|Home | About | Journals | Submit | Contact Us | Français|
Improper glucose-stimulated insulin secretion from pancreatic β cells is a major factor in the onset of type 2 diabetes. We now report that HMGN3, a nuclear protein that binds to nucleosomes and affects chromatin function, is highly expressed in β cells and that in mice, loss of HMGN3 impairs glucose-stimulated insulin secretion and leads to a diabetic phenotype. In pancreatic β cells, loss of HMGN3 affects the transcription of several genes involved in glucose-stimulated insulin secretion, including that of the Glut2 glucose transporter. Chromatin immunoprecipitation reveals that HMGN3 and the transcription factor PDX1 mutually reinforce their specific binding to the chromatin in the promoter of the Glut2 gene, thereby regulating GLUT2 protein levels in pancreatic islets and in β cells. Our results identify a new regulator of glucose homeostasis and demonstrate a link between the activity of a nucleosome binding structural protein and the regulation of insulin secretion.
Proper secretion of insulin from the β cells located in pancreatic islets is a key element in maintaining glucose homeostasis and preventing the onset of diabetes mellitus, a severe and prevalent metabolic disorder (17, 28, 29). Despite intensive studies, the full spectrum of genetic factors leading to diabetes is still not fully understood (43). β cells secrete insulin in response to elevated levels of glucose which can occur after eating, a response named glucose-stimulated insulin secretion (GSIS). The entry of glucose into β cells is mediated by type 2 glucose transporters (GLUT2) (15, 23). The internalized glucose is phosphorylated by glucokinase, leading to an increase in ATP levels, closure of the K+ channels, plasma membrane depolarization, and activation of voltage-gated Ca2+ channels, thereby triggering exocytosis of insulin-storing granules from these cells (28, 32).
Inadequate progression of the GSIS pathway, due to impaired signaling or improper protein functionality, may lead to diabetes mellitus (12, 16, 22, 41, 42). Accordingly, transcription factors such as HNF1α, HNF4α, PDX1, FoxA2, and NeuroD, which regulate the expression of the protein components involved in maintaining specific β-cell functions, have been shown to play an important role in insulin secretion and in the pathogenesis of diabetes (9). The transcription network of β cells functions in the context of chromatin and is affected by changes in chromatin structure (11); therefore, nucleosome binding proteins that affect the structure and activity of the chromatin fiber may also play a role in β-cell function and insulin secretion. We now report that the nucleosomal binding protein HMGN3 is highly expressed in pancreatic islets and that loss of this protein impairs insulin secretion, thereby leading to a diabetic phenotype.
HMGN (high-mobility group N) proteins are a family of structural proteins that bind specifically to the 147-bp nucleosome core particle, the building block of the chromatin fiber, without any obvious specificity for the underlying DNA sequence (6). The binding of these proteins to nucleosomes induces structural changes in chromatin, alters the levels of posttranslational modifications in the tail of the nucleosomal histones, and affects transcription, replication, and DNA repair of chromatin templates (6, 36). In living cells, the interaction of HMGN with chromatin is highly dynamic: the proteins move continuously among nucleosomes, and their residence time at any specific site is short. HMGN proteins compete with other architectural proteins for chromatin binding sites and therefore can be viewed as functioning within a dynamic network of proteins that continuously move among nucleosomes, constantly reconfiguring their chromatin binding sites (8). The three members of this family, HMGN1, HMGN2, and HMGN3, have a highly conserved domain structure and very similar biochemical properties (6); however, their expression levels vary among tissues (26, 40).
Here we demonstrate that HMGN3 is highly expressed in the insulin-producing cells of the pancreatic islets in both mice and humans. To test whether the protein affects the function of pancreatic β cells, we generated Hmgn3−/− mice. These mice are viable and appear normal; however, compared to their Hmgn3+/+ littermates, their blood glucose levels are elevated, their serum insulin levels are reduced, and their glucose tolerance is impaired. To test the molecular mechanisms leading to this diabetic phenotype, we analyzed the role of HMGN3 in the function of MIN6 cells, a mouse insulinoma cell line that has been extensively used for functional studies of pancreatic β cells (27). We found that HMGN3 modulates the expression of several genes involved in insulin secretion, including that for the glucose transporter GLUT2. In both MIN6 cells and Hmgn3−/− mice, loss of HMGN3 protein reduces the levels of GLUT2. HMGN3, but not the HMGN1 or HMGN2 variant, binds specifically to chromatin in the promoter region of the Glut2 gene and enhances its expression by facilitating the recruitment of PDX1 and additional transcription factors to the Glut2 promoter. Chromatin immunoprecipitation (ChIP) experiments reveal that HMGN3 enhances the acetylation of H3 at the Glut2 promoter and suggest that HMGN3 and PDX1 mutually reinforce their interaction with the chromatin at the Glut2 promoter but not throughout the chromatin or in the nucleoplasm. Given that the human and mouse HMGN3 amino acid sequences are identical and that this protein is also highly expressed in the β cells of humans, our findings may have relevance to the understanding of the molecular mechanisms regulating the transcriptional network of human pancreatic β cells. Our studies provide insight into the mechanism of action of HMGN proteins and identify a new factor which, either by itself or in conjunction with additional genetic defects, may play a role in the pathogenesis of type 2 diabetes.
The following antibodies were used: anti-PDX1 (Upstate and Chemicon), anti-HNF1α (Santa Cruz), anti-HNF4α (Santa Cruz), anti-FoxA2 (Santa Cruz), anti-Glut2 (Millipore), anti-actin (Sigma), anti-rabbit insulin (Santa Cruz), anti-mouse insulin (Zymed), and Alexa Fluor 488 or Alexa Fluor 568 goat anti-rabbit immunoglobulin G (IgG; Molecular Probes). Affinity-purified antibodies to HMGN variants and to histone H3 elicited in either rabbits or goats were prepared as described previously (7). The antibody to HMGN3 was previously characterized (40). Prepared solutions of insulin (10 mg/ml) and glucose (45%) were purchased from Sigma.
The targeting vector to conditionally remove the gene for HMGN3 was constructed by a recombinogenic cloning strategy (25) with a murine bacterial artificial chromosome clone, RP24-388H18. The vector was constructed to remove exons III and IV, which code for the nucleosomal binding domain of HMGN3 (Fig. (Fig.1D).1D). A 21-kb fragment containing the gene for HMGN3 (except exon I) was retrieved from the bacterial artificial chromosome clone and inserted into the targeting vector PL253 by recombination in bacterial strain DY380. The neo gene with the phosphoglycerate kinase 1 promoter (pGKneo) was employed as a positive selectable marker, and the pGK-thymidine kinase cassette was used as a negative selectable marker (34). The loxP/Frt-flanked positive selectable marker and the loxP site for conditional deletion of the gene for HMGN3 were inserted as described in Fig. Fig.1.1. Electroporation and selection were performed by using the v6.4 embryonic stem (ES) cell line as described elsewhere (34). DNAs derived by G418/1,2′-deoxy-2′-fluoro-1-β-d-arabinofuranosly-5-iodo-uracil-resistant ES cell clones were screened by diagnostic BclI or NheI restriction enzyme digestion by using the 5′ and 3′ probes external to the targeting vector sequence. Two independent ES cell clones targeted to the gene for HMGN3 and injected into C57BL/6 blastocysts generated chimeras that transmitted the mutated allele to the progeny (5). The neomycin resistance (neo) cassette was removed by crossing with FLP mice, and the genomic fragment containing exons III and IV of the Hmgn3 gene was removed by crossing with EIIA-Cre mice. The mice containing the targeted allele were backcrossed into the C57BL/6 background for at least five generations. Mice were bred in a specific-pathogen-free facility with food and water ad libitum.
The immunofluorescence assay was performed as previously described (13). The primary antibodies used were rabbit anti-HMGN3 (prepared in our laboratory), rabbit anti-insulin (0.4 μg/ml; Santa Cruz Biotechnology), and mouse anti-insulin (0.2 μg/ml; Zymed) antibodies. Alexa Fluor 488 or Alexa Fluor 568 goat anti-rabbit IgG (Molecular Probe) secondary antibodies were used at 8 μg/ml.
Blood glucose levels were measured with an Ascensia CONTOUR blood glucose meter (Bayer). Blood was collected from the tail vein. Serum insulin levels were measured with a rat/mouse insulin enzyme-linked immunosorbent assay kit (Millipore).
Mice were fasted for 16 h before injection. Glucose (2 g/kg) or insulin (0.75 U/kg) was injected intraperitoneally, and glucose levels were measured at 0, 15, 30, 60, 120, and 180 min for the glucose tolerance test and at 0, 15, 30, 45, and 60 min for the insulin tolerance test.
MIN6 cells (obtained from A. L. Notkins, National Institute of Dental and Craniofacial Research, NIH) were maintained in Dulbecco modified Eagle medium (DMEM) containing 15% fetal bovine serum and 25 mM HEPES. ON-target plus SMART pool siRNAs for HMGN1, HMGN2, and HMGN3 and control siRNA were purchased from Dharmacon. An additional set of siRNAs for HMGN3 was purchased from Qiagen. The siRNA (final concentration, 20 nM) was transfected into MIN6 cells, cultured in DMEM containing 15% (vol/vol) fetal bovine serum, by using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Four days posttransfection, the medium was changed and then the cells were retransfected and cultured for an additional 4 days. Efficiency of HMGN knockdown was monitored by Western blotting (protein level) and quantitative real-time PCR (Q-RT-PCR; RNA level). Cell viability was monitored with a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay kit (Roche) according to the manufacturer's instructions.
MIN6 cells transfected with siRNA were cultured in DMEM containing 15% fetal bovine serum and 2 mM glucose overnight. Cells were washed twice with phosphate-buffered saline and cultured in KRBH (25 mM HEPES [pH 7.4], 125 mM NaCl, 1.3 mM CaCl2, 5 mM NaHCO3, 5.9 mM KCl, 1.2 mM MgCl2, 0.1% [wt/vol] bovine serum albumin) containing 2 mM glucose for 2 h. After 2 h, the cells were washed twice with phosphate-buffered saline and induced to secrete insulin by exposure to various concentrations of either glucose or arginine in KRBH. After 10 min, supernatants were collected. Insulin concentrations were measured by enzyme-linked immunosorbent assay (Millipore).
Total RNA from MIN6 cells or isolated islets was purified with an RNeasy mini plus or an RNeasy micro kit (Qiagen), respectively. cDNA was prepared from purified RNA with an iScript cDNA synthesis kit (Bio-Rad). Q-RT-PCR analysis was performed with the ABI PRISM 7900 system and Power SYBR green PCR master mix (ABI) according to the manufacturer's recommendations. The sequences of the primers used are shown in Table Table11.
The polyclonal antibody to mouse HMGN3 was made in goats (Rockland) and affinity purified on immobilized HMGN3. Cell lysates were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, transferred to polyvinylidene difluoride membrane (Millipore), and then stained with goat anti-HMGN3 antibody. Bound antibody was detected with horseradish peroxidase-conjugated anti-goat IgG antibody (Pierce) and ECL plus reagent (GE Healthcare).
Cells were fixed with 0.6% formaldehyde for 10 min, sonicated to produced ~500-bp DNA fragments, and then subjected to ChIP assay. ChIP assays were performed with a ChIP IT express kit (Active Motif) according to the manufacturer's instructions. ChIP experiments were performed with affinity-purified anti-HMGN3 and anti-PDX1 antibodies (Upstate). Immunoprecipitated DNA was analyzed, and the relative enrichment value was calculated as described previously (24). For coimmunoprecipitation, 0.35 M NaCl extracts from either MIN6 cells or MIN6 cells that were cotransfected with plasmids expressing HMGN3 and PDX1 were prepared. The extracts were dialyzed to 0.15 M NaCl-20 mM Tris (pH 7.5)-0.2% NP-40-1 mM EDTA and treated with antibodies to either HMGN3 or PDX1.
The expression of HMGN3 is tissue specific and differs from that of the closely related but more abundant and ubiquitous HMGN1 and HMGN2 variants (40). In the pancreas, HMGN3 is specifically enriched in the islets (Fig. (Fig.1A)1A) while HMGN2 (Fig. (Fig.1C)1C) and HMGN1 (data not shown) are more broadly expressed throughout the entire pancreas, including the islet cells. In the islets, antibodies to HMGN3 prominently stain the nuclei of cells expressing insulin, an indication that β cells are highly enriched in HMGN3 protein (Fig. (Fig.1A).1A). Western blot analyses verified that the expression of HMGN3 in mouse islets is significantly higher than that in the surrounding exocrine cells (Fig. (Fig.1B),1B), and double-immunofluorescence analysis reveals that while HMGN3 is highly expressed in the islets, the HMGN2 variant is expressed both in the islets and in the exocrine pancreatic cells (Fig. (Fig.1C).1C). High HMGN3 expression in β cells of pancreatic islets is not unique to mice, since the expression of HMGN3 is also highly elevated in the β cells of the human pancreas (data not shown). The high level of HMGN3 in the insulin-producing cells of the pancreatic islets raises the possibility that this nucleosomal binding protein affects β-cell function.
To test whether the HMGN3 protein could affect β-cell function, we generated Hmgn3−/− mice (Fig. 1D to H). To minimize genomic alterations, we excised from the Hmgn3 gene only the region that codes for the nucleosomal binding domain of HMGN3, which is highly conserved among all HMGN variants (35). Genomic analysis with appropriate primers verified complete loss of exons III and IV of Hmgn3 (Fig. (Fig.1E),1E), and Western blot assays of extracts from several tissues obtained from Hmgn3+/+ and Hmgn3−/− mice revealed complete loss of the HMGN3 protein in the knockout mice (Fig. (Fig.1F).1F). Immunostaining of pancreatic sections verified the absence of HMGN3 from the pancreatic islets and from the nuclei of insulin-producing cells (Fig. 1G and H). The Hmgn3−/− mice are viable and fertile.
A comparison of male Hmgn3−/− and Hmgn3+/+ mice, obtained by crossing Hmgn3+/− mice, did not reveal significant differences in the body weights of the mice (Fig. (Fig.2A).2A). However, the glucose levels of Hmgn3−/− mice were 12% higher during feeding, but not at fasting, than those of their Hmgn3+/+ littermates (Fig. 2B and C), suggesting possible effects of HMGN3 on glucose metabolism. The insulin levels in the serum of fasting mice were not significantly affected by HMGN3 (data not shown), while the insulin levels in the serum of feeding Hmgn3−/− mice were approximately 50% of those of their Hmgn3+/+ littermates (Fig. (Fig.2D),2D), providing a possible explanation for the elevated glucose levels in their blood. Further analyses revealed that glucose tolerance (Fig. (Fig.2E),2E), but not insulin sensitivity (Fig. (Fig.2F),2F), was impaired in the Hmgn3−/− mice. Taken together, these findings suggest that the altered glucose homeostasis in feeding Hmgn3−/− mice could be due to the effects of HMGN3 on β-cell function, perhaps due to faulty insulin secretion in response to increased levels of blood glucose during feeding.
To examine the molecular mechanism underlying the phenotype of Hmgn3−/− mice, we analyzed the function of HMGN3 in mouse insulinoma MIN6 cells, which have been extensively used to study the properties of pancreatic β cells, including their pathway of insulin synthesis and secretion (27). Like pancreatic β cells, MIN6 cells also had a high content of HMGN3 and the amount of HMGN3 in these cells is significantly higher than that of somatic cells such as mouse embryonic fibroblasts (Fig. (Fig.3F).3F). Treatment of these cells with siRNA targeted to Hmgn3 (siN3) specifically downregulated Hmgn3 transcripts (Fig. (Fig.3A)3A) and reduced the levels of HMGN3a protein and its splice variant HMGN3b by >80% (Fig. 3A and B) without affecting cell viability (not shown). As a control, cells were treated with siRNAs targeted to Hmgn1 (siN1) or Hmgn2 (siN2). These siRNAs were specific for their target genes and had no effect on Hmgn3 transcript levels (Fig. (Fig.3A3A).
To test whether HMGN3 affects β-cell function, we exposed MIN6 cells to either glucose or arginine; two compounds known to stimulate insulin secretion by different pathways (28). In both pathways, insulin release from β cells is associated with membrane depolarization; however, only the glucose-stimulated pathway involves facilitated import of glucose into the cells (33). We found that siN3-mediated depletion of HMGN3 decreased GSIS from MIN6 cells significantly more than arginine-stimulated insulin secretion (Fig. 3D and E). Reduction in GSIS was also observed with another siN3 sequence (data not shown), an indication that the effects are HMGN3 specific and not due to off-target action of the siRNAs. Furthermore, control experiments with siRNAs to the Hmgn1 and Hmgn2 variants, which efficiently reduced the levels of the corresponding proteins (Fig. 3A and B), also did not affect glucose-mediated insulin secretion (Fig. (Fig.3C).3C). We therefore conclude that HMGN3 specifically enhances GSIS from β cells.
HMGN variants are nucleosome binding proteins known to affect the structure and function of chromatin, including transcription. Hence, reduction in HMGN3 protein may alter the levels of the various components involved in the GSIS pathway. Q-RT-PCR revealed that, indeed, reduction in the levels of HMGN3 protein led to a decrease in the transcript levels of insulin, Pax6, and Glut2 (glucose transporter 2 gene) and to an increase in the levels of Kir6.2, all of which are known to be involved in GSIS (Fig. (Fig.4A).4A). The changes in the transcript levels of several components of the GSIS pathway suggest that the diabetic phenotype of the Hmgn3−/− mutant mice and the impaired insulin secretion from MIN6 cells result from HMGN3-mediated effects on the transcription of specific genes.
The initial phase of GSIS involves the export of insulin-storing granules and is dependent on the amount of stored insulin in the granules rather than on the steady-state levels of insulin transcription (31). We found that the total insulin content of MIN6 cells treated with siN3 was indistinguishable from that of cells treated with control siRNAs (Fig. (Fig.4D).4D). Thus, the decreased level of insulin secretion from the glucose-stimulated MIN6 cells depleted of HMGN3 is not majorly due to the decreased levels of insulin mRNA. Likewise, since the transcription factor Pax6 is not directly involved in the early phases of GSIS, it is unlikely that the HMGN3-mediated decrease in Pax6 transcripts is a major factor in the observed effects on insulin secretion. Kir6.2, a component of the ATP-sensitive potassium channel, plays a role in several physiological processes, and genetic mutations in this protein affect insulin secretion (19). Our analysis of arginine-stimulated insulin secretion in MIN6 cells suggests that, by themselves, the changes in Kir6.2 levels do not have major effects on insulin secretion. GLUT2 plays an important metabolic function; however, its role in GSIS is not fully understood (23, 30). Given the metabolic significance of this gene and our observation that downregulation of HMGN3 affected GSIS more profoundly than arginine-stimulated insulin secretion (Fig. 3D and E), we focused on the role of HMGN3 in regulating Glut2 expression.
Western blot analyses of MIN6 cell extracts verified that the decrease in Glut2 transcripts resulted in a significant reduction in GLUT2 protein levels (Fig. (Fig.4C).4C). Furthermore, the level of Glut2 mRNA was not affected by the siRNA-mediated downregulation of either Hmgn1 or Hmgn2 (Fig. (Fig.4B).4B). This finding supports our observation that siRNA-mediated knockdown of HMGN1 or HMGN2 has no effect on insulin secretion (Fig. (Fig.3C).3C). Taken together, these results support a specific role for the HMGN3 variant in GSIS and in β-cell function.
Since our analysis of MIN6 cells indicates that loss of HMGN3 downregulates GLUT2, we tested whether loss of HMGN3 also affects Glut2 expression in mice. To this end, we determined the levels of Glut2 mRNA and GLUT2 protein in pancreatic islets (Fig. (Fig.4E)4E) isolated from Hmgn3−/− and Hmgn3+/+ littermates. Q-RT-PCR (Fig. (Fig.4F)4F) and Western blot analyses (Fig. (Fig.4G)4G) revealed that loss of HMGN3 leads to an 80% reduction in the amount of Glut2 transcripts and a 50% reduction in the amount of GLUT2 protein. Taken together, our results indicate that HMGN3 enhances GLUT2 expression both in pancreatic islets and in MIN6 cells. We therefore investigated the mechanism whereby HMGN3 affects Glut2 expression.
To gain insights into the mechanism whereby HMGN3 affects Glut2 transcription, we first tested whether HMGN3 interacts with the chromatin in the regulatory region of the gene. ChIP analyses of an 11-kb region (from −9 kb to +2.5 kb relative to start of transcription) with antibodies to HMGN3 revealed a two- to threefold enrichment of HMGN3 over a 3-kb region 5′ to the start of transcription (Fig. (Fig.5A).5A). Consistent with the finding that only the HMGN3 variant affects Glut2 transcription and insulin secretion, this region is not enriched in either HMGN1 or HMGN2. Taken together, the findings indicate that HMGN3 enhances Glut2 transcription by binding to the chromatin regions containing regulatory sequences for this gene. Since HMGN proteins are known to affect chromatin structure, we first tested whether depletion of HMGN3 alters the chromatin structure of the Glut2 promoter. Detailed analysis of the kinetics of DNase I digestion of this region did not reveal any major HMGN3-dependent effects on the chromatin structure of the Glut2 chromatin (Fig. 5B to D).
Although HMGN3 did not affect the chromatin structure of the Glut2 promoter, it had major effects on the binding of several transcription factors known to affect Glut2 transcription, such as HNF1α, FoxA2, HNF4α, and PDX1 (2, 10, 38, 39), to the promoter region of this gene (Fig. (Fig.6A).6A). Depending on the transcription factor tested, siRNA-mediated downregulation of HMGN3, which did not affect the transcript levels of these factors (Fig. (Fig.4A),4A), decreased the relative amounts of the factors bound to the Glut2 promoter by three- to eightfold. These data indicate that HMGN3 facilitates the recruitment of transcription factors to the Glut2 promoter and is the first example of an HMGN variant facilitating the binding of specific transcription factors to a specific chromatin region. PDX1, the factor most prominently affected by the siRNA treatment, plays a major role in the maintenance of adult β cells, where it regulates the transcription of several genes, including Glut2 (9). We therefore focused on the possible mechanism whereby HMGN3 enhances the binding of PDX1 to the Glut2 promoter.
To explore the mechanism by which HMGN3 promotes the interaction of PDX1 with Glut2 chromatin, we first tested whether the binding of HMGN3 precedes that of PDX1. If this were the case, downregulation of HMGN3 would affect the chromatin binding of PDX1 (Fig. (Fig.6A)6A) but the downregulation of PDX1 would not affect the interaction of HMGN3 with the Glut2 promoter. Surprisingly, we found that the efficient siRNA-mediated downregulation of PDX1 (Fig. (Fig.6C)6C) downregulated the binding of HMGN3 to the promoter region but did not alter the background signals at an upstream region (region P17) of Glut2 (Fig. (Fig.6B).6B). Thus, PDX1 and HMGN3 mutually reinforce each other's interactions with the Glut2 promoter.
A possible explanation for the mutual effects of PDX1 and HMGN3 is that these two proteins preassemble into a regulatory macromolecular complex prior to their binding to chromatin. We found that anti-HMGN3 antibody efficiently precipitated HMGN3 from a nuclear extract of MIN6 cells (Fig. (Fig.6D,6D, lower panel); however, the precipitate did not contain PDX1 (Fig. (Fig.6D,6D, upper panel). Likewise, anti-PDX1 antibody efficiently immunoprecipitated PDX1, but these precipitates did not contain HMGN3 (Fig. (Fig.6D,6D, right panels). Thus, the two proteins do not assemble into a macromolecular complex in the nucleoplasm. We next tested whether the two proteins interact globally throughout the entire chromatin. ChIP analyses indicate that antibodies to both HMGN3 and PDX1 efficiently immunoprecipitate chromatin containing the respective proteins (Fig. (Fig.6E);6E); however, the chromatin enriched for HMGN3 is not enriched for PDX1 (Fig. (Fig.6E,6E, upper left panel) and the chromatin enriched in PDX1 does not contain HMGN3 (Fig. (Fig.6E,6E, lower right panel). Thus, HMGN3 and PDX1 do not interact with each other in nuclear extracts and globally do not share a chromatin binding site.
In contrast, sequential ChIP analysis revealed that chromatin that was first precipitated with anti-HMGN3 antibody and then reprecipitated with anti-PDX1 antibody was significantly enriched in DNA of the Glut2 proximal promoter (Fig. (Fig.6F,6F, left). Likewise, the Glut2 proximal promoter DNA was enriched in chromatin that was sequentially immunoprecipitated with anti-PDX1 and anti-HMGN3 antibodies (Fig. (Fig.6F,6F, right). Thus, the two proteins colocalized specifically and mutually reinforce their binding to the chromatin in the promoter region of Glut2. Although this interaction does not lead to detectable changes in the chromatin structure, as assessed by DNase I digestion (Fig. (Fig.5),5), ChIP analysis indicates that the presence of HMGN3 increases the overall acetylation levels of histone H3 in the promoter region of the gene (Fig. (Fig.6G).6G). HMGN proteins were shown to modulate histone modifications in a variant-specific fashion (36); the exact modification(s) affected by HMGN3 remains to be determined.
The regulation of glucose and insulin levels in the blood involves the coordinated action of several organs and numerous regulatory factors (9, 17, 28). Our studies indicate that the chromatin binding protein HMGN3 impairs glucose homeostasis by affecting GSIS from pancreatic β cells, thereby raising the possibility that misexpression of, or specific mutations in, HMGN3 may be a contributing factor in diabetes.
Several types of experiments link the expression of HMGN3 to GSIS. First, the protein is highly expressed in pancreatic islets, where it localizes to the nuclei of insulin-producing β cells. The expression pattern of HMGN3 is clearly different from that of the major members of the HMGN protein family HMGN1 and HMGN2, which are highly expressed both in the islets and in the exocrine cells of the pancreas (Fig. (Fig.1C)1C) and ubiquitously detected in most tissues (6). Furthermore, while the expression of HMGN1 and HMGN2 is generally downregulated during differentiation (4, 13, 18), our preliminary analyses indicate that HMGN3 protein is first expressed in a few pancreatic cells only after embryonic day 14.5 and the levels of the protein increase during differentiation (data not shown). HMGN3 is highly expressed in the terminally differentiated β cells of both mice and humans. The relative abundance of HMGN3 in adult β cells suggests that the protein plays a role in the specialized function of these cells.
Second, loss of HMGN3 increases the glucose levels and impairs the glucose tolerance of feeding, but not fasting, mice. Nutritional signals are known to affect the levels of insulin in the blood, a key regulator of glucose homeostasis (28). The decreased insulin levels in the serum of feeding Hmgn3−/− mice provide a direct link between HMGN3 expression and glucose homeostasis. Third, loss of HMGN3, but not loss of either HMGN1 or HMGN2, impairs GSIS. Each of the specific siRNAs siN1, siN2, and siN3 efficiently downregulated the expression of its target protein, but only siN3 treatments reduced the GSIS from MIN6 cells (Fig. (Fig.3).3). The levels of HMGN3 expression in β cells and in MIN6 cells are high relative to those in other mouse tissues, but they are comparable to those of HMGN1 or HMGN2 (Fig. (Fig.1C1C and and3B).3B). Yet, in spite of similar protein levels, only HMGN3 affects insulin secretion, suggesting a specific role for HMGN3 in β-cell function. Fourth, while loss of HMGN3 impairs GSIS, it does not impair arginine-stimulated insulin secretion. Arginine stimulates insulin secretion by affecting membrane polarization directly, independently of glucose sensing and import into β cells (28). Taken together, the findings indicate that HMGN3 plays a role in glucose secretion both in mice and in isolated β cells.
GLUT2 is a membrane protein that facilitates the entry of glucose into various mammalian cells, including insulin-secreting pancreatic β cells (15, 23, 30, 38). Although loss of GLUT2 has severe metabolic consequences and affect insulin secretion, several studies, including analysis of Glut2−/− mice, suggest that partial downregulation of GLUT2 levels in β cells does not lead to a strong diabetic phenotype and does not significantly impact insulin secretion from isolated cells (15, 16, 21, 37). Thus, by itself, the partial downregulation of GLUT2 levels in β cells does not fully explain the diabetic phenotype of Hmgn3−/− mice. HMGN3 is expressed in all of the various pancreatic islet cell types, including glucagon-secreting alpha cells (data not shown), as well as in several other tissues, including the brain (20, 40), all of which impact insulin levels and glucose homeostasis. Furthermore, we also observed that the siRNA-mediated downregulation of HMGN3 in MIN6 cells alters the expression of several genes involved in GSIS (Fig. (Fig.4A),4A), suggesting that loss of HMGN3 may have a widespread effect on the transcription profile of β cells. While the transcription level of a single gene may not be severely affected, the cumulative changes in the expression levels of several genes impair the normal function of β cells and impact the GSIS pathway. Recent studies on the role of the histone methyltransferase Set7/9 in β-cell function demonstrate that changes in histone modification and chromatin structure alter the expression of several genes and lead to impaired GSIS (11). Members of the HMGN protein family have been shown to affect the structure of chromatin and the modification levels of histone tails. Thus, HMGN3 may affect the expression levels of β-cell genes by altering their chromatin structure and affecting the recruitment of transcription factors to specific sites.
Our ChIP analyses (Fig. (Fig.5A)5A) indicate that HMGN3 enhances Glut2 transcription directly by binding to the chromatin in the promoter region of this gene. Significantly, only HMGN3, and not the closely related proteins HMGN1 and HMGN2, is enriched in the Glut2 promoter, a finding consistent with the observation that these two HMGN variants do not affect Glut2 transcription (Fig. (Fig.4B)4B) or GSIS (Fig. (Fig.3C).3C). Interestingly, HMGN3 does not have significant effects on the kinetics of DNase I digestion of Glut2 (Fig. 5B to D), an indication that the HMGN3-mediated transcriptional enhancement of this gene does not involves major changes in chromatin structure. On the other hand, HMGN3 alters the levels of histone modification in this gene (Fig. (Fig.6G).6G). Previous studies demonstrated that HMGN1 and HMGN2 affect the levels of several histone modifications in a variant-specific manner (36). Given the large number of histone modifications, it remains to be seen which specific modification is affected by HMGN3.
We found that loss of HMGN3 significantly decreased the binding of several transcription factors, including PDX1, to Glut2 chromatin (Fig. (Fig.6A),6A), thereby providing a mechanism whereby HMGN3 regulates Glut2 transcription. PDX1 is a key transcription factor involved in the development and maintenance of mature β cells, where it affects the transcription of key genes, including Glut2 (9). In considering the mechanism whereby HMGN3 promotes the binding of PDX1 to Glut2, we found that these two proteins specifically associate on the chromatin of the Glut2 promoter, as indicated by our reciprocal sequential precipitation experiments (Fig. (Fig.6F).6F). Furthermore, we found that while HMGN3 affects the binding of PDX1, PDX1 in turn affects the binding of HMGN3 to the same region (Fig. (Fig.6B).6B). Our studies clearly demonstrate that the two proteins do not form a complex in the nucleoplasm and do not bind to the same binding sites throughout the entire chromatin (Fig. 6D and E). It follows that the two proteins mutually strengthen their interaction with a specific chromatin locus. This mode of binding is reminiscent of that of the glucocorticoid receptor (GR) and the chromatin binding protein HMGB1, which have been shown to mutually stabilize each other's chromatin interactions (1).
Stochastic assembly of transcription factors at their target sites has been described in several systems (14). For HMGB1 and GR, Agresti et al. argue that in living cells the stochastic assembly of rapidly moving chromatin binding protein at specific binding sites mutually reinforces their chromatin interactions (1). It is well documented that, just like GR and HMGB1, HMGN and various transcription factors move rapidly through the nucleus, continuously sampling nucleosomes for optimal binding sites (8). We therefore propose that the stochastic assembly of HMGN3 with PDX1 and perhaps other transcription factors at the Glut2 promoter reinforces their chromatin interaction, thereby enhancing Glut2 transcription.
Together with recent studies on the effect of HMGN1 on estrogen-mediated transcriptional activation (44) and with previous analyses of the role of HMGN in gene expression, the emerging picture suggests that HMGN proteins modulate the interaction of specific transcription factors at distinct genes in the context of chromatin (3, 18, 24).
In summary, we identify HMGN3 as a chromatin binding protein that modulates the interaction of a transcription factor with chromatin and in pancreatic β cells affects the expression of several genes, including Glut2. The protein is present in all of the cell types of pancreatic islets and in several tissues known to be involved in regulating glucose homeostasis. In mice, loss of HMGN3 leads to an impaired GSIS response and to changes in glucose and insulin levels. The human and mouse HMGN3 amino acid sequences are very similar, and human β cells express significant levels of HMGN3. Our findings raise the possibility that impaired function of HMGN3 is a contributing factor in the etiology of diabetes.
We thank Jung Ho Ko for preparing the targeting vector, the confocal imaging facility, LEC, CCR, for help with microscopy, and the NIH fellows editorial board for editorial assistance.
This research was supported by the Center for Cancer Research, intramural program of the NCI, NIH.
Published ahead of print on 3 August 2009.