A human antibody against pathologic IAPP aggregates protects beta cells in type 2 diabetes

In patients with type 2 diabetes, pancreatic beta cells progressively degenerate and gradually lose their ability to produce insulin and regulate blood glucose. Beta cell dysfunction and loss is associated with an accumulation of aggregated forms of islet amyloid polypeptide (IAPP) consisting of soluble prebrillar IAPP oligomers as well as insoluble IAPP brils in pancreatic islets. Here, we describe a novel human monoclonal antibody selectively targeting IAPP oligomers and neutralizing IAPP aggregate toxicity by preventing membrane disruption and apoptosis in vitro. Antibody treatment in rats and mice transgenic for human IAPP, and human islet-engrafted mouse models of type 2 diabetes triggered clearance of IAPP oligomers resulting in beta cell protection and improved glucose control. These results provide new evidence for the pathological role of IAPP oligomers and suggest that antibody-mediated removal of IAPP oligomers could be a pharmaceutical strategy to support beta cell function in type 2 diabetes. Slides imaged on Leica laser scanning microscope. Image analysis conducted on all islets present on three sections (~ 50 µm interval) from each islet preparation using Image J software. Apoptosis was counted as the number of TUNEL and DAPI double-positive nuclei relative to total number of DAPI-positive nuclei. Amyloid deposition was computed as the islet area occupied by ThioS staining.


Introduction
Type 2 diabetes (T2D) is a chronic metabolic disorder characterized by insulin resistance and progressive dysfunction and loss of insulin-producing pancreatic beta cells, resulting in insulin de ciency and elevated blood glucose. Beta cell death is accompanied by the accumulation and aggregation of the 37residue peptide hormone islet amyloid polypeptide (IAPP or amylin) that is co-secreted with insulin and forms amyloid deposits in a majority of pancreatic islets from T2D patients 1 . In its physiological monomeric conformation, IAPP acts as a regulator of glucose homeostasis through satiety control and inhibition of gastric emptying 2,3 . As IAPP is highly amyloidogenic under conditions of increased secretory demand it can readily misfold and aggregate into soluble oligomers and insoluble amyloid brils that are thought to contribute to beta cell dysfunction and death in T2D [4][5][6] . The involvement of IAPP aggregates in beta cell decline and T2D progression is supported by an increasing body of evidence. First, amyloid severity inversely correlates with beta cell area in pancreatic islets of T2D patients 4 . Second, a sporadic mutation in the human IAPP coding sequence leading to a S20G amino acid substitution is associated with a higher peptide propensity for aggregation, an increased risk for developing T2D and a more severe form of the disease [7][8][9] . Third, cat and primates producing amyloidogenic variants of IAPP and forming pancreatic islet amyloid deposits naturally develop signs of T2D 10 . In non-human primates, islet amyloid severity was also shown to correlate with beta cell loss and T2D progression 11,12 . Fourth, while rodent IAPP is unable to aggregate into amyloid, transgenic mice and rats expressing human IAPP (hIAPP) spontaneously develop a T2D phenotype characterized by islet amyloidosis and decreased beta cell mass [13][14][15][16] . Likewise, humanization of the non-amyloidogenic porcine IAPP using CRISPR/Cas9 gene editing leads to T2D in miniature pigs 17 . Of note, IAPP aggregation is also linked to beta cell deterioration in human islets cultured in high glucose or transplanted into mice or humans with type 1 diabetes [18][19][20] . Synthetic IAPP aggregates, primarily oligomers produced at an early stage of amyloid bril formation, induce beta cell dysfunction and apoptosis in vitro 21 . Their cytotoxicity presumably results from membrane permeabilization 22,23 , induction of oxidative and ER stress 24,25 , and pro-in ammatory cytokine release 26,27 . However, the contribution of IAPP oligomers to beta cell loss and T2D progression remains controversial and has not been yet clearly established in vivo. Monoclonal antibodies are approved therapeutic agents that offer the unique advantage of neutralizing and facilitating the removal of speci c disease-related target antigens. We hypothesized that accumulation of IAPP oligomers in pancreatic islets is responsible for dysfunction and degeneration of beta cells observed in T2D, and that passive immunization with monoclonal antibodies directed against IAPP oligomers may protect beta cells and provide therapeutic e cacy. To this end, we developed a human monoclonal antibody highly selective for toxic IAPP oligomer species that was evaluated for its ability to clear IAPP oligomers, preserve beta cell function and prevent disease progression in rodent models of T2D.

Results
A human monoclonal antibody with high a nity and selectively for aggregated hIAPP.
We identi ed and recombinantly cloned a monoclonal antibody of IgG1 subclass (termed α-IAPP-O) selectively targeting pathologic human IAPP (hIAPP) aggregates by analyses of complements of human memory B cells derived from a clinically selected human population composed of healthy elderly donors. The α-IAPP-O antibody selectively immunoreacted at a low nanomolar concentration with extracellular hIAPP aggregates present on amyloid-positive pancreatic islets from type 2 diabetic subjects (Fig. 1a,b), with absence of binding to native physiological hIAPP monomers within insulin-producing pancreatic beta cells and to unrelated disease-associated amyloidogenic proteins and β-amyloid plaques in the brain of Alzheimer's disease patients ( Supplementary Fig. 1). α-IAPP-O binding kinetics revealed high a nity for hIAPP aggregates (K D =1.54 nM) likely consisting of a heterogenous mixture of aggregated species, as compared to monomeric hIAPP (biotin-hIAPP, K D =2.84 µM) using biolayer interferometry (Fig. 1c, The α-IAPP-O epitope was mapped to an N-terminal sequence conserved among hIAPP and less amyloidogenic IAPP orthologues and binding was impaired by proline substitution in the amyloidogenic region responsible for amyloid bril formation, C-terminal truncation and SDS-induced denaturation ( Supplementary Fig. 2).
Human IAPP rapidly aggregated into amyloid brils visualized by transmission electron microscopy, and amyloid growth monitored by thio avin-T (ThioT) uorescence followed a sigmoidal kinetics characterized by a short lag phase, an exponential growth phase and an equilibrium phase (Fig. 2a). To elucidate the nature of aggregated hIAPP species recognized by α-IAPP-O, we performed time-resolved immunoblot analysis of fractions collected over the course of amyloid bril formation. α-IAPP-O was shown to preferentially immunoreact with transient pre brillar oligomers that are produced during the lag and the growth phases of amyloid formation (Fig. 2b). In contrast to a non-selective IAPP antibody (α-IAPP), α-IAPP-O bound neither to monomeric hIAPP nor to non-amyloidogenic rodent IAPP (rIAPP) and detergent-denatured hIAPP aggregates. α-IAPP-O recognized high molecular weight hIAPP species (>200 kDa) but not monomers or chemically cross-linked aggregates generated in vitro and resolved by nonreducing SDS-PAGE and Western blotting (Fig. 2c). We next studied the effect of α-IAPP-O on hIAPP aggregation using kinetic modelling of molecular events underlying amyloid bril formation [28][29][30][31][32] .
Aggregation kinetics of hIAPP, unseeded or seeded by adding preformed brillar hIAPP, were compatible with models describing either fragmentation or secondary nucleation-dominated mechanisms ( Supplementary Fig. 3). In both cases, model analysis indicates that α-IAPP-O concentration-dependently delayed bril formation by inhibiting primary nucleation, with minimal effects on bril-dependent processes such as elongation, secondary nucleation and fragmentation (Fig. 2d,e and Supplementary Incubation of beta cells with hIAPP oligomers formed in vitro reduced the viability by 95% and increased apoptosis identi ed by TUNEL staining to 85% (Fig. 3a,b). Cytotoxicity was not observed when applying amyloid brils ( Supplementary Fig. 7a,b). α-IAPP-O neutralized cytotoxicity and apoptosis in a concentration-dependent manner as compared to an IgG control antibody. Beta cell apoptosis was accompanied by cell membrane deposition of hIAPP oligomers and ThioS-positive amyloid brils (  Supplementary Fig. 7d) and prevented permeabilization of liposome membranes induced by hIAPP oligomers (Fig. 3f,g). We next determined the effects of α-IAPP-O on human islets isolated from obese donors at risk for diabetes (Supplementary Table 2). Human islets were exposed to high glucose leading to the accumulation of extracellular ThioS-positive amyloid deposits and to beta cell apoptosis (Fig. 3h). Co-incubation with α-IAPP-O (0.5 µM) reduced ThioS-positive amyloid load and apoptotic beta cell death compared to IgG control. Further, α-IAPP-O improved beta cell function evaluated by insulin response to elevated glucose using islet perifusion, similar to the amyloid-inhibiting compound Congo red 33 at high concentration (25 µM) ( Fig. 3i and Supplementary Fig. 7e).
α-IAPP-O improves key pathological features of type-2 diabetes in rats and mice.
The effects of α-IAPP-O were next evaluated in a transgenic rat model with beta cell-speci c expression of hIAPP. Transgenic rats were shown to spontaneously develop a diabetic phenotype characterized by extensive islet amyloid formation resulting in progressive beta cell dysfunction and loss, ultimately leading to insulin depletion and hyperglycemia 15,16 . In this model, α-IAPP-O speci cally engaged extracellular hIAPP oligomers surrounding insulin-producing beta cells in transgenic rat islets (Fig. 4a). The antibody did not bind to extracellular ThioS-positive amyloid, nor to monomeric IAPP constitutively expressed within transgenic and wild-type beta cells (  Supplementary Fig. 9c). ch α-IAPP-O e cacy was target-related as it had no glucose-lowering effect in wild-type rats, and an isotype control antibody had no impact on glycemia in transgenic or wild-type rats ( Supplementary Fig. 9d). In a separate study, rats with marked glucose intolerance and hyperglycemia at baseline (Supplementary Fig. 10a,b) received weekly dosing of ch α-IAPP-O (1, 3 and 10 mg/kg i.p.). ch α-IAPP-O treatment was associated with signi cant improvements in glucose tolerance, glucose-stimulated insulin response and beta cell function at 1 and 10 mg/kg compared to vehicle (Fig. 4d-f and Supplementary Fig. 10c). These effects were accompanied by reduced glycemia, increased circulating insulin levels and normalized body weight ( Fig. 4g-i and Supplementary Fig. 10d). ch α-IAPP-O treatment was already effective after 7 weeks and effects became more apparent as the phenotype progressed ( Supplementary Fig. 10e,f). Slowing of disease progression was associated with preserved islet size and beta cell content in the pancreas of rats treated with ch α-IAPP-O relative to vehicle (Fig. 4j,k and Supplementary Fig. 10g). ch α-IAPP-O treatment was also associated with an increase in soluble IAPP levels and a decrease in insoluble IAPP aggregates in pancreas homogenates (Fig. 4l). Further, we con rmed the therapeutic effects of α-IAPP-O in hIAPP transgenic mice characterized by glucose intolerance, hyperglycemia and insulin de ciency together with islet amyloidosis, oligomer deposition and beta cell loss ( Supplementary Fig. 11). In this independent model, weekly administration of mouse chimeric ch α-IAPP-O (10 mg/kg i.p.) but not an isotype-matched control antibody improved glycemia and protected pancreatic beta cells ( Supplementary Fig. 11b-h).  Fig. 13a,b). Likewise, ch α-IAPP-O decreased total pancreatic IAPP aggregates measured by ELISA ( Supplementary Fig. 13c). Oligomer removal was paralleled by recruitment of CD68positive islet resident macrophages to hIAPP oligomers and brils, without affecting the total number of islet macrophages ( Fig. 5c and Supplementary Fig. 13d). In addition, clearance of hIAPP oligomers was associated with increased insulin-immunoreactive beta cell area and reduced levels of the diseaserelevant pro-in ammatory cytokine IL-1β 26,27 in the pancreas ( Fig. 5d and Supplementary Fig. 13e,f). In contrast, inert ch α-IAPP-O did neither signi cantly affect hIAPP oligomer deposition, macrophage recruitment and IL-1β levels, nor preserve beta cell content despite a small reduction in islet amyloid load α-IAPP-O prevents diabetes in human islet-engrafted mouse models.
IAPP aggregation and amyloid deposition has been reported in transplanted human pancreatic islets where it might contribute to graft failure and recurrence of hyperglycemia [37][38][39] . To evaluate the effect of α-IAPP-O on graft function in vivo, immunode cient NSG mice rendered diabetic by streptozotocin (STZ) injection or Rag2 -/mice fed a high-fat diet (HFD) 38,40 were both transplanted with human islets from nondiabetic and pre-diabetic donors ( Fig. 6a and Supplementary Table 2). While NSG recipient mice weekly administered with ch IgG (10 mg/kg i.p.) rapidly returned to hyperglycemia, ch α-IAPP-O (10 mg/kg i.p.) treatment maintained normoglycemia and delayed the recurrence of diabetes (Fig. 6b,c and Supplementary Fig. 15a). Human islet-engrafted Rag2 -/mice fed a HFD two weeks post-transplant for twelve weeks developed glucose intolerance, hyperglycemia and hyperinsulinemia accompanying obesity, as opposed to non-obese recipients fed a control diet ( Fig. 6d-g). Treatment with ch α-IAPP-O (10 mg/kg i.p., once weekly) along with HFD normalized glycemia, and consistently reduced plasma insulin and human C-peptide levels ( Fig. 6d-f and Supplementary Fig. 15b,c), pointing towards an improved function and adaptation of engrafted human islets. Furthermore, ch α-IAPP-O treatment (10 mg/kg i.p., once weekly) initiated in obese diabetic mice previously fed a HFD for six weeks reversed abnormal glucose tolerance, while glucose tolerance deteriorated because of graft dysfunction in recipients receiving ch IgG (10 mg/kg i.p., once weekly) ( Fig. 6h and Supplementary Fig. 15d). Graft analysis revealed a high number of in ltrating macrophages associated with a decrease in oligomeric hIAPP but not amyloid deposits upon ch α-IAPP-O treatment (Fig. 6i), in line with stimulation of macrophage-mediated phagocytic clearance of toxic hIAPP oligomers.

Discussion
Clinically, the progressive nature of T2D is linked to beta cell dysfunction and loss, with patients lacking the ability to produce su cient endogenous insulin to counteract insulin resistance and to control blood glucose levels 41 . While the primary cause of beta cell failure in T2D is unknown, the accumulation of aggregated forms of the beta cell peptide hormone hIAPP in pancreatic islets is a likely contributor to decreased beta cell function and mass in an early stage of the disease. We describe here that hIAPP oligomers can cause beta cell dysfunction and death during T2D development which can be prevented by applying a human-derived antibody selective for these toxic hIAPP species in cultured beta cells and isolated human islets, as well as in transgenic rodents and human islet-engrafted mouse models.
The data provide evidence that hIAPP pre brillar oligomers applied to beta cells induce apoptosis in vitro, in contrast to monomers and mature brils. Neutralization of hIAPP oligomers by α-IAPP-O, a human monoclonal IgG1 antibody selectively binding extracellular hIAPP deposits in the pancreas of T2D patients and hIAPP intermediates produced during amyloid formation, prevented cell accumulation of amyloid brils, lipid membrane disruption and beta cell toxicity. This is in line with previous studies indicating that toxic hIAPP oligomers deposit at the cell surface and permeabilize the cell membrane via pore formation and/or elongation into amyloid brils 42 . We have also shown that α-IAPP-O interacts with the N-terminus of IAPP when self-assembled into oligomers and inhibits the formation of amyloid bril end-product, supporting a key role of IAPP N-terminal residues in the initial aggregation process facilitating membrane interaction and permeation 43 . Treatment with α-IAPP-O and with a general inhibitor of aggregation reduced islet amyloid content and beta cell toxicity in isolated human islets exposed to elevated glucose levels, strengthening the role of hIAPP oligomers and islet amyloidosis in human beta cell deterioration under diabetic conditions such as hyperglycemia. Of note, beta cell toxicity solely caused by hIAPP aggregates and independently of hyperglycemia was also reported in human and hIAPP-expressing mouse islets 44,45 . In these studies, beta cell apoptosis has been associated with oxidative stress, Fas upregulation and caspase-8 activation. Additional mechanisms by which hIAPP aggregates could mediate islet beta cell apoptosis and that are characteristic of human T2D include membrane disruption 46 , endoplasmic reticulum (ER) stress 24 , defects in autophagy 47 , activation of the receptor for advanced glycation end-products (RAGE) 48 and in ammation 26,27,36 . These pathways potentially contribute to hIAPP-induced beta cell loss and development of diabetes. This has been extensively studied in hIAPP-expressing transgenic rats and mice recapitulating features of human T2D [13][14][15][16] . In these animal models, both intracellular and extracellular hIAPP oligomers have been reported to trigger beta cell dysfunction and loss. Our data support a direct role for extracellular oligomers in beta cell pathogenesis, diabetes onset and progression in vivo. First, we have demonstrated that α-IAPP-O selectively engages extracellular oligomers in overtly diabetic hIAPP transgenic rats and mice after a single intraperitoneal injection. Oligomers bound by α-IAPP-O were extensively deposited around islet beta cells with a distribution distinct from amyloid brils. Second, chronic administration of α-IAPP-O in prediabetic and diabetic transgenic animals reduced beta cell loss and improved insulin secretion and glycemia. α-IAPP-O also prevented beta cell failure and diabetes development in human islet-engrafted mice, ruling out any confounding effects of hIAPP expression in transgenic models.
The mechanism of action of α-IAPP-O was dependent on the phagocytic clearance of extracellular hIAPP oligomers, but not amyloid, by islet macrophages in vivo. Although macrophages were recruited at sites of amyloid deposition, we did not observe any effect on amyloid load, consistent with the selective Together, our ndings indicate that antibody-mediated removal of extracellular hIAPP oligomers in an early phase of events leading to beta cell exhaustion can preserve beta cell function and limit the progression of T2D. This is of particular relevance since the effectiveness of currently available treatments for T2D are limited in time and likely impacted by the continuous deterioration of beta cell function over the course of the disease 51 . Novel treatment strategies to delay disease progression by restoring and durably preserving beta cell function are needed 52,53 . Anti-diabetic medications offering blood glucose control by improving peripheral insulin sensitivity such as metformin, or by increasing insulin secretion such as dipeptidyl peptidase-4 (DPP-4) inhibitors and glucagon-like peptide-1 receptor agonists (GLP-1 RAs) have demonstrated bene ts in beta cell adaptation to high-fat diet-induced insulin resistance in hIAPP transgenic rodents and short-term improvements in human islet graft function in diabetic mouse recipients 34,54-56 . However, evidence for a long-term impact of these drugs on the progressive deterioration of beta cell function are lacking [57][58][59][60] .
The therapeutic development of evolutionarily optimized human antibodies directed against misfolded and aggregated endogenous proteins delivered promising preclinical and clinical ndings in degenerative brain diseases such as Alzheimer's disease 61 , Parkinson's disease 62 , amyotrophic lateral sclerosis (ALS) and fronto-temporal dementia (FTD) 63,64 . Our data with the human-derived antibody α-IAPP-O targeting toxic IAPP oligomers expand this therapeutic concept towards T2D and opens a new avenue for beta cell protective therapies.

Antibody generation
Antibodies were generated from a de-identi ed blood lymphocyte library collected from healthy elderly subjects by screening for high a nity binding toward aggregated IAPP and absence of cross-reactivity toward unrelated amyloid-forming proteins 61 . Antibody sequences were cloned from corresponding memory B cells using cDNA cloning of IgG heavy and kappa or lambda light chain variable region sequences, and sub-cloned into human IgG1 expression constructs using Ig framework-speci c primers for human variable heavy and light chain families in combination with human J-H segment-speci c primers. Chimeric analogs ( ch α-IAPP-O and ch α-IAPP-O/F) were engineered to contain mouse IgG2a or rat IgG2b backbones. Recombinant antibodies were expressed in CHO-S cells and puri ed by protein A or protein G a nity chromatography. Fab fragments were generated by enzymatic digestion of human IgG1 antibody followed by puri cation on an IgG-CH1 a nity column (GingisKHAN Fab kit, Genovis). The study was approved by the local ethics committee, written informed consent was obtained prior to the investigations.

Aggregation assay and kinetic analysis
Spontaneous aggregation of hIAPP, biotin-hIAPP and rIAPP peptides (Bachem) in the absence and presence of different concentrations of antibodies was assessed by monitoring amyloid bril formation via the increase of uorescence of the amyloid-speci c dye thio avin-T (ThioT, Sigma) over time. constants were calculated using the AmyloFit platform 65 . Microscopic events underlying hIAPP bril formation that are inhibited by antibodies were identi ed by comparing rate constants obtained in the absence and presence of antibodies in unseeded and seeded reactions.

Transmission electron microscopy
Samples were adsorbed onto glow-discharged carbon-coated copper grids (S162-3, Plano). Grids were stained with 2% (w/v) uranyl acetate for 1 min, washed with ddH 2 O, air-dried and imaged using a Philips CM100 transmission electron microscope with an acceleration voltage of 80 to 100 kV.

Dynamic light scattering
The average size of particles present in samples was measured by dynamic light scattering (DLS) at a xed angle of θ=173 ° and a laser source of 633 nm on a Zetasizer Nano (Malvern, UK). Additionally, samples were centrifuged at 10'000 g for 15 min and supernatants (soluble fractions) were analyzed.

INS-1 cells and cell-based assays
Rat  (ThermoFisher Scienti c) by uorescence emitted at 517 nm (495 nm excitation) every 3 minutes after plate was shaken for 9 sec at 400 rpm. At the end of the experiment, maximum uorescence leakage was induced by addition of 1 μL 10% Triton-X100 (Sigma). Percentage membrane leakage was calculated by the equation (F-F 0 )/(F max -F 0 )x100, with F corresponding to uorescence measured over time, F 0 to initial uorescence, and F max to maximum uorescence.

Human islet culture
Human pancreas tissue was harvested from three obese adult brain-dead donors at the Centre Hospitalier Régional Universitaire de Lille (France) and islets were isolated as previously described 66 . Donor characteristics and islet information are provided in Supplementary Table 2 was supplied in drinking water (3 to 3.8 g/L). Daily water intake was estimated by weighing the water bottles and metformin concentration was adjusted accordingly to reach a target dose of 200 mg/kg/day. Treatments started at 12 weeks of age and were blinded until full completion of the studies.
Oral glucose tolerance test (oGTT) was performed on fasted rats (12h overnight fasting with free access to water). Rats were orally administered with 2 g/kg glucose (50% solution, B. Braun Medical AG) and insulin-and hIAPP-immunoreactive beta cell content were analyzed on all islets (> 2500 µm 2 ) identi ed on para n-embedded rat pancreas head, core and tail (four sections each per rat). α-IAPP-O-bound hIAPP aggregates, ThioS-positive amyloid, and CD68-immunoreactive macrophages present within islets were quanti ed on rat pancreas and human islet graft cryosections (four and three to six sections per tissue, respectively). Data were computed as the uorescence area above a predetermined threshold using Image-Pro Premier software (Media Cybernetics) and expressed as percentage of corresponding islet and tissue area. Colocalization between CD68-immunoreactive macrophages and α-IAPP-O-bound hIAPP aggregates or ThioS-positive amyloid was analyzed using ImageJ/FiJi software. Beta cell mass (mg) was calculated as follows: (Σ insulin-positive area / pancreas area) x pancreas weight (mg).
Lyophilized hIAPP peptide was reconstituted in 0.1 M sodium bicarbonate buffer (pH 8.4) to a nal peptide concentration of 4 mg/ml, incubated with pHrodo green STP ester dye (20 mg/ml in DMSO; P35369, ThermoFisher Scienti c) for 30 min at room temperature in the dark, lyophilized with an Alpha 1-2 LDplus freeze dryer (Christ) and stored at -20°C until use. Lyophilized pHrodo-labeled hIAPP and and phagocytosis was analyzed using a FACS Aria II ow cytometer equipped with BD FACS Diva software (BD Biosciences). Intracellular pHrodo green was excited using a 488 nm laser and the uorescence emission was collected using a 530/30 nm lter (FITC). A total of 10'000 events were acquired from each sample and data were exported as Flow Cytometry Standard format 3.0 les (FCS les) and analyzed with FlowJo software (Tree Star Inc.). Gating was done on single macrophages with high forward and side scatter (FSC-A and SSC-A) levels, and pHrodo-hIAPP-positive macrophages with uorescence emission above cytochalasin D-treated macrophages (negative control) were counted.

Statistical analysis
Data are expressed as means ± s.e.m and results between groups were analyzed using Student's t-test, one-way and two-way ANOVA with post hoc tests for multiple comparisons. Statistical analyses were conducted using GraphPad PRISM 7 (GraphPad Software, USA) and signi cance was set at *p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001. All authors approved the nal version of the manuscript.