Matrix Metalloproteinase-9 Protects Islets from Amyloid-induced Toxicity*

Background: MMP-9 cleaves hIAPP, the major constituent of islet amyloid deposits. Results: MMP-9-cleaved hIAPP fragments are largely non-amyloidogenic and non-cytotoxic. MMP-9 overexpression in amyloid-prone islets reduces amyloid deposition and the associated toxicity. Conclusion: MMP-9 protects islets from amyloid-induced toxicity. Significance: MMP-9 activation specifically in islets might be a novel strategy to limit β-cell loss in type 2 diabetes. Deposition of human islet amyloid polypeptide (hIAPP, also known as amylin) as islet amyloid is a characteristic feature of the pancreas in type 2 diabetes, contributing to increased β-cell apoptosis and reduced β-cell mass. Matrix metalloproteinase-9 (MMP-9) is active in islets and cleaves hIAPP. We investigated whether hIAPP fragments arising from MMP-9 cleavage retain the potential to aggregate and cause toxicity, and whether overexpressing MMP-9 in amyloid-prone islets reduces amyloid burden and the resulting β-cell toxicity. Synthetic hIAPP was incubated with MMP-9 and the major hIAPP fragments observed by MS comprised residues 1–15, 1–25, 16–37, 16–25, and 26–37. The fragments 1–15, 1–25, and 26–37 did not form amyloid fibrils in vitro and they were not cytotoxic when incubated with β cells. Mixtures of these fragments with full-length hIAPP did not modulate the kinetics of fibril formation by full-length hIAPP. In contrast, the 16–37 fragment formed fibrils more rapidly than full-length hIAPP but was less cytotoxic. Co-incubation of MMP-9 and fragment 16–37 ablated amyloidogenicity, suggesting that MMP-9 cleaves hIAPP 16–37 into non-amyloidogenic fragments. Consistent with MMP-9 cleavage resulting in largely non-amyloidogenic degradation products, adenoviral overexpression of MMP-9 in amyloid-prone islets reduced amyloid deposition and β-cell apoptosis. These findings suggest that increasing islet MMP-9 activity might be a strategy to limit β-cell loss in type 2 diabetes.

Amyloid deposits are found in the islets of most patients with type 2 diabetes (1), with the degree of deposition being inversely associated with both ␤-cell mass and function (2,3). Further, amyloid deposition following islet transplantation is associated with graft failure (4 -6). The 37 amino acid peptide constituent of these amyloid deposits, human islet amyloid polypeptide (hIAPP, 5 also known as amylin) is co-secreted with insulin by the ␤ cell (7) and has been shown to be toxic to ␤ cells in culture (8). Thus, reducing amyloid burden to improve ␤-cell survival has been the focus of a number of studies using different approaches.
Several natural products and small molecule inhibitors of hIAPP-derived amyloid formation have been investigated, as have protein and polypeptide-based inhibitors (reviewed in Ref. 9). However, the therapeutic potential of many of these molecules has not been delineated as some of these studies were carried out in cell-free systems. In addition, many of these compounds are not drug-like; they are chemically labile polyphenols, compounds which are expected to have poor bioavailability or large polypeptides, some of which contain non-genetically coded amino acids. A second approach to limit the deleterious consequences of amyloid deposition is to exploit enzymes that are capable of degrading hIAPP, such as neprilysin and insulin-degrading enzyme (IDE). We and others have shown that up-regulation of neprilysin in hIAPP transgenic mouse islets reduces amyloid deposition and ␤-cell apoptosis, while its inhibition increases amyloid deposition and ␤-cell loss (10,11). Similarly, inhibition of IDE results in increased amyloidogenesis in ␤-cell lines and reduced cell viability when cells are incubated with synthetic hIAPP (12).
Matrix metalloproteinases (MMPs) are proteolytic enzymes that are involved in the breakdown of extracellular matrix proteins (13). We recently demonstrated that two members of the MMP family, MMP-2 and MMP-9 (also known as gelatinase A and B) cleave hIAPP (14). Inhibition of MMP-9, but not MMP-2, increased amyloid deposition and ␤-cell apoptosis in mouse islets, suggesting that MMP-9 plays a physiological role in limiting islet amyloid deposition (14). Further, MMP-9 expression was found to be reduced in islets of subjects with type 2 diabetes (14), suggesting that deficiency of this protease in islets could predispose toward amyloid deposition in humans.
hIAPP is secreted from the ␤ cell and aggregates to form amyloid extracellularly (15). As MMP-9 is localized to the extracellular space (16), this makes it a prime candidate to degrade hIAPP in vivo. However, sequence differences exist between species that render some forms of IAPP incapable of aggregating. In particular hIAPP, but not rodent IAPP (rIAPP), is amyloidogenic and cytotoxic (17). These two polypeptides differ at six residues, five of which are located within amino acids 20 through 29. Additional studies using targeted mutations have established that the sequence in this region is an important determinant of the molecule's amyloidogenicity (17)(18)(19)(20). hIAPP contains two potential MMP-9 cleavage sites, one between amino acids 15 and 16 and another between amino acids 25 and 26 (14). Cleavage at the latter position should yield fragments with reduced amyloidogenicity and sequential cleavage between amino acids 15 and 16 followed by cleavage between amino acids 25 and 26 would be predicted to also be anti-amyloidogenic. Thus, we sought to investigate the amyloidogenic and cytotoxic potential of MMP-9-cleaved hIAPP fragments. Given that MMP-9 expression is reduced in type 2 diabetes, we also tested whether up-regulation of MMP-9 activity in primary islets can protect from amyloid-induced toxicity.

Experimental Procedures
Peptide Synthesis and Purification-IAPP peptides were synthesized on a 0.25 mmol scale using a microwave peptide synthesizer, by 9-fluorenylmethoxycarbonyl (Fmoc) chemistry as previously described (21). Fmoc-protected pseudoproline dipeptide derivatives were incorporated at positions 9 -10, 19 -20, and 27-28 to facilitate the synthesis of full-length hIAPP and of a variant containing the fluorescent amino acid p-cyanophenylalanine (p-cyanoPhe). 5-(4Ј-Fmoc-aminomethyl-3Ј, 5-dimethoxyphenol) valeric acid resin was used to generate an amidated C terminus for full-length hIAPP and the fragments 16 -37 and 26 -37. Fmoc-Phe-PEG-PS resin and Fmoc-Ala-PEG-PS resin were used to form a free C terminus for fragment 1-15 and fragment 1-25, respectively. The disulfide bond was formed via oxidation by DMSO (22). The peptides were purified by reverse-phase HPLC. The identity of the pure products was confirmed by MS using a Bruker MALDI-TOF MS: full-length hIAPP, expected 3903.3, observed 3902.  An excitation wavelength of 450 nm and emission  wavelength of 485 nm was used for these thioflavin-T  experiments. hIAPP-F15 p-cyanoPhe is a variant of full-length hIAPP in which the phenylalanine at position 15 was replaced with the non-genetically coded amino acid p-cyanophenylalanine. The fluorescence of p-cyanoPhe is sensitive to environmental changes and can be used to monitor the kinetics of fibril formation. This substitution does not perturb the kinetics of amyloid formation and the measured time course of fibril formation is identical to the one probed using wild-type hIAPP (23,24). hIAPP-F15 p-cyanoPhe can therefore be used to monitor fibril formation by full-length hIAPP in the presence of another amyloid forming peptide (e.g. MMP-9-cleaved hIAPP fragments). Since the fluorescence of p-cyanoPhe is high when the cyano group is hydrogen bonded to water (25), the fluorescence signal decreases when hIAPP-F15 p-cyanoPhe aggregates. p-cya-noPhe flurorescence was excited at 240 nm, and emission was monitored at 296 nm.
Transmission EM (TEM)-At the end of the kinetic reactions, 4 l of the peptide solution was removed and blotted on a carbon-coated Formvar 300-mesh copper grid for 1 min and then negatively stained with saturated uranyl acetate for 1 min.
Circular Dichroism (CD)-CD experiments were performed using an Applied Photophysics Chirascan CD spectrometer. Spectra were recorded from 190 to 260 nm at 1 nm intervals in a quartz cuvette with 0.1 cm path length at 25°C. Spectra were background subtracted and presented as unsmoothed curves. CD spectra were measured from aliquots removed at the end of the kinetic (thioflavin-T) experiments.
MALDI-TOF and LC-MS-Unless otherwise specified, 10 M full-length hIAPP was incubated with or without 0.2 M recombinant human MMP-9 (EMD, Billerica, MA) at 37°C for 22 h. Samples were desalted using C18 ZipTips (EMD) and analyzed with MALDI-TOF MS with ␣-cyano-4-hydroxycinnamic acid as the matrix. For the hIAPP/MMP-9 time course analysis, 10 M full-length hIAPP was incubated with 0.2 M MMP-9 and a sample was analyzed by LC/MS after the indicated incubation time. The raw intensity of each peak was normalized to its extinction coefficient.
Cell Line Experiments-The ␤ cell line INS-1 was cultured in RPMI 1640 medium containing 10% fetal bovine serum, 1 mM sodium pyruvate, 100 units/ml penicillin, 100 g/ml streptomycin, and 11.1 mM glucose (complete medium). 8 ϫ 10 4 cells/ well were plated into a gelatin-coated 96-well plate and incubated for 24 h, after which the medium was replaced with complete medium containing freshly dissolved peptides. After a 24 h incubation, 50 l of XTT reagent (26) (Roche Applied Science, Madison, WI) was added, and 24 h later absorbance at 492 nm with a reference wavelength of 650 nm was measured. The absorbance of a blank sample (no cells) was used to determine the assay background and was subtracted from every experimental sample. Non-amyloidogenic rIAPP was used as a negative control. Each sample was run in triplicate and normalized to buffer-treated cells.
Primary ␤-Cell Experiments-Islets from non-transgenic mice were isolated by collagenase digestion and dispersed into single cells (ZymeFreeTM Enzyme Free Cell Dissociation Reagent, VWR, Radnor, PA). Islet cells were then incubated with anti-CD31-FITC antibody (ab33858, Abcam, Cambridge, MA) to label endothelial cells and sorted into a ␤-cell enriched ("␤ cell") and a ␤-cell depleted ("non-␤ cell") population based on the cell's auto-fluorescence (FACSARIA, BD Biosciences, San Jose, CA). A sample from each population was used for RNA analysis (TaqMan, Applied Biosystems, Carlsbad, CA). 18S was used as housekeeping gene. 1 ϫ 10 5 cells/well were plated into a 96-well plate, and freshly dissolved peptides added 24 h later. Viability was assessed 24 h later using the XTT assay as described above.
Islet Experiments-Islets from 10 week old transgenic mice with ␤-cell expression of hIAPP (27) on a F1 C57BL/6 x DBA/2 background and their littermate non-transgenic controls were isolated by collagenase digestion, purified, and transduced with adenoviruses expressing luciferase or human MMP-9 (28) for 24 h (multiplicity of infection ϭ 50, given that an average islet contains 1000 cells) in complete medium. Thereafter, islets were transferred into complete medium containing 16.7 mM glucose and cultured for 144 h with medium being renewed every 48 h. Histological samples were prepared as described elsewhere (29). Briefly, islets were formalin-fixed, paraffin-embedded, sectioned and labeled with thioflavin-S to visualize amyloid deposits and insulin antibody to visualize ␤ cells. Total islet, amyloid, and ␤-cell areas were calculated from at least 30 islets using a computer-based quantitative method (29). ␤-cell apoptosis was determined on at least 40 islets by insulin and propidium iodide staining (29). The observer was blinded to the genotype and culture condition of the islets. These studies were approved by the VA Puget Sound Health Care System Institutional Animal Care and Use Committee in Seattle.
MMP-9 Activity-Human MMP-9 activity was measured in islet supernatants collected 144 h after transduction according to the manufacturer's instructions (Human Active MMP-9 Fluorokine E Kit, R&D Systems, Minneapolis, MN).
Statistical Analysis-Data are presented as mean Ϯ S.E. for the number of experiments indicated. Statistical significance was determined using ANOVA with Tukey's Multiple Comparison Test as the post-hoc test or a paired t test when only two groups were compared (GraphPad Software, San Diego, CA). To compare the dose responses of full-length hIAPP and hIAPP 16 -37, the slope of the linear regression of each experiment was calculated and compared between the two groups using the Mann-Whitney U non-parametric test. A p Ͻ 0.05 was considered statistically significant.

Results
Human MMP-9 Degrades Monomeric but Not Aggregated hIAPP-The primary sequence of hIAPP includes two likely MMP-9 cleavage sites: the Phe-15-Leu-16 peptide bond and the Ala-25-Ile-26 bond (Fig. 1A). A model of the structure of the hIAPP amyloid fibril is displayed in Fig. 1 and is based on crystallographic studies of small fragments of hIAPP (30).
To test the susceptibility of monomeric hIAPP and aggregated hIAPP fibrils to cleavage by MMP-9, freshly dissolved full-length hIAPP was incubated with MMP-9 or buffer and analyzed by TEM and MS 22 h later. A dense mesh of fibrils was evident in the buffertreated sample ( Fig. 2A). In contrast, when full-length hIAPP was incubated with MMP-9, TEM showed the existence of only a few aggregates (Fig. 2B) and MS detected products consistent with MMP-9 cleavage of hIAPP between residues 15 and 16 as well as between residues 25 and 26 (Fig. 2C).
We next tested whether MMP-9 is able to cleave pre-aggregated hIAPP fibrils. For this experiment, full-length hIAPP was aggregated for 24 h, a time which is longer than required to form amyloid fibrils, MMP-9 was added and a sample removed for analysis 24 h later. TEM of buffer or MMP-9-treated samples were indistinguishable and both revealed dense mats of amyloid fibrils (Fig. 2, D and E). No degradation products were detected by MS following incubation of aggregated full-length hIAPP with MMP-9 (Fig. 2F).
Identification and Time-dependent Abundance of MMP-9cleaved hIAPP Degradation Products-Incubation of fulllength hIAPP with MMP-9 and subsequent analysis of the degradation products by LC/MS as a function of time confirmed  DECEMBER 18, 2015 • VOLUME 290 • NUMBER 51 that MMP-9 cleaves between residues 15 and 16 as well as 25 and 26 (Fig. 2G). All four fragments yielded by a single cleavage (1-15, 16 -37, 1-25, and 26 -37) were detected early on and their relative amount remained constant over time, except for hIAPP 26 -37 whose amount increased over the course of the experiment. In contrast, the amount of full-length hIAPP decreased over time, as expected, and was barely detectable at 8 h. hIAPP 16 -25 was not detected at 2 h but appeared at 4 h and increased thereafter, consistent with it being the only product resulting from two distinct cleavages.

MMP-9 Protects Islets from Amyloid-induced Toxicity
hIAPP 16 -37 Readily Aggregates into Amyloid Fibrils and Can Be Cleaved Further by MMP-9 -We synthesized fulllength hIAPP as well as the degradation products hIAPP 1-15, 16 -37, 1-25, and 26 -37 and tested their ability to form amyloid fibrils using thioflavin-T binding assays and TEM. As expected, the kinetic curve observed for full-length hIAPP exhibited a distinct lag phase on the order of 15 h followed by a growth phase until the final plateau was reached (Fig. 3A). TEM images collected at the end of the reaction revealed a dense mat of amyloid fibrils (Fig. 3C) and CD spectra indicated the pres-ence of ␤-sheets (Fig. 3D). In contrast, hIAPP 1-15, 1-25, and 26 -37 did not exhibit any increase in thioflavin-T fluorescence over the entire time course of the study (Fig. 3A), even when these fragments were incubated in the presence of 20% glycerol (Fig. 3B) or 2% HFIP (data not shown), both conditions known to promote amyloid formation. The absence of fibrils was confirmed by TEM (Fig. 3C), and CD spectra showed that hIAPP 1-15, 1-25 and 26 -37 remained in random coil confirmation (Fig. 3D). In contrast to the other fragments, thioflavin-T fluorescence of hIAPP 16 -37 increased rapidly without an initial lag phase leading to a maximal signal, which was similar to that of full-length hIAPP (Fig.  3A). TEM and CD showed that hIAPP 16 -37 forms fibrils (Fig.  3C) and adopts a ␤-sheet structure (Fig. 3D).
Non-Amyloidogenic hIAPP Fragments Are Not Cytotoxic and hIAPP 16 -37 Is Less Cytotoxic Than Full-length hIAPP-We next determined whether the hIAPP fragments produced by  DECEMBER 18, 2015 • VOLUME 290 • NUMBER 51 MMP-9 cleavage are cytotoxic. Full-length hIAPP is known to be cytotoxic (8) and was used as a positive control, while nontoxic rIAPP (17) was used as negative control. Viability of INS-1 ␤ cells following 24 h exposure to 60 M of the hIAPP fragments 1-15, 1-25 and 26 -37 was indistinguishable from cells treated with either buffer alone or 60 M rIAPP (Fig. 4A). In contrast, hIAPP fragment 16 -37 was cytotoxic, but appeared less cytotoxic than full-length hIAPP at this same dose.

hIAPP 16 -37 Accelerates Amyloid Formation by Full-Length hIAPP, But the Other hIAPP Fragments Do
Not-Some hIAPPderived fragments and structurally related peptides have been shown to be able to inhibit while others have been shown to be able to enhance fibril formation (31)(32)(33)(34). To test whether hIAPP 16 -37 is able to modulate the aggregation kinetics of full-length hIAPP, mixtures of full-length hIAPP and hIAPP 16 -37 at ratios of 1:0.5, 1:1, 1:5, and 1:10 were analyzed. hIAPP 16 -37 accelerated fibril formation by full-length hIAPP in a dose-dependent manner (Fig. 5A). The maximal fluorescence was increased and T50, the time to reach 50% of the maximal fluorescence signal, was reduced in a dose-dependent manner (Fig. 5B). Since the thioflavin-T fluorescent signal does not discern between fibrils derived from full-length hIAPP or hIAPP 16 -37, we used the fluorescent hIAPP-F15 p-cyanoPhe analog to monitor fibril formation by full-length hIAPP, independent of any contribution of hIAPP 16 -37 to the total fibril formation. Addition of a 5-fold excess of hIAPP 16 -37 accelerated hIAPP-F15 p-cyanoPhe fibril formation and a much shorter lag phase was observed (Fig. 5C).
The other non-amyloidogenic fragments were also tested for their effect on amyloid formation by measuring the resulting thioflavin-T kinetics and by obtaining TEM images of the mixtures. Addition of hIAPP 1-15 (Fig. 6A), hIAPP 1-25 (Fig. 6B) or hIAPP 26 -37 (Fig. 6C) to full-length hIAPP did not alter thioflavin-T kinetics at any of the ratios tested. TEMs obtained at the end of the kinetic analysis confirmed the finding that fibril formation was not inhibited (Fig. 6D). T50 was slightly greater with the mixtures of 1:1 and 1:5 full-length hIAPP to  4). The arbitrary value of the buffer control in each experiment was set to 100. The non-amyloidogenic rIAPP was included as a negative control. ***, p Ͻ 0.001 versus buffer control. hIAPP 1-25, but was not different with the other peptide ratios (Fig. 6E). Further, T50 was not altered when hIAPP 1-15 or 26 -37 were mixed with full-length hIAPP (Fig. 6E).

MMP-9 Overexpression Reduces Amyloid-induced ␤-Cell
Apoptosis in Mouse Islets-Since the MMP-9 cleavage products, with the exception of hIAPP 16 -37, are not toxic and the latter is less toxic than full-length hIAPP, and is further cleaved into non-toxic fragments by MMP-9, we hypothesized that upregulation of human MMP-9 activity in islets protects against amyloid formation and the resulting ␤-cell toxicity. Amyloidprone hIAPP transgenic mouse islets were transduced and cultured at 16.7 mM glucose to induce amyloid deposition and ␤-cell apoptosis (35,36). MMP-9-transduction increased human MMP-9 activity in islet supernatants (Fig. 7A). As expected, non-transgenic mouse islets did not develop amyloid deposits, while hIAPP transgenic islets contained amyloid (Fig.  7B). MMP-9 overexpression significantly reduced amyloid deposition in hIAPP transgenic mouse islets (Fig. 7B). Finally, the number of apoptotic ␤-cells was increased 5-fold in amyloidcontaining islets compared with non-transgenic control islets. MMP-9 overexpression did not significantly change the rate of ␤-cell apoptosis in non-transgenic islets but reduced amyloidinduced ␤-cell apoptosis in hIAPP transgenic islets by 47% (Fig.  7C).

Discussion
The data presented here demonstrate that MMP-9 decreases amyloid formation in cell-free systems and cultured, isolated islets. With the exception of the 16 -37 fragment, the MMP-9 cleavage products of hIAPP are neither amyloidogenic nor toxic to ␤ cells. hIAPP 16 -37 is toxic, but is significantly less so than full-length hIAPP. In addition, hIAPP 16 -37 itself is further cleaved into non-amyloidogenic fragments by MMP-9. Further, MMP-9 overexpression reduces amyloid deposition and amyloid-induced ␤-cell apoptosis in isolated mouse islets. Thus, up-regulation of MMP-9 might be a valid strategy to protect human islets from amyloid-induced damage.
The unique peptide constituent of the islet amyloid deposits observed in humans with type 2 diabetes is the ␤ cell secretory product hIAPP. hIAPP is known to be secreted into the extracellular space (15) where MMP-9 is present (16). Thus, MMP-9 is ideally positioned to degrade hIAPP and limit fibril formation. We have previously reported that inhibition of MMP-9 protease activity increases amyloid deposition and amyloid-induced ␤-cell apoptosis in isolated hIAPP transgenic mouse islets (14), suggesting that MMP-9 is part of a physiological clearance mechanism that protects the islet from excess hIAPP.  (n ϭ 3). B, T50 (time to reach 50% maximal fluorescence) for the kinetic curves in A was calculated and is shown for the indicated peptide ratios. Note that the aggregation with 5-and 10-fold excess of hIAPP 16 -37 was too fast to accurately measure T50. *, p Ͻ 0.05 versus full-length hIAPP alone (n ϭ 3). C, fluorescence kinetics of 16 M full-length p-cyanoPhe-hIAPP with or without 5-fold excess hIAPP 16 -37 are shown. Note p-cyanoPhe fluorescence decreases with increased peptide aggregation as the quantum yield is lower in the aggregated state (n ϭ 2). DECEMBER 18, 2015 • VOLUME 290 • NUMBER 51

MMP-9 Protects Islets from Amyloid-induced Toxicity
Islet amyloid deposits are present in the majority of patients with type 2 diabetes (3) and interestingly, we found reduced MMP-9 expression levels in islets from patients with type 2 diabetes (14). These data suggest that under physiological conditions MMP-9 acts to reduce amyloid accumulation and that this clearance mechanism is likely to be dysfunctional in humans with type 2 diabetes.
Our data show that MMP-9 is able to degrade monomeric hIAPP, but not aggregated hIAPP fibrils. A possible explanation for this is that MMP-9 can bind monomeric hIAPP but not hIAPP in amyloid deposits. This is supported by two high resolution structure models of hIAPP aggregation (30,37). As depicted in Fig. 1, the model of Wiltzius and colleagues places the MMP-9 cleavage site between residues Phe-15 and Leu-16 in the ordered N-terminal ␤-strand and residues Ala-25 and Ile-26 in the ordered core on the inside of the fibril. This suggests that, unlike monomeric hIAPP, the aggregated amyloid fibril structure protects the potential cleavage sites from attack by MMP-9. This finding is consistent with another model of hIAPP fibril formation (37), which, in contrast to the first model, places residues 25 and 26 in a partially ordered loop/ turn that links the two ␤-strands, similarly suggesting that hIAPP fibrils are protected from MMP-9 attack.
All but one of the tested MMP-9-cleaved hIAPP fragments did not induce ␤-cell toxicity in vitro. The absence of toxicity is likely explained by the lack of amyloidogenicity of these peptides. Although hIAPP residues 8 -20, 10 -19, and 30 -37 have been shown to have the potential to form amyloid when incubated at high concentrations (38,39), we found the MMP-9cleaved hIAPP fragments 1-15, 1-25, and 26 -37 did not form amyloid even when they were incubated under conditions that promote amyloid formation. It is likely that the disulfide bond in hIAPP 1-15 and the charged groups at both termini (Lys and Arg) are not compatible with ␤-sheet structures. Computer algorithms that predict amyloidogenicity and aggregation regions in unfolded polypeptide chains (40,41) confirm the absence of potential to aggregate in the hIAPP fragments 1-15 and 26 -37. The 8 -20 fragment of hIAPP can form amyloid in isolation, but the constraints induced by the disulfide bond and the charged groups at the N terminus and in the middle of the peptide likely overcome the aggregation prone residues 8 -20 in hIAPP 1-25. While one of the computer algorithms calculates no amyloidogenicity for hIAPP 1-25, the other predicts a potential aggregation site at residues 12-17, although with much lower aggregation potential than full-length hIAPP. It is of course possible that hIAPP 1-25 forms amyloid if incubated for several days at a high concentration, but these conditions are not relevant to the situation in vivo. Thus, consistent with our experimental observations, these computational analysis suggest that hIAPP fragments 1-15, 1-25, and 26 -37 have no or very modest potential to aggregate.
We found one of the MMP-9-cleaved hIAPP fragments, namely 16 -37, readily forms amyloid fibrils. This fragment contains amino acids 20 -29 of the full-length peptide. These ten residues were previously found to influence the potential of full-length hIAPP to aggregate and they were also shown to be able to independently assemble into fibrils (17,42). Further, hIAPP 16 -37 was toxic to ␤ cells (both primary and transformed) and accelerated the aggregation of full-length hIAPP. This suggests that the production of hIAPP 16 -37 is associated with a pro-apoptotic phenotype. However, we do not believe that hIAPP 16 -37 is a major determinant of amyloid-induced ␤-cell apoptosis in vivo. First, hIAPP 16 -37 has an MMP-9 cleavage site between residues 25-26 and co-incubation of hIAPP 16 -37 with MMP-9 leads to non-amyloidogenic and non-cytotoxic fragments. Second, it has been speculated that early aggregates (oligomers) of hIAPP are the major cytotoxic species and that fully aggregated hIAPP is inert (43)(44)(45)(46)(47). We observed that hIAPP 16 -37 aggregates much faster than fulllength hIAPP and one could speculate that it therefore reaches the inert aggregated state more quickly. In line with this hypothesis, we found that hIAPP 16 -37 was less cytotoxic than full-length hIAPP when cultured with either primary or transformed ␤ cells. A third reason why hIAPP 16 -37 probably does not contribute substantially to amyloid formation in vivo is that, based on the abundance of the cleavage products, our data suggest that Ala25-Ile26 is MMP-9's preferred cleavage site and that the yield of hIAPP 16 -37 per se is likely to be rather low. Overall, these data suggest that even though hIAPP 16 -37 is cytotoxic, its abundance in vivo is probably insignificant and its pro-amyloidogenic/cytotoxic effects negligible.
Hexapeptides derived from the amyloidogenic region 20 -29 as well as other hIAPP fragments and structurally related peptides have been shown to be modulators of hIAPP fibrillogenesis (31)(32)(33)(34). For MMP-9 cleavage of hIAPP to be a target to reduce amyloid formation and cytotoxicity, it is important that its cleavage products do not aggravate full-length hIAPP's toxicity. Mixtures of full-length hIAPP with up to 10-fold excess hIAPP 1-15, 1-25, or 26 -37 did not influence the aggregation kinetics of full-length hIAPP, suggesting that these fragments do not exacerbate amyloid-induced damage.
Finally, we tested these findings from cell-free and in vitro systems in primary islets. MMP-9 up-regulation reduced amyloid deposition in isolated islets by 25% and amyloid-induced ␤-cell apoptosis by 50%. These observations suggest that MMP-9 may have therapeutic potential. First, MMP-9 activa-FIGURE 7. MMP-9 overexpression reduces amyloid deposition and ␤-cell apoptosis in hIAPP transgenic mouse islets. Human MMP-9 activity in supernatants of mouse islets transduced with adenoviruses that express luciferase (luc) or MMP-9 is shown in A. *, p Ͻ 0.05 versus luc (n ϭ 3). Amyloid area/islet area (B) and % apoptotic ␤ cells (C) of transduced islets are shown. *, p Ͻ 0.05 and ***, p Ͻ 0.001 versus non-transgenic controls. †, p Ͻ 0.05 transgenic luciferase versus transgenic MMP-9 (n ϭ 5). DECEMBER 18, 2015 • VOLUME 290 • NUMBER 51 tion in the islet could be used to reduce amyloid-induced cytotoxicity in type 2 diabetes. This concept is supported by the fact that islets from patients with type 2 diabetes have reduced MMP-9 expression (14). A second potential area of interest for MMP-9 activation could be islet transplantation where amyloid formation is associated with loss of ␤ cells, transplant failure, and the recurrence of hyperglycemia (4,5,48,49).

MMP-9 Protects Islets from Amyloid-induced Toxicity
In summary, our findings show that MMP-9 cleavage of hIAPP largely ablates hIAPP aggregation and amyloid-induced toxicity. Thus, increasing MMP-9 activity in vivo could be a strategy to protect the islet from amyloid-induced damage in human diabetes.
Author Contributions-D. T. M. contributed to the study design, performed research, analyzed data, and wrote the manuscript. L. H. T. contributed to the study design, performed research, analyzed data, and reviewed/edited the manuscript. M. F. H., A. T. T., and R. L. H. helped interpret the data and reviewed/edited the manuscript. S. Z., D. P. R., and S. E. K. contributed to the study design, helped interpret the data and reviewed/edited the manuscript. D. T. M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the work as a whole.