Role of Proinsulin Self-Association in Mutant INS Gene–Induced Diabetes of Youth

Proinsulin (PI) misfolding, endoplasmic reticulum (ER) stress, and decreased β-cell mass are key features of β-cell failure and diabetes in the autosomal-dominant diabetes known as mutant INS gene–induced diabetes of youth (MIDY) (1–3). All MIDY patients carry a mutation in only one of two INS alleles (3–6), and, in principle, one normal INS allele should be sufficient to maintain normoglycemia (7). The underlying mechanism of MIDY is thought to involve abnormal interactions of misfolded mutant PI with bystander PI-WT in the ER, impairing the ER export of PI-WT and decreasing production of bioactive mature insulin (8,9). However, to date, how these interactions are initiated remains largely unknown. Interestingly, the clinical spectrum of MIDY ranges from neonatal-onset severe insulin-deficient diabetes to relatively late-onset mild diabetes (1,2), suggesting that different PI mutants may behave differently in their folding processes as well as in their ability to interact and obstruct bystander PI-WT in the ER.

In this study, we report that a proline (Pro) substitution for tyrosine residue at the 16th position of the PI B-chain tyrosine-16 (Tyr-B16) [i.e., PI-Y(B16)P] results in dramatic PI misfolding in the ER. Yet, it is unable to interact with bystander PI-WT in the ER and is unable to impair the ER export of PI-WT or to block its ability to form mature insulin. Interestingly, Y(B16)P disrupts the main α-helix in the PI B-domain. Normally, PI uses this B-domain α-helix in order to homodimerize in the ER (10). Thus, we hypothesize that cross-dimerization between MIDY mutant and PI-WT may be required to initiate their abnormal interaction that impairs the intracellular transport of PI-WT, resulting in insulin deficiency that leads to diabetes (11).

Structural evidence strongly implicates Tyr-B16 as a key residue for normal PI dimerization (12). With this in mind, we have mutagenized this residue to either aspartic acid [Y(B16D)] or alanine [Y(B16)A]; both substitutions were entirely innocuous when introduced into PI-WT. Notably, however, when these substitutions were introduced into the MIDY-causing Akita PI-C(A7)Y, each could function as an intra-allelic suppressor, limiting abnormal interactions with PI-WT; alleviating dominant-negative behavior on PI-WT intracellular trafficking; and increasing insulin production in pancreatic β-cell lines and human islets. This study provides insight that protecting the PI-PI contact surface might be a potential druggable target to limit abnormal PI interactions and prevent/delay the development and progression of MIDY.

Guinea pig polyclonal antiporcine insulin, human insulin-specific radioimmunoassay (RIA) kit, human PI-specific RIA kit, and mouse insulin ELISA kit were from Millipore (Billerica, MA). Guinea pig polyclonal anti-human insulin was from Novus (Centennial, CO). Rabbit polyclonal anti-mouse PI C-peptide (EVEDPQVAQLELGGGPGAGDLQTLALEVAQQ) was raised by Bioworld (Nanjing, China). Mouse monoclonal anti-human PI C-peptide–A-chain junction (GSLQKRGIVE) was raised by Abmart (Shanghai, China) as previously described (13). Rabbit anti-Myc and anti-GFP were from Immunology Consultants Laboratories (Portland, OR). Rabbit anti-BiP was a gift from Dr. P.S. Kim (U. Cincinnati). The ³⁵S-amino acid mixture was from PerkinElmer (Waltham, MA). Dithiothreitol, protein A-agarose, digitonin, N-ethylmaleimide, brefeldin A, and all other chemical reagents were from Sigma-Aldrich (St. Louis, MO). The 4–12% gradient SDS-PAGE used NuPage gels, and Met/Cys-deficient DMEM and all other tissue culture reagents were from Invitrogen (Thermo Fisher, Waltham, MA). Tris-tricine-urea-SDS-PAGE was performed as described previously (14).

Single-chain Tyr^B16-DesDi and Asp^B16-DesDi insulin analogs were synthesized using a Tribute Peptide Synthesizer with preprogrammed solid-phase Fmoc protocol (15). The synthesis was carried out on a 0.1-mmol scale using H-Asn(Trt)-HMPB-ChemMatrix resin (0.46 mmol/g loading). Side chain protection was Asn(Trt), Arg(Pbf), Asp(OtBu) Cys(Trt), Gln(Trt), Glu(OtBu), His(Trt), Lys(Boc), Ser(tBu), Thr(tBu), Tyr(tBu). DIC/6-chloro-HOBt in 1:2 dichloromethane/DMF was used for the coupling and 20% piperidine in DMF for N^α-Fmoc-deprotection. The product peptide was then cleaved from the resin and simultaneously deprotected by subjecting it to trifluoroacetic acid (TFA)/TIPS/water/DODT/anisole (90:2.5:2.5:2.5:2.5 v/v) for 2.5 h at ambient temperature. The crude product was precipitated with ice-cold ether, washed, and dried. For folding, 100-mg Asp^B16-DesDi crude polypeptide (corresponding to 18 µmol), cysteine hydrochloride (43 mg, 2 mmol/L), and glycine (270 mg, 20 mmol/L) were placed in a 600-mL glass beaker, and distilled, deionized water was added to a 0.1 mmol/L final concentration. The pH of the solution was adjusted to 10.5, and the reaction was stirred open to the air at 4°C for 16 h. The desired folded product was purified both by preparative high-performance liquid chromatography (HPLC) and by analytical reverse-phase HPLC on a C8 Proto (4.6 × 250 mm) 300 Å, 5 μm, Higgins Analytical Inc. column, using 25–50% Solvent B (0.1% TFA in acetonitrile) in Solvent A (0.1% TFA in water) over 35 min at a flow rate of 1.0 mL/min (1525; Waters). Protein elution was monitored at 215 and 280 nm (2489 absorbance detector; Waters), and masses were confirmed by matrix-assisted laser desorption/ionization-time of flight mass spectrometer (MALDI-TOF). Normal insulin helical structure was confirmed by far-ultraviolet circular dichroism spectra from 190 to 250 nm (16), and normal thermodynamic stability was established by guanidine hydrochloride-induced denaturation (monitored by circular dichroism at 222 nm). ¹H–nuclear magnetic resonance (¹H–NMR) spectra were acquired on protein solutions (0.5 mg in 300 μL D₂O; direct meter reading pD 7.6) at 25°C using a Bruker AVANCE 700 MHz spectrometer.

The nonstandard amino acid ornithine (Orn) was used at position B29 in place of lysine to preserve the side-chain amino group without susceptibility to tryptic cleavage (16). Two-chain insulin analogs ([Tyr^B16, Orn^B29] insulin and [Asp^B16, Orn^B29] insulin) were prepared by trypsin-catalyzed cleavage of human insulin or Asp^B16-DesDi single-chain insulins (and purification by reverse-phase HPLC) to generate Des-Octapeptide insulins, followed by tryptic-mediated peptide bond formation between Arg^B22 and a synthetic octapeptide bearing Orn^B29 (17,18); The final two-chain products were purified by analytical reverse-phase HPLC with molecular mass verified by MALDI-TOF (4700; Applied Biosystems). Self-association properties of the two insulin analogs were determined using HPLC size-exclusion chromatography in the absence of zinc (16). In brief, insulin analogs were made 0.6 mmol/L in 10 mmol/L Tris-HCl (pH 7.4), 140 mmol/L NaCl and 0.02% sodium azide, and 20 µL were loaded on an Enrich SEC70 column (10 mm × 300 mm with fractionation range 3–70 kDa) and were run in the same buffer. Elution time (Supplementary Fig. 1A) was monitored by absorbance at 280 nm.

The plasmids encoding mouse Ins2 PI-WT or C(A7)Y, untagged human PI-WT, Myc-tagged human PI-WT or C(A7) mutant, hPro-CpepSfGFP bearing Cys(A7) or the Tyr(A7) mutation were described previously (10,14). These plasmids were used as templates to introduce additional Y(B16)D and Y(B16)A mutations using the QuikChange site-directed mutagenesis kit (Stratagene). All resulting plasmids encoding corresponding PI mutations were confirmed by direct DNA sequencing.

HEK293 cells, INS1 cells, or Min6 cells were plated into 6- or 12-well plates 1 day before transfection. A total of 1–2 µg plasmid DNA was transfected using Lipofectamine (Invitrogen). For metabolic labeling, transfected 293T cells or Min6 cells were pulse-labeled with ³⁵S-Met/Cys for 30 min with chase for different times as indicated. The chase media and lysed cells were immunoprecipitated (IP) with anti-insulin or anti-Myc, and analyzed using tris-tricine-urea-SDS PAGE or 4–12% gradient SDS-PAGE under reducing and nonreducing conditions as described previously (14). For coprecipitation (co-IP), the cells were lysed with co-IP buffer containing 100 mmol/L NaCl, 25 mmol/L Tris pH 7.0, 0.1% Triton X-100, 5 mmol/L EDTA, and protease inhibitors mixture. Of the total lysates, 90% were IP with anti-GFP followed by Western blotting with anti-PI antibodies as indicated. The remaining 10% of the total lysates were used to determine expression levels of untagged and GFP-tagged PI. For confocal imaging, INS1 cells transfected to express hPro-CpepSfGFP-WT or mutants were permeabilized with 0.5% Triton X-100, fixed with 3.7% formaldehyde and prepared for indirect immunofluorescence, imaged by confocal fluorescence microscopy. The BiP promoter-driven firefly luciferase assay, human PI-specific RIA, and human insulin-specific RIA were described previously (8).

A total of 2500 IEQ human islets (Prodo Laboratories, Aliso Viejo, CA) were transduced with 1.5 × 10¹¹ particles of adenoviruses expressing Myc-tagged murine PI-WT, C(A7)Y, or C(A7)Y/Y(B16)D and were incubated in a humidified incubator at 37°C with 5% CO₂ for 12 h. Media were then changed, and islets were harvested 48 h after transduction. Islets were lysed in a buffer containing 25 mmol/L Tris, 0.15 mol/L NaCl, 1 mmol/L EDTA, 1% NP40, 5% glycerol; pH 7.4 (c-Myc Kit IP Lysis Buffer) and protease inhibitors (# 78425; Thermo Fisher). The lysates were resolved by SDS-PAGE and electrotransfered followed by immunoblotting using primary antibodies to human PI, Myc, and GAPDH, with appropriate peroxidase-conjugated secondary antibodies and enhanced chemiluminescence. For human insulin content, islet lysates were assayed using a human insulin ELISA that does not cross-react with PI (# 10–1132–01; Mercodia), following manufacturer’s instructions (n = 3 in each sample).

Statistical analyses were carried out by two-tailed Student t test (for Fig. 4B and D) or one-way ANOVA followed by Dunnett multiple comparisons (for Fig. 2B, D, F, and G, Fig. 5D, and Fig. 6D–F) using GraphPad Prism 7. A P value of <0.05 was taken as statistically significant.

The data sets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request.

In the ER, PI dimerization (10,19) involves nonpolar association of apposed B-chain residues 8–29 (encompassing the central B9-B19 α-helix), which continue to remain associated in mature insulin (20,21). The insulin crystal structure highlights that B16Y contributes more surface area to the dimerization interface than any other residue (Fig. 1A and B), and a substitution of this residue yields a high-potency monomeric insulin (12). We therefore engineered both two-chain and single-chain insulin derivatives bearing the Y(B16)D substitution, both of which showed impaired dimerization in dilute solution as judged by size-exclusion chromatography and one-dimensional ¹H-NMR spectroscopy, respectively (Fig. 1C and D). To determine the role of Tyr-B16 in PI folding and dimerization in the ER, we expressed PI-Y(B16)D and Y(B16)A in both 293T and Min6 β-cells. We demonstrated that these substitutions affect neither PI ER oxidative folding (Fig. 2A, B, E, and F) nor ER export (Fig. 2C–E and G), suggesting that impaired dimerization alone does prevent the intracellular trafficking of PI. However, Y(B16)P that disrupted the PI B9-B19 α-helix indeed impaired formation of proper disulfide bonds and blocked secretion of PI (Fig. 2), suggesting that the central α-helix of the PI B-chain indeed plays an important role in PI folding in the ER.

We hypothesize that the PI-PI contact surface initiates the attack by misfolded MIDY PI mutants on coexpressed PI-WT. We therefore decided to explore the role of PI-Y(B16) in the interactions between the MIDY-causing Akita PI-C(A7)Y and PI-WT. In 293 cells, we coexpressed untagged human PI-WT with Myc-tagged human PI-WT or C(A7)Y [Myc-PI-C(A7)Y] with or without Y(B16)D substitution (Fig. 3A, lanes 4–6 and 4′-6′). By nonreducing SDS-PAGE and anti-PI immunoblotting, the overexpression of either untagged PI-WT or Myc-PI-WT resulted in detectable formation of homotypic disulfide-linked complexes (Fig. 3A, lanes 2 and 3). Recent evidence indicates that Akita PI is similarly predisposed to form a ladder of aberrant disulfide-linked complexes even at low expression levels (13,22,23), and Akita PI can heterotypically corecruit PI-WT into these aberrant complexes (9).

Upon coexpression, although monomeric forms of Myc-PI-WT, C(A7)Y, and C(A7)Y/Y(B16)D were expressed at comparable levels, Myc-PI-C(A7)Y clearly formed more disulfide-linked dimers (untagged PI dimers and Myc-tagged dimers were marked as D and D′, respectively), heterodimers (marked as H), and higher-molecular-weight PI complexes (Fig. 3A, compare lanes 4 and 5). Upon treatment with the reducing agent dithiothreitol, all of these complexes collapsed into a single band of reduced untagged or Myc-tagged-PI, indicating that the bands seen upon nonreducing SDS-PAGE are disulfide-mediated PI dimers and cross-dimers, as well as disulfide-linked PI complexes (DLPCs). Importantly, the formation of these abnormal PI complexes was notably decreased in cells expressing the C(A7)Y/Y(B16)D double mutant (Fig. 3A, compare lanes 5 and 6).

To examine whether these disulfide-mediated PI interactions occur in β-cells, we expressed Myc-PI-WT or mutants in Min6 β-cells. As shown in Fig. 3B, Myc-PI-C(A7)Y not only formed significantly more abnormal disulfide-linked PI dimers (marked as D′) compared with that of Myc-PI-WT, but also formed disulfide-linked cross-dimers with endogenous PI (noted by an arrow). The high-molecular-weight DLPC were also increased in the cells expressing Myc-PI-C(A7)Y (Fig. 3B, compare lanes 2 and 3). However, once again, introducing the intra-allelic Y(B16)D significantly decreased these abnormal PI complexes (Fig. 3B, compare lanes 3 and 4). Together, these results suggest that limiting PI-PI interactions by introducing Y(B16)D can decrease formation of abnormal DLPCs between PI-C(A7)Y and PI-WT.

To further confirm this point, we expressed GFP-tagged PI-C(A7)Y with or without the Y(B16)D substitution in both 293 cells and Min6 cells. We found that Y(B16)D inhibits co-IP of PI-WT by GFP-PI-C(A7)Y in both 293 cells (Fig. 4A and B) and Min6 cells (Fig. 4C and D) under steady-state condition. Thus, the Y(B16)D substitution decreases the strength of interaction of MIDY PI with PI-WT in the ER.

Next, we examined folding and secretion of PI-C(A7)Y with or without B16 substitutions, and we found that the introduction of Asp (D), Ala (A), or Pro (P) had no effect on the impaired disulfide maturation of PI-C(A7)Y, and did not rescue secretion of the MIDY mutant (which was retained intracellularly) in both 293 cells (Fig. 5A and B) and Min6 cells (Fig. 5C). We then checked whether B16 substitutions could limit the dominant-negative effect of PI-C(A7)Y on bystander PI-WT. We coexpressed mouse PI-C(A7)Y ± Y(B16)D or Y(B16)P substitutions with human PI-WT in 293T cells. We used a species-specific immunoassay to follow the secretion of coexpressed human PI-WT in the presence of mutant PI. Indeed, the expression of mouse PI-C(A7)Y blocked the secretion of PI-WT by ∼70% (Fig. 5D), which is consistent with previous reports (8,9,14,24). However, the Y(B16)D substitution [or Y(B16)P] functioned as an intra-allelic suppressor of the dominant-negative effect of PI-C(A7)Y on the secretion of coexpressed PI-WT (Fig. 5D).

These immunoassay results were then confirmed using pulse-chase experiments. Once again, PI-Y(B16)A and PI-Y(B16)D were secreted normally, in parallel with the secretion of coexpressed PI-WT (Fig. 5E, lanes 15 or 21). Further, both misfolded mutants PI-C(A7)Y and PI-Y(B16)P were fully blocked in their secretion. However, the Akita PI-C(A7)Y conferred a clear dominant-negative effect on coexpressed PI-WT that was not apparent for PI-Y(B16)P (Fig. 5E, lanes 12 vs. 18), indicating that not all misfolded PI mutants have the ability to block secretion of bystander PI-WT. Moreover, we found that neither PI-C(A7)Y/Y(B16)A nor C(A7)Y/Y(B16)D could block the secretion of coexpressed PI-WT (Fig. 5E, lanes 12 vs. 24 or 30), indicating that these Tyr-B16 substitutions reversed the dominant-negative effect of Akita mutant PI.

To further investigate the effect of B16 substitutions on endogenous PI in β-cells, we transfected GFP-tagged PI-WT or mutants, and we used three independent approaches to examine transdominant effects. First, using confocal immunofluorescence, we found that while GFP-PI-WT and Y(B16)D showed a punctate insulin secretory granule pattern, PI-C(A7)Y appeared in a typical ER pattern and caused a profound loss of endogenous insulin (Fig. 6A middle, white arrows). However, although the double mutant PI-C(A7)Y/Y(B16)D also appeared to be retained in the ER, it did not detectably decrease endogenous insulin in β-cells (Fig. 6A bottom, white arrowheads). Second, using human insulin-specific RIA, we confirmed that introducing Y(B16)D in PI-C(A7)Y alleviated the transdominant effect of the MIDY mutant as demonstrated by increased insulin production (Fig. 6B). Further, in human islets, adenoviral expression of mouse PI-C(A7)Y significantly decreased human PI and insulin content, whereas comparable expression of the C(A7)Y/Y(B16)D mutant (Supplementary Fig. 3) again demonstrated intra-allelic suppression of the dominant-negative effect of the MIDY mutant on both human PI and insulin (Fig. 6D). Finally, the enhanced ER export of PI-WT correlated with decreased ER stress in β-cells coexpressing PI-C(A7)Y versus C(A7)Y/Y(B16)D, as measured in a transcriptional BiP promoter reporter assay (Fig. 6D). Taken together, these results strongly suggest that Tyr-B16, which is a major component of the PI-PI interaction surface, is involved in the physical interaction between MIDY mutant PI and PI-WT, and this interaction is necessary for the mutant protein to block PI-WT exit from the ER, impair insulin production, trigger ER stress, and promote the development of diabetes.

To date, more than 50 insulin gene mutations have been reported to cause MIDY, an autosomal-dominant form of diabetes (also known as maturity-onset diabetes of the young type 10) (1,11). More than 60% of the identified mutant alleles have been predicted or experimentally confirmed to impair oxidative folding of PI in the ER (1,25). Importantly, these misfolded MIDY mutants not only fail to exit from the ER, but also physically attack coexpressed PI-WT, impairing the folding and ER export of coexpressed PI-WT, decreasing insulin production, and initiating insulin-deficient diabetes (11,14,26). Heretofore, the molecular mechanism underlying these dominant-negative effects have been completely unknown. As the PI-PI contact surface might possibly engage even before completion of the folding of individual PI monomers, the present work focuses attention on this surface as a possible site initiating the attack of misfolded mutant PI onto bystander PI in the ER.

In this study, we found at least two substitutions of the key Tyr-B16 residue that had no observable impact on PI monomer folding or ER export but that could limit PI-PI contact and thereby decrease abnormal interactions and alleviate transdominant effects of observed in MIDY. It is interesting to note that among the roughly 30 known PI sites/residues that trigger MIDY, none has yet been described that introduces a charged or highly polar residue into the B9-B19 α-helix. We postulate that such substitutions may be relatively ineffective in propagating misfolding onto bystander PI-WT molecules. This hypothesis is supported by data from the PI-Y(B16)P mutant, which disrupts the B9-B19 α-helix to cause severe PI misfolding and ER retention (Fig. 1), but it cannot impose a dominant-negative effect on bystander PI-WT (Fig. 5). These data support the notion that PI-PI contact is an early folding event and initiates the transdominant effects of MIDY mutants.

Increasing evidence indicates that PI misfolding and ER stress not only are the molecular basis of MIDY but may also play an important role in the development and progression of type 1 and type 2 diabetes (2,27–30). Increased PI misfolding has been recently reported in β-cells with either defects in the ER folding environment (31–33), defective ER export machinery (13), or increased PI synthesis due to insulin resistance in rodent and human type 2 diabetes islets (23). These data suggest that PI misfolding is an early event in the progression to type 2 diabetes. As the most abundant protein synthesized in β-cells, PI is predisposed to misfold (11,14,34) from which it may propagate this misfolding onto bystander PI (1,8,9). This becomes an important driving force for β-cell ER stress and insulin deficiency (32,35–37). The data in this report provide the first evidence pointing to the idea that initial PI-PI B-chain contact may be a site worth targeting, to block cross-dimerization of misfolded PI with PI bystanders, enhance insulin production, and offer a new approach to preventing β-cell insulin deficiency and diabetes (Fig. 7).

Acknowledgments. The authors thank Nelson Phillips at Case Western Reserve University for advice regarding size-exclusion chromatography.

Funding. This work was supported by National Institutes of Health (NIH) (R01-DK-48280, R24-DK-110973, DK-111174, and R01-DK-040949), Natural Science Foundation of China (81830025, 81620108004, and 81570699), and National Key R&D Program of China (2019YFA0802502). We acknowledge support from Tianjin Municipal Science and Technology Commission (17ZXMFSY00150), from the Michigan Diabetes Research Center Morphology Core (NIH P30-DK-020572), and from the Protein Folding Diseases Initiative of the University of Michigan. Human pancreatic islets were provided by the National Institute of Diabetes and Digestive and Kidney Diseases–funded Integrated Islet Distribution Program (IIDP) at City of Hope (NIH 2UC4-DK-098085) and JDRF-funded IIDP Islet Award Initiative. N.R. is a Predoctoral Fellow of the NIH Medical Scientist Training Program at Case Western Reserve University (5T32GM007250-38) supported by an NIH fellowship (1F30-DK-112644).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. J.S., Y.X., X.L., L.H., W.C., S.A.M., L.L., R.M., D.L., A.A., B.D., N.R., N.P.W., P.I.-A., and D.A.O. generated research data. M.A.W., P.I.A., R.J.K., and D.A.O. contributed to discussion and reviewed/edited the manuscript. P.A. and M.L. initiated and designed the research project, reviewed the data, and wrote the manuscript. M.L. and P.A. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.