Impaired Cleavage of Preproinsulin Signal Peptide Linked to Autosomal-Dominant Diabetes

Guinea pig anti-porcine insulin, rat insulin (#RI-13 K), and human insulin-specific (#HI-14 K) and proinsulin-specific radioimmunoassay (#HPI-15 K) were from Millipore; rabbit anti-Myc and anti-GFP were from Immunology Consultants Laboratories; Zysorbin was from Zymed; ³⁵S-amino acid mixture (Met+Cys) and pure ³⁵S-Met were from ICN; dithiothreitol (DTT), protein A-agarose, digitonin, N-ethylmaleimide (NEM), and radioimmunoassay (RIA)-grade bovine serum albumin were from Sigma-Aldrich; Met/Cys-deficient Dulbecco’s modified Eagle’s medium (DMEM) and all other tissue culture reagents were from Invitrogen. Plasmids encoding human Ero1α and Ero1β were from Dr. P. Scherer (University of Texas Southwestern), and mouse protein disulfide isomerase (PDI) was from Dr. M. Green (St. Louis University).

The human INS cDNA and mouse Ins2 cDNA with or without Myc tag as described (21) were subcloned into pTarget and pCMS-GFP vectors. Mutations were introduced using the QuikChange site-directed mutagenesis kit (Stratagene) and confirmed by direct DNA sequencing.

INS1 (rat) cells, Min6 (mouse) cells, or 293T (human) cells were plated onto 12-well plates 1 day before transfection with Lipofectamine (Invitrogen) using 1 to 2 μg plasmid DNA. At 48 h after transfection, cells were pulse-labeled with ³⁵S-Met/Cys or pure ³⁵S-Met and chased as indicated. Cells used for analysis of (pre)proinsulin oxidative folding were preincubated with 20 mmol/L NEM in PBS on ice for 10 min before lysis. Lysis buffer (1% Triton X-100, 0.1% SDS, Tris, pH 7.4) contained 2 mmol/L NEM and a proteinase inhibitor cocktail. For immunoprecipitation, samples were precleared with Zysorbin and immunoprecipitated as described previously (21). Nonreducing or reducing Tris-Tricine urea–SDS-PAGE was performed as described (18). For Western blotting, 20 μg total lysate protein was boiled in SDS sample buffer with or without 100 mmol/L DTT, resolved by 4–12% NuPage, electrotransferred to nitrocellulose, and blotted with rabbit anti-Myc followed by anti-rabbit–HRP conjugate, with development by enhanced chemiluminescence. The BiP promoter-driven firefly luciferase assay for Min6 cells 48 h posttransfection to express human preproinsulin-WT or mutants (normalized to Renilla luciferase activity) was performed as described (21) using the dual-luciferase reporter kit (Promega).

After labeling with ³⁵S-Cys/Met, 293T cells transfected with human preproinsulin-WT or mutant were washed, resuspended, and incubated in 50 μl of 150 mmol/L NaCl, 2 mmol/L CaCl₂, 50 mmol/L HEPES, pH 7.5, ± 0.01% digitonin on ice for 10 min. The cells were spun at 14,000 rpm for 10 min at 4°C; the supernate was transferred to a new tube containing 450 μL lysis buffer and the pellet lysed in 500 μL lysis buffer. Both supernate and pellet were immunoprecipitated with anti-insulin or anti-GFP and analyzed using 4–12% NuPage.

293T cells were cotransfected with human preproinsulin-WT and either mouse preproinsulin-WT or mutants, plus empty vector or a vector expressing PDI or ER–oxidoreductin-1 (ERO1), at a DNA molar ratio of 1:2:3. At 40 h posttransfection, medium-bathing transfected cells were collected for an additional 16 h and used to measure secreted human proinsulin using the human proinsulin-specific RIA.

Min6 cells were transfected with vectors expressing human preproinsulin-WT or mutants. At 40 h posttransfection, cells were incubated with DMEM containing 25.5 mmol/L glucose and 10% FBS for 16 h. The media were then collected and cells lysed with acid/ethanol (20) and human insulin measured by human insulin-specific RIA. In parallel, duplicate wells of each transfection were used for isolating total RNA. Human insulin mRNA levels were determined by RT-PCR using forward primer: 5′-GAACCAACACCTGTGCGGCTCAC-3′ and reverse primer: 5′-CGAGGCTTCTTCTACACACCCA-3′. In this way, human insulin protein level was normalized to human insulin mRNA level.

Statistical analyses were carried out by ANOVA followed by Bonferroni multiple comparison test using GraphPad Prism 5. Data are presented as means ± SD. A P value of <0.05 was taken as statistically significant.

Evolutionarily, the preproinsulin SP diverges more in sequence than does the insulin B-chain or A-chain (28). In tetrapods, residue A(SP24) predominates at the cleavage site (residue “−1”), whereas there is considerable natural variation at the penultimate “−2” position (Fig. 1A). In chimpanzee, S(SP23) predominates (Fig. 1A), and there is a polymorphic variation in this position even within the human genome (29). Thus, it was of interest to further investigate the physiological significance of both A(SP23)S and A(SP24)D substitutions.

In MIN6 pancreatic β-cells transfected to express Myc-tagged human preproinsulin [preProCpepMyc (21)], A(SP24)D was predominantly uncleaved, with a minor cleaved band migrating near ProCpepMyc-WT (Fig. 1B). Impaired efficiency of SP cleavage (Fig. 1B) can directly account for decreased human insulin production (Fig. 1C). By contrast, A(SP23)S showed no defect of SP cleavage or human insulin production (Fig. 1B and C). Moreover, A(SP24)D induced ER stress in Min6 cells, whereas the ER stress response for A(SP23)S was similar to that for preproinsulin-WT (Fig. 1D). In 293T cells, newly synthesized A(SP23)S was rapidly processed to proinsulin and secreted, exactly like WT, whereas A(SP24)D showed little preproinsulin processing or secretion (Fig. 2A). From multiple such experiments, ≤10% of A(SP24)D underwent SP cleavage (Fig. 2B). We examined the subsequent fate of newly synthesized mutant preproinsulin for up to 4 h of chase. Although secretion of either uncleaved or SP-cleaved A(SP24)D was negligible, there was intracellular loss of A(SP24)D, some of which was clearly blocked by the proteasome inhibitor MG132 (Fig. 2C). Although the data suggest ER-associated degradation (ERAD) of A(SP24)D at steady state, A(SP24)D showed substantial accumulation in β-cells (Fig. 1B). Thus, A(SP23)S exhibits no defect of SP cleavage or insulin production, whereas A(SP24)D exhibits impaired SP cleavage, intracellular retention, at least partial ERAD, activation of ER stress response, and blocked insulin production.

Conceivably, failed SP cleavage could reflect failure of preproinsulin translocation across the ER membrane (30). We used digitonin to partially permeabilize the plasma membrane while keeping organelle membranes intact (31) in cells coexpressing recombinant preproinsulin and cytosolic green fluorescent protein (GFP); upon permeabilization and centrifugation, >50% of cytosolic GFP was relocated from pellet to supernate, whereas both WT and A(SP24)D remained exclusively in the pellet (Fig. 3A). Thus, A(SP24)D is not free in the cytosol, implying its association with the ER. It is known that the preproinsulin SP plus the first 14 residues of proinsulin B-chain are sufficient to direct the downstream translation product into the ER lumen (32). To confirm that this is the case for A(SP24)D, we engineered the A(Cpep18)N substitution into the C-peptide (Fig. 3B, top) to create an artificial N-linked glycosylation site that can be used to establish that at least 76 residues of nascent preproinsulin polypeptide have accessed the luminal side of the ER. When combined with the fact that human proinsulin has no methionine residues (i.e., pure ³⁵S-Met can label only preproinsulin; ³⁵S-proinsulin labeling requires ³⁵S-Cys), we found that essentially all molecules of both proinsulin-WT (³⁵S-Cys/Met) and A(SP24)D (either ³⁵S-Met or ³⁵S-Cys/Met) became available in the ER lumen, as established by N-linked glycosylation that could be deglycosylated with PNGase F (Fig. 3B). Thus, essentially all molecules of A(SP24)D undergo ER translocation.

To our knowledge, no studies have previously examined the relationship between SP cleavage and oxidative folding of preproinsulin within the ER. We used nonreducing Tris-Tricine urea–SDS-PAGE (16) to examine disulfide bond formation in recombinant newly synthesized preproinsulin. Uncleaved preproinsulin-WT (labeled with ³⁵S-Met) appears fully reduced (Fig. 4A, lane 2). Upon SP removal, proinsulin-WT (labeled with ³⁵S-Cys/Met) may begin as a fully reduced species (Fig. 4A, lane 1, arrow) but proceeds to native and isomeric oxidized proinsulin (lower bands). Within 10 min, reduced (pre)proinsulin nearly vanished, and oxidized native proinsulin became enriched (Fig. 4B, lane 2), indicating both rapid SP cleavage and downstream proinsulin folding.

For A(SP24)D, the sequence of early events appeared altered: newly synthesized preproinsulin-A(SP24)D (³⁵S-Met–labeled) started to become oxidized (Fig. 4A, lanes 7 and 8), and this increased after 10 min (Fig. 4B, lane 8). By comparing the recovery of fully reduced preproinsulin-A(SP24)D by nonreducing gel versus total preproinsulin recovered under reducing conditions, we calculated (from multiple experiments) that 42 ± 2.7% of preproinsulin-A(SP24)D had oxidized (either properly or improperly, see below) at the zero chase time. The oxidized fraction of preproinsulin-A(SP24)D was 61% by 10 min of chase; 86% by 30 min of chase; and 90% under steady-state conditions by Western blotting (Fig. 4C, right). By contrast, proinsulin-WT was >90% oxidized at the zero chase time, and there was essentially no fully reduced proinsulin-WT remaining after 10 min (Fig. 4B). Under steady-state conditions, ≤10% oxidized A(SP24)D remains as monomer recovered under nonreducing conditions (Fig. 4C, compare lanes 3 and 4), indicating that the majority forms disulfide-linked protein complexes. Most importantly, in β-cells, A(SP24)D coimmunoprecipitated coexpressed proinsulin-WT, which was detected only under reducing conditions; thus, this proinsulin-WT must also have been recruited into disulfide-linked complexes (Fig. 4D). Altogether, the data in Figs. 3 and 4 establish that impaired SP cleavage triggers defective oxidative folding within the ER, leaving exposed thiols on preproinsulin that could create cellular problems by participating in inappropriate intermolecular thiol attack.

Using the SignalP-3.0 computer-based algorithm, we noted that whereas human preproinsulin-WT uses the AAA-proinsulin sequence to generate proinsulin beginning with the appropriate F(B1) residue, the A(SP24)D sequence predicts SP cleavage at low efficiency, generating a minor portion of AD-Proins that includes two extra residues preceding the proinsulin B-chain (Fig. 5A). We therefore designed three additional preproinsulin mutants (AAAAD-Proins, AAAMD-Proins, and AMAD-Proins) intended to interrogate the location of the preproinsulin cleavage site while finding a high-efficiency means to generate the aberrant AD-Proins cleavage product (Fig. 5A). By SDS-PAGE/autoradiography after metabolic labeling with ³⁵S-Cys/Met, unlike the inefficient SP cleavage of A(SP24)D (Fig. 5B, lanes 3 and 4), other bioengineered preproinsulins underwent efficient (but aberrant) SP cleavage. In Fig. 5B, a comparison of lanes 2, 4, 6, 8,and 10 (each including ³⁵S-Cys label) showed the presence of a cleaved band in every instance. However, when comparing lanes 3, 5,and 7 (Fig. 5B) labeled with pure ³⁵S-Met, the cleaved band became undetectable except in lane 7 (AAAMD-Proins), when methionine was still found to be contained within the product that must have been cleaved upstream of the Met residue. By contrast, in lane 9 (AMAD-Proins; Fig. 5B) the cleaved product had lost the Met residue (i.e., a cleavage site less than three residues upstream from the normal site but containing at least two residues from the normal site). These observations precisely match the cleavage sites predicted. Regardless of whether the cleavage reaction occurred at low efficiency or high efficiency, the aberrant proinsulins bearing two extra NH₂-terminal amino acids in every instance failed to be secreted normally (Fig. 5C). Thus, fidelity of SP cleavage, exposing the critical proinsulin NH₂-terminal arm, including phenylalanine-B1 (33), is crucial to normal proinsulin folding and export from the ER.

Because A(SP24)D forms aberrant disulfide-linked protein complexes (Fig. 4C) that include coexpressed proinsulin-WT (Fig. 4D), we made A(SP24)D-DelCys (in which all six cysteines were mutated) to test for dominant-negative behavior. Interestingly, most of the blockade of WT proinsulin secretion was lost when thiols were eliminated from the A(SP24)D mutant (Fig. 6A). These data support the concept of thiol attack as a mechanism of dominant-negative behavior of A(SP24)D (i.e., precisely the danger mechanism that has been proposed for MIDY) (21,25).

Heterozygous patients bearing INS gene mutations that cause decreased insulin biosynthesis from the mutant allele do not develop diabetes (5). Thus, decreased insulin production from A(SP24)D itself cannot account for diabetes in heterozygous patients. To explore this further, individually epitope-tagged and preproinsulin partners were coexpressed in β-cells. A preProCpepMyc construct, either bearing or not bearing the A(SP24)D mutation, was coexpressed in INS1 cells along with preProCpepGFP, a reporter of WT proinsulin trafficking and processing (20,24). Although new synthesis of preProCpepGFP was unaffected by coexpression of A(SP24)D, the production of mature human insulin, as measured by accumulation of the processed CpepGFP over 20 h of continuous metabolic labeling, was clearly inhibited upon coexpression of the mutant (Fig. 6B). These data link mutant A(SP24)D expression to insulin deficiency derived from the WT proinsulin allele.

Whereas the AAAAD-Proins construct exclusively forms miscleaved AD-Proinsulin (Fig. 5), we also engineered a preproinsulin (PFD-Proins, Fig. 5A) that cannot be cleaved at all by signal peptidase as measured either by pulse-labeling (Fig. 5B, lane 11) or at steady state (Fig. 6C). Individually, these two constructs should be able to test the relative dominant-negative contributions derived from the actual A(SP24)D mutant which, after 1–4 h chase, can be detected as both forms (Fig. 2C).

First, to test whether miscleaved AD-Proins was a potent dominant-negative, we coexpressed progressively decreasing doses of AAAAD-Proins with bystander preProCpepMyc-WT to emulate the small quantities of AD-Proinsulin typically synthesized from A(SP24)D. At all expression levels, untagged AD-Proinsulin itself was impaired in the secretory pathway, but only at the highest expression levels did AD-Proinsulin inhibit export of coexpressed bystander ProCpepMyc-WT (Fig. 6D, top). However, at low expression levels [all of which are higher than that generated from A(SP24)D], AD-Proinsulin could no longer block secretion of ProCpepMyc-WT (Fig. 6D, bottom).

Next, we found that newly synthesized PFD-Proins also could not be secreted; but more importantly, uncleaved PFD-Proins efficiently blocked in trans the ER export of coexpressed proinsulin-WT (Fig. 6E). Thus, PFD-Proins completely emulated the dominant-negative effects of A(SP24)D. Taken together, the data in Fig. 6 strongly suggest that it is the improperly oxidized full-length preproinsulin-A(SP24)D, and not the miscleaved AD-Proins product derived from A(SP24)D, that is primarily responsible for dominant-negative blockade of coexpressed proinsulin-WT in patients with MIDY.

We have previously demonstrated that MIDY mutants selectively interfere with the oxidative folding to the native state of innocent bystander proinsulin-WT molecules, and promote their ERAD (21). We reasoned that conditions accelerating native disulfide bond formation in proinsulin-WT might help these molecules escape blockade caused by A(SP24)D (and other MIDY mutants). Recent reports indicate that members of the ERO1 family promote formation of proinsulin disulfide bonds (34,35). We therefore explored whether overexpression of ERO1 [or PDI (36)] could help to rescue human proinsulin-WT from blockade by either mouse Akita proinsulin [another MIDY mutant (20,21)] or mouse preproinsulin-A(SP24)D. Using a human-proinsulin specific RIA, we found that increased expression of ERO1 (either α or β), but not PDI, showed rescue activity for export of proinsulin-WT (Fig. 7A). To confirm this rescue effect in β-cells, we coexpressed human preproinsulin-WT with mouse preproinsulin-A(SP24)D, which lowered the human insulin content of transfected INS1 cells by >70% (Fig. 7B). However, in cells with simultaneously increased expression of ERO1α or ERO1β, human insulin content doubled (Fig. 7B). To independently validate these RIA results, we performed pulse-chase experiments like those shown in Fig. 6E. Once again, A(SP24)D blocked in trans the export of coexpressed proinsulin-WT (Fig. 7C, lane 7). However, in cells with increased expression of ERO1α, the behavior of A(SP24)D itself was unaffected, whereas export of coexpressed proinsulin-WT was partially rescued (Fig. 7C, lane 13). Similar results were obtained upon increased expression of ERO1β (not shown).

To explore further the mechanism of ERO1-mediated rescue of proinsulin-WT, we examined the recovery of the native disulfide isoform of newly synthesized proinsulin-WT relative to total proinsulin-WT recovery under reducing conditions (21). In the presence of A(SP24)D, the oxidative maturation of proinsulin-WT to the native state was significantly decreased (Fig. 8A, open bars). However, upon increased expression of ERO1β, the recovery of native disulfide-bonded proinsulin-WT was restored despite the presence of coexpressed A(SP24)D (Fig. 8A, far right black bar). Moreover, final recovery of proinsulin-WT was greatly increased in cells with increased ERO1β expression (Fig. 8B). Thus, despite presence of a coexpressed MIDY mutant, in cells with increased expression of ERO1, coexpressed proinsulin-WT exhibited improved folding and decreased ERAD (Fig. 8), explaining its secretion rescue (Fig. 7).

Point mutations adjacent to the SP cleavage site of preproteins may impair SP cleavage (37–39) and downstream folding (40) and can cause ER stress and cytotoxicity (41). In this report, we have investigated the biological behaviors of preproinsulin-A(SP23)S and A(SP24)D. These mutants provide a window into understanding the earliest steps of insulin biosynthesis. Unfortunately for patients expressing A(SP24)D, a series of pathological events go well beyond the defect in SP cleavage with failure to provide the critical free NH₂-terminal phenylalanine-B1 (33), impairing oxidation of the natural insulin disulfide bonds and creating a miscleaved SP, resulting in the ER retention of both forms of the mutant protein, blocking in trans the coexpressed proinsulin-WT from the normal INS gene allele and thereby profoundly inhibiting β-cell insulin production as well as induction of ER stress. On the positive side, we have developed a therapeutic maneuver that may minimize the dominant-negative effects of MIDY mutants.

All available evidence suggests that A(SP23)S is cleaved with normal efficiency (Fig. 2A and B) and fidelity to generate authentic proinsulin (Fig. 1B) and insulin (Fig. 1C) with no net ER stress beyond that seen for preproinsulin-WT (Fig. 1D). As serine is the native residue in this position in chimpanzee (Fig. 1A) and is the site of polymorphic variation in humans (29), we propose that in spite of our earlier report (12), impaired preproinsulin SP cleavage and proinsulin misfolding are not operative for this particular variant. We cannot exclude that A(SP23)S might cause diabetes through a different mechanism. By contrast, although A(SP24)D is successfully translocated across the ER membrane (Fig. 3), it is inefficiently processed at an inappropriate SP cleavage site, generating a small subpopulation of miscleaved AD-Proins molecules (Figs. 1, 2, and 5–7). Both uncleaved and miscleaved A(SP24)D are retained intracellularly (Fig. 2A) and undergo ERAD (Fig. 2C), suggesting that they are misfolded. Nevertheless, at steady state, a major subpopulation of uncleaved A(SP24)D (along with the smaller subpopulation of SP cleaved) molecules accumulates in the ER (Fig. 1B), activating an ER stress response (Fig. 1D).

The half-life of preproinsulin-WT is exceedingly short, making it technically difficult to follow uncleaved preproinsulin in live cells (Fig. 4B). Cleavage by signal peptidase generally requires that 60–80 amino acids of the nascent polypeptide are cotranslationally translocated (42), which represents most of the preproinsulin molecule. What the A(SP24)D demonstrates is that preproinsulin Cys residues can be exposed to the oxidative environment of the ER even before SP cleavage (Fig. 4). To our knowledge, this is the first study to directly examine oxidative folding of preproinsulin in the ER. Normally, uncleaved preproinsulin-WT is fully reduced, and, upon SP removal, proinsulin begins to quickly form both native and isomeric disulfide bonds (Fig. 4A and B). However, under conditions in which normal SP cleavage is interrupted [e.g., in A(SP24)D], uncleaved preproinsulin exhibits delayed and abnormal oxidative folding (Fig. 4A and B), causing the formation of disulfide-linked complexes (Fig. 4C), which also entraps coexpressed proinsulin-WT (Fig. 4D).

We have previously shown that acute ER stress is not itself sufficient to block export of coexpressed WT INS gene product (21). Moreover, recent studies indicate that in MIDY, insulin deficiency precedes a net loss of β-cells (22,23). In this report, we demonstrate that A(SP24)D blocks the export of coexpressed proinsulin-WT (Fig. 6) and thereby decreases insulin production (Figs. 6A and 7B, open bar), which may be sufficient to account for initiation of insulin deficiency in patients with MIDY. The evidence suggests that it is primarily the uncleaved preproinsulin subfraction rather than the minor misfolded AD-Proins subfraction (Fig. 6Dand E) that drives the dominant-negative insulin secretory behavior in patients bearing A(SP24)D.

Studies in humans suggest that one fully functional INS gene allele may be sufficient to maintain normoglycemia (5). We therefore reasoned that rescuing proinsulin-WT from dominant-negative blockade might restore insulin production to rescue from diabetes in MIDY. With this in mind, we have made a first attempt at a molecular therapeutic approach to rescue proinsulin-WT export in the presence of A(SP24)D; specifically searching for ways, without repairing the misfolding of MIDY mutants, that might enhance proinsulin-WT folding and ER export in order to limit its participation in aberrant protein complex formation in the ER (20). Formation of proinsulin’s three disulfide bonds may occur via transfer of reducing equivalents to ER oxidoreductases and from there to the cytosol via ERO1; recently, Zito et al. (34) established that the diabetic phenotype of Akita mice is exacerbated upon loss of ERO1β. Conceivably, increased ERO1 or PDI expression might promote oxidative folding of proinsulin-WT. However, whereas increased PDI expression has been found to diminish glucose-stimulated insulin secretion and induce ER stress with accumulation of proinsulin in the ER (36), increased ERO1 expression promotes proinsulin-WT export through the secretory pathway (Fig. 7A and C), increasing insulin production (Fig. 7B). The simplest explanation for the increased insulin production is enhancement of oxidative folding of proinsulin-WT (Fig. 8A), decreasing its participation in disulfide-linked complexes (Fig. 8A), decreasing its ERAD (Fig. 8B), and increasing export of proinsulin-WT from the ER (Fig. 7).

Little progress has been made in the development of treatments for ER stress-related pancreatic β-cell failure. The ideas developed in this report suggest that enhancing the oxidative folding and export of proinsulin-WT may be a feasible way to restore insulin production in heterozygous MIDY patients. Several laboratories around the world have been working on small molecules that can manipulate the level and/or activity of ERO1β, and this may be a useful way forward. Moreover, it will be important to further explore the ratio of misfolded to total proinsulin forms, to understand the threshold of misfolding that is needed to trigger onset of insulin-deficiency and diabetes (23), and the means to raise that threshold by protecting bystander proinsulin export for insulin production and glucose homeostasis.

This work was supported primarily by National Institutes of Health (NIH) grants R01-DK-48280 (to P.A.) and R01-DK-088856 and by March of Dimes 6-FY11-357 (to M.L.) as well as National Nature Science Foundation of China Grant 81070629 (to M.L.). We also acknowledge assistance from the Molecular Biology and DNA Sequencing Core of the NIH-funded Diabetes Research and Training Center (P60-DK-20572).

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

M.L., R.L.-L., J.W., L.H., H.G., and D.L. researched data. M.L., S.-o.S., J.W., L.H., F.B., and P.A. contributed to discussion. M.L. and P.A. wrote the manuscript. M.L., J.W., F.B., and P.A. reviewed and edited the manuscript. P.A. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors thank Bill and Dee Brehm for helping to create the Brehm Center for Diabetes Research to advance diabetes research at the University of Michigan.