Novel Human Insulin Isoforms and Cα-Peptide Product in Islets of Langerhans and Choroid Plexus

The human insulin (INS) gene underwent significant variations during evolution as its tissue expression migrated from neurons of invertebrates to islets of Langerhans in pancreas of vertebrates, and the functional spectrum of its peptide products diversified over time from growth and development, learning and memory, and reproduction to primarily metabolism and regulation of anabolism, including its well-known function of regulating cellular uptake of glucose (1). The primary source for brain insulin is presumed to be insulin secreted from β-cells in islets (2). However, this is not the sole source, because it is now established that mouse choroid plexus (ChP) epithelium cells (EChP) also express insulin. Mouse ChP contains several Ins2 isoforms (Ins2-V1, -V2, -V3, and -V4) that are processed into mature insulin that is then secreted into cerebrospinal fluid (CSF) proportional to CSF serotonin concentrations; glucose does not regulate insulin secretion from isolated EChP (3). Because of ongoing interest in actions of insulin in brain and the fact that insulin is produced in ChP, we undertook further investigation of the human INS isoforms and their protein products.

Human INS is typically described as having three exons that encode a preproinsulin of 110 amino acids (AA) that undergoes extensive posttranslational modifications. Proinsulin of 86 AA, a 24-AA signal peptide having been removed, exits the endoplasmic reticulum and is delivered to the Golgi apparatus for packaging in secretory granules. As granules become mature and acidify, proinsulin is Remain A-chain and B-chain typically then cleaved, freeing a 30-AA B-chain, a 31-AA C-peptide, and a 21-AA A-chain. In the process, four AA (RR and RK) are lost from the ends of C-peptide. Secretory granules in β-cells therefore contain various proportions of proinsulin, partially processed insulin, C-peptide, and mature insulin. Biologically active human and mouse insulin in circulation is composed of single A- and B-chains connected by disulfide bonds and small amounts of cleaved and noncleaved proinsulin with varying degrees of biological activity (4).

Humans possess one INS gene, while mice have two: Ins1 and Ins2. Ins1 is a retrogene derived from partially spliced Ins2 (5). Mouse islets express both genes in a three-to-two ratio (6), but only Ins2, the evolutionarily older gene, is present in ChP, while its expression is barely if at all detectable in other brain areas; Ins1 is not detected in any brain area (3). Invertebrate genomes encode multiple insulin-related peptides in neurons, including pro–molluscan insulin-related peptides, and posttranslational processing produces A- and B-chains and long and short C-peptides (7). Multiple human INS mRNA isoforms, all of which encode the known 110-AA preproinsulin, are reported to be regulated by glucose in pancreatic islets (8), but interestingly, insulin secretion from ChP is not controlled by glucose (3). However, in an extensive literature search, we found no reports of human-specific INS upstream open reading frames (uORF) or spliced isoforms encoding novel polypeptides (8,9). At least 50% of human genes contain uORF that add layers of gene regulation and protein products for functional diversity (10). As INS evolved, expansion of actions of its isoform products in islets, ChP, and brain areas may have resulted in divergence of posttranscription and translational products and divergence of functions. We uncovered novel human-specific uORF isoforms and a novel intraexon 3 splicing variant (here designated INS3B) that encodes a short 19-AA Cα-peptide at both mRNA and protein levels. We found that protease cleavage of INS3B was reduced in islets from individuals with type 2 diabetes mellitus (T2DM). We also found that INS isoforms and Cα-peptide are present in human ChP, including in those whose β-cells were presumed absent because of β-cell destruction (11).

Islets from human nondiabetic (n = 13) and T2DM (n = 9) postmortem pancreata (human islet checklist provided in Supplementary Table 1) were provided by the National Institute of Diabetes and Digestive and Kidney Diseases–funded Integrated Islet Distribution Program at City of Hope (National Institutes of Health grant 2UC4DK098085) (https://iidp.coh.org/). A total of 1,000–2,000 islets were ordered for each sample, and ∼200 islets were handpicked for proteomics and the rest for gene expression experiments. Frozen human ChP sections and blood samples (nondiabetic n = 5; T1DM n = 6; T2DM n = 4) were from the Human Brain Collection Core, National Institute of Mental Health (12) (Supplementary Table 2). Mouse pancreata, islets, and ChP were dissected as we previously described (3).

We measured insulin levels in plasma of adults without diabetes after overnight fast (fasting glucose) (n = 7) and 2 h after continuous intravenous glucose (2h-IVG) (n = 15) administration (glucose clamp methodology as previously described (13), whereby circulating glucose was clamped at fasting levels plus 98 mg/dL glucose during glucose infusion). The protocol has institutional review board approval (National Institutes of Health Institutional Review Board 15-AG-0063). This was undertaken to confirm that our new mass spectrometry (MS)-based assay was capable of quantifying endogenous INS products in controlled settings and for comparison with measurements in the same samples by ELISA.

We used the Sequencher 5.4.6 software (Gene Codes Corporation, Ann Arbor, MI) to assemble pancreatic islet and insulinoma expressed sequence tag sequences (https://www.ncbi.nlm.nih.gov/dbEST/) that are homologous to human INS gene. Human Splicing Finder (HSF) (https://www.umd.be/HSF/HSF.shtml) was used to predict potential INS exon 3 intraexonal splicing events (14). We used GlobPlot 2 (https://globplot.embl.de/) (15) to predict intrinsic protein disorder domain of INS isoforms. IMAGE cDNA clones were ordered from Source Bioscience (https://www.sourcebioscience.com/), and the inserts were sequenced by the Sanger sequencing service of Eurofins Genomics (Louisville, KY).

RNA isolation, cDNA synthesis, quantitative RT-PCR, and RNAscope in situ hybridization (ISH) procedures were described previously (3). We designed splicing junction–specific TaqMan probes (Supplementary Table 3) for primary ORF (pORF) isoforms (INS1A, -1B, -1C, -I1, and -3B). FAM-labeled TaqMan probes of INS exons 2 and 3A junction (INS3A; Hs00355773_m1) and β-cell autoantigen and transcription factor probes were ordered from Thermo Fisher Scientific (Waltham, MA). We designed the uORF-specific forward primer overlapping the INSU isoform translation initiation methionine and the TaqMan probes at the splicing junctions of exon 1UA/exon 2 (INSUA), exon 1UB/exon 2 (INSUB), exon 1UC/exon 2 (INSUC), and exon 1UC/intron 1 (INSU1) (Fig. 1) to measure expression of uORF in human tissues. Because the exon-intron junction size of INSU2 (intron 2 retention) is larger than the amplicon limit for the TaqMan assay, INSU2 transcription levels were measured by averaged values of INSU1 and EX2-I2 probes (Fig. 1A and Supplementary Table 3). The duplex fluorescent TaqMan assay was performed in replicates (StepOnePlus Real-Time PCR System), and the relative fold change was calculated using the formula: 2^−ΔΔCt normalized with GAPDH (VIC labeled; catalog number 4326317E) (16). Droplet digital PCR (ddPCR) absolute values were derived from Poisson distribution of positive and negative droplets (QX200 ddPCR System; Bio-Rad Laboratories, Philadelphia, PA) that were normalized with endogenous control β2 microglobulin (VIC labeled; catalog number 4326319E).

RNAscope ISH probes were custom designed for INSU isoform (11 ZZ pairs targeting 2–272 of upstream transcription start site [uTSS] and 273–732 of primary TSS nucleotide sequences of MT335691 in C1 or C2 channel). Transthyretin (TTR) (18 ZZ pairs targeting 2–917 nucleotides of NM_000371.3 in C1 channel), islet amyloid polypeptide (IAPP) (16 ZZ pairs targeting 424–1,947 nucleotides of NM_000415.2 in C3 channel), and RBFOX3 (NEUN) (20 ZZ pairs targeting region 720–2,217 nucleotides of NM_001082575.2 in C3 channel) were from cataloged probes from Advanced Cell Diagnostics, Inc. (Hayward, CA). Human islets were individually handpicked and embedded with Shandon M-1 Embedding Matrix (Thermo Fisher Scientific), frozen in ethanol/dry ice bath, and stored at −80°C. Pretreatment of ChP sections, probe reactions, and labeling were performed according to the RNAscope Multiplex Fluorescent Detection Kit v2 protocol (17).

Thioflavin T (ThT) was from MilliporeSigma (catalog number T3516; Rockville, MD). IAPP (37 AA) (catalog number LT110006) (amidated C-terminal and disulfide bridge between Cys2 and Cys7) was from LifeTein (Hillsborough, NJ), and C-peptide (catalog number AS-61127) was from AnaSpec (Fremont, CA). Cα-peptide was custom made (Genemed Synthesis, Inc., San Antonio, TX). Black 96-well microplates (chimney well) were from Greiner Bio-One (Frickenhausen, Germany). Soluble and monomeric IAPP was made according to published protocols (18,19), and the AnaSpec Manual of SensoLyte Thioflavin T Aggregation Kit (catalog number AS-72214) was used for the IAPP aggregation assay, with final concentration of 1% DMSO and 25 μmol/L ThT in component A buffer. The buffer was used as blank and IAPP without inhibitors, and the final concentration of IAPP and C- and Cα-peptides was 50 μmol/L each in the inhibition assays. Amyloid kinetics were measured by increasing ThT fluorescent amyloid binding intensity (λex 440 nm; λem 485 nm; height 3 mm; flashes 12) for 36 cycles, 5 min per cycle, at 37°C with 15 s of shaking (100 rpm) between the reads in the EnSpire Multimode Plate Reader (PerkinElmer, Inc., Waltham, MA).

For a full overview, see Supplementary Material for ELISA and selected reaction monitoring (SRM)–MS, Supplementary Tables 4–6, and Supplementary Fig. 1A–K.

GraphPad Prism v9 software was used for statistical analysis, and data are presented as means ± SEM. For the ThT amyloid dye assay, area under the curve was calculated using the trapezoidal rule, and the data were analyzed using linear regression and one-way ANOVA with Tukey multiple comparison correction (20). The normalized expression values of INS isoforms in TaqMan fold changes and ddPCR percentages of target/β2 microglobulin droplets, ELISA, and SRM quantitative signals were analyzed with one-way ANOVA, two-way ANOVA (repeated measures; subject matching with Sidak correction), and unpaired Student t test, and P < 0.05 was considered significant.

The data sets and resources (TaqMan probes and SRM-MS peptides) generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request. National Center for Biotechnology Information GenBank accession numbers: INSU1 (MT335687), INSU2 (MT335688), INSUA (MT335689), INSUB (MT335690), INSUC (MT335691), and INS3B (MT335692)

Using the HSF bioinformatic tool (14), we found that the third exon contains a potential intraexonal splicing acceptor (>85% HSF matrix) site 36 bp downstream of the conventional exon 3A splicing acceptor site. We named the intraexon 3 spliced isoform INS3B (Fig. 1A, purple lettering), which translates to a 74-AA proinsulin (i.e., CαK-peptide), within which resides a 19-AA Cα-peptide instead of the classical 86-AA proinsulin with its 31-AA C-peptide encoded by INS3A (Fig. 1A and Fig. 2A). The Cα-peptide (EAEDLQGSLQPLALEGSLQ) does not contain the β-sheet (GQVEL) and hairpin (GGGPG) motifs that are present in the middle potion of the 31-AA C-peptide (21); however, they have identical estimated pI of 3.34 (22). Additionally, the 74-AA proinsulin of INS3B is predicted to have a soluble globular structure (15) without an intrinsically disordered protein region (VELGGGPGAGSLQP) (Fig. 2C). The classical insulin is transcribed from exons 1A (INS1A), 1B (INS1B), and 1C (INS1C), including one with intron 1 retention (INS1I) (Fig. 1A, green lettering). The combination is named INS3A and translates to a 110-AA preproinsulin and can be detected by TaqMan probe INS3A (Fig. 1A, green bar and bracket, and Supplementary Table 3). INS3A (31-AA C-peptide) and INS3B (19-AA Cα-peptide) expression in islets made up 85.2 ± 7.2% (n = 22) and 14.6 ± 2.7% (n = 22), respectively, of the total INS expression (Fig. 1B) measured by using the Exon-2 TaqMan Probe (Fig. 1A, blue bar under exon 2 that exists in all INS isoforms).

Exons 1 and 3 of INS are alternatively spliced, while exon 2 is a constitutive exon, within which a translation initiation site resides (9) (Fig. 1A, green asterisk). We assembled 3,544 human insulinoma and islet expressed sequence tags and found that many of them retain introns 1 and 2. In addition, several cDNA clones contain alternative uTSS upstream of the conventional INS 5′-cap site (21). Furthermore, we found that the uTSS nucleotide sequence contains a consensus human Kozak ribosomal binding site (23) (TGGGAGATGGGC) for an alternative translation initiation 45 bp upstream of the 5′-cap site (Fig. 1, red asterisk). The uORF of INSU1 is in frame with the pORF with retention of intron 1 and could be potentially translated to a 204-AA polypeptide, while the uORF of INSU2 (insulinoma BioSample: SAMN00164222) retains introns 1 and 2, potentially producing a frameshifted 198-AA polypeptide (Fig. 1A, red lettering, and Fig. 2B). The presence of INSU1 clones was validated by Sanger sequencing of IMAGE clones (https://www.imageconsortium.org/) of IMAGp998A2012483Q (BM85746) and IMAGp998O0113413Q (BU782803) in both directions. The uORF containing exons 1UA, 1UB, and 1UC (Fig. 1A, blue lettering) could be potentially spliced to exon 2, generating INSUA, INSUB, and INSUC, encoding polypeptides of 53, 153, and 73 AA, respectively (Fig. 2C).

We detected pORF isoforms that had similar expression levels in both control and T2DM islets (Fig. 3A) by quantitative RT-PCR. All the INS pORF were expressed in both islets and ChP, indicating that ChP contains INS precursor mRNA splicing machinery similar to that of β-cells (Fig. 3B). The total INS (Exon-2 TaqMan Probe) mRNA level was >2,000-fold higher in islets than in ChP. The abundant INS1A, INS1C, and INS3B mRNA levels were ∼1,000-fold greater in islets than in ChP, while the lower-expressed INS1B and INS1I mRNA levels were ∼30-fold higher in islets than in ChP (Fig. 3B). We also found that postmortem ChP from insulin-deficient T1DM and T2DM contained INS isoforms at similar levels (Fig. 3C); we verified by ELISA in the stored blood samples of those who were receiving insulin therapy that C-peptide was absent. Additionally, by ddPCR, we detected uORF spliced isoforms (INSUA, -UB, -UC, and -U1) in both islets and ChP. On the other hand, the nonspliced isoform INSU2 was not detected in ChP (Fig. 3D).

By ddPCR, we compared expression of β-cell autoantigens and transcription factors between islets and ChP. Except for islet cell autoantigen 1 (ICA1), expression of which was 10-fold higher in ChP than in control islets, expression of other known autoantigens (GAD2, PTPRN, and SLC30A8) was 253-, 346-, and 339-fold higher in islets (Fig. 4A), respectively. Endocrine differentiation factor neurogenin-3 (NEUROG3) mRNA levels were similar, while expression levels of several other critical β-cell transcription factors (MAFA, NEUROD1, ISL1, and NKX6–1) were 33-, 344-, 66-, and 27-fold higher in islets (Fig. 4B), respectively. Pancreatic and duodenal homeobox 1 transcription factor (PDX1) and paired box gene 4 transcription factor (PAX4) were absent from ChP (Fig. 4B). Using a triplex RNAscope ISH assay, we found that INS mRNA colocalized with TTR (a thyroid hormone transporter and present in all EChP) in control, T1DM, and T2DM ChP samples (Fig. 5, white arrows), as in mice (3); the IAPP gene was not expressed in ChP. This observation contrasts with pancreatic islets where INS colocalized with IAPP, as expected in β-cells, but did not colocalize with TTR (Fig. 5, yellow arrow); in islets, TTR was present in α-cells, as previously reported (24).

To evaluate the robustness of our SRM-MS assay, we measured A-chain, B-chain, and Cα-peptide in plasma under controlled conditions (i.e., after an intravenous hyperglycemic glucose clamp), during which circulating glucose levels were held constant at each participant’s fasting level plus 98 mg/dL; none had diabetes. Fasting plasma glucose and glucose after 2h-IVG were 88 ± 2 and 179 ± 4 mg/dL, respectively (Fig. 6A). There were significant increases after 2h-IVG, when measured by SRM, in both A- and B-chains (P < 0.0001) (Fig. 6B). ELISA measures total insulin, and as expected, it was also significantly increased after 2h-IVG, in concordance with the SRM results (P < 0.0001) (Supplementary Fig. 2A). The changes in Cα-peptide and CαK-peptide after 2h-IVG were much smaller (P = 0.0579) (Fig. 6C). However, there was a significant difference with ELISA measurement of C-peptide (P = 0.002) (Supplementary Fig. 2B), again as expected, demonstrating that Cα-peptide and 74-AA proinsulin were not as responsive to glucose as the classical C-peptide. We then calculated the sample-matched ratio of CαK-peptide to Cα-peptide and found it to be significantly increased after 2h-IVG (P = 0.019) (Fig. 6D), demonstrating that 2 h after continuous hyperglycemia, a significant amount of 74-AA proinsulin is released, likely from immature granules.

We did not find significant changes in A- or B-chain or uORF U-peptide levels extracted from islets between control and T2DM islets (Supplementary Fig. 3A–C). However, we did find that the amount of Cα-peptide was significantly lower in T2DM islets compared with control islets (P = 0.047) (Fig. 7A). The amounts of the 74-AA proinsulin represented by CαK-peptide were similar in control and T2DM islets (Fig. 7B), although the sample-matched ratio of CαK-peptide to Cα-peptide was significantly increased (P = 0.035) (Fig. 7C) in T2DM islets, implying the proinsulin protease process is compromised in T2DM β-cells. There were no significant changes in Cα-peptide, CαK-peptide, or ratio of CαK-peptide to Cα-peptide in nondiabetic, T1DM, or T2DM ChP samples (Fig. 7D–F). SRM detection of Cα-peptide again demonstrated insulin presence in postmortem ChP from individuals treated with lifelong exogenous insulin and whose peripheral blood lacked C-peptide.

Figure 8A is a representative experiment of three replicates showing IAPP fibrillation dynamics with lag time of 90 min. We observed that Cα-peptide inhibited IAPP fibrillation more efficiently than C-peptide by one-way ANOVA (P = 0.032) (Fig. 8B) and by regression analysis (P = 0.002) (Fig. 8C) in the range of 90–130 min. The ThT reporter was not affected by C- and Cα-peptides themselves (data not shown). Cα-peptide, but not C-peptide, also inhibits β-amyloid fibrillation (Supplementary Fig. 5).

We uncovered novel uORF in specific INS isoforms that are transcribed from an alternative INS upstream promoter and uncovered a shorter 19-AA connecting peptide, Cα-peptide, in islets and ChP. Furthermore, this is the first report of the presence of insulin, either its mRNA or protein, in human ChP of insulin-deficient patients. Using our newly developed SRM-MS method, we validated translation of the novel human INS isoforms, and we quantified the distinctive uORF- and pORF-derived A- and B-chains and processed Cα- and nonprocessed CαK-peptides at low nanogram to picogram per milliliter biofluids. Commercial ELISA cannot distinguish between A- and B-chains, nor between Cα- and C-peptides. The ratio of CαK-peptide to Cα-peptide reflects islet proinsulin processing status and confirms insulin presence in ChP. The limit of quantitation of subnanogram per milliliter of the uORF- and pORF-derived peptides allows for detection of low quantities of insulin that cannot be accurately quantified via ELISA in CSF (25).

The conventional 86-AA proinsulin is prefolded into its correct conformation in the Golgi apparatus, together with C-peptide (26). The N-terminus of the 31-AA C-peptide (EAEDLQVGQ) acts as a chaperone (27) to promote correct insulin folding, while its C-terminus (EGSLQ), in circulation, is known to bind the orphan G-protein–coupled receptor GPR146 (28). Furthermore, and intriguingly, its middle region contains β-sheet (GQVEL) and hairpin (GGGPG) motifs that prevent proinsulin fibrillation (29,30). The novel human Cα-peptide has the same estimated isoelectric point as that of C-peptide (pI 3.34), but it is devoid of those motifs and therefore acts more efficiently than C-peptide as a trans-chaperone against IAPP amyloid formation. The shortened isoform might also even function as a cis-chaperone to mitigate any amyloidogenicity of insulin itself (30). The increased ratio of nonprocessed to processed Cα-peptide in hyperglycemic plasma and T2DM islets implies more proinsulin secretion, a phenomenon already described (31). The expression of proinsulin-processing enzymes (convertase PC2, carboxypeptidase E, and especially convertase PC1/3) is reduced in T2DM islets (32).

Of no less interest, IAPP is not present in ChP, nor is PDX1. PDX1 is a necessary transcription factor for glucose-mediated INS regulation and for IAPP gene expression (33), and its absence explains why glucose does not regulate INS (3) and why IAPP is lacking in ChP. IAPP packaging in granules with insulin within β-cells points to its later evolution in islets, most likely at approximately the same time as INS expression in the periphery became confined to β-cells. Teleologically, it is indeed a good thing that IAPP is absent in ChP, because IAPP-based amyloidosis, which was described as far back as 1901 by Opie (34) as being pathological to islets of patients with T2DM, would in all likelihood be detrimental to neuronal function. It is a highly fibrillogenic peptide, and it is not known exactly how its fibril formation is inhibited within granules. Within granules, as seen on electron micrographic images, insulin is stored in the core and IAPP in the halo region. C-peptide within proinsulin is also in the halo and, like insulin itself, inhibits IAPP fibril formation (35). The alignment of Cα-peptide would place it in the halo, and therefore, it likely has a role in preventing not only insulin but also IAPP-based fibril formation. Decreasing amounts of Cα-peptide in T2DM might therefore make IAPP deposition more likely in β-cells. Cα-peptide therapeutics could possibly prevent IAPP-based islet amyloidosis.

Over the span of evolution, and with the enlargement and increased sophistication of neural networks, neurotrophic factors such as BDNF and GDNF that are not present in invertebrates (36,37) made their appearance and likely superseded and/or added to the many original insulin growth functions in the brain. Insulin gene duplications and deletions also occurred many times (38) and were permissive to the appearance of the gene in vertebrate endodermal cells such as β-cells (39), with residual remains in the ectoderm (40) such as EChP. We have already reported three isoforms in mice, Ins2-V1, -V2, and -V3, and we recently uncovered an additional mouse Ins2 isoform, Ins2-V4 (Supplementary Fig. 4A), in which uTSS (without uORF) is encoded by exon 1A, which is located almost 20 kb upstream of the classical exon 1B. The Ins2-V4 isoform mRNA was not detected in mouse islets but was seen in mouse ChP and total pancreas (Supplementary Fig. 4B), providing us with another example of selectivity in isoform expression. It is now accepted that during evolution, the endocrine and nervous systems overlapped. Phylogenetically, neurons preceded typical endocrine glands; neurons first emerged in cnidarians, while endocrine glands were first seen in primitive vertebrates, such as the hagfish, the insulin of which is surprisingly similar to that of humans (41). Vertebrates benefitted greatly in evolution by separation of metabolic properties of islet-derived insulin in peripheral organs, thereby allowing increases in body size and storage of energy supplies, such as glycogen and fat, for later use by the evolving brain. However, remnant brain INS is still provided by ChP, but its regulation, unlike islet-derived insulin, is not by glucose, because ChP is devoid of PDX1 and PAX4 transcription factors; ChP insulin synthesis and secretion are regulated by a neurotransmitter, serotonin, as we stated earlier, through the 5-HT2C receptor (3).

If ChP-derived insulin is critical for certain postnatal brain functions (which have not been adequately determined) or has any biological relevance independent of peripherally derived insulin, it seems reasonable to assume that loss of this source of insulin, should insulin-containing EChP be subject to autoimmune destruction or other critical insults, would have detrimental effects. Additionally, EChP, which are terminally differentiated cells (42), have many essential functions, not least of which are transportation of thyroid hormone into brain via TTR and formation of CSF. Yet, distinct central nervous system effects other than what can be ascribed to classical disordered metabolism are not hallmarks of T1DM, not even prior to initiation of insulin therapy or during the insulitis phase of the disease. Moreover, genetic susceptibility to T1DM does not confer certainty of its development, and it is presumed that environmental factors are also required; EChP might be protected, even though they express low levels of triggering antigens. Regardless, it seems clear that EChP still maintain insulin expression when there is peripheral insulin deficiency.

In summary, we now report that insulin is produced in human ChP. We uncovered a novel cis-chaperone and nonamyloidogenic Cα-peptide, and we conclude that proinsulin is produced and processed in EChP in a similar manner to β-cells. However, IAPP and PDX1 are not expressed by EChP. We also introduce a novel MS-based assay for quantifying low amounts of proinsulin products.

Acknowledgments. The authors thank Drs. Ruin Moaddel, Zhixing Yuan, and Maire E. Doyle, Laboratory of Clinical Investigation, National Institute on Aging, National Institutes of Health, for critical review and input for the liquid chromatography with tandem MS–SRM methods, results, and images. The authors thank Professor Daniel Raleigh, Stony Brook University, for invaluable advice on the IAPP ThT fibrillation assay, without which we could not have carried out the experiments in Fig. 8 and Supplementary Fig. 5.

Funding. Q.-R.L., M.Z., P.Z., C.H.M., N.S.H., D.L.L., Q.C., J.F.O., L.F., and J.M.E. are supported by the Intramural Research Program (IRP) of the National Institute on Aging, and P.A. and S.M. are supported by the IRP of the National Institute of Mental Health (ZIC MH002903). Postmortem pancreata were provided by the National Institute of Diabetes and Digestive and Kidney Diseases–funded Integrated Islet Distribution Program at City of Hope (National Institutes of Health grant 2UC4DK098085) (https://iidp.coh.org/). The tissue/data used in this research were obtained from the Human Brain Collection Core, IRP, National Institute of Mental Health (https://www.nimh.nih.gov/hbcc).

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

Author Contributions. Q.-R.L. and J.M.E. conceptualized the study and wrote the manuscript. Q.-R.L. performed bioinformatic search and molecular biological studies. M.Z., P.Z., and L.F. performed proteomic SRM experiments and data analysis. C.H.M. dissected mouse ChP and performed mouse gene expression assay. N.S.H. and D.L.L. performed RNAscope ISH of human islets and ChP. Q.C. performed ELISA experiments. P.A. and S.M. dissected human postmortem ChP samples, extracted clinical information, and organized Supplementary Table 1. J.F.O. arranged human islet collection and dissected mouse pancreatic islets. C.W.C. performed the hyperglycemic clamps and plasma collection. All authors reviewed and approved the publication of this manuscript. J.M.E. 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.