A mutation in the insulin 2 gene induces diabetes with severe pancreatic β-cell dysfunction in the Mody mouse (original) (raw)
A missense mutation in the Ins2Mody allele. Diabetes of the Akita mouse is caused by a single locus, Mody, and is inherited in an autosomal dominant manner (5). The Mody locus maps to the telomeric region of Chromosome 7, and this area includes an important diabetogenic gene Ins2. Therefore, a possible genomic alteration of Ins2 was initially examined. The three exons and the 5′-flanking region of Ins2 were amplified from genomic DNA from control C57BL/6J and Mody mice. Mody mice had a G→A transition at nucleotide 1907 in exon 3 on one of the two Ins2 alleles (Fig. 1). This mutation changed amino acid Cys96 (TGC) to Tyr (TAC). Cys96 corresponds to the seventh amino acid in the A chain (A7) of mature insulin and forms one of the three intramolecular disulfide bonds with Cys31 located at B7. Disruption of a disulfide bond between the A and B chains is likely to induce a major conformational change in insulin 2 molecules. To determine whether this mutation is responsible for the diabetic phenotype of Mody mice, 10 diabetic and 10 control mice from each line of three Mody mutant C3H/He background congenic lines were examined. Taking advantage of the fact that the mutation disrupts a Fnu 4HI site of the Ins2 gene, RFLP analysis was performed. The PCR products digested with Fnu 4HI were 140 bp (wild-type) and 280 bp (mutant), as expected (Fig. 2a). The genotype of Ins2 and the phenotype for diabetes completely matched in 60 mice from three congenic lines (examples shown in Fig. 2b). Thus, we concluded that this Ins2 mutation induces diabetes in Mody mice.
Mutation of the Ins2 gene in the Mody mouse. Ins2 exon 3 was amplified using PCR from genomic DNA. The PCR products derived from either control C57BL/6J (upper) or Mody (lower) mice were directly sequenced from both directions. A single G→A transition at nucleotide 1907 of mouse Ins2 gene (8) on one of the two alleles distinguished the Mody allele.
Genotyping of the Ins2 gene by RFLP analysis. Ins2 exon 3 was amplified using PCR from genomic DNA. The left lane shows ϕX174/Hae III-digested DNA markers. (a) The size of PCR products derived from C57BL/6J (C1, C2, and C3) or Mody mice (A1, A2, and A3) was 280 bp. The mutation found in Mody mice, described in Fig. 1, disrupts an Fnu 4HI site in the exon 3 of Ins2. Digestion with Fnu 4HI did not change the size of the PCR products from the mutated allele (280 bp) but decreased that of the wild-type allele to 140 bp. (b) Representative genotyping of 16 offspring derived from three Mody congenic lines with C3H/He background is shown. Mice with diabetes are shown as “+” under the lane number. The genotype of Ins2 was completely matched with the phenotype in each individual.
Insulin transcription in the islets of Mody mice. Mice have another functional insulin gene, Ins1. Sequence analysis of the coding region of Ins1 revealed that Mody mice did not have any mutations in Ins1 (data not shown). Therefore, the heterozygous mice should have normal insulin molecules derived from three alleles (one Ins2 and two Ins1). To semiquantify the expression levels of the whole insulin, Ins1 and Ins2 cDNAs were amplified from islet RNA using primers derived from the common sequences of both cDNAs. The total insulin levels in the islets of Mody mice were a little lower than those of the control mice when the amounts of RNA from both mice were adjusted by the levels of β-actin (Fig. 3a). Because Ins1 and Ins2 transcripts should be amplified with equal efficiency by the same primers, the transcript from each allele can be quantified if there are unique restriction enzyme sites (17). Bst EII digestion of the amplified insulin cDNAs separated Ins2 transcripts (111 bp) from Ins1 transcripts (257 bp; Fig. 3b, lanes 3 and 4) and revealed that Ins1 transcripts represent approximately 25% of the total insulin transcripts in both C57BL/6J and Mody mice. Ins2 transcripts from wild-type and mutant alleles in Mody mice were similarly quantified by Fnu 4HI digestion. The transcription levels of wild-type (174 bp) and mutant alleles (263 bp) were similar (Fig. 3b, lanes 5 and 6). These results indicate that there is no gross defect in the transcription from Ins1 and Ins2 alleles in the islets of Mody mice.
Insulin transcription in the islets. (a) Total RNA from islets was reverse transcribed to cDNA using oligo-(dT)17 primer. β-actin and insulin cDNAs were then amplified by PCR. The cycle numbers used for the PCR were 18 (lanes 1 and 2), 21 (lanes 3 and 4), 24 (lanes 5 and 6), 27 (lanes 7 and 8), and 30 (lanes 9 and 10), respectively. The amounts of cDNA from control C57BL/6J (odd lanes) and Mody mice (even lanes) were adjusted by the levels of amplified β-actin (upper panels). Note that both Ins1 and Ins2 transcripts should be amplified with equal efficiencies because primers were derived from the common sequences between them. The PCR products of Ins2 (263 bp) and Ins1 (257 bp) were not resolved in this gel system. The total insulin levels in the islets of Mody mice were approximately 85%–90% of those of the control mice (lower panels). (b) The insulin transcripts amplified from islet RNA of either C57BL/6J (lanes 1, 3, and 5) or Mody mice (lanes 2, 4, and 6) were run without digestion (lanes 1 and 2). They were then digested with Bst EII for discrimination between Ins1 (257 bp) and Ins2 (111 bp) transcripts (lanes 3 and 4). Similarly, they were digested with Fnu 4HI to separate Ins1 (167 bp), wild-type_Ins2_ (174 bp), and mutant Ins2 (263 bp) transcripts (lanes 5 and 6). The left lane shows radiolabeled _φ_X174/_Hin_f I digested DNA markers. Because the PCR products were labeled with an end-labeled 5′ primer, the radioactivity of each band corresponds to the expression level, irrespective of its size. The measurement of the radioactivity of each band revealed that 27% and 73% of the total insulin transcripts in C57BL/6J mice are derived from Ins1 and Ins2, respectively. Similar values, 24% for Ins1 and 76% for Ins2, were obtained from Mody mice. Furthermore, 39% of the total insulin transcripts, which is approximately half the value of total Ins2, were derived from the mutant Ins2 allele in Mody mice, suggesting that both normal and mutant Ins2 alleles are transcribed similarly.
Morphologic studies of the pancreatic islets of Mody mice. Expression of insulin protein levels was then characterized using immunofluorescence staining (Fig. 4, a and b). The overall size of the islets of Mody mice tended to be smaller, and the intensity of the insulin immunoreactivity was remarkably weaker than that of C57BL/6J mice. The decrease was so prominent that it cannot simply be ascribed to the loss of immunoreactivity of insulin derived from the mutant allele. This indicates that the content of wild-type insulin was also dramatically decreased. A small population of β cells expressed insulin levels comparable to that of normal mice. Immunofluorescence analysis using anti–C-peptide antibodies revealed similar results: a decrease and heterogeneous staining pattern in the islets of Mody mice (Fig. 4, c and d). The amount appeared relatively higher than that of insulin. As anticipated from the nature of the antigen oligopeptide, this antibody recognized both wild-type and mutant proinsulin (see Figs. 7, 8c, 9, and 10), whereas anti-insulin antibody whose antigen is the whole insulin protein did not recognize the mutant (see Fig. 8b). Nevertheless, much weaker C-peptide immunoreactivity in β cells of Mody mice strongly suggests a decrease in the amount of proinsulin, including the wild type.
Immunofluorescent detection of insulin, C-peptide, and BiP in the islets. Pancreatic islets of C57BL/6J (a, c, and e) and Mody (b, d, and f) mice were incubated with anti-insulin (a and b), anti-C-peptide (c and d), and anti-BiP (e and f) antibodies. Note that the positive staining for insulin, C-peptide, and BiP was found exclusively in the cytoplasm, as revealed by the lissamine rhodamine sulfonyl chloride–labeling method (red). Bar, 50 μm.
Immunoblotting analysis of the islet proteins. An equal amount of the islet protein (50 μg) from either C57BL/6J (lanes 3 and 7) or Mody mice (lanes 4 and 8) was loaded in each lane onto 16.5% polyacrylamide gel with (lanes 5_–_8) or without (lanes 1_–_4) 100 mM DTT. Human proinsulin (lanes 1 and 5) and human insulin (lanes 2 and 6) were loaded as standards. The human C-peptide ran off this tricine–SDS-PAGE system (data not shown). Immunoblotting was performed using anti–C-peptide antibodies (lanes 1_–_8). On the same membranes, similar analyses were performed using anti-insulin (lanes 9_–_12, corresponding to lanes 1_–_4 in the C-peptide immunoblot), anti-PDI (lanes 13 and 14, corresponding to lanes 3 and 4), and anti-BiP antibodies (lanes 15_–_18, corresponding to lanes 3, 4, 7, and 8). The islet protein lysed by either sample buffer containing 3% SDS (16) or acid-ethanol (26) revealed similar proteins immunoreactive to anti–C-peptide antibodies on nonreducing gels (data not shown). DTT, dithiothreitol; PDI, protein disulfide isomerase.
CHO cell lines expressing either wild-type or mutant insulin 2. (a) Northern blot analysis of insulin. Total RNA (20 μg) from CHO (lane 1), CHO-Ins2wt (clone w9; lane 2), and CHO-Ins2Mody cells (clone a7; lane 3) were electrophoresed, transferred to a nylon membrane, and hybridized with a mouse Ins2 cDNA probe. (b) Insulin secretion of CHO-Ins2wt (clone w9) and CHO-Ins2Mody (clone a7). Both cells were seeded at a density of 2 × 105 cells/6-cm dish. After 24 h, the cells were incubated with serum-free media for the indicated times. Insulin stored in the cells (black bars) and that released into the media (open bars) were measured using anti-insulin antibodies. Although cell density and time course were different, similar data were obtained from another independent experiment. (c) Proinsulin content and secretion. CHO-Ins2wt (lanes 3 and 8 for clone w7; lanes 4 and 9 for clone w9), and CHO-Ins2Mody cells (lanes 5 and 10 for clone a1; lanes 6 and 11 for clone a3; lanes 7 and 12 for clone a7) were incubated with serum-free media for 24 h. The cells (lanes 3_–_7) were then solubilized, and the media (lanes 8_–_12) were concentrated by 10% trichloroacetic acid. These samples were resolved with tricine–SDS-PAGE (16.5% polyacrylamide gel) in a reducing condition (100 mM DTT). Immunoblotting analysis was performed using anti–C-peptide antibodies. Lanes 1 and 2 contain human proinsulin standard and the islet protein from normal C57BL/6J mice, respectively. CHO, Chinese hamster ovary.
Pulse-chase labeling of CHO cells that express wild-type and mutant insulin. CHO-Ins2wt (clone w9, upper) and CHO-Ins2Mody (clone a7, lower) were labeled for 30 min with [35S]methionine. After chasing for indicated times, immunoprecipitation was performed from either cell extracts (C) or media (M), using a mixture of anti-insulin and anti–C-peptide antibodies. Immune complexes were analyzed with tricine–SDS-PAGE (16.5% polyacrylamide gel).
Coprecipitation of proinsulin and BiP. Proinsulin was immunoprecipitated from CHO (lane 1), CHO-Ins2wt (clone w9; lane 2), and CHO-Ins2Mody cells (clone a7, lane 3). Immunoprecipitates were loaded on glycine–SDS-PAGE gel (8% polyacrylamide gel), transferred onto an Immobilon-P membrane, and immunoblotted with anti-BiP antibodies. The faint band of BiP immunoprecipitated from CHO cells (lane 1) might be due to its weak affinity to immunoglobulin.
Electron microscopy of control β cells revealed that a large number of dense-core secretory granules filled the entire cytoplasm (Fig. 5, a and b). On the other hand, the size and number of secretory granules were remarkably smaller in Mody mice (Fig. 5, c and d). This accounts for the significant reduction of the β-cell size in mutant mice. Instead, the cytoplasm was filled with the ER. ER without ribosomes was especially prominent, which represents transitional ER located between the Golgi apparatus and the rough ER (Fig. 5d, arrowhead). The lumina of the transitional ER were markedly enlarged and had a more electron-dense appearance. These data suggest a block in the transport of proinsulin from the ER to the Golgi apparatus and an accumulation in the ER.
Ultrastructural morphology of the islets. Electron micrographs of the β cells were taken from either C57BL/6J (a and b) or Mody (c and d) mice. Parts b and d represent higher magnification of a and c, respectively. Arrows indicate the rough ER, arrowheads represent the transitional ER, and G indicates the Golgi apparatus. Bar, 1 μm (a and c); Bar, 0.25 μm (b and d). ER, endoplasmic reticulum.
This interpretation was further supported by the findings of immunoelectron microscopy using the two-face, double-labeling method. When the ultrathin sections were immunolabeled with gold particles, most of the insulin and C-peptide were found in the secretory granules in the control islets (Fig. 6a). In contrast, significant labeling of C-peptide was observed in the ER in Mody mice, although residual granules exhibited weak immunolabeling of insulin and C-peptide (Fig. 6b). Because the bulk of proinsulin conversion occurs in the immature, clathrin-coated granule (18), the C-peptide immunoreactivity in the ER should represent that of proinsulin.
Immunolocalization of insulin and C-peptide. β cells from either C57BL/6J (a) or Mody (b) mice were double immunolabeled for insulin (large gold particles) and C-peptide (small gold particles). Although insulin and C-peptide signals are profoundly decreased in the secretory granules in Mody mice compared with the control strain (arrows), a significant amount of C-peptide immunoreactivity exists in the ER (arrowheads) in Mody mice. Bar, 0.25 μm.
Insulin processing in the islets of the Mody mice. Insulin processing was examined using immunoblotting analysis of islet proteins. Because our anti-insulin antibody has a very weak affinity to insulin on the reducing gels that separate the A and B chains (data not shown), the immunoblotting was performed on nonreducing gels. Insulin migrated as a 5.5-kDa protein close to the position of human insulin in the control islets (Fig. 7, lanes 10 and 11). The amount of insulin was dramatically decreased in the islets of Mody mice, consistent with the results of immunostaining (Fig. 7, lane 12). The immunoblotting by anti–C-peptide antibody on reducing gels revealed 8.6-kDa and 7.6-kDa proteins in the normal islets, whereas only the 8.6-kDa protein was detected in the islets of Mody mice (Fig. 7, lanes 7 and 8). The 8.6-kDa protein was considered proinsulin because it comigrated with human proinsulin (Fig. 7, lane 5). The nature of the 7.6-kDa protein remains unknown, but probably represents some form of proinsulin metabolite with a different structure that cannot be attained in the mutants. A notable reduction in proinsulin levels, but less prominent than that of insulin, was again observed in the islets of Mody mice as in the immunostaining study. Immunoblotting using the same antibody was also performed on nonreducing gels. Proinsulin was detected around the position corresponding to that found on reducing gels in the control (Fig. 7, lane 3). In addition, bands whose molecular weights are approximately twice that of monomeric proinsulin were evident. If they represent dimers, it suggests that proinsulin dimers are physiologically formed during folding through intermolecular disulfide bonds. In addition to these discrete bands, there was a high level of immunoreactivity at the boundary between the stacking and separating gels. This immunoreactivity is clearly specific because it was not detected by the same antibody on reducing gels (Fig. 7, lanes 7 and 8) nor was it detected by different antibodies on nonreducing gels (Fig. 7, lanes 11 and 12). It likely corresponds to complexes with other proteins or aggregates of proinsulin molecules. Alternatively, sodium dodecyl sulfate (SDS) may cause artificial dimers or aggregates of proinsulin in nonreducing gels. Similar protein that cannot enter a resolving gel has been identified as aggregates of expanded polyglutamine-containing ataxin-3 protein that causes Mashado-Joseph disease (19). This high-molecular-weight form was also present in the islets of Mody mice (Fig. 7, lane 4). No other discrete bands immunoreactive to anti–C-peptide antibodies were detected in mutant mice. These findings indicate that little proinsulin is processed to insulin but is accumulated as high molecular weight complexes or aggregates in the β cells of Mody mice.
Molecular chaperones and enzymes involved in proinsulin folding. The existence of high molecular-weight forms of proinsulin prompted us to examine the possible proteins involved in the folding of proinsulin. There are at least two classes of proteins involved in this process: enzymes that catalyze specific isomerization steps and chaperones that stabilize unfolded or partially folded structures and prevent the formation of inappropriate intra- or interchain interactions (20). The first class of protein includes PDI. PDI, an ER-resident enzyme, catalyzes the isomerization of protein disulfide bonds and thereby facilitates formation of the correct set of disulfide bonds (20). Because the Mody mutation changes a cysteine residue that normally forms an intramolecular disulfide bond of insulin, possible stimulation of PDI expression was examined. PDI was overexpressed in the islets of Mody mice (Fig. 7, lanes 13 and 14).
One of the representative ER chaperones that directly determine the folding of the polypeptide is BiP (21). BiP synthesis is induced by the accumulation of secretory precursors or mutated proteins in the ER or by a number of different stress conditions that increase aberrant protein folding (20). BiP was expressed weakly both in exocrine and endocrine cells in the control pancreas, although some islet cells had focal overexpression (Fig. 4e). In contrast, it was distinguishably overexpressed in the islets of Mody mice (Fig. 4f). Double immunostaining with anti–C-peptide antibodies revealed that BiP overexpression occurred specifically in the β cells (data not shown). Immunoelectron microscopy revealed that BiP was confined to the ER of the control islets and also detected in the abnormally enlarged lumen of the ER in β cells of Mody mice (data not shown). Immunoblotting study confirmed that BiP was overexpressed in the islets of mutant mice (Fig. 7, lanes 15_–_18). A portion of BiP showed a slower mobility on nonreducing gels, suggesting a formation of complexes with proinsulin (Fig. 7, lanes 15 and 16). These results indicate that PDI and BiP, thought to be involved in the folding of proinsulin, are overexpressed in the islets of Mody mice.
Analysis in CHO cells expressing wild-type and mutant insulin 2. To examine the intracellular metabolism of proinsulin in details, CHO cell lines were established that constitutively express either wild-type (CHO-Ins2wt) or mutant insulin 2 (CHO-Ins2Mody). There are two major advantages to the use of these cell lines. First, they overcome the difficulty of using a limited number of murine islets. Second, in contrast to the islets of Mody mice, it is not necessary to consider coexisting wild-type insulin 2 or insulin 1 molecules. Among several clones established, CHO-Ins2wt (clone w9) and CHO-Ins2Mody (clone a7) were analyzed in detail. Northern blot analysis revealed that CHO-Ins2Mody cells expressed approximately two times the insulin message than those of CHO-Ins2wt cells (Fig. 8a).
Next, the amount of insulin secreted into the media and the insulin content in cells were determined using anti-insulin antibodies (Fig. 8b). In CHO-Ins2wt cells, a significant amount of IRI was found in the cell extracts, and the amount secreted in the media was increased with time. In contrast, IRI was not detected in either CHO-Ins2Mody cell extracts or their media. As shown in Fig. 9, little proinsulin was processed to insulin in CHO cells, probably because non-endocrine CHO cells lack hormone-specific conversion endopeptidases such as PC1/PC3 or PC2. Although anti-insulin antibodies we used could not detect proinsulin on immunoblots (Fig. 7), they have a significant affinity to human proinsulin in solution in the immunoassay (data not shown). Therefore, most IRI found in these cells should represent proinsulin. The lack of IRI in CHO-Ins2Mody cells indicates either loss of the immunoreactivity or absence of mutant proinsulin. To discriminate these possibilities, immunoblotting analysis of these cell extracts and the media was performed using anti–C-peptide antibodies. Proinsulin was detected in several CHO-Ins2Mody cell lines independently isolated (Fig. 8c, lanes 5_–_7), and the protein levels in cells were proportional to the RNA levels (data not shown). This finding indicates an actual synthesis and intracellular accumulation of mutant proinsulin in CHO-Ins2Mody cells. It also indicates that mutant proinsulin lost the immunoreactivity to anti-insulin antibodies probably because of the conformational change but preserved that to anti–C-peptide antibodies. Although the media of CHO-Ins2wt cells contained a significant amount of proinsulin (Fig. 8c, lanes 8 and 9), there was no detectable proinsulin in those of CHO-Ins2Mody cells (Fig. 8c, lanes 10_–_12). These results indicate that mutant proinsulin is not efficiently secreted in the media, but is degraded intracellularly. Proinsulin detected in and secreted from these CHO cell lines comigrated with the 8.6-kDa protein, but not the 7.6-kDa protein, of control islets (Fig. 8c), suggesting that the 7.6 kDa protein possibly corresponds to a conversion intermediate specific to the β cells. The high-molecular-weight form of proinsulin was not found in the extracts of these CHO cell lines on nonreducing gels (data not shown).
To examine how mutant proinsulin is metabolized in these cells, pulse-chase experiments were performed. Following the pulse-labeling of cells with [35S]methionine for 30 minutes and chasing for the indicated period, immunoprecipitation was performed using a mixture of anti-insulin and anti–C-peptide antibodies (Fig. 9). The amount of proinsulin initially synthesized in CHO-Ins2Mody cells was approximately two times that in CHO-Ins2wt cells, consistent with their insulin mRNA levels. Virtually no wild-type or mutant proinsulin was converted to insulin in CHO cells during the 24-hour period. Following the chase, both wild-type and mutant proinsulin disappeared from cells at similar rates. Although the amounts secreted in the media during the first two hours were similar in wild-type and mutant proinsulin (approximately 8% of the initial proinsulin), the secretion rate of proinsulin differed thereafter. Whereas wild-type proinsulin was gradually secreted and accumulated in the media, the amount of mutant proinsulin in the media decreased as time passed. The lower amounts recovered at the later time points in CHO-Ins2wt cells is probably due to the competitive inhibition for immunoprecipitation by accumulated unlabeled proinsulin secreted in the media, as shown in Fig. 8, b and c. This competitive inhibition should not occur in CHO-Ins2Mody cells because there was little proinsulin in the media (Fig. 8c). The decreased secretion of proinsulin into the media of CHO-Ins2Mody cells suggests no further secretion of synthesized mutant proinsulin after two hours. Thus, some portions of mutant proinsulin could be secreted in a relatively short time, but once trapped in the ER, they are destined to be degraded intracellularly. Immunoblotting analysis also confirmed that the amounts of proinsulin secreted in the media during two hours were comparable in these two cell lines (data not shown), whereas those during 24 hours were far different (Fig. 8c).
To obtain further evidence for the role of BiP in the folding of proinsulin, interaction of proinsulin and BiP was directly examined in the CHO cell lines described above. Proinsulin was first immunoprecipitated with a mixture of anti-insulin and anti–C-peptide antibodies, and then the immunoprecipitates were immunoblotted using anti-BiP antibodies. In CHO-Ins2wt cells, BiP formed a complex with wild-type proinsulin (Fig. 10, lane 2). Moreover, increased amounts of the complex were found between BiP and mutant proinsulin in CHO-Ins2Mody cells (Fig. 10, lane 3).