNA; but no protein can be detected in the α-cell lines, suggesting that post-transcriptional regulation contributes to the restriction of Nkx6.1 to β-cells. To investigate the regulator of Nkx6.1 expression, we outlined the structure of the mouse nkx6.1 gene, and we identified regions that direct cell type-specific expression. Thenkx6.1 gene has a long 5′-untranslated region (5′-UTR) downstream of a cluster of transcription start sites.nkx6.1 gene sequences from −5.6 to +1.0 kilobase pairs have specific promoter activity in β-cell lines but not in NIH3T3 cells. This activity is dependent on sequences located at about −800 base pairs and on the 5′-UTR. Electrophoretic mobility shift assays demonstrate that homeodomain transcription factors PDX1 and Nkx2.2 can bind to the sequence element located at −800 base pairs. In addition, dicistronic assays establish that the 5′-UTR region functions as a potent internal ribosomal entry site, providing cell type-specific regulation of translation. These data demonstrate that complex regulation of both Nkx6.1 transcription and translation provides the specificity of expression required during pancreas development.

Among the known pancreatic transcription factors, the homeodomain factor Nkx6.1 is unique in its absolute restriction in the mature pancreas to the ␤-cells. In the developing fetus, however, Nkx6.1 is initially expressed in almost all the epithelial cells of the pancreatic buds. Starting around embryonic day 13 (E13), Nkx6.1 expression becomes restricted to ␤-cells and ␤-cell precursors (13). Targeted disruption of the nkx6.1 gene causes a severe defect in ␤-cell differentiation in mice. The nkx6.1 null mutants have normal numbers of insulin-expressing cells through E12.5, but new ␤-cell formation is blocked after E12.5. Hence, Nkx6.1 is a necessary component of the signals triggering the major wave of ␤-cell differentiation and proliferation after E12.5. 1 Nkx6.1 functions at one step in the hierarchy of transcription factors controlling pancreatic development and differentiation. Although Nkx6.1 represses the transcription of target genes (14), its downstream genetic targets have not been identified. Upstream of Nkx6.1, two homeodomain transcription factors are known to control its pancreatic expression, Nkx2.2 in the fetal pancreas after E12.5 and PDX1 in adult ␤-cells (9,15). However, the mechanisms by which these factors control Nkx6.1 expression and the potential roles of other factors are unknown.
Control of cell type-specific gene expression frequently operates at the level of gene transcription, but post-transcriptional mechanisms including controls at the level of translation initiation may play essential roles as well. Generally, cap-dependent ribosomal scanning identifies translation start sites and initiates translation on the majority of cellular mRNAs. This process is severely hampered on long 5Ј-untranslated regions (5Ј-UTR) 2 containing multiple upstream reading frames and secondary structure (16). Translation of such mRNAs may initiate through a cap-independent mechanism utilizing an internal ribosomal entry site (IRES) in the 5Ј-UTR. Cellular mRNAs containing IRESs can be very specifically regulated, providing a post-transcriptional mechanism to control their expression (17,18). By using this mechanism, the translation of a number of mammalian growth factor RNAs is specifically regulated during differentiation or cell growth (19 -22). Furthermore, during Drosophila embryonic development, the 5Ј-UTRs of mRNAs encoding homeodomain transcription factors Antp and Ubx are known to regulate protein expression in a spatiotemporal manner, although there are no reports of the existence of IRESs in any mammalian homeodomain transcription factor genes (23,24).
To understand the mechanisms that regulate ␤-cell-specific expression of Nkx6.1, we outlined the structure of the mouse nkx6.1 gene and identified a promoter that directs cell type-specific expression in ␤-cells. A promoter element found approximately 800 bp upstream of the transcription initiation sites contains binding sites for Nkx2.2 and PDX1 and functions as an important transcriptional enhancer. In addition, a potent IRES in the 5Ј-UTR further restricts Nkx6.1 expression to ␤-cells. These findings establish that gene regulation through an IRES plays a similar role in development of the mammalian pancreatic islet as in Drosophila development.
For transient mammalian cell transfections, ␤TC3 cells, ␣TC1.6 cells, and NIH3T3 cells were plated in 6-well tissue culture plates 24 h before transfection. For the standard reporter gene analysis, 1.8 g of each luciferase reporter plasmid and 0.2 g of the CMV␤-Gal plasmid were co-transfected into the cells using Superfect® (Qiagen) under conditions recommended by the manufacturer. Forty eight hours after transfection, cells were harvested, and luciferase and ␤-galactosidase assays were performed as described previously (14). Luciferase activity was corrected for transfection efficiency by use of the co-transfected CMV␤-Gal plasmid. For evaluation of PDX1 and Nkx2.2 effects on the reporter gene constructs, 50 ng of expression vector (pBAT12-IPF1, pBAT12hNkx2.2, or expression vector without insert, pBAT12 (14)) were co-transfected into NIH3T3 cells with 2.0 g of each luciferase reporter plasmid. Cells were harvested 48 h later and assayed for luciferase activity. All reporter gene analyses were performed on at least three occasions, and data are expressed as mean Ϯ S.E.
RNA Isolation and Northern Blot Analyses-Total RNA from cell lines was isolated using TRIzol® (Life Technologies, Inc.) per the manufacturer's protocol. Northern blots were performed by standard procedures using 10 g of total RNA (25). A fragment of hamster nkx6.1 cDNA was used as a probe for Northern analysis and was prepared by digesting pBAT12-Nkx6.1 (26) with KpnI and NotI and labeling the liberated fragment with [ 32 P]dCTP.
5Ј-Rapid Amplification of cDNA Ends (RACE)-The 5Ј end of mouse nkx6.1 cDNA was identified by 5Ј-RACE, using a modification of the protocol from the 5Ј-RACE system, version 2.0 (Life Technologies, Inc.). For mouse cDNA, 2.5 pmol of specific primer HW8 (5Ј-GCG TTC GCT TTG ATG TAG GA-3Ј) was annealed to 1 g of total RNA from ␤TC3 cells. Reverse transcription was carried out using SuperScript II reverse transcriptase (Life Technologies, Inc.). After first strand cDNA synthesis, the original mRNA template was removed by treatment with RNase, and homopolymeric dCTP tails were then added to the 3Ј end of the cDNA using terminal deoxynucleotidyltransferase. By using these products as a template, we carried out 35 cycles of PCR using the 5Ј-RACE Abridged Anchor Primer (Life Technologies, Inc.) and HW9 (5Ј-CGC CTG GGG TAG CTT CAA AG-3Ј) as primers. For the nested PCR, we used Abridged Universal Amplification Primer (Life Technologies, Inc.) and HW11 (5Ј-GCG GAT CCG CCT CTG ATC TCG CTC GGA -3Ј) as primers, and we performed 35 cycles of PCR. The PCR products were subcloned into pBluescript KS(ϩ) and sequenced.
RNase Protection Assay-The fragment from nucleotide Ϫ159 to nucleotide ϩ100 of the nkx6.1 gene was amplified by PCR and subcloned into pBluescript. Labeled antisense RNA probe was generated using this fragment as a template. RNase protection assays were carried out using HybSpeed RPA ® kit (Ambion) per manufacturer's protocol. Hybridization of the riboprobe to RNA was performed in a 10-l reaction containing 8 ϫ 10 4 cpm of probe and 10 g of total RNA from ␤TC3 cells and 40 g of yeast tRNA. The control sample contained 50 g of yeast tRNA alone.
Cloning of the Mouse nkx6.1 Gene Promoter-A DASH mouse genomic library was screened for the nkx6.1 gene using a mouse nkx6.1 partial cDNA probe corresponding to the coding region of exon1. The DASH clone encoding the longest 5Ј region of the nkx6.1 gene was subcloned into the EcoRI site of pBluescript KS(ϩ). This plasmid contains an approximately 10-kb fragment of nkx6.1 gene (pBSNkx6.1-10 kb). This clone was characterized by restriction enzyme analysis and sequencing.
Reporter Gene Constructs and Assay-To generate reporter plasmids, fragments of the 5Ј region of the nkx6.1 gene (obtained either by restriction digestion or PCR) were ligated upstream of the luciferase gene in the plasmid pFOXLuc1 (14). Mutagenesis of the reporter gene constructs was performed using the Quick Change® mutagenesis kit (Stratagene). All constructs were confirmed by sequencing.
In Vitro Transcription and Translation and Electrophoretic Mobility Shift Assay (EMSA)-Nkx2.2 and PDX-1 proteins were produced in vitro using T7TNT Quick Coupled Lysate System® (Promega). Singlestranded wild-type oligonucleotides (5Ј-GATCTAGCCCCTCATAAGT-GATAATGATCTAGGGG-3Ј), corresponding to the sequence between nucleotides Ϫ817 and Ϫ788 (B1) and (5Ј-CGGAAGAGACGCACTTA-AACTGCTTTTC-3Ј) corresponding to the sequence between Ϫ478 and Ϫ441 nucleotides, were 5Ј-end-labeled with [␥-32 P]ATP using T4 polynucleotide kinase. The labeled oligonucleotide was column-purified and annealed to an excess of complementary strand. EMSA buffers and electrophoresis conditions were as described previously (27). One l of the in vitro reaction mixture was used for the EMSAs. The following oligonucleotides were used as competitors in EMSA reactions (top strands shown): M1, GATCTAGCCCCTCATAAGTGATGGTGATC-TAGGGG, and M2, GATCTAGCCCCTCATGGGTGATAATGATCTA-GGGG.
Western Blotting Analyses-Expression of Nkx6.1 in nuclear extracts was measured by performing Western blot analysis using polyclonal anti-Nkx6.1 antibody (14). Western blots were visualized by using the ECL Plus® system (Amersham Pharmacia Biotech).
Dicistronic Plasmids and Dicistronic Assay-For generating the basic dicistronic construct, the RSV promoter region driving the CAT gene was inserted upstream of the luciferase gene in pFOXLuc1 to obtain pFoxRSV-CAT-Luc. The 5Ј-UTR of nkx6.1 was inserted bi-directionally into the region between the CAT and luciferase genes to obtain pFoxRSV-CAT-5Ј-UTR-Luc and pFoxRSV-CAT-5Ј-UTR-R-Luc. Next, by using PCR-based site-directed mutagenesis, the cloning sites and 5Ј- UTR of the luciferase gene were removed. The constructs were confirmed by sequencing. These dicistronic reporter genes were co-transfected with pBluescript KS(ϩ) into mammalian cells, and 48 h after the transfection cells were harvested and assayed for luciferase and CAT activities, as described previously (28). Luciferase enzyme activity from each transfection was normalized to the activity of CAT and was used as an index of IRES activity.

Expression of Nkx6.1 in Cell
Lines-To identify cell lines that express Nkx6.1, we performed Northern blot analysis with a hamster nkx6.1 cDNA probe and Western blot analysis with antiserum directed against the carboxyl-terminal end of the hamster Nkx6.1 protein. Total RNA and nuclear extracts for these assays were prepared from ␤-cell lines ␤TC3 and HIT-T15, from ␣-cell lines INR-1 and ␣TC1.6, and from non-pancreatic cell lines NIH3T3 and COS7. As shown in Fig. 1, nkx6.1 mRNA is expressed in all four pancreatic islet cell lines but not in NIH3T3 cells. However, the Nkx6.1 protein can be detected only in ␤TC3 cells. Although the expression level of nkx6.1 mRNA is higher in ␤TC3 cells than the other cell line, this difference alone cannot explain the greater difference in the expression of Nkx6.1 protein. These results suggest that the expression of Nkx6.1 is regulated both transcriptionally and post-transcriptionally.
Structure of the Mouse nkx6.1 Gene-As an initial step toward characterizing the 5Ј end of the nkx6.1 gene, we cloned and sequenced exon 1 and the 5Ј-flanking region of the mouse nkx6.1 gene. 5Ј-RACE performed on RNA from ␤TC3 cells identified a cluster of seven transcription start sites clustered at ϳ1 kb upstream from the translation start site with no additional intervening introns (Fig. 2B). RNase protection analysis with ␤TC3 RNA confirmed the position of four of these start sites.
The strongest band by RNase protection was defined as ϩ1 bp, and other nucleotides are numbered accordingly. The mouse genomic sequence lacked a TATAA box upstream of the transcription start sites, but consensus CCAAT box sequences are located at Ϫ220 bp and at Ϫ150 bp (29).
Deletion Analysis of the Mouse nkx6.1 Promoter-When ligated upstream of the luciferase reporter gene and transfected into mammalian cell lines, a large fragment of the mouse nkx6.1 gene including 5.6 kb of 5Ј-flanking sequence and 973 bp of the 5Ј-UTR was sufficient to direct the expression of luciferase in the ␤-cell line ␤TC3 but not in the fibroblast cell line NIH3T3 (Fig. 3), demonstrating that this fragment of the gene contains sequences that are important for ␤-cell-specific expression. Deletion of the sequence between Ϫ5600 and Ϫ2570 bp increases luciferase activity modestly in both cell lines. This increase is lost, however, when equal molar quantities of the plasmid are used for the transfections (data not shown).
Further deletion of the region between Ϫ893 and Ϫ645 bp causes a significant decrease in promoter activity in ␤TC3 cells but not in NIH3T3 cells, implicating this region in ␤-cell-specific expression. Deletion of sequences within the proximal 334 bp of the promoter causes the progressive diminution of promoter activity in both cell types showing that this region is important for basal promoter activity, although the greater decrease in activity in ␤-cells suggests some degree of ␤-cellspecific function for these sequences.
To map more precisely the ␤-cell-specific enhancer sequences within the region between Ϫ893 and Ϫ645 bp, we generated a series of small deletions within this region. As shown in Fig.  4A, the sequences between Ϫ840 and Ϫ771 bp are necessary for this ␤-cell specific activity. This sequence also can function weakly as a ␤-cell-specific enhancer when linked to a heterologous promoter (Fig. 4B). This sequence contains two potentially important binding sites for ␤-cell transcription factors, the Nkx2-2-binding site core sequence TAAGTG (30) and the PDX-1-binding site core sequence TAAT (31). Mutation of the potential PDX-1-binding site causes a significant fall in promoter activity, although mutation of the Nkx2-2-binding site causes a more modest decrease in promoter activity (Fig. 4C).
PDX-1 and Nkx2-2 Bind to the nkx6.1 Promoter-To address whether PDX-1 and Nkx2-2 bind to these sites, we performed an EMSA using a double-stranded oligodeoxynucleotide corresponding to nucleotide Ϫ817 to Ϫ788 (B1) as a probe. In vitro translated Nkx2-2 and PDX-1 can bind to this site (Fig. 5A), and they are competed by the unlabeled oligonucleotide. An unlabeled oligonucleotide containing a mutation in the Nkx2-2-binding core (mutant M2) can still compete for PDX-1 binding but not for Nkx2-2 binding. Interestingly, an unlabeled oligonucleotide containing a mutation in the PDX-1-binding core (mutant M1) cannot compete for either PDX-1 or Nkx2-2 binding.
When co-transfected into NIH3T3 cells, PDX-1 can activate the ␤-cell-specific enhancer linked to the rat prolactin promoter (Fig. 5B). Nkx2-2, however, cannot activate the mini-enhancer Relative luciferase activities are calculated with the activity of cells transfected with the pFOXLuc1 plasmid alone set at 1. The asterisks indicate a p value Ͻ0.01 for the comparison of the activities of the Ϫ771 promoter with the Ϫ840 promoter and Ͻ0.05 for the comparison of the activities of the Ϫ771 promoter with the Ϫ893 promoter in ␤TC3 cells as calculated by the paired Student's t test. B, a reporter plasmids was constructed with the Ϫ840 to Ϫ6450bp fragment of the nkx6.1 promoter inserted upstream of the minimal rat prolactin promoter and the luciferase gene and then was co-transfected with a CMV promoter-driven ␤-galactosidase expression plasmid into NIH3T3 cells (filled bars) or ␤TC3 cells (hatched bars). Relative luciferase activities are calculated with the activity of cells transfected with the pFOXLuc1 plasmid containing only the prolactin promoter set at 1. The asterisk indicates a p value Ͻ0.05 for the comparison of the activities of the pFOXluc.prl.Nkx6.1(Ϫ840/Ϫ645) plasmid with the pFOXluc.prl plasmid in ␤TC3 cells as calculated by the paired Student's t test. C, reporter plasmids containing the Ϫ893-bp promoter fragment and complete 5Ј-UTR with or without the mutations shown upstream of the luciferase gene were co-transfected with a CMV promoter-driven ␤-galactosidase expression plasmid into ␤TC3 cells. Relative luciferase activities are calculated with the activity of cells transfected with the pFOXLuc1 plasmid alone set at 1. The asterisk indicates a p value Ͻ0.05 for the comparison of the activities of the M1 mutant promoter with the wildtype promoter as calculated by the paired Student's t test. All data are shown as mean Ϯ S.E. by itself or in combination with PDX-1 (Fig. 5B), although it can activate the intact nkx6.1 promoter (data not shown). Neither factor affects the expression of luciferase from the parent vector containing the prolactin promoter alone. Recently Sepulveda et al. (32) demonstrated that the closely related cardiac homeodomain factor Nkx2.5 cooperates with GATA-4, a zinc finger transcription factor, to activate the ␣-actin promoter. Interestingly, for this interaction, a GATA4-binding site is not necessary. To test the possibility that Nkx2.2 also cooperates with GATA factors in pancreatic ␤-cells to activate the nkx6.1 promoter, we co-transfected vectors expressing either GATA4 or GATA6 along with the Nkx2.2 expression vector and a reporter plasmid containing either the ␤-cell-specific minienhancer linked to the rat prolactin promoter or the intact nkx6.1 promoter driving luciferase. However, neither GATA4 nor GATA6 produced any additional activation of the nkx6.1 promoter or mini-enhancer (data not shown).
There are additional PDX-1 and Nkx2-2-binding sites in the nkx6.1 promoter outside of the ␤-cell-specific enhancer region. There are multiple TAAT sequences that fit the PDX-1 binding consensus within the proximal 900 bp of the promoter (see boldface sequences in Fig. 2B) as well as several copies of the (C/T)AAG sequence that forms the core of the Nkx2-2-binding sequence. One of these sites, located at Ϫ460 bp, can function as a high affinity Nkx2-2-binding site (Fig. 5C).
It should be noted as well that there are other potential binding sites for ␤-cell transcription factors in the nkx6.1 promoter, including two copies of the HNF6-binding site consensus sequence (33) (see Fig. 2B). The functional importance of these sites is difficult to ascertain since they fall within a region of the proximal promoter that is also important for expression in NIH3T3 cells.
Complex Function of the nkx6.1 Gene 5Ј-UTR-Whereas its promoter plays a critical role in expression of the nkx6.1 gene, Fig. 3 demonstrates that sequences within the 5Ј-UTR are at least as important. In ␤TC3 cells, the deletion of the 5Ј-UTR causes a nearly complete loss of luciferase expression from the nkx6.1 promoter constructs. In NIH3T3 cells, however, removal of the 5Ј-UTR increases luciferase activity.
When moved from its normal position downstream of the transcription start site, the function of the 5Ј-UTR changes. As shown in Fig. 6, when positioned downstream of the nkx6.1 or herpes simplex virus thymidine kinase (TK) minimal promoter, the 5Ј-UTR enhances the expression of luciferase in ␣and ␤-cell lines but not in the NIH3T3 cell line. In contrast, the 5Ј-UTR produces no activity in ␣and ␤-cells and significant repression in NIH3T3 cells when placed upstream of the nkx6.1 or TK promoters.
These results demonstrate that the 5Ј-UTR can function as a position-independent repressor in non-islet cells, and as an activator in islet cells when located in its normal position downstream of the promoter. Cell type-specific function ap- 5. PDX-1 and Nkx2.2 binding to the nkx6.1 promoter. A, EMSA using in vitro translated Nkx2.2 and PDX-1 is shown. 32 P-Labeled oligonucleotides encoding the B1 enhancer element (sequences are shown in (Fig. 4B)) were incubated with 1 l of each in vitro translated protein for 15 min at room temperature and then subjected to electrophoresis on a 5% polyacrylamide gel. Unlabeled competitor oligonucleotides (sequences are shown in (Fig. 4B)) were added at 20-and 200-fold molar excess. B, a reporter plasmid containing five tandem copies of the B1 enhancer element upstream of the prolactin minimal promoter driving luciferase and pBAT12 expression plasmids expressing the Nkx2.2 and PDX-1 cDNAs under the control of the CMV promoter were co-transfected into NIH3T3 cells. Relative luciferase activities are calculated with the activity of cells transfected with the pBAT12 expression vector without cDNA insert set at 1. All data are shown as mean Ϯ S.E. C, an EMSA using in vitro translated Nkx2.2 is shown. 32 P-Labeled oligonucleotides encoding the B1 enhancer element (lanes 1-3) or the related sequence at Ϫ460 (lanes 4 and 5) in the nkx6.1 promoter (see "Experimental Procedures" for sequence) were incubated with 1 l of the in vitro translated protein for 15 min at room temperature and then subjected to electrophoresis on a 5% polyacrylamide gel. The control lanes (2 and 4) contain in vitro translated luciferase protein. N.S. indicates a nonspecific protein-DNA complex produced by proteins present in the rabbit reticulocyte lysate mix. pears to be dependent on an intact 5Ј-UTR, since the 5Ј-UTR loses all specificity when cut in half (Fig. 6).
Identification of an IRES in the 5Ј-UTR-Several features of the 5Ј-UTR suggest that it may provide a poor template for protein synthesis after cap-dependent scanning: it is long (973 bp), G/C-rich (67.3%), and contains out of frame ATG codons with reasonable Kozac consensus sequences: These limitations could be overcome by an IRES. In addition, the presence of an IRES that functions in a cell type-specific manner could explain the functional characteristics of the 5Ј-UTR.
To test for this possibility, a dicistronic gene (pFoxRSV-CAT-Luc) was constructed placing the CAT gene and the luciferase gene in series under the control of RSV promoter. The 5Ј-UTR of the nkx6.1 was inserted between the two cistrons of this plasmid (pFOX-CAT-5Ј-UTR-Luc). In addition, the 5Ј-UTR was inserted in an inverted orientation (pFOX-CAT-5Ј-UTR-R-Luc) as a nonspecific control. These plasmids were transfected into two ␤-cell lines, ␤TC3 and INS1, the ␣-cell line ␣TC1.6, and two non-islet cell lines NIH3T3 and COS7. CAT and luciferase activity were assayed 48 h after transfection. The ratio of luciferase activity to CAT activity provides a gauge of IRES function. As shown in Fig. 7, the 5Ј-UTR can function as an IRES when placed in its native orientation, and this activity is consistently greater in islet cell lines. Furthermore, ␤-cell lines showed modestly higher activity than ␣-cell lines. These data demonstrate that the IRES function of the 5Ј-UTR contributes to the tissue-specific expression of nkx6.1.

DISCUSSION
In the present study, we have characterized the nkx6.1 promoter and mapped a region involved in its cell type-specific expression. In addition, we found that the expression of Nkx6.1 also is controlled at the post-transcriptional level, and an IRES in the 5Ј-UTR plays an important role in directing its expression to islet cells.
Like many transcription factor genes, the 5Ј-flanking region of the mouse nkx6.1 gene lacks a classic TATA box. The TATA box is typically located 30 bp upstream of the transcription initiation site and helps specify the transcription initiation site by directing the binding of TFIID. Characteristic of genes that lack TATA boxes, the nkx6.1 gene has multiple transcription initiation sites as mapped by 5Ј-RACE and RNase protection assay. Also characteristic of TATA-less genes, the transcription initiation sites lie just downstream of two CCAAT boxes, at least one of which is functional in both islet and non-islet cells.
Complex interactions among a number of transcription factors control the temporal expression of genes during the development of the pancreas. Tight control over the temporal and spatial expression of these factors is essential for proper development of the endocrine cells. Nkx6.1 is expressed in at least three different cell types during mouse pancreatic development as follows: initial broad expression in the epithelial cells that compose the dorsal and ventral buds, restricted expression after embryonic day 13 in islet cell precursors, and finally in mature ␤-cells. Studies of mice that lack an intact nkx2.2 gene demonstrate that Nkx2.2 is required for Nkx6.1 expression in the pancreas after E13; specific inactivation of the pdx-1 gene in insulin-expressing cells demonstrates that PDX-1 is required for maintaining Nkx6.1 expression in differentiated ␤-cells (9,15). It can be concluded from these prior studies that Nkx6.1 expression is regulated directly or indirectly by these two factors. The promoter studies reported here support the conclusion that both factors drive Nkx6.1 expression directly, by binding to the nkx6.1 gene promoter.
Although there are several binding sites for Nkx2.2 and PDX-1 in the proximal Nkx6.1 promoter, only the sites located at Ϫ800 are required for expression specifically in the ␤-cell line. The more proximal sites may also contribute, but removal of more proximal sequences affects expression in NIH3T3 cells as well. The presence of binding sites for both factors in the ␤-cell-specific enhancer located at Ϫ800 bp is intriguing given the essential role of both factors in Nkx6.1 expression. Despite the juxtaposed binding sites, however, Nkx2.2 does not activate the ␤-cell-specific enhancer even in the presence of co-expressed PDX-1.
We recently found that Nkx2.2 by itself cannot activate transcription even from a construct with 7 tandem repeats of an ideal Nkx2.2-binding site. The NK2 domain just downstream of the homeodomain in Nkx2.2 inhibits the activation domain, and some modification of the NK2-specific domain may be required to allow Nkx2.2 to activate transcription (30). When Nkx2.2 is overexpressed as in the co-transfection experiments reported here, the non-islet cells may lack specific modifiers of Dicistronic reporter plasmid containing the 5Ј-UTR of nkx6.1 bidirectionally inserted between the CAT and luciferase genes were transfected into the cell lines shown. CAT activity and luciferase activity were measured, and the luciferase activity divided by CAT activity was used as an index of IRES activity. All data are shown as mean Ϯ S.E. The ratio of CAT/luciferase activity for the control dicistronic constructs without the 5Ј-UTR is set at 1.0. the NK2 domain, or the capacity of the cells to modify the NK2 domain may be exceeded. Hence, we cannot rule out the possibility that Nkx2.2 cooperates with PDX1 in regulating the Nkx6.1 promoter in the normal cellular context.
In addition, a different array of factors may control Nkx6.1 expression at different points in development. The cells used for these experiments are probably more representative of mature ␤-cells than undifferentiated pancreatic epithelial cells or islet cell progenitors. During the differentiation of islet cells in the fetal mouse pancreas, the islet cell progenitors transiently express the basic helix-loop-helix transcription factor neuroge-nin3 prior to further maturation and expression of PDX-1 (34). Some of these early neurogenin3-expressing cells co-express Nkx6.1, but not PDX-1, demonstrating that cells at this stage in differentiation do not require PDX-1 for Nkx6.1 expression (34). Other homeodomain proteins capable of binding to the TAAT sites or other factors such as HNF6 (35) may fill the PDX-1 role at this stage.
In addition to controls at the level of the promoter, expression from the nkx6.1 gene is regulated by its long, complex 5Ј-UTR, which has similarities to the 5Ј-UTRs found in some Drosophila homeobox genes. For example, ubx contains a 968-bp 5Ј-UTR and 2 upstream ATGs, and antp contains a 1735-bp 5Ј-UTR and 15 upstream ATGs. Both of these 5Ј-UTRs contain IRESs that promote developmentally regulated translation. Similarly, the nkx6.1 5Ј-UTR functions as an IRES; while it can function in all the cells types tested, its activity is significantly higher in islet cell lines. These results demonstrate that the IRES activity of the nkx6.1 5Ј-UTR shows cell type specificity. In addition, the 5Ј-UTR also inhibits transcription in the NIH3T3 cells, but not in islet cells, in a positionindependent fashion. Taken together, the 5Ј-UTR contributes significantly to the cell type-specific expression of Nkx6.1.
The identified functions of the 5Ј-UTR, however, cannot completely explain the differences in expression of Nkx6.1 between ␣and ␤-cell lines. No Nkx6.1 protein can be detected in ␣-cell lines, despite the presence of nkx6.1 mRNA, suggesting that post-transcriptional regulation of Nkx6.1 expression contributes to its restriction from ␣-cells. Although the IRES in the 5Ј-UTR provides a mechanism for cell type-specific translation of nkx6.1 mRNA, the activity of the IRES is similar in ␣TC1 cells and ␤-TC3 cells. It is possible that in the context of the intact gene, the IRES may function in a more tightly restricted fashion, or other posttranscriptional mechanisms must play additional roles in the cell type-specific expression of Nkx6.1.
It should be noted that further controls provide additional limits on the function of Nkx6.1 once the protein is expressed. When it binds to target genes, Nkx6.1 is a potent transcriptional repressor; but a sequence in the carboxyl-terminal end of the molecule prevents DNA binding by the homeodomain (14). Presumably only when this binding inhibition is relieved by interactions provided in the appropriate cellular environment can Nkx6.1 then turn off target genes. Together, several layers of regulation ensure that gene targeting by nkx6.1 is tightly restricted temporally and spatially.