# 28. Excess PARP1 catalysis contributes to terminal loss

**Best judgment: about 24%; skeptical–favorable sensitivity: 1.5%–65%.** These are subjective probabilities of the complete human proposition, not predicted treatment efficacy. Diabetic animal skin preservation supports a causal role, while early human localization, selective intervention and durable magnitude remain uncertain. Reviewed 14 September 2026. [Probability method](probability_method.md) · [Collection index](README.md).

## The causal claim and its scope

In adults with type 2 diabetes diagnosed within five years and no clinical DPN at entry, excessive **PARP1-dependent ADP-ribosylation** in the connected peripheral sensory unit contributes to physical peptidergic C-fiber terminal loss. Selectively limiting that excess from entry would avert at least **20% of five-year net terminal loss**, in a baseline-defined group accounting for at least **10% of reference-care incident DPN**.

The route includes NAD consumption and the effects of ADP-ribosylation on metabolism, inflammatory signaling and cellular maintenance. An obligatory collapse of NAD or ATP is not part of the proposition. PARP1 protein abundance, cleaved PARP during apoptosis and accumulated poly(ADP-ribose), or PAR, are distinct from catalytic flux.

**Connected compartments.** The finite sensory unit comprises the identified peptidergic neuron, its directly ensheathing nonmyelinating Schwann cells, its somal satellite glia, and endothelial cells of the microvessels supplying that mapped unit. Quantify PARP1 reactions in each cell class, including the neuronal nucleus. An axonal NAD deficit need not originate from PARP1 physically located in the terminal. Blood, spinal-cord or unrelated skin-cell activity alone does not satisfy the claim.

Identify adult peptidergic unmyelinated neurons through combined molecular and anatomical classification, then track their distal cutaneous arbors physically. Stable subtype assignment must survive changes in CGRP, PGP9.5 and reporter expression. For each person, net terminal loss is destruction minus successful replacement over five years, divided by baseline terminal amount. Average people equally and define:

`B = (mean reference net loss − mean net loss with Q) / mean reference net loss.`

Reference mean loss must be positive. Reference care is contemporary diabetes care without Q. Pain, conduction and soma survival are additional measurements, not substitutes for physical terminal persistence, disappearance and replacement.

**Finite candidate groups.** Evaluate the whole eligible population, males, females, baseline BMI ≥30 kg/m², baseline HbA1c ≥7.5%, and a baseline PARP1-activity-positive group. The last requires at least one hour of qualifying activity during a 30-day entry observation in at least 10% of mapped units. No additional intersections or groups defined by future response qualify.

Incident DPN means new bilateral length-dependent peripheral nerve signs with objective small-fiber loss or abnormal conduction, excluding another cause. Coverage is `C = P(group membership | incident DPN under reference care)`. A single same group must satisfy the whole chain.

**Qualifying excess.** Measure PARP1-specific incorporation of ADP-ribose from NAD+ into cellular acceptors, in amount per relevant cellular volume per hour. Track newly formed mono- and polymeric products separately from their removal. A qualifying rate is at least twice the positive age-, sex- and site-matched nondiabetic median and above its 95th percentile for that cell class and compartment.

In at least 10% of mapped units, with equal person weighting, one same cell class must accumulate at least **24 hours of qualifying activity during year one**. The activity must precede the connected arbor's first **≥5% net terminal deficit from entry maintained for 30 days**. Different classes can qualify in different candidate groups, but observations in disconnected tissues cannot be assembled into one route. These activity, duration and timing boundaries are operational choices, not measured injury thresholds.

**Q — limit excess catalytic flux.** An ideal local feedback intervention caps PARP1 catalytic flux at the matched nondiabetic 95th percentile whenever the unrestricted rate under the cell's current state would exceed it. Lower rates remain unchanged. Q operates from entry for five years in the four named cell classes, with at least 95% of the above-bound excess removed.

Q changes catalytic reaction probability without directly changing PARP1 abundance, DNA binding, other PARPs, NAD synthesis or PAR removal. It does not impose drug-induced DNA trapping. Any altered DNA retention or repair resulting from changed catalysis remains a downstream consequence and counts. The ideal cap requires capabilities beyond current systemic drugs or constitutive deletion.

Systemic glycemia, lipids, medications, pressure and other initiating exposures remain comparable. DNA damage, NAD supply, inflammation, blood flow, autophagy and repair may respond downstream. Q does not selectively retain whatever later proves beneficial: the cap also limits high-rate repair responses, and their costs belong in the structural outcome.

## Biological reasoning

PARP1 uses NAD+ to modify itself and other proteins. The immediate bookkeeping is:

`NAD change = synthesis + net transport − PARP1 consumption − other consumption.`

`PAR change = polymer formation − degradation and removal.`

A high PAR signal can reflect increased formation, slower removal, or both. High NAD consumption need not deplete the pool if synthesis compensates; a low pool does not identify its consumer. Restoring a redox ratio does not establish restoration of absolute NAD concentration.

The finite causal comparison is:

`DNA and other cellular stresses → excess PARP1 catalysis → changes in nucleotide supply, protein modification and signaling → destruction minus replacement.`

The intermediate effects can coexist. The proposition is the total structural effect of the specified catalytic cap, not a claim that every intermediate must occur. Thus a failure to find GAPDH ADP-ribosylation does not exclude every PARP1 route. Conversely, no untested intermediate can rescue the proposition if a complete, selective Q comparison has B below the stated boundary.

Nuclear NAD consumption can alter somal maintenance and axonal supply; glial or endothelial effects can alter the neuron's environment. None of those connections is instantaneous equilibration with distal axonal NAD. NAD resynthesis also consumes resources, with costs depending on precursor and recycling routes. Bulk nerve phosphocreatine, ATP and calculated redox ratios cannot identify those compartmental fluxes.

PARP1 participates in repair and transcription. Excess activity might be damaging, adaptive, or both. Limiting it could spare metabolic resources yet impair recovery from ongoing DNA damage. Other NAD consumers, including SARM1 and CD38, may sustain injury after PARP1 is capped. These possibilities lower confidence in durable net preservation rather than becoming mandatory independent discounts.

## Evidence that moves the judgment

**Diabetic skin structure supplies the strongest positive.** Obrosova's study gave male STZ-diabetic rats GPI-15427, 30 mg/kg/day, for ten weeks after two untreated weeks. Skin arms had **6–9 animals**. Results reported 47% diabetic fiber loss and treatment to 94% of control; the discussion instead said 44% loss. Either implies approximately **86–87% deficit reduction**, calculated from rounded prose, not an animal-level interval.

A parallel constitutive Parp1 knockout experiment had **8–11 mice per skin arm** and reported preserved density against roughly 46% loss in diabetic wild types. Knockouts required two or three extra STZ injections and started heavier; final glucose was 24.6 versus 27.7 mmol/L. The rat glucose comparison was 24.5 versus 26.1 mmol/L. Persistent severe hyperglycemia strengthens the pharmacological result, but nonsignificant differences do not prove identical metabolic exposure. Global deletion, different backgrounds and systemic drug effects do not isolate Q. PGP9.5 crossing density also does not identify peptidergic terminal survival independently of marker expression. [Obrosova et al., 2008](https://pmc.ncbi.nlm.nih.gov/articles/PMC3057075/).

Drel's related GPI-15427 study supported large myelinated tibial-axon caliber and conduction in STZ rats. It did not supply another epidermal C-fiber rescue result. Similar investigators and protocols should not be counted as an independent laboratory replication; cohort independence across the related reports was not established here. [Drel et al., 2010](https://pmc.ncbi.nlm.nih.gov/articles/PMC2875829/).

**The early metabolic experiment does not directly show NAD depletion.** Li's male Wistar study used PJ34, 30 mg/kg/day, in weeks three and four after STZ. Biochemical groups had **6–9 animals**. Whole-nerve ATP was 0.800 μmol/g untreated and 0.807 treated, whereas phosphocreatine rose from 2.45 to 3.19 μmol/g: **+0.9% versus +30.2%**. The reported cytosolic and mitochondrial NAD+/NADH ratios were calculated from other metabolites; absolute NAD, ADP, AMP and phosphate were not measured. Near-normalized conduction is a functional result. It cannot establish five-year terminal preservation or assign the metabolic effect to axonal PARP1. [Li et al., 2004](https://doi.org/10.1007/s00125-004-1356-0).

**Human localization is vascular and cross-sectional.** Szabó studied 21 controls, 22 people with parental diabetes history, 23 with impaired glucose tolerance and 21 with type 2 diabetes. Forearm skin staining showed more PAR-positive endothelial nuclei in the parental-history and diabetic groups. That supports an early human vascular association, not directly measured PARP1 flux in a connected distal sensory unit, NAD depletion in C fibers or incident-DPN prediction. [Szabó et al., 2002](https://pubmed.ncbi.nlm.nih.gov/12438293/).

**Cell experiments support a conditional metabolic effect.** In immortalized mouse IMS32 Schwann cells, rucaparib restored measured NAD/NADH pools, GAPDH activity and glycolysis under high glucose without added pyruvate. The dependence on pyruvate matters: glucose alone with pyruvate did not cause the same rapid death, and osmotic controls did not reproduce it. These are culture experiments, not independent human donors. [Yako et al., 2021](https://pmc.ncbi.nlm.nih.gov/articles/PMC8460646/).

The follow-up used 1 μM rucaparib. Viability assays used 15 mM glucose, whereas dense-cell ATP-rate assays used **100 mM** glucose; those conditions cannot be silently combined into one human exposure. Rucaparib restored glycolytic but not mitochondrial ATP production or PDH activity. The investigators did not find increased GAPDH ADP-ribosylation or the expected PAR/AIF pattern for parthanatos. ATP-rate results represented 9–10 reported experiments, not donors. These observations support a metabolic branch while challenging an obligatory GAPDH-modification or parthanatos chain. [Yako et al., 2024](https://pmc.ncbi.nlm.nih.gov/articles/PMC11508270/).

**Newer DPN reports do not close the structural gap.** Yuan's db/db study reports PARP-inhibitor effects on DRG mitophagy, mitochondrial injury and pain. The accessible primary abstract does not establish epidermal terminal preservation; full methods and animal counts were unavailable for this review. The same access limit applies to the 2025 intrathecal PARP1-siRNA/chitosan study: its abstract reports functional and molecular improvements after STZ, without a quantified epidermal endpoint. These support possible routes, with less weight than the accessible skin experiment. [Yuan et al., 2022](https://doi.org/10.1111/jnc.15606), [MoqbelRedhwan et al., 2025](https://pubmed.ncbi.nlm.nih.gov/39565138/).

A 2026 nerve-constriction study adds PARP1 perturbation and GPX4-associated sensory-neuron evidence, but its methods specify **Nestin-Cre**, not an adult DRG-exclusive manipulation. Comparing DRG with heart, liver and lung does not establish exclusion of other neural tissues. Drug arms had six mice each; reported outcomes were pain, excitability, iron and oxidation markers, without quantified terminal preservation or directly tracked ferroptotic death. This supports an adjacent neural mechanism without closing the diabetic structural gap. [Guo et al., 2026](https://www.nature.com/articles/s41420-026-03307-4).

**Human pharmacology is not uniformly neuroprotective.** The GOLD gastric-cancer trial compared olaparib plus paclitaxel with placebo plus paclitaxel. Posted nonserious-event counts were 41/262 versus 42/259 for “neuropathy peripheral”: risk ratio **0.97, approximate 95% interval 0.65–1.43**. The separate sensory-neuropathy code was 41/262 versus 29/259: **1.40, 0.90–2.18**. Do not sum these potentially overlapping codes.

This is a randomized human caution against universal protection, not an engaged test of selective PARP1 capping in early diabetes. It includes cancer, chemotherapy, differing exposure time and nonstructural adverse-event ascertainment. No accessed study supplied an early-T2D structural intervention with local PARP1 engagement; the targeted ClinicalTrials.gov search found no matching DPN intervention trial. [GOLD posted results](https://clinicaltrials.gov/study/NCT01924533?tab=results).

## Probabilities of the logical claims

Each row after R1 assumes **all preceding rows are true** for the same group. These are conditional beliefs, not independent marginal probabilities.

| ID | Additional proposition | Skeptical / best / favorable | Best cumulative |
|---|---|---:|---:|
| R1 | Qualifying excess PARP1 flux occurs in the specified connected unit before sustained terminal deficit. | 35 / **70** / 90% | 70% |
| R2 | Given R1, Q produces five-year structural benefit B≥5%. | 35 / **70** / 90% | 49% |
| R3 | Given R1–R2, benefit reaches B≥20%. | 40 / **75** / 95% | 36.75% |
| R4 | One same group satisfying R1–R3 accounts for C≥10% of reference-care incident DPN. | 30 / **65** / 85% | 23.8875% |

R1 is more likely than not because diabetic tissue perturbations and early human vascular staining align, but direct flux, compartment and timing remain inferred. R2 credits large diabetic skin effects conditional on real early excess; it remains uncertain because existing interventions also change systemic and noncatalytic biology, and repair costs can offset benefit.

R3 is relatively favorable once a structural effect exists: the rodent effect is much larger than the chosen 20% boundary. Duration, subtype and continued human metabolic injury justify a wide range. R4 reflects a potentially shared stress response without measured case coverage; knowing the route exists in a group does not establish that group's share of incident disease.

A B of 10% passes R2 and fails R3. B=30% with C=3% passes R3 and fails R4. Neither NAD depletion nor identifying a particular intermediate is an extra requirement of this total-effect proposition.

## Ideal experiments that would resolve each claim

Assume noninjurious longitudinal human reaction and structure measurements, durable subtype identification and selective adult catalytic control. Independent people are the inference units. Complete measurement at unlimited precision defines the truth conditions; finite estimates spanning a boundary remain unresolved.

**R1 — excess activity before loss.** Sample eligible adults and matched nondiabetic references representatively. During entry and year one, directly quantify PARP1-specific ADP-ribose incorporation and removal in the mapped neuron, ensheathing Schwann cells, satellite glia and supplying endothelium. Resolve enzyme identity with rapid selective catalytic interruption or equivalent reaction tracing; measure absolute NAD and other consumers separately. Establish the rate boundary, 24-hour cumulative duration, ≥10% unit fraction and ordering before the sustained 5% deficit. YES requires one candidate group and cell class meeting all conditions; NO means none does. Accumulated PAR without formation/removal rates, late activation alone, disconnected tissue or inadequate temporal sampling remains insufficient. This establishes exposure, not its structural harm. Existing studies supply staining and perturbation evidence, not the complete human measurement.

**R2 — net structural causation.** Assuming R1, randomize people in the same baseline groups to Q or reference care for five years. Verify that above-bound PARP1 flux is capped with the stated fidelity in every named class, lower rates and other enzymes remain directly unaltered, and no independent DNA-trapping action occurs. Compare an independent cap implementation and restoration of intervention-resistant catalytic activity at endogenous abundance to distinguish pathway effects from intervention artifacts. Keep initiating metabolic exposures comparable while allowing all downstream metabolism, vascular changes, DNA repair and replacement to respond. Count physical terminal destruction and successful replacement, not merely staining. YES is positive reference mean loss and B≥5% in a group passing R1; NO is B<5% in every such group. Incomplete local engagement, altered systemic glucose or insufficient five-year precision leaves it unresolved. Existing animal comparisons support structure but do not implement this human cap.

**R3 — material magnitude.** Assuming R1–R2, use the same randomized human comparison and denominator. Estimate B precisely enough to distinguish 20%, including delayed losses and replacement throughout five years. YES requires B≥20% in at least one still-qualifying group; NO means every such group has 5%≤B<20%. A large short-term metabolic rescue cannot settle this claim. Follow-up loss, marker-dependent classification or an interval spanning 20% remains unresolved. Existing rat percentages support a large possible effect but provide neither this time window nor the intended subtype.

**R4 — case coverage.** Assuming a group passes R1–R3, follow a representative reference-care population for five years and ascertain incident DPN uniformly. Estimate that group's fraction of all incident cases, retaining the original entry definitions and resolving outcomes after dropout. YES requires C≥10% for at least one group already passing the preceding rows; NO means all such groups have C<10%. Exposure prevalence, cancer-trial neuropathy and responder-enriched recruitment do not answer this denominator. An interval crossing 10% remains unresolved. No accessed PARP study directly supplies this population quantity.

## Uncertainty, sensitivity and overlap

Best inputs **[70,70,75,65]%** multiply to **23.8875%**. Skeptical **[35,35,40,30]%** and favorable **[90,90,95,85]%** give **1.47% and 65.4075%**. The range covers coordinated interpretations of the same evidence, not an effect-size interval, calibrated posterior or sampling distribution.

A restricted, model-dependent interpretation, [50,50,55,45]%, gives **6.19%**. A common local-stress route, [85,85,85,80]%, gives **49.13%**. The interpretations differ jointly in actual early flux, intervention fidelity, durable magnitude and coverage. No empirical scenario weights or probability variance are available.

Hypothetically, compelling early human flux evidence moving R1 to 90%, with other judgments fixed, raises the headline to **30.71%**. Representative negative measurements moving R1 to 30% lower it to **10.24%**. Settling R1 true leaves **34.125%**, not certainty. Merging R1 and R2 must preserve their **49% joint probability**; splitting optional metabolic intermediates must not manufacture extra discounts.

Activity thresholds and the 20%/10% materiality definitions are separate sensitivity choices. Changing them changes the proposition and requires re-elicitation; the displayed range keeps them fixed.

PARP1 can overlap with [bioenergetic insufficiency](07_bioenergetic_insufficiency.md), [glial support](17_schwann_metabolic_support.md), [inflammatory signaling](19_tnf_tlr4_inflammation.md), [autophagy](24_autophagy_mitophagy.md) and [SARM1](27_sarm1_axon_destruction.md). None follows automatically from another: PARP1 may matter through protein modification without ATP collapse, and NAD depletion may have another cause. These probabilities are not additive attributable fractions or an intervention ranking.

[Reproducible calculations](../../../probes/damage_calibration_parp_checks.py) retain the rounded structural-value discrepancy, distinguish bulk pools from rates and analyze the two human adverse-event codes separately.