Long-Range Enhancers Are Required to Maintain Expression of the Autoantigen Islet-Specific Glucose-6-Phosphatase Catalytic Subunit–Related Protein in Adult Mouse Islets In Vivo

The construction of IGRP-LacZ fusion genes containing mouse promoter sequence between −306, −911, and −3911 and +3, relative to the transcription start site (TSS), is described in the online appendix (available at http://dx.doi.org/10.2337/db07-0092), as is the use of bacterial recombination to introduce a LacZ-containing cassette into a bacterial artificial chromosome (BAC) clone containing the IGRP gene. The isolation and ligation of the putative IGRP enhancers A, B, C, E, and F (Figs. 1 and 2) to a minimal IGRP-luciferase fusion gene is also described in the online appendix.

The animal housing and surgical facilities used for the mice in these studies met American Association for the Accreditation of Laboratory Animal Care standards. All animal protocols were approved by the Vanderbilt University Medical Center Animal Care Committee. Food (Purina Mouse Chow 5001; Ralston-Purina, St. Louis, MO) and water were provided ad libitum.

Transgenic mice were generated by the joint Vanderbilt Cancer Center/Diabetes Research and Training Center Transgenic Animal/Embryonic Stem Cell Core Facility. The generation of transgenic mice expressing the −306 IGRP-LacZ fusion gene has been described (28). The −911 IGRP-LacZ and −3911 IGRP-LacZ fusion genes and the IGRP-BAC transgene were isolated as described in the online appendix. Transgenic mice were produced by microinjection of 3–10 pl of a 3 ng/μl solution of the transgenes into the pronuclei of one-cell embryos derived from B6D2F1 females. Embryos were then transferred into the oviducts of pseudopregnant ICR females. Transgenic founders were bred with B6D2F1 mice (The Jackson Laboratory, Bar Harbor, ME) and offspring genotyped using real-time PCR in conjunction with the following primers: 5′-ACGCTGATTGAAGCAGAAGCC-3′ and 5′-ATCGGTCAGACGATTCATTGGC-3′ and the Bio-Rad iQ SYBR Green Supermix (Hercules, CA). The detection of LacZ expression by 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-gal) staining of mouse tissues was performed at 4°C as previously described (28).

Mouse pancreatic islet β-cell–derived βTC-3 cells were cultured and cotransfected with 0.5 μg of an expression vector encoding Renilla luciferase (Promega) and 2 μg of the indicated firefly luciferase plasmids using the lipofectamine reagent (Gibco BRL), as previously described (22). Following transfection, βTC-3 cells were harvested by trypsin digestion and then resuspended in passive lysis buffer (Promega). After two cycles of freeze/thawing, firefly and Renilla luciferase activity were then assayed using the Dual Luciferase Assay Kit (Promega). To correct for variations in transfection efficiency, results are expressed as the ratio of firefly to Renilla luciferase activity. Fusion gene expression was assessed in three separate experiments each using an independent preparation of each plasmid, transfected in triplicate.

Previous studies (28) with transgenic mice in which LacZ gene expression was under the control of the −306 to +3 IGRP promoter region demonstrated that this region was sufficient to drive LacZ expression in newborn mouse islets. As seen with the endogenous IGRP gene (2,9), transgene expression initiated at the expected time in development, around embryonic day 14, and was detected predominantly in β-cells (28). However, unlike the endogenous IGRP gene (3), LacZ gene expression was also detected in the cerebellum (28). In addition, in adult mice, expression of the endogenous IGRP gene is maintained in all β-cells (4,9), whereas LacZ expression was almost completely absent in the transgenic adult mouse islets (28). Figure 3 shows a more detailed analysis of the time course of this loss of LacZ expression in these animals. Clear pancreatic LacZ expression was readily detected in newborn mice, but only mosaic islet expression was detected by 3 weeks (Fig. 3). This mosaic expression was maintained at 6 weeks but was very rarely detected at 6 months (Fig. 3). In contrast, LacZ expression in the cerebellum was maintained in adult animals (Fig. 3). A second independent founder line of mice also demonstrated a loss of islet transgene expression in adult animals (data not shown).

Transfection experiments in βTC-3 cells (28) have shown that the −911 to +3 IGRP promoter region confers ∼25% higher fusion gene expression than the −306 to + 3 region. In addition, the promoter sequence between −500 and −306 is well conserved between mouse and human IGRP (data not shown). We therefore generated transgenic mice in which LacZ reporter gene expression was under the control of the −911 to +3 IGRP promoter region. As with the −306/+3 promoter region, the −911/+3 promoter region drove LacZ expression in newborn mouse islets (Fig. 4) . Moreover, as with the endogenous IGRP gene, LacZ expression was not detected in the brain (Fig. 4). Nevertheless, transgene expression was again markedly reduced in adult mouse islets (Fig. 4). Figure 4 shows an analysis of the time course of this loss of LacZ expression. Clear pancreatic LacZ expression was detected in newborn mice, but only mosaic islet expression was detected by 3 weeks (Fig. 4). This mosaic expression was maintained at 6 weeks but was not apparent at 6 months (Fig. 4). Although a faint blue color was detected following whole-mount staining of pancreata isolated from 6-month-old mice, this appeared to be nonspecific since pancreatic sections were negative (Fig. 4). A second independent founder line of mice similarly showed a loss of transgene expression in adult animals (data not shown).

These data suggest that the proximal IGRP promoter between −911 and +3 is not sufficient for the maintenance of IGRP gene expression in adult mice and that cis-acting elements elsewhere in the IGRP gene are required. Only limited genomic sequence was available when the −306 and −911 IGRP-LacZ transgenic mice were generated, but BAC clones containing the entire mouse (AL929170) and human (AC069137) IGRP genes were subsequently described. Sequence alignment of the mouse and human IGRP promoter regions between −5000 and +1 using the IntelliGenetics IFIND program (Mountain View, CA) identified a conserved region between −1800 and −2500 (data not shown). We therefore generated transgenic mice in which LacZ gene expression was under the control of the −3911 to +3 IGRP promoter region that encompassed the conserved −1800/−2500 region. Five independent founder lines of mice were analyzed. As with the −306/+3 and −911/+3 promoter regions, the −3911/+3 promoter drove clear LacZ expression in newborn mouse pancreas (Fig. 5A). Figure 5 shows an analysis of LacZ expression in the −3911 IGRP-LacZ transgenic mice over time. In contrast to the −306 and −911 IGRP-LacZ transgenic mice, transgene expression was clearly detected in adult mouse islets (Fig. 5A and B). However, with the exception of line 38, only mosaic islet expression was detected at 3 weeks, and the proportion of cells expressing the transgene decreased over time (Fig. 5B). Figure 5B shows two panels of pancreas sections isolated from the same line 38 mouse. These demonstrate that line 38 islets were detected in which transgene expression remained on in most cells (left panel); however, even with this founder line, islets were also detected in which a mosaic pattern of expression was apparent (right panel). These results suggest that the presence of the conserved −1800/−2500 region and the proximal promoter are not sufficient for the maintenance of transgene expression in adult mouse islets. Interestingly, this demonstration of mosaic transgene expression in multiple animals reinforces the concept that the β-cell population within adult islets is not homogeneous (29).

In some lines of the −3911 IGRP-LacZ mice, transgene expression was also detected in the brain, mainly in the cerebellum but also in other areas (Fig. 5A). The loss of brain expression in the mice expressing the −911/+3 IGRP-LacZ transgene initially suggested that elements required for restricting appropriate IGRP tissue-specific expression were located between −911 and −306. The reemergence of brain transgene expression in some lines of −3911/+3 mice indicates that this is not the case. Instead, brain expression may simply reflect the effect of random integration. The tendency to detect expression in the cerebellum may be explained by the high expression of Pax-6 in this tissue (30), a factor that is critical for IGRP promoter activity (27).

To identify candidate regions of the IGRP gene that might contain elements required for the maintenance of IGRP gene expression, another sequence alignment was performed but this time using the VISTA program (31,32) in conjunction with mouse and human IGRP gene sequence 10 kbp 5′ and 3′ of the IGRP TSS (Fig. 6). This analysis identified two conserved regions in the mouse IGRP promoter, located between approximately −1800 and −2500 and −3700 and −4700, and a conserved nonpromoter region, located 3′ of exon 3 between approximately +3050 and +3830 (Fig. 6). These regions are found in the same relative location in the human IGRP gene except for the conserved −3700/−4700 region, which is located ∼6,000 bp 5′ of the TSS in the human IGRP gene. This explains why an alignment of the proximal mouse and human 5,000-bp promoter regions using the IntelliGenetics IFIND program missed this region. We hypothesized that these regions might conceivably represent transcriptional enhancers required for the maintenance of IGRP gene expression in adult mice. To address this hypothesis, these regions, designated enhancers A, B, and C (Fig. 6), were isolated and ligated 5′ of an IGRP-luciferase fusion gene containing the proximal human IGRP promoter sequence between −324 and +3 (Fig. 1). Luciferase expression directed by these fusion genes was then analyzed by transient transfection of βTC-3 cells. Figure 1 shows that all three regions enhanced reporter gene expression beyond that driven by the −324/+3 IGRP-luciferase fusion gene alone. In addition, this effect was independent of orientation consistent with the definition of an enhancer (33). However, when the +3050 to +3830 region was ligated 3′ of the −324/+3 IGRP-luciferase fusion gene, it did not enhance reporter gene expression, which is inconsistent with the strict definition of an enhancer whose actions are orientation and location independent (33).

Based on these results, further transgenic mice were generated to assess whether the combination of the proximal IGRP promoter with enhancers A, B, and C were sufficient to maintain IGRP transgene expression in adult mice. Because the results of the fusion gene analyses suggested that spacing affected enhancer C function (Fig. 1), a transgene was generated by using bacterial recombination to introduce a cassette into exon 5 of the IGRP gene such that the spacing between the IGRP TSS and enhancer C was the same as that found in the native IGRP gene (Fig. 7A). The IGRP gene was located within a BAC clone, and the cassette comprised an enhanced green fluorescent protein (EGFP) cDNA, an internal ribosome entry site (IRES), a cDNA encoding a β-galactosidase–neomycin resistance fusion protein, an FLP recombination target (FRT) site, a tetracycline-resistance cassette, and a second FRT site (Fig. 7A). The EGFP cDNA was inserted in frame with the coding sequence of IGRP exon 5.

After recombination, the tetracycline-resistance cassette was removed using FLP recombinase. Bacterial recombination was then used to shuttle the region of the IGRP gene from −5000 to +10051 from the modified BAC clone to the plasmid pGEM7 (Fig. 7A). This region contains enhancers A, B, and C, in addition to the proximal promoter, all exons and introns, and the targeting cassette (Fig. 7A). Transgenic mice containing this region, minus the pGEM7 plasmid backbone, were then generated.

Three independent founder lines of mice were analyzed. EGFP fluorescence was not detected, suggesting that the EGFP domain within the IGRP-EGFP fusion protein may not fold correctly. However, as with the −306, −911, and −3911 IGRP-LacZ fusion genes, in all three mouse lines the IGRP-BAC transgene drove clear LacZ expression in newborn pancreata (Fig. 7B). However, variable results were seen in adult mice with either no (line 12), mosaic (line 15), or strong (line 16) LacZ expression detected in islets by 3 weeks of age (Fig. 7B and C). These variable results probably reflect the consequences of random integration, but they suggest that the presence of the proximal promoter in combination with enhancers A, B, and C are still not sufficient for the maintenance of transgene expression in adult mouse islets.

Based on these results, we expanded the search for additional putative enhancers that might contribute to IGRP gene expression by using the University of California Santa Cruz Genome Browser (available at http://genome.ucsc.edu/index.html) to locate conserved regions present in the entire ∼13,900- and ∼12,100-bp intervals between the neighboring Spbc25 (5′) and Abcb11 (3′) genes, respectively. This analysis identified one additional region, designated region D (data not shown), of substantial sequence conservation between mice and rats 5′ of the IGRP gene; however, this region is not conserved in humans, and the IGRP gene is not expressed in rats (3). In contrast, four blocks of conserved sequence were identified 3′ of the IGRP gene (data not shown). Two of these conserved regions are located close to the Abcb11 gene; therefore, we hypothesized that they are likely to be involved in the regulation of Abcb11 gene expression. In contrast, the other two conserved regions are located closer to the IGRP gene, ∼900 and ∼2,500 bp 3′ of the last exon, exon 5, between approximately +8771 and +9372 and +10622 and +11022, respectively, relative to the TSS (Fig. 6). We hypothesized that these regions, designated E and F, might represent transcriptional enhancers required for the maintenance of IGRP gene expression.

To address this hypothesis, these regions were isolated and ligated 5′ of an IGRP-luciferase fusion gene containing the proximal human IGRP promoter region located between −324 and +3 (Fig. 2). Luciferase expression directed by these fusion genes was then analyzed by transient transfection of βTC-3 cells. Figure 2 shows that regions E and F enhanced reporter gene expression beyond that driven by the −324/+3 IGRP-luciferase fusion gene alone. However, this effect was dependent on orientation and location (Fig. 2). Thus, region E only enhanced luciferase expression in one orientation, and when region F was ligated 3′ of the −324/+3 IGRP-luciferase fusion gene it did not enhance reporter gene expression (Fig. 2). Thus, neither enhancer E nor F meets the strict definition of an enhancer whose actions are orientation and location independent (33). Intronic enhancers have previously been identified in other genes whose expressions are enriched in islets, including glucagon (34), islet amyloid polypeptide (35), hepatocyte nuclear factor-6 (36), and Pax-6 (37). However, to the best of our knowledge this is the first report of an enhancer located 3′ of an islet-enriched gene.

Sequence comparisons (Fig. 6) and transfection studies (Figs. 1 and 2) have identified five putative enhancers in the IGRP gene. Transgenic mice containing IGRP promoter sequence from −306 to +3 (Fig. 3) or −911 to +3 (Fig. 4), which lack all of these enhancers, and mice containing IGRP promoter sequence from −3911 to +3, which encompasses only enhancer A (Fig. 5), all demonstrated strong LacZ expression in newborn mouse islets but mosaic expression in adult mouse islets. Interestingly, this decrease in transgene expression between newborn and adult islets coincides with the major changes in pancreatic gene expression that occur at this time (38). Strong LacZ expression was also observed in islets in newborn IGRP-BAC mice, in which the transgene contains enhancers A, B, C, and E, but again expression was not consistently maintained in adult islets (Fig. 7). Region F is absent from the IGRP-BAC transgene, and while region E is present its spacing relative to the IGRP TSS is altered (Fig. 7A). We hypothesize that this may explain why expression of the IGRP-BAC transgene is not uniformly maintained in adult mouse islets in vivo despite the presence of enhancers A, B, and C. However, we cannot rule out the existence of additional enhancers located 5′ or 3′ beyond the Spbc25 and Abcb11 genes that flank the IGRP gene.

Ideally, we would like to generate transgenic mice containing all five enhancers and then determine whether this is now sufficient for the maintenance of transgene expression in adult islets. However, transfection studies (Figs. 1 and 2) suggest that this will require the insertion of a reporter gene into the IGRP locus in a manner such that the distance between enhancers C, E, and F and the TSS is precisely maintained. Unfortunately, the IGRP gene only spans ∼8,000 bp (21), and VISTA analyses (Fig. 6) reveal multiple other conserved areas within the IGRP gene in addition to these five enhancers. Therefore, although the technology is available to insert an IRES-LacZ cassette within the IGRP gene without altering the distance between enhancers C, E, and F and the TSS (Fig. 7), it does not appear that this can be done without disrupting conserved sequences. Based on this limitation, we can only conclude that because enhancers A, B, and C are not sufficient for the uniform maintenance of IGRP gene expression in adult mouse islets, long-range enhancers, presumably including enhancers E and F, must also be required. Far-upstream enhancers, located >100 kbp from the TSS, have been shown to be important for the expression of several genes, including Sox9 and Bmp2 (39,40). The importance of such long-range enhancers has been emphasized from the results of recent genome-wide association studies that identified genetic variants that increase the susceptibility to developing type 2 diabetes (41–44). One such variant is located in a region that contains no known genes or microRNAs. This region is located 125 kbp upstream of the CDKN2A gene and may represent an enhancer that modulates CDKN2A expression (41–44).

The observation that the IGRP promoter sequence from −306 to +3 fails to drive transgene expression in all adult islet β-cells is surprising in that this region of the IGRP promoter binds neuroD/BETA2 (23), Pax-6 (27), pancreatic-duodenal homeobox factor-1 (27), and MafA (C.C.M., R.M.O., unpublished observations), the same factors that bind the −668 to +1 region of rat insulin II promoter (45). As with the −306/+3 IGRP promoter (Fig. 3), this region of the rat insulin II promoter confers islet-enriched transgene expression (46,47), but in contrast with the IGRP promoter, it confers sustained transgene expression in adult mice (46). The molecular explanation for this difference is unclear, especially since insulin and IGRP fusion genes that incorporate these promoter regions confer similar levels of reporter gene expression in islet-derived cell lines (22).

Research in the laboratory of R.O. was supported by National Institutes of Health (NIH) Grant DK061645. Research in the laboratory of J.C.H. was supported by the American Diabetes Association (9901-116) and the Barbara Davis Center Diabetes and Endocrinology Research Center (P30 DK57516). The Vanderbilt Diabetes Research and Training Center is funded by NIH Grant P60 DK20593-24, and the Vanderbilt-Ingram Cancer Center is funded by NIH Grant P30 CA68485-06. C.C.M. was supported by the Vanderbilt Viruses, Nucleic Acids, and Cancer Training Program (5T32 CA09385-17).

We thank Peer B. Jacobson, David R. Powell, Hong Chen, Kenneth A. Platt, Clayton Matthews, Barbara Bergman, Karl Hilliker, and Karen L. Rufus for assistance with this project and Maureen Gannon for helpful comments on the manuscript. We thank Doug Mortlock, Ron Chandler, and Kelly Chandler for extensive help with the bacterial recombination method. We also thank Cathleen Pettepher, Managing Director of the Vanderbilt Cancer Center/Diabetes Research and Training Center Transgenic Animal/Embryonic Stem Cell Core Facility, for generating the transgenic mice and Shimon Efrat for providing the βTC-3 cell line.