Cell-Surface Autoantibody Targets Zinc Transporter-8 (ZnT8) for In Vivo β-Cell Imaging and Islet-Specific Therapies

The exquisite specificity of autoimmune β-cell destruction in type 1 diabetes (T1D) is attributed to islet infiltration by autoreactive T and B lymphocytes (1). These adaptive immune cells use cell-surface receptors to recognize specific autoepitopes or autoantigens on the cell surface of β-cells. Although identifying T-cell epitopes has been crucial to understanding β-cell recognition by autoreactive T cells, cell-surface antigenic markers for B-cell recognition remain elusive. Cell surface–directed autoreactive B cells can differentiate into plasma cells that secrete islet cell-surface autoantibody (ICSA) (2). Therefore, identifying an ICSA-secreting autoreactive B cell could identify a cognate B-cell receptor (BCR) that forms an immunological synapse with a cell-surface autoantigen. This autoantigen can then be extracted from the surface membrane of β-cells (3) and processed as T-cell epitopes to be presented or cross-presented to CD4⁺ or CD8⁺ T cells, leading to the activation of autoreactive T cells (4,5). Thus, cloning an ICSA-secreting autoreactive B cell could provide a critical tool to understand B cell–mediated autoimmunity in T1D.

A major cell-surface autoantigen targeted by ICSA is ZnT8 (6). It is a specialized zinc transporter dedicated to sequestering zinc into the insulin biosynthetic pathway of β-cells, where zinc is required for proinsulin processing and insulin packaging (7). ZnT8 expression is exclusively limited to the pancreatic islets, which are mainly composed of β-cells (8–10). The subcellular distribution of ZnT8 is in a dynamic equilibrium between the cell-surface membrane, the endoplasmic reticulum, and the insulin secretory granules (11–13). Glucose-stimulated insulin secretion (GSIS) promotes ZnT8 trafficking to the cell surface (14), a process that exposes the surfaced ZnT8 to BCRs of ZnT8-specific B cells and to serum ICSA directed to the extracellular ZnT8 epitope, termed ZnT8ecA (6,15,16). Importantly, seroconversion of ZnT8ecA occurs in the earliest phase of autoimmunity progression in T1D (6), suggesting that ZnT8-specific autoreactive B cells may contribute to islet autoimmunity initiation. ZnT8ecA is a natural molecular decoy of its cognate BCR. Decoy receptors have been used as a therapeutic modality to suppress T-cell autoimmunity in T1D (17,18). Cloning ZnT8ecA will enable further research into its immunoregulatory effect on ZnT8 autoreactive B cells, which may translate to new islet-targeted immunotherapies that suppress local B-cell autoimmunity against islet ZnT8.

Only a minuscule fraction of the total B-cell pool is expected to recognize cell-surface ZnT8 and secrete ZnT8ecA. Until now, cloning cell-surface ZnT8 antibodies has been inherently difficult (19) because the extracellular surface of ZnT8 is membrane flush and therefore lacks an ectodomain to engage antibodies (Fig. 1A). Furthermore, three short extracellular loops (ECL1–3) of ZnT8 are largely invariant in mammals, rendering them antigenically inert across species (Fig. 1B).

In this report, we establish a novel immunization strategy to induce autoreactive antibodies that recognize ECLs in natively folded ZnT8 and clone a high-affinity antibody (mAb43) with specificity for both human and rodent ZnT8. Furthermore, we reveal a functional coupling of GSIS with mAb43 surface binding, establish ZnT8-mediated mAb43 uptake in β-cells, and demonstrate mAb43-guided islet homing of an imaging payload in mice.

NOD, C57BL/6, Lepr^db/db, and MIP-GFP mice were purchased from the Jackson Laboratory, and ZnT8-knockout (KO) mice were purchased from Taconic Biosciences. All animal procedures were approved by the Institutional Animal Care and Use Committees of Johns Hopkins University School of Medicine and the University of Colorado.

Human ZnT8 and its C-terminal domain (CTD) were transiently expressed in FreeStyle 293-F cells and purified as described previously (20). The purified ZnT8 protein was reconstituted at a ZnT8-to-lipid ratio of 1:20 (wt/wt) into proteoliposomes composed of DOPC, DOPE, and DOPG at a 2:1:1 ratio with Lipid-A adjuvant at 10% of total lipid content. Liposomes were prepared without adding ZnT8.

Seven-week-old male and female homozygous ZnT8-KO mice (n = 4 for each sex) and 10-week-old NOD females (n = 5) were injected intraperitoneally weekly with ZnT8 proteoliposome emulsion (100 μL). ZnT8-KO (n = 2) and NOD female (n = 3) littermates were used for control liposome immunization. All mice were euthanized 5 weeks post-injection. Drained lymph nodes and spleens were collected to generate hybridomas by electrofusion. Monoclonal fused cells were identified by comparative ELISAs. Mouse/human chimeric mAb43 and mAb20 were generated using the human immunoglobulin IgG1 (IgG1) Fc sequence to replace the corresponding mouse Fc sequence. mAb43-mScarlet, mAb20-mScarlet, mouse interleukin-2 (IL-2), and mutated IL-2 were generated by appending the mScarlet, IL-2, or mutated IL-2 protein sequence to the Fc C-terminus (21). The mAb43–IL-2 mutein was generated by introducing a D123N point mutation to mouse IL-2. All recombinant antibodies and their derivatives were transiently expressed in 293-F cells and purified using standard protein A/G affinity chromatography, followed by size-exclusion high-performance liquid chromatography (HPLC) in PBS to generate highly purified monoclonal antibodies (mAbs).

Both solution- and proteoliposome-based ELISAs were used to screen mAbs based on their reactivities to full-length ZnT8 (flZnT8) and CTD, as described previously (22).

The human EndoC-β1H β-cell line (23) and both wild-type (WT) and ZnT8-KO rat INS-1E β-cell lines (15) were used to examine cell-surface and intracellular immunofluorescence (IF) labeling by mAb43 and cell-surface marker antibodies. Commercial antibody information is provided in the Supplementary Key Resource Table. Cell culture conditions and immunolabeling protocols are detailed in the Supplementary Methods.

Excised mouse pancreata were fixed in 4% paraformaldehyde (PFA), paraffin embedded, sectioned (4 μm), and immunolabeled using mouse/human chimeric mAb43 or mAb20 (1:50). Biotinylated anti-human IgG antibody was applied and then bound to an avidin-peroxidase complex using the HRP-ABC Kit (Vector Laboratories). Peroxidase activity was detected by DAB (Vector Laboratories). Flash-frozen human pancreas cryosections (5 μm) were PFA fixed and then indirectly fluorescently immunolabeled using mouse mAb43 (1:1,000) and anti-insulin (1:250).

Stably transfected INS-1E cells expressing ZnT8-GFP or ZnT8_FLAG-GFP were solubilized and examined by an analytical sizing HPLC equipped with a fluorescence detector (488/510 nm) as described previously (14).

Fab43 was generated by papain digestion of purified mAb43, bound to purified ZnT8 in detergent-lipid micelles, and analyzed by negative-statin electron microscopy (EM) single-particle analysis as described previously (22).

Excised mouse pancreata were accutase dispersed, DyeCycle Violet (DCV) stained for intact cells, and immunolabeled with chimeric mAb43, followed by phycoerythrin (PE)-conjugated anti-human IgG. Chimeric mAb20 was used as an isotype control. Labeled cells were immediately sorted on a MoFlo XDP cell sorter, as described previously (14). Gated R0 or R1 cells were grown in a 1:100 Matrigel-coated microwell dish and then PFA fixed, permeabilized, and immunolabeled by chimeric mAb43, followed by anti-human IgG-PE, anti-insulin allophycocyanin (APC), and antiglucagon–Alexa Fluor 488 (A488). After washing and DCV counterstaining, IF images were acquired using a Zeiss LSM 700. In addition, pancreatic islets were dispersed, PFA fixed, permeabilized, cytospun, and immunolabeled for imaging, as described above.

Chimeric mAb43 was prebound with anti-human–IgG-A647 at a 1:2 molar ratio. Live EndoC-β1H cells in Krebs buffer were exposed to the antibody complex (0.03 mg/mL), in the presence of 2 or 20 mmol/L glucose, at 8 or 20°C as indicated, and then washed free of unbound antibody complex for IF imaging.

Various tissues from mAb43- or mAb20-injected mice were dried by a brief spin on a strainer, weighed, homogenized in PBS with DNase and protease inhibitors, solubilized with SDS, and then loaded onto SDS-PAGE gels. Chimeric mAb43 or mAb20 in each tissue was detected by anti-human–IgG immunoblotting and quantified using an internal human IgG standard. Tissue uptake was corrected for tissue weight and total administered mAb dose; the amount of antibody retained was calculated as a percentage of injected mAb per gram of each tissue collected (%mAb_injected/g).

Whole pancreata from C57BL/6 mice or MIP-GFP mice were excised, flattened between a pair of microscope slides, PFA fixed, and optically cleared as described (24). One 10-week-old male and one female MIP-GFP mouse received mAb43-mScarlet (15 mg/kg) intravenously. Three days postinjection, pancreata were excised and processed as above without detergent treatment.

Wholemount pancreas images were acquired on an ImageXpress Micro high-content analysis system using the 488- and 585-nm filter sets for GFP and mAb43-mScarlet fluorescence, respectively. Transmission light scanning was recorded simultaneously to produce a brightfield image. The Fiji stitching plugin in ImageJ was used to combine tiled images. For each fluorescence channel, background level was measured and subtracted numerically across the entire image, and then fluorescence values were displayed without further modification. Mander overlap coefficients were computed across the whole pancreas or regions of interest using all pixels above auto thresholds for GFP and mScarlet fluorescence without background correction.

Mouse islets were isolated and manually picked according to published protocols (25,26). mAb43-mScarlet or mAb20-mScarlet was added to the islet culture medium to a final concentration of 0.01 mg/mL. Islets were imaged as described above for wholemount pancreata.

Human pancreas sections were obtained from the Human Pancreas Analysis Consortium (27).

All values are expressed as the mean ± SEM. Two-tailed Student t test was used to compare groups. Significance indicated in the figures is denoted as *P < 0.01.

All data generated during this study are included in the published article and its online supplementary files. Reagents generated are available upon reasonable request. Johns Hopkins University requires material transfer agreements for all inbound or outbound research materials. Further information and requests for resources and reagents should be directed to and will be fulfilled by the corresponding author D.F.

To overcome immune self-tolerance to highly conserved ZnT8 epitopes, we used two different immunization strategies to elicit mouse antibody responses to human ZnT8: 1) deleting the ZnT8 gene to avoid negative selection in immunologically intact mice; and 2) stimulating autoreactivity to ZnT8 in immunologically compromised mice with defective immune tolerance. Accordingly, we immunized ZnT8-KO mice and NOD female mice, the latter of which are prone to developing spontaneous autoimmune diabetes. To preserve the native folding of ZnT8 antigen post-injection, we developed a liposome-reconstituted ZnT8 formulation (16,20). ZnT8 is a two-module protein consisting of a transmembrane domain (TMD) and a cytosolic CTD (Fig. 1A). The mouse antibody response to the TMD was interrogated by comparative ELISAs against flZnT8 and its CTD. Both mouse strains showed robust anti-flZnT8 (TMD+CTD) and anti-CTD responses compared with mice receiving empty liposome injections. Compared with ZnT8-KO mice (Fig. 1C), NOD mice exhibited a significantly higher serum reactivity toward flZnT8 at lower serum dilutions (Fig. 1D), suggesting the presence of anti-TMD reactivity in proteoliposome-injected NOD mice. Next, we generated hybridoma cells from immunized ZnT8-KO and NOD mice. Only NOD mice produced an anti-TMD mAb (mAb43) that reacted exclusively to flZnT8 (TMD+CTD), with no detectable CTD reactivity (Fig. 1E and F). Reconstitution of detergent-solubilized human ZnT8 into proteoliposomes increased mAb43 reactivity by 6.29-fold, demonstrating a preferential recognition of the natively folded membrane-flush TMD (Fig. 1G). A validated antibody recognizing the ZnT8 CTD (mAb20) was used as a binding control (22). No difference was observed in mAb20 reactivity to three different antigen formats: detergent-solubilized ZnT8, CTD, and liposome-reconstituted ZnT8 (Fig. 1E–G). mAb43 and mAb20 titrations to ZnT8 proteoliposomes yielded binding affinities at 0.42 ± 0.05 and 0.57 ± 0.07 nmol/L, respectively.

To determine whether the observed anti-TMD reactivity of mAb43 was directed to the extracellular surface of the TMD, we compared IF labeling of a live human β-cell line (EndoC-β1H) by mAb43, mAb20, and an antibody against the abundant cell-surface marker CD71. All experiments were performed at 8°C to arrest antibody endocytosis and in the presence of 20 mmol/L glucose to stimulate ZnT8 surfacing (14). mAb43 and anti-CD71 yielded strong IF punctation on the cell surface, whereas mAb20 did not produce a detectable signal (Fig. 2A). In contrast, both mAb20 and mAb43 strongly labeled intracellular ZnT8 in permeabilized EndoC-β1H cells (Fig. 2B). The latter signal resulted from mAb43 binding to the ZnT8 epitope inside vesicular lumina. We further examined mAb43 cross-reactivity to a rat β-cell line (INS-1E) in comparison with a rodent-reactive antibody against the abundant cell-surface marker Na⁺/K⁺ ATPase. mAb43 and anti-Na⁺/K⁺ ATPase yielded strong IF punctation on the cell surface (Fig. 2C). In contrast, immunolabeling of permeabilized INS-1E cells revealed vesicular and nuclear labeling by mAb43 and Na⁺/K⁺ ATPase antibody, respectively (Fig. 2D). CRISPR/Cas9-mediated ZnT8 ablation in INS-1E cells abolished mAb43 binding at the cell surface and in intracellular vesicles, validating the mAb43 specificity for ZnT8 in rodent β-cells (Fig. 2C and D). Quantifying the differences in mAb43 or mAb20 IF labeling of EndoC-β1H cells and mAb43 IF labeling of WT or ZnT8-KO INS-1E cells further demonstrated distinct subcellular profiles of ZnT8 immunolabeling by mAb20 and mAb43, in agreement with their targeting of different ZnT8 epitopes (Fig. 2F). Finally, we examined competitive ZnT8 binding by mouse mAb43 and human serum from a T1D donor positive for ZnT8ecA (6). Exposing live EndoC-β1H cells to either mouse mAb43 or human serum yielded strong mouse (red) or human (yellow) IgG punctation on the cell surface (Fig. 3E). By comparison, exposing live EndoC-β1H cells to both mouse mAb43 and human serum predominantly yielded mouse IgG punctation, regardless of serum or mAb43 preblocking (Fig. 3E). Imaging quantification indicated that mAb43 displaced >80% of serum IgG punctation on the cell surface (Fig. 3G). This finding suggests that the polyclonal serum ZnT8ecA from a patient with diabetes is predominately directed to a cell-surface ZnT8 epitope shared by mAb43.

To map the mAb43 epitope to ZnT8 ECLs, we inserted a FLAG octapeptide into individual ECLs to perturb their local conformation and then compared mAb43 binding with native ZnT8 and ZnT8_FLAG. Among nine insertion constructs, only an ECL2 insertion resulted in ZnT8_FLAG expression in INS-1E cells (14). An EGFP was appended to the ZnT8 C-terminus to monitor the formation of a binary mAb-ZnT8 complex by fluorescence size-exclusion HPLC. mAb43 binding shifted the ZnT8-GFP peak leftward, indicating a stable mAb43-ZnT8-GFP complex (Fig. 3A). The FLAG tag abolished mAb43 binding to ZnT8_FLAG-GFP but added anti-FLAG binding that formed a stable anti-FLAG–ZnT8_FLAG–GFP complex (Fig. 3B). Therefore, mAb43 and the anti-FLAG antibody competed for ECL2 and neighboring ECLs on the TMD surface. Moreover, mAb43 was not reactive to SDS-denatured ZnT8 on immunoblots (Fig. 3C), further demonstrating the conformation specificity of mAb43. Finally, negative-stain EM single-particle analysis of a protein complex between an antigen-binding fragment of mAb43 (Fab43) and a detergent-solubilized ZnT8 revealed only 1 Fab43 molecule binding to 1 ZnT8 homodimer (Fig. 3D). The point of Fab43 attachment to the ZnT8 homodimer density was ∼18° off the twofold homodimer axis, in alignment with a splayed TMD. Because the two ZnT8 protomers in a ZnT8 homodimer adopt distinct conformations (28), Fab43 seems to recognize either an outward- or inward-facing conformation. Taken together, our biochemical data indicate that mAb43 forms a stable complex with ZnT8 through conformation-specific binding to ECLs.

mAb43 immunolabeling of paraffin-embedded mouse pancreas sections revealed specific mAb43 binding to islets, but mAb20 exhibited a lack of cross-reactivity to mouse ZnT8 (Fig. 4A). Co-immunolabeling of dispersed detergent-permeabilized mouse islet cells with anti-insulin, anti-glucagon, and mAb43 showed that mAb43 recognizes both α- and β-cells (Fig. 4B). Co-immunolabeling of human pancreatic cryosections from 2 donors, 1 control and 1 with T2D, revealed colocalization of anti-insulin and mAb43 IF signals, demonstrating the specificity of mAb43 for human islets (Fig. 4C). Next, we used FACS to examine mAb43 immunolabeling of live β-cells in a single-cell suspension of enzymatically-dispersed mouse pancreata. mAb43-labeled pancreatic cells were gated against large cell debris and granular vesicles based on positive staining by a cell-permeable DNA dye, DCV. Only a small fraction (1.7%) of whole pancreatic cells fell into the DCV⁺/mAb43-PE⁺ quadrant (Fig. 4D). The sorted cells were permeabilized and immunolabeled for ZnT8, insulin, and glucagon. Almost all sorted DCV⁺/mAb43-PE⁺ cells were positive for both ZnT8 (red) and insulin (yellow) but negative for glucagon (green) (Fig. 4E). Quantification of insulin and ZnT8 immunolabeling revealed a clear enrichment of β-cells in the mAb43⁺ R1 cell population (Fig. 4F). Thus, flow sorting of mAb43-labeled cells allowed separation of β-cells from the bulk of the pancreas mass (98.3%). The mAb43 specificity for primary mouse islets is consistent with the highly selective nature of T1D autoimmunity against β-cells.

To track cell-surface capture of mAb43 and the ensuing ZnT8-mediated mAb43 endocytosis, we used a fluorescent A647 secondary antibody to label mAb20 and mAb43 and a CellMask green stain to demarcate the cell boundary. Live EndoC-β1H cells were monitored for antibody surface binding and internalization. mAb43-A647 was rapidly internalized at 37°C, whereas mAb20-A647 exposure yielded no detectable signal (Fig. 5A). When EndoC-β1H cells were chilled at 8°C and mAb43-A647 endocytosis arrested, but cell-surface binding of mAb43-A647 persisted (Fig. 5B). Importantly, lowering glucose concentration from 20 to 2 mmol/L markedly reduced both mAb43 cell-surface binding at 8°C and mAb43-A647 uptake at 37°C (Fig. 5A and B). Imaging quantification suggested that glucose stimulation (20 mmol/L) increased total mAb43-A647 IF labeling by 22.1- and 15.0-fold at 37 and 8°C, respectively (Fig. 5C). The difference in mAb43-A643 IF signals between 37 and 8°C approximated to the net mAb43-A647 uptake. Glucose stimulation increased ZnT8-mediated mAb43 uptake by 30.9-fold (Fig. 5C). Glucose-dependent mAb43 capture and internalization were also observed using a fusion of mAb43 with a monomeric red fluorescent protein, mScarlet (28) (Supplementary Fig. 1).

To characterize in vivo delivery of mAb43, we generated a mouse-Fab/human-Fc chimeric mAb43, injected 4 male C57BL/6 mice (C1–4) at 5 mg/kg, and then used anti-human IgG immunoblotting to detect the chimeric mAb43 in a panel of excised organs. C1–C3 received mAb43 intravenously and C4 intraperitoneally. Plasma mAb43 was eliminated within a day (Fig. 6A), in agreement with the mouse pharmacokinetic model of target-mediated antibody clearance for low-dose administration (29). From 1 to 6 days post-injection, mAb43 was detected predominantly in the pancreas, and its biodistribution profile remained unchanged regardless of the administration route (Fig. 6B). The pancreas-to-serum ratio of mAb43 ranged from 24.6 to 66.2. Control experiments with a mouse/human chimeric mAb20 yielded no detectable signal in the pancreas by 3 days post-injection (Fig. 6A–C). By comparison, the half-life of mAb43 in the pancreas was ∼1 week, with an initial pancreas concentration of 21.1 ± 0.9% mAb_injected/g tissue weight 1 day post-injection, tapering down to 14.3 ± 1.5 and 11.1 ± 1.0% mAb_injected/g tissue weight at 5 and 6 days post-injection, respectively (Fig. 6D). The pancreas-specific mAb43 biodistribution demonstrates the feasibility of targeting mAb43 to the pancreas through systemic administration.

Four NOD females (N1–4) and 4 Lepr^db/db males (D1–4), both 18 weeks of age, were given a single chimeric mAb43 intraperitoneal injection (5 mg/kg); then, mAb43 tissue uptake was measured 48 h post-injection. At 18 weeks of age, the lymphocytic infiltration into pancreatic islets of female NOD mice is well established (30), and overt obesity has developed in male Lepr^db/db mice. Both mouse strains exhibited biodistribution profiles like that of C57BL/6 mice, with mAb43 accumulation predominately in the pancreas (Fig. 6E). Pancreatic mAb43 uptake was compared among individual mice of different mouse strains, all of which exhibited both normoglycemic and hyperglycemic fasting blood glucose (FBG) concentrations (Fig. 6F). On average, C57BL/6 mice had a modestly higher mAb43 uptake than NOD and Lepr^db/db mice (Fig. 6G). One NOD and 2 Lepr^db/db mice were diabetic (FBG >250 mg/dL), and these mice exhibited notably reduced pancreatic mAb43 uptake compared with their respective nondiabetic counterparts.

To demonstrate the utility of mAb43 for targeted drug delivery, we generated a fusion protein of mAb43 and IL-2. Low-dose IL-2 reverses established T1D in NOD mice (31–33) and preserves insulin production in T1D clinical trials (34,35). The therapeutic effect of IL-2 is attributed to bolstering of T regulatory cells (Tregs), which are known to suppress overactive immune responses. However, IL-2 can also stimulate effector T cells and NK cells, limiting the IL-2 dose to a narrow range for selective Treg expansion. The mAb43–IL-2 fusion protein administered by intravenous injection to a C57BL/6 mouse was broadly distributed in various organs (Fig. 6H) because the IL-2 receptor (IL-2R) on the surface of organ-resident immune cells has a higher binding affinity for IL-2 (K_d ∼10⁻¹¹) than pancreatic ZnT8 for mAb43 (K_d ∼10⁻¹⁰). Point mutations in human IL-2 reduce its IL-2R affinity while retaining the Treg expansion efficacy (36,37). Likewise, we generated a mouse IL-2 mutein with a reduced IL-2R affinity. Systemically administered mAb43–IL-2 mutein exhibited a pancreas-selective biodistribution (Fig. 6H), demonstrating mAb43-mediated delivery of a well-established T1D therapeutic to the pancreas. This targeted delivery should reduce systemic IL-2 toxicity and increase IL-2 pancreatic concentration to achieve optimal therapeutic efficacy.

To evaluate the feasibility of mAb43 for in vivo delivery of imaging payloads, we injected mAb43-mScarlet into mice that express GFP specifically in β-cells (MIP-GFP) (24). Wholemount pancreas imaging revealed a high degree of global colocalization between GFP and mScarlet with Mander overlap coefficients of 0.93 and 0.79 for the fraction of mScarlet overlapping GFP (M1) and GFP overlapping mScarlet (M2), respectively (Fig. 7A). The 21% unmatched GFP signal was primarily from erythrocyte GFP autofluorescence in pancreatic arteries and their branches (M1 ≈ M2 ≈ 0.0) (Fig. 7B). In contrast, GFP-mScarlet colocalization was nearly absolute in islet clusters that surrounded large blood vessels (M1, M2 >0.9) (Fig. 7B). High-power magnification confirmed colocalization of individual GFP and mScarlet puncta (Fig. 7C). In some regions, scattered minor mScarlet signals without overlapping GFP signals were observed (Fig. 7D); these signals are probably small β-cell clusters, the GFP signals of which could be detected when the pancreas was treated with detergent during tissue clearing (Supplementary Fig. 2). Finally, isolated mouse islets were examined for mScarlet uptake ex vivo. mAb43-mScarlet exposure of isolated mouse islets resulted in intense mScarlet fluorescence, whereas exposure with mAb20-mScarlet resulted in no detectable uptake (Fig. 7E and F). These findings further demonstrate the specificity of mAb43-mediated mScarlet uptake through ZnT8 binding to the islet cell surface.

Our data indicate that the generation of mAb43 depends on self-tolerance breakdown in NOD mice, in which CD4⁺ T cells autoreactive to ZnT8 spontaneously occur, but they are only weakly pathogenic (38). Accordingly, ZnT8-proteoliposome immunization was required to boost autoreactivity to ECLs. ZnT8 gene ablation in ZnT8-KO mice is insufficient to induce antibodies against ECLs of human ZnT8, probably because the extracellular ZnT8 epitope is conserved across species in other ZnT homologs (ZnT1–10). In particular, part of the ZnT signature sequence is located in ECL1 (39). Individual ECLs contain only 4-7 residues and therefore are too short to fold independently (Fig. 1B). Multiple ECLs and their interactions are required to make a recognizable conformation. The multiloop folding of the ZnT8 extracellular epitope therefore results in a distinctly conformation-specific mAb43 autoantibody with the hallmark in vivo β-cell specificity of T1D. This specificity is demonstrated by the effective blockade of the cell-surface ZnT8 from serum ZnT8ecA immunolabeling by mAb43 binding to a largely overlapping epitope shared by the polyclonal ZnT8ecA (Fig. 2E).

The pancreas-specific biodistribution of mAb43, in conjunction with its islet-specific immunolabeling of pancreas sections, suggests that systemically administrated mAb43 could be delivered specifically to pancreatic islets in vivo. Wholemount pancreas imaging revealed regional mAb43-mScarlet enrichment in islet clusters on the periphery of the pancreas (Fig. 7A). These highly vascularized islets allow rapid insulin release into the circulation. In addition, local GSIS activity is functionally coupled with ZnT8 surfacing and subsequent capture of circulating mAb43. Islet-targeted accumulation of mAb43 is retained in both T1D and T2D mouse models, but the level of mAb43 uptake is decreased compared with that of healthy controls, reflecting the loss of β-cell mass and/or function in diseased mice (Fig. 6E–G). The in vivo islet specificity of mAb43 is consistent with the islet-specific expression of ZnT8. Antibodies recognizing specific markers on the β-cell surface could be used to target β-cells for the delivery of imaging agents or drugs that may deleteriously affect non-islet tissues (Fig. 6H). Previous β-cell surface markers targeted for delivery include ZnT8 with an antibody (Ab31) raised against a peptide sequence of ECL2 (19), sphingomyelin patches, NTPDase3, and GLP1R (40–42). Thus far, only mAb43 has demonstrated a pancreas-specific biodistribution profile that supports its utility for the targeted delivery of imaging payloads and disease-modifying therapies through systemic administration.

Acknowledgments. The authors thank Hao Zhang from the Flow Cytometry and Immunology Core Facility at Johns Hopkins Bloomberg School of Public Health for assistance in flow cytometry data acquisition and analysis, and they also thank Hoku West-Foyle from the Microscope Facility of Johns Hopkins University School of Medicine for assistance in wholemount pancreas data acquisition and analysis.

Funding. This study was supported by the National Institutes of Health (NIH) and the National Institute of Diabetes and Digestive and Kidney Diseases grants R56 DK123435 and R01DK125746 (D.F.), P30DK116073 (L.Y.), R01DK110183 (M.L.G.), and RO1DK084171 (G.W.W.) and the National Institute of General Medical Sciences grant R01 NS127292 (Hui. Li). The Zeiss confocal microscope was funded through NIH shared instrumentation grant S10OD016374. The MoFlo XDP cell sorter was funded through NIH grants S10OD016315 and S10RR13777001. This manuscript used data and tissue acquired from the Human Pancreas Analysis Program (HPAP-RRID:SCR_016202) Database (https://hpap.pmacs.upenn.edu), a Human Islet Research Network (RRID:SCR_014393) Consortium (UC4-DK-112217, U01-DK-123594, UC4-DK-112232, and U01-DK-123716).

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

Author Contributions. Z.G., D.K., C.X., G.W.W., and D.F. were responsible for antibody biodistribution. Z.G., D.K., M.L.G., and D.F. performed mouse pancreas immunohistology and flow cytometry as well as mouse islet and wholemount pancreas imaging. Z.G., C.M., M.N., L.Y., and D.F. were responsible for antigen preparation and mouse immunization. Z.G., C.M., and D.F. performed hybridoma screening. Z.G. and D.F. were responsible for mAb43 cloning and epitope mapping as well as recombinant antibody production. Hua Li and Hui. Li performed EM single-particle analysis. L.Y. and D.F. conceptualized the study. M.L.G. and D.F. wrote the manuscript. All authors read and edited the manuscript. D.F. 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 data and the accuracy of the data analysis.