Amylin aggregation

In adults within five years of type 2 diabetes diagnosis and without DPN at entry, circulating pancreatic human islet amyloid polypeptide (hIAPP, or amylin) supplies aggregates to connected peripheral sensory compartments, or supplies peptide that aggregates there. Selectively removing these local assemblies would reduce five-year physical loss of peptidergic C-fiber terminals by at least 20% in a baseline-defined group accounting for at least 10% of incident DPN cases in this reference population.

Amylin is a peptide hormone secreted with insulin. The specified molecular family is assemblies containing at least two copies of mature, full-length human IAPP, with its native disulfide bond and amidated end, including soluble oligomers, protofibrils and mature fibrils. “Aggregate” does not mean “toxic”: harmless assemblies are included. Free monomer, proIAPP, peptide fragments and synthetic analogues such as pramlintide are outside this intervention. Distinct aggregate conformations may have different effects; success does not require every member to be harmful.

The relevant locations are the axon and terminal, their dorsal-root-ganglion parent neuron, associated satellite and Schwann cells, the innervated skin interstitium, and the tissue side of supplying microvessels. A deposit in the dermis qualifies only if its exposure connects to the axonal territory under assessment; proximity somewhere in a biopsy is insufficient.

Q operates from enrollment through year five. It continuously removes at least 95% of these local assemblies relative to the person's untreated trajectory, including new arrivals and assemblies forming locally. Removal destroys rather than fragments the assemblies into smaller oligomers. Free-monomer availability is matched to reference, and Q directly changes neither pancreatic secretion, systemic glycemia, insulin, weight nor other amyloid proteins. It is a defined selective-clearance comparison, not an assumed effect of an antibody that also improves pancreatic function. Local downstream inflammatory, vascular, metabolic and repair responses can change, and their benefits and costs count in the terminal outcome. Maintaining normal free hormone exposure is not the same as presuming aggregates have no receptor effects.

The finite baseline groups are: all eligible adults; body mass index at least 30 kg/m²; triglycerides at least 150 mg/dL; HbA1c at least 8%; fasting total plasma IAPP at least 20 pM; or local cutaneous aggregate burden above the age- and site-matched nondiabetic 95th percentile. The plasma threshold is a chosen descriptive boundary, not a toxicity threshold. Group membership does not depend on later treatment response. A single group must satisfy all requirements; success in different groups at successive steps cannot complete the claim.

For exposure, use at least 1 pM in IAPP monomer equivalents within assemblies, also above the matched nondiabetic 95th percentile, sustained for at least 30 days in year one. It must occur in connected compartments serving at least 10% of sampled terminal units in at least 10% of that group's participants, before any additional 5% net terminal decrease over a 30-day interval. These are operational boundaries for the priced hypothesis, not established biological thresholds. The quantity is local aggregate-associated peptide, not total circulating IAPP or antibody-positive spots per millimeter.

Let L0 be mean five-year physical terminal loss in a group under reference care, and LQ the corresponding loss with Q. Benefit B = (L0 − LQ)/L0, evaluated only where reference net loss is positive. Follow the same baseline territories and count disappearance and replacement; stable lineage identity and anatomy must distinguish peptidergic C-terminals from changing peptide or PGP9.5 expression. The independent participant is the inference unit. Deaths or missing territories cannot be counted as zero loss or silently excluded to select survivors; if unresolved outcomes could reverse the materiality decision, the result remains unresolved.

The coverage fraction is the share of reference-arm incident DPN cases belonging to the successful baseline group. It is not the group's prevalence among all people with diabetes, the proportion of positive biopsies, or the proportion responding to treatment.

Source assessment

Best judgment: about 13%; sensitivity range: 0.6–59%. The probability concerns the complete human causal proposition below, not the fraction of patients affected or the benefit of an available drug. Chronic hIAPP-expressing mice provide a substantial structural positive without hyperglycemia. The principal uncertainty is whether naturally formed aggregates expose human sensory tissue before denervation at a dose whose removal would materially preserve terminals.

Why the mechanism could operate

A circulating hormone can become locally damaging if aggregation, delivery and clearance create persistent assemblies at vulnerable membranes or support cells. Membrane disruption, altered receptor signaling, inflammatory responses and organelle dysfunction are alternative routes. They are not all necessary serial links.

The exposure could also be incidental: deposits may accumulate after tissue injury, mark diabetes generally, or be relatively inert end products. Pancreatic aggregate removal could benefit nerves indirectly through improved metabolic control. Conversely, weak blood associations would not exclude concentrated or retained tissue exposure. Neither total plasma concentration nor the mere presence of amyloid settles the relevant comparison.

Conditional probabilities and uncertainty

Each probability after row 1 assumes the entire preceding prefix holds for at least one of the same baseline groups. A later row asks whether a group satisfying that prefix also satisfies the new condition.

See the claims and probabilities table above.

R1

Human skin localization is cross-sectional, unquantified as a local dose and confounded by age and anatomical site.

R2

Chronic transgenic and local aggregation-control experiments support injury; selective removal at natural human exposure is untested.

R3

Short experimental denervation can recover; duration, compensation and exact human fiber identity remain open.

R4

Six selected established-DPN biopsies do not estimate early exposure prevalence or attributable incident disease.

Best inputs [60,65,55,60]% give 0.60 × 0.65 × 0.55 × 0.60 = 12.87%. Lower [30,25,25,30]% and upper [90,90,85,85]% interpretations give 0.5625% and 58.5225%. The range reflects jointly skeptical or favorable interpretations of exposure, causal transfer and coverage. It is not a confidence interval, a probability distribution or a variance estimate.

The favorable interpretation treats chronic mouse exposure and human deposits as evidence of a persistent local human toxin. The skeptical interpretation gives more weight to unmatched human tissue, distinct aggregate preparation conditions, and the absence of selective chronic rescue. These uncertainties can move together. Merging rows 2–3 into a single conditional at 35.75% leaves the complete best probability unchanged; the row count itself carries no penalty.

Evidence for and against the complete claim

The chronic mouse result survives a donor-level check. Albariqi and colleagues studied non-obese mice expressing hIAPP in pancreatic beta cells. Their weight, glucose and insulin were similar to wild-type controls, while circulating IAPP increased. At 16–18 weeks, the six hIAPP mice had approximately 25.4 versus 47.1 epidermal fibers/mm in six controls, a 46% lower endpoint abundance. Reading all plotted points gives an unadjusted difference of −21.7 fibers/mm, approximate 95% CI −33.8 to −9.6. A two-pixel uncertainty in every point changes the mean difference by at most 0.5 fibers/mm. This is a substantial structural positive, not merely a pain result. It is an endpoint comparison rather than a measured five-year loss fraction. The mice came from separate transgenic and non-transgenic breeding lines, and no adult selective aggregate-removal intervention reverses this chronic phenotype in the study. Albariqi et al., 2023, Figure 2 and supplemental Figure 2; graphical check (background note outside this collection).

Adding hIAPP to obese diabetic mice worsened mechanical hypersensitivity, but the hIAPP-obese animals also had more severe diabetes. That comparison does not cleanly attribute additional structural loss to peptide aggregation at matched metabolic exposure. The normoglycemic comparison is the more informative part of the chronic evidence.

Local injection partly closes the dose gap, but not the aggregate-formation gap. The intravenous experiment used 40 µg/kg, calculated by the paper as about 125 nM in blood, roughly 3,000 times its cited human T2D concentration of 42 pM. Its measured endpoint was mechanical sensitivity. The decisive local structural experiments instead used 1,000 femtograms in 5 µL, equivalent to about 51 pM in the injectate. Calling this a calculated plasma concentration is incorrect: distributing that entire mass into 2 mL blood would give only 0.128 pM. Neither calculation measures concentration at a nerve membrane. Local injection reduced IENFD, with no detectable residual deficit at day 27 when mechanical hypersensitivity had resolved. Albariqi et al., Figures 3–5 and Methods.

The aggregation-inhibitor experiment incubated hIAPP at 50 µM overnight with or without 1 mM anle145c, then diluted it one million-fold to 50 pM before injection. The inhibitor prevented the subsequent pain and IENFD effects. This supports dependence on the prepared aggregate state. It does not demonstrate that the same assemblies form from freely circulating peptide at 50 pM in human tissue, or that treating an established chronic exposure clears them. Stock handling, dilution, surfaces and local retention matter to that inference.

Nonaggregating mouse IAPP and the less amyloidogenic analogue pramlintide supplied useful comparators. They did not reproduce the hIAPP structural effect. In culture, 100–1,000 nM hIAPP reduced neurite growth by about 12–13%, without neuronal death at 24 hours. Those concentrations are far above the cited bulk human range. Outgrowth experiments used eight donor cultures; the mitochondrial reactive-oxygen panel used three donor cultures, despite showing 551–654 cells per arm. Cell counts do not provide hundreds of independent biological replicates. Primary experiments.

The human comparison does not isolate DPN-specific aggregate exposure. Six T2D participants with neuropathy had lower IENFD and more double-positive IAPP/oligomer spots than nine nondiabetic controls. Median ages were 59 versus 47 years. All DPN tissue came from feet; controls included hands and feet, and both groups mixed plantar and dorsal locations. There was no diabetes-without-neuropathy group, no prospective timing and no measured local molar aggregate concentration. The donor/site identities needed to adjust these differences are unavailable. Blinded double staining supports localization, but does not by itself identify an aggregate's molecular composition, circulating origin or toxicity. Primary study and supplemental Table 1.

Receptor blockade is informative but incomplete. CGRP8–37 blocked CGRP-induced hypersensitivity while failing to block the hIAPP response. This tests the antagonist-sensitive pathway under those conditions. It does not establish equivalent blockade of every amylin receptor subtype: primary receptor pharmacology finds weaker, subtype-dependent antagonism of amylin responses. The aggregation comparisons support the claim without requiring a blanket assertion of receptor independence. Albariqi experiment; Hay et al., receptor pharmacology.

Selective aggregate targeting is possible, but nerve preservation has not been demonstrated with it. Wirth and colleagues developed an antibody preferentially binding hIAPP oligomers, with aggregate affinity around 1.54 nM versus 2.84 µM for monomer. In transgenic rats, treatment reduced pancreatic oligomer staining by up to 68%, with improved beta-cell preservation and glucose control. The active antibody recruited macrophage clearance; an Fc-inert version did not reproduce these effects. This supports a selective targeting concept, while also showing that a real antibody changes inflammatory and metabolic pathways. The experiments do not supply a peripheral-nerve endpoint or the local-only Q comparison. Wirth et al., 2023.

A publicly posted 2026 poster abstract from the Albariqi/Eijkelkamp research program adds cell stress and ganglion transcriptomics, including 19 differentially expressed genes. It provides no sample sizes or new controlled human structural contrast. It therefore cannot be counted as independent clinical replication. Chan et al., public IASP abstract.

The 225-participant APOLLO trial establishes that suppressing a different amyloid precursor, transthyretin, can improve human neuropathy: the 18-month modified neuropathy-score difference was −34 points. This is a capability anchor for amyloid-associated neuropathy, not an estimate for hIAPP, a small-fiber structural endpoint, or a diabetes-prevention result. Adams et al., 2018.

Ideal experiments that would resolve each claim

These define truth conditions for the proposition. The common ideal observation follows independent people in the stated early-T2D population through five years, with repeated measurements in the same anatomical territories and a selective Q-versus-reference comparison. It identifies aggregate-associated mature hIAPP independently of a nonspecific oligomer antibody, verifies free-monomer preservation and 95% local clearance, and measures physical peptidergic terminal turnover.

For each row, YES requires a precise result satisfying every stated subcondition in at least one group extending the earlier prefix. NO requires excluding it for every group that satisfies that prefix. Boundary-crossing intervals or missing dose, timing or endpoint information are unresolved. Later conditional claims cannot be estimated simply by observing that an earlier requirement failed.

1 — Source, local exposure and timing. Establish pancreatic-to-blood-to-sensory-tissue delivery of mature hIAPP, distinguishing imported assemblies from local assembly of imported monomer and from peptide produced within nerve tissue. Quantify local aggregate-associated peptide and matched nondiabetic reference levels. YES requires the 1 pM, 95th-percentile, 30-day and 10%-territory/participant conditions before the stated additional-loss boundary in year one. NO requires adequate measurements excluding this conjunction across all six candidate groups. A late deposit, total plasma assay or unquantified double-positive spot is unresolved. The pancreatic transgene supports source direction in mice; the human skin samples address only late localization.

2 — A nontrivial causal structural effect. In groups meeting row 1, compare Q with an otherwise identical reference: verify local aggregate depletion, unchanged free hormone exposure and no direct pancreatic or systemic metabolic manipulation. Follow actual terminal disappearance and replacement for five years. YES requires B≥5%; NO requires a precise upper bound below 5% in every exposure-positive candidate group despite engagement. A fall in staining caused by epitope masking, analgesia without retained terminals, or benefit explained solely by improved glycemia remains unresolved for Q. Premixed anle145c and peptide-variant comparisons support aggregation dependence at the imposed exposure, while chronic transgenic loss supplies a complementary structural result. None directly resolves selective human clearance.

3 — Materiality over five years. Use the same intervention, population and physical outcome as row 2, now testing B≥20%. YES requires a lower uncertainty bound at or above 20% for a group satisfying the prefix; NO requires upper bounds below 20% for all such groups. This is a stronger magnitude claim, not another mechanism. A precise 10% benefit could satisfy row 2 while failing row 3. Reference loss too small to identify a relative effect, unresolved deaths, or effects confined to short experimental denervation leave the comparison unresolved. The roughly 46% chronic mouse endpoint difference is not a human preservation fraction, and recovery after local injection limits a simple permanent-injury extrapolation.

4 — Coverage of incident disease. Identify incident DPN using a consistent clinical definition in a representative reference cohort and determine what fraction of those cases belongs to a baseline group satisfying rows 1–3 under Q. YES requires at least 10%; NO requires every structurally responsive group to account for less than 10%. A post hoc responder group or the fraction of selected DPN biopsies staining positive cannot substitute for this measurement. Existing human tissue studies provide no usable incident-case denominator.

Evidence updates and remaining uncertainty

A hypothetical age- and site-matched human series demonstrating quantified pre-loss exposure, followed by selective local clearance with durable physical preservation, would raise several links together. Inputs [85,85,80,80]% would give 46.24%. Finding more deposits in established DPN without timing or a causal comparison would warrant a much smaller update.

A hypothetical result showing that matched early-DPN tissue lacks the specified exposure, and that assemblies produced under natural conditions are inert despite verified removal, would favor [25,30,30,35]%, giving 0.7875%. A null transcript signature or absent public trial report would not supply that evidence.

The available primary evidence supports a real experimental route from aggregation-prone hIAPP to denervation. It does not determine natural human aggregate dose, onset timing, prevention effect or case coverage. The most useful currently missing evidence is a matched, prospective human exposure–structure comparison or selective chronic nerve rescue; the public poster and pancreatic antibody studies do not provide it. Five-point differences in this probability are not reliable distinctions.

The mechanism may act through redox injury, organelle maintenance, inflammation or calcium entry. These are overlapping routes, not additive probabilities. Its aggregate-clearance intervention differs from C-peptide replacement and insulin trophic support, even though all concern pancreatic peptides.