Distinct states of proinsulin misfolding in MIDY

A precondition for efficient proinsulin export from the endoplasmic reticulum (ER) is that proinsulin meets ER quality control folding requirements, including formation of the Cys(B19)–Cys(A20) “interchain” disulfide bond, facilitating formation of the Cys(B7)–Cys(A7) bridge. The third proinsulin disulfide, Cys(A6)–Cys(A11), is not required for anterograde trafficking, i.e., a “lose-A6/A11” mutant [Cys(A6), Cys(A11) both converted to Ser] is well secreted. Nevertheless, an unpaired Cys(A11) can participate in disulfide mispairings, causing ER retention of proinsulin. Among the many missense mutations causing the syndrome of Mutant INS gene-induced Diabetes of Youth (MIDY), all seem to exhibit perturbed proinsulin disulfide bond formation. Here, we have examined a series of seven MIDY mutants [including G(B8)V, Y(B26)C, L(A16)P, H(B5)D, V(B18)A, R(Cpep + 2)C, E(A4)K], six of which are essentially completely blocked in export from the ER in pancreatic β-cells. Three of these mutants, however, must disrupt the Cys(A6)–Cys(A11) pairing to expose a critical unpaired cysteine thiol perturbation of proinsulin folding and ER export, because when introduced into the proinsulin lose-A6/A11 background, these mutants exhibit native-like disulfide bonding and improved trafficking. This maneuver also ameliorates dominant-negative blockade of export of co-expressed wild-type proinsulin. A growing molecular understanding of proinsulin misfolding may permit allele-specific pharmacological targeting for some MIDY mutants. Supplementary Information The online version contains supplementary material available at 10.1007/s00018-021-03871-1.

Distinct MIDY mutant alleles are associated with diabetes onset ranging from neonatal life through adulthood [2,8]. In the current study, we sought to determine whether Cys residues from the A6/A11 tandem might contribute to proinsuiln failure in MIDY. Here, we provide evidence for two main subsets of misfolded MIDY mutants that we propose are distinguished by their competence in forming the Cys(A6)-Cys(A11) disulfide. In one subset, impaired closure of the Cys(A6)-Cys(A11) bridge leads directly to aberrant intermolecular disulfide-linked complexes, blocking proinsulin export from the ER and dominantly retaining co-expressed WT proinsulin.

Materials, Antibodies, and Human Insulin Immunoassay
See Resources Table. We make special note that the manufacturer indicates that the human insulin STELLUX ELISA assay (ALPCO 80-INSHU-CH01) has a cross-reactivity with human insulin of 100%; human proinsulin undetectable; mouse insulin undetectable; rat insulin-1 0.49%, and rat insulin-2 undetectable.

Plasmids encoding mutant proinsulins
Plasmids encoding myc-tagged human proinsulin, myc-tagged human "lose-A6/A11" proinsulin and hPro-CpepSfGFP (Resources Table) were used as templates for mutagenesis using the QuikChange site-directed mutagenesis kit, with resulting plasmids confirmed by direct DNA sequencing.  Table) were cultured in DMEM supplemented with 10% calf serum and penicillin/streptomycin. Cells were transfected with lipofectamine (Thermo-Fisher Scientific).

SDS-PAGE and Western blotting
At 24 or 48 h post-transfection, cells were washed with PBS and lysed in Laemmli sample buffer or RIPA buffer (10 mM Tris pH 7.4, 150 mM NaCl, 0.1% SDS, 1% NP40, 2 mM EDTA) including protease/phosphatase inhibitor cocktail (Sigma-Aldrich). Samples were resolved by SDS-PAGE in 4-12% Bis-Tris NuPAGE gels under either nonreducing or reducing conditions. Completed nonreducing gels were incubated in 30 mM dithiothreitol (DTT) for 10 min at room temperature prior to electrotransfer to nitrocellulose. Development of immunoblots used enhanced chemiluminescence, captured with a Fotodyne gel imager, quantified using ImageJ.

Live cell Microscopy of GFP-tagged proinsulins
Transfected INS832/13 cells were cultured in LabTek-II coverglass chambers (Nunc). At 24 h posttransfection live cell fluorescence was examined using a Nikon A1 Confocal Microscope. Multiple image/fields were captured for each construct; raw images were processed using NIS-Elements Viewer 5.21 software.

ER stress response as measured by BiP-Luciferase
Cells were co-transfected with plasmids encoding human proinsulin construct, BiP promoter-firefly luciferase reporter, and renilla luciferase, at a DNA ratio of 100:10:1. At 48 h post-transfection, cells were lysed and analyzed for firefly luciferase normalized to renilla luciferase activity (Promega Dual Luciferase assay).

Statistical Analysis
Results were calculated as mean ± s.d.. Statistical analyses employed two-tailed unpaired Student's t-test

Structural Modeling
Spatial relationships in native insulin were examined in the wild-type zinc insulin hexamer (Protein Databank [PBD] entry 4INS, from the insulin hexamer) [22,23]. This structure contains an asymmetric dimer (TT') in the asymmetric unit. Structural elements were visualized using InsightII and Pymol.

Data and Resource Availability
Data generated and analyzed during this study are included in the published article (and its online supplementary files). Additional data generated during the current study are available from the corresponding author upon reasonable request.

The effect of MIDY mutations on proinsulin folding and export
In this study, we have examined MIDY mutants producing autosomal-dominant neonatal diabetes, including H(B5)D [7], a previously unreported G(B8)V, Y(B26)C and L(A16)P [24], and the commonly reported R89C [which has a polypeptide position R(Cpep+2)C] [25], as well as two mutants producing diabetes at somewhat later ages including V(B18)A [26] and E(A4)K [27]. Most MIDY missense mutants fall within the established primary structure of mature human insulin. Fig. 1A highlights the degree of natural sequence variation at each position [28]; Figs. 1B, C show the site of these MIDY residues in the 2-chain insulin structure. Unsurprisingly, MIDY mutations tend to occur at sites of residue conservation, indicating evolutionary intolerance for substitution. Although the C-peptide and its flanking cleavage sites are not shown, we note that R(Cpep+2)C replaces a highly conserved Arg at the C-peptide/A-chain cleavage site with an extra (7 th ) cysteine that, by definition, has no natural disulfide partner.
Each MIDY mutant was engineered into the hPro-CpepMyc cDNA encoding human proinsulin bearing myc-tagged C-peptide [12]. When expressed heterologously in 293T cells, the construct bearing the WT insulin chains was secreted to the medium, in contrast with any of the 7 MIDY mutants ( Fig. 2A; quantified in Fig. 2C). Moreover, WT human proinsulin expression allowed for secretion of co-expressed WT mouse proinsulin (detected with a species-specific proinsulin mAb), whereas each of the MIDY mutants yielded a "bystander effect" [11,29], inhibiting the secretion of co-expressed WT proinsulin (Fig.   2B).
The abundance of expression of WT proinsulin, or other factors ongoing in β-cells, can subvert the efficiency of ER quality control, leading to escape of a subset misfolded proinsulin molecules from the ER [14]. With this in mind, we expressed MIDY proinsulins in Min6 (mouse) β-cells -using the myc-tag to unequivocally distinguish transfected from endogenous proinsulin. WT human proinsulin was abundantly exported in Min6 cells, being secreted as unprocessed proinsulin (Fig. 3A) and being converted intracellularly to human insulin (Fig. 3B). Whereas all MIDY mutants were significantly impaired in export (Fig. 3C), there was demonstrable low-level secretion of several MIDY mutants, notably including proinsulin-V(B18)A (Fig. 3A, C). We could independently confirm some intracellular transport of proinsulin-V(B18)A by detection of processed human insulin in these mouse β-cells (Supplemental Fig.   S1), indicating that V(B18)A was in small part able to reach immature secretory granules for proinsulin-toinsulin conversion [30,31].

MIDY proinsulins engage in intermolecular disulfide mispairing in the ER of β-cells
We recently found that proinsulin in pancreatic β-cells is susceptible to forming intermolecular disulfidelinked complexes, and an increased abundance of these aberrant complexes precedes the onset of diabetes [32]. In lysates of Min6 cells transfected to express myc-tagged human proinsulin or MIDY mutants, we looked for the presence of proinsulin disulfide-linked complexes (resolved by nonreducing SDS-PAGE and detected by human-specific proinsulin immunoblotting).
Reducing SDS-PAGE (in which intermolecular disulfide-linked proinsulin runs as reduced monomers) was used to detect total human proinsulin levels in the cells. By nonreducing SDS-PAGE, whereas WT myc-tagged human proinsulin was predominantly monomeric, the various MIDY mutants formed an increased fraction of disulfidelinked dimers and higher molecular weight complexes (Fig. 5A). By analysis using two different proinsulin antibodies, we were able to calculate a misfolding index indicating increased disulfide-linked complex formation for at least three (and probably all) of the MIDY proinsulin mutants (Supplemental Fig.   S2). Moreover, there was a suggestion of increased ER stress induced by most MIDY mutants [response measured by BiP-luciferase assay, although only L(A16)P generated statistical significance (Supplemental

Impact of cysteines A6 and A11 on MIDY proinsulin folding and export
A proinsulin double mutant known as lose-A6/A11 (bearing Cys-to-Ser substitutions) exhibits essentially normal intracellular trafficking in β-cells; thus, MIDY mutants retained by virtue of an unpaired Cys(A11) [19] might be rescued in a lose-A6/A11 background. We immediately observed significantly rescued When expressed either in pancreatic β-cell lines or in β-cells in vivo, a construct known as hPro-CpepSfGFP (bearing a superfolder-GFP within the C-peptide of human proinsulin) tends to accumulate in the juxtanuclear Golgi region and becomes stored as processed fluorescent C-peptide in mature secretory granules [13,33]. With this in mind, we examined each of the previously noted MIDY mutants in an otherwise WT or lose-A6/A11 proinsulin background in INS832/13 β-cells. In the absence of mutations causing MIDY, both proinsulin and lose-A6/A11 were delivered to punctate secretory granules at the β- In conjunction with this, the latter three MIDY mutants also yielded processed GFP-tagged C-peptide ( Fig.   7B lower panel, green arrows).
[As an aside, the common R(Cpep+2)C mutation mutates the C-A junctional cleavage site, and the cleavage intermediate band generated from this construct exhibited a unique, abnormal mobility, suggesting an atypical cleavage (Fig. 7B).] Additionally, the V(B18)A mutation yielded a hint of processed C-peptide in both the WT and lose-A6/A11 background (Fig. 7B) and in counting multiple transfected cell images, some green fluorescent secretory granules were detected, albeit decreased by > 70%. Taken together, these morphological and biochemical findings support the characterization of two main subgroups of MIDY mutants that can either be rescued or not-significantlybenefitted by the lose-A6/A11 substitutions, respectively, while V(B18)A behaves as a slightly better tolerated MIDY mutation.
As the G(B8)V mutation had not been described previously (we learned of this spontaneous mutation in a New Zealand girl who presented with diabetes at 6 weeks of age), we directly compared newly-

Impact of cysteines A6 and A11 on dominant-negative MIDY behavior
To test dominant-negative blockade of WT proinsulin, we co-expressed in 293T cells each human myctagged MIDY mutant in the lose-A6/A11 background, with untagged WT mouse proinsulin. We could then look independently by Western blotting at the export of the lose-A6/A11 MIDY constructs (anti-myc immunoblotting) and WT proinsulin (mAb anti-rodent proinsulin). We observed negligible secretion of G(B8)V, Y(B26)C, and L(A16)P in the lose-A6/A11 background ( Fig. 8A; quantified in Fig. 8C), accompanied by suppressed escape of co-expressed WT proinsulin (Fig. 8B). In contrast, we detected significant secretion of H(B5)D, R(Cpep+2)C, and E(A4)K in the lose-A6/A11 background ( Fig. 8A; quantified in Fig. 8C) accompanied by secretion of co-expressed WT proinsulin (Fig. 8B). These data suggest that the same proinsulin mutants that are themselves less susceptible to ER quality control are also less effective in trans-dominant blockade of co-expressed WT proinsulin. Indeed, when comparing to Figure 2, none of the lose-A6/A11 MIDY constructs except for L(A16)P are as efficient in blocking coexpressed WT proinsulin, including V(B18)A which itself appears blocked yet cannot efficiently block the trafficking of co-expressed WT proinsulin. Thus the data highlight the potential importance of unpaired Cys(A11) and Cys(A6) in adverse proinsulin folding and trafficking.

DISCUSSION
MIDY is disease of protein misfolding causing autosomal dominant diabetes. MIDY mutations threaten native proinsulin disulfide bond formation [12,19]. ER quality control recognizes and prevents export of misfolded proinsulin, thus MIDY mutants have loss of function in generating insulin [4,5]. Moreover, natural self-association properties of proinsulin enable the defective MIDY gene product to associate with and dominantly block export of WT proinsulin [10,11,34]. Nevertheless, different MIDY mutations lead to different ages of diabetes onset, suggesting greater or lesser severity of proinsulin misfolding [4,35]. In this study, we've examined seven MIDY proinsulin mutants by testing for the importance of Cys(A6) and Cys(A11) because, curiously, these residues normally engage in an intramolecular disulfide bond that is nonessential for proinsulin trafficking [18,19] yet in MIDY they may participate in intermolecular disulfide bonds that could render mutant proinsulin molecules unacceptable to ER quality control.

Proinsulin-V(B18)A is linked to diabetes onset beyond the neonatal period [ages 3.3 -10.5 years [26]].
Although interchange of Val-to-Ala is a conservative substitution that is generally well-tolerated [36], protein context is important in determining pathogenicity, as some similar missense substitutions are linked to disease [37]. Equally important is cellular context: we did not detect anterograde trafficking of proinsulin-V(B18)A from 293T cells even in the lose-A6/A11 background (Fig. 8), yet low-level anterograde trafficking was observable in pancreatic β-cells (Fig. 3), even leading to the formation of small quantities of mature human insulin (Supplemental Fig. S1). These data suggest that the V(B18)A mutant exhibits a slightly less severe proinsulin misfolding than a number of other MIDY mutants.

Defective Cys(A6)-Cys(A11) pairing defines two subgroups of mutations
The MIDY mutants G(B8)V, Y(B26)C, and L(A16)P exhibited significantly less anterograde trafficking than WT proinsulin, and even less than the MIDY mutant proinsulin-V(B18)A. The trafficking defect of the newly-described proinsulin-G(B8)V is severe and very similar to that of the G(B8)S mutation (Supplemental Fig. S6), which has been recently described to account for spontaneous diabetes in the KINGS diabetic mouse [38], as well as being one of the first described human MIDY mutants [25] -its folding defect is considered further, below. Additionally, it might be predicted that proinsulin-Y(B26)C and L(A16)P (also discussed further, below) would be catastrophically detrimental to folding as they (in addition to position B8) fall among the more conserved residues within insulin (Fig. 1A) -and patients bearing these heterozygous mutations have been found to be insulin-requiring within the first 6 months of life [24]. In 293T cells, these mutants could not be secreted either in the proinsulin or lose-A6/A11 background (Figs. 2, 8); moreover, secretion was profoundly low in pancreatic β-cells (Fig. 3), and could not be statistically increased in the lose-A6/A11 background in either Min6 or INS832/13 cells (Figs. 6, 8).
In contrast with the foregoing proinsulin mutants, we find that MIDY mutations encoding H(B5)D, R(Cpep+2)C, or E(A4)K fall into a distinct group. Specifically, whereas these mutants cannot be secreted in the proinsulin backgound, they are all exported to a greater degree in the lose-A6/A11 backgroundable to be secreted (Fig. 6) and to be processed to mature C-peptide in pancreatic β-cells (Fig. 7).
Recovery of export in the lose-A6/A11 background can be accounted for by diminished formation of intermolecular disulfide-linked complexes (Fig. 5). As this particular subgroup of MIDY mutants develops improved folding in the lose-A6/A11 background, leading to improved trafficking in β-cells, we conclude that for this subset of MIDY proinsulin mutants, burying these cysteines in the Cys(A6)-Cys(A11) disulfide bond makes them unavailable for intermolecular disulfide mispairing [19]. Rescue in the lose-A6/A11 background then allows unperturbed initial formation of the Cys(B19)-Cys(A20) disulfide, leaving driving the remaining Cys(B7) and Cys(A7) free thiol groups into disulfide partnership [19], eliminating or diminishing the predisposition for intermolecular disulfide bond formation, and that is one of the major factors in ER quality control of proinsulin export.
Additionally, in the solution structure of proinsulin, R(Cpep+2) and E(A4) appear to form a salt bridge and may contribute to the 'CA knuckle' [42], which may also help to stabilize the proximal A-chain α-helix [43]. In turn, there is a close relationship between proximal A-chain conformation and the Cys(A6)-Cys(A11) disulfide bond [44]. (Additionally it seems likely that the extra unpaired cysteine of R(Cpep+2)C might also engage in intermolecular disulfide crosslinking -for which further study is still and relaxins, lacks a side chain but exhibits a critical positive ϕ dihedral angle ordinarily forbidden to Lamino acids [23]. From in vitro folding studies, any L-amino-acid substitution in this position predisposes to impaired disulfide pairing [45] as observed for the G(B8)S MIDY mutant [46], triggering permanent neonatal diabetes [25,38]. We posit that G(B8)V similarly enforces an inappropriate negative B8 ϕ dihedral angle, in turn misorienting the side chain of Cys(B7), which cannot be rescued by the lose-A6/A11 template. The side chain of Y(B26) is inserted into a solvent-exposed inter-chain crevice (Supplemental Fig. S7B), and its packing against the conserved nonpolar surface of the central-B-chain αhelix bearing L(B11), V(B12) and L(B15) stabilizes a nascent native-like supersecondary structural "Uturn" [23] to promote proper formation of the essential Cys(B19)-Cys(A20) disulfide bond [9,47], which is likely to be disrupted in the Y(B26)C MIDY mutant -and this too cannot be rescued in the lose-A6/A11 background. L(A16) fits snuggly within a potential internal cavity at the confluence of insulin's three native α-helices (Supplemental Fig. S7C), and the A16-A19 α-helical turn is thought to facilitate alignment with the B-chain α-helix to enable initial Cys(B19)-Cys(A20) pairing [47]. The imino-side chain of P(A16) is expected to distort local α-helical conformation [48], incurring a local steric clash and destabilizing the internal cavity, leading to an insuperable block to formation of the Cys(B19)-Cys(A20) pairing within proinsulin's folding nucleus [9].
In a slightly milder category from the others, V(B18)A would create a packing defect (Supplemental Fig.   S7D) that would be expected to enlarge the range of possible positions/orientations accessible to the adjoining Cys(B19) and its Cys(A20) partner. Evidently, this mutation does not completely preclude Cys(B19)-Cys(A20) pairing (Fig. 3C, Supplemental Fig. S1), with slight secretory improvement by the lose-A6/A11 template ( Fig. 6A-C; Supplemental Fig. S5).
Finally, we note that the lose-A6/A11 background resulted in notable diminution of dominant-negative behavior, particularly for the subgroup of MIDY mutants that are themselves rescued in folding and intracellular transport [a modest effect was also observed for Val(B18)A]. These data strongly support the hypothesis that it is the recognition and retention of the misfolded mutant partner by ER quality control that confers ER retention upon the co-expressed WT partner [5]. Indeed there is a definite proinsulinproinsulin partnership, as indicated by the importance of proinsulin dimerization propensity to dominantnegative behavior [34]. Altogether, the findings in this report highlight an improved structural understanding by which proinsulin misfolding might be attacked pharmacologically, a) to block exposure of unpaired cysteines, and b) to limit the increased aberrant proinsulin-proinsulin associations [49] which are a hallmark of MIDY and perhaps also of type 2 diabetes [32].  shown in gold, and sites of mutation within the insulin moiety of proinsulin considered in this study denoted (the C a carbon of Gly B8 is shown as a red ball).      Figure).