# 27. SARM1 catalytic activity contributes to terminal destruction

**Best judgment: about 36%; skeptical–favorable sensitivity: 2.5%–77%.** These are subjective probabilities of the complete human proposition, not the fraction of fibers a treatment would save. Diabetic structural perturbations and human-neuron experiments support this executor more strongly than many individual initiating routes, while human timing and durable attributable loss remain unresolved. 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 SARM1-dependent nucleotide cleavage in peptidergic sensory axons contributes to physical terminal loss. Selectively suppressing SARM1 catalysis 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**.

This prices a downstream execution route. NMNAT2 depletion, altered NMN/NAD balance, calcium, oxidative injury and posttranslational regulation are alternative inputs. No particular upstream trigger, SIRT3 activity or K641 modification is mandatory. Neither total protein deletion nor a change in pain is the intervention being priced.

**Cells and structure.** Identify adult peptidergic unmyelinated sensory neurons by combined anatomical and molecular classification. Map their somas, peripheral axons and distal cutaneous arbors, and track physical structures independently of changes in CGRP, PGP9.5 or reporter intensity. The qualifying SARM1 activity must occur in the peripheral axon or terminal connected to the arbor being followed; somal, glial, blood or spinal-cord activity alone is insufficient.

For each person, five-year net terminal loss is destruction minus replacement, divided by baseline terminal amount. Newly formed terminals subsequently lost contribute to both quantities. Average humans equally and define:

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

The reference mean must be positive. Reference care is contemporary diabetes care without Q. Persistent fibers, their loss and successful replacement are counted physically. Functional recovery and relief of pain are separate outcomes.

**Finite groups.** Candidates are the whole eligible population, males, females, baseline BMI ≥30 kg/m², baseline HbA1c ≥7.5%, and a baseline SARM1-activity-positive group. The last has the qualifying activity below in at least 10% of mapped neurons during a 30-day entry observation. Additional intersections and groups defined by future response do not 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)`. The same group and connected neurons must remain compatible across all requirements.

**Activity rather than expression.** Measure endogenous SARM1-attributable NAD+ cleavage flux in cytosol of the mapped peripheral axon, in amount per axoplasmic volume per hour. Distinguish substrate consumption from the concentrations of NAD, NMN, ADPR and cADPR. Use direct enzyme-specific reaction tracking, calibrated against rapid selective catalytic suppression while preserving the instantaneous substrate condition. Separately measure synthesis, transport and other consumers, including CD38 and PARPs.

A qualifying flux is **at least twice the age-, sex- and site-matched nondiabetic median and above its 95th percentile**. The normal median must be positive for this fold boundary. At least 10% of mapped neurons, with equal person weighting in a candidate group, must accumulate **at least 60 minutes of qualifying activity during year one**.

The qualifying activity must precede the first **≥5% net physical terminal deficit from entry maintained for 30 days** in the connected arbor. Continuous activation for months is not required: an execution process can act in brief episodes. The activity, duration, neuron-fraction and structural-landmark boundaries are chosen operational definitions, not known biological thresholds.

**Q — catalytic suppression.** Starting at entry and continuing for five years, Q makes SARM1 catalytically inactive in the named neurons, while retaining its protein abundance, localization and noncatalytic interactions. Suppress its NAD+/NADP+ cleavage reactions, including hydrolysis, cyclization and base exchange; measure residual activity directly rather than infer it from a drug label. At least 95% suppression under the endogenous range of substrate and activation conditions is the practical fidelity boundary for the ideal comparison.

This ideal intervention requires selective molecular capabilities beyond a global knockout or an incompletely characterized inhibitor. It must not have an activating interval as concentrations change, and must not directly inhibit CD38, PARPs, sirtuins or NAD synthesis. It acts throughout the target neurons so that ongoing somal supply can respond, while R1 still requires peripheral axonal activity.

Q does not directly correct diabetes, restore NMNAT2 delivery, provide NAD precursors, eliminate damaged mitochondria, stop firing or alter glial signaling. Those processes may change downstream. All consequences of losing basal SARM1 catalysis, including impaired disposal, altered signaling, repair or later bypass degeneration, count in B. Suppression is defined biochemically, not by preservation of the desired axons.

## Reasoning from the biology

SARM1 is an inducible nucleotide-cleaving enzyme. Its activation can accelerate NAD consumption, change ADPR/cADPR signaling and disrupt energy and calcium homeostasis. The relevant balance is:

`change in axonal NAD = local production + net import − SARM1 cleavage − other consumption.`

Enzyme activation does not by itself establish net depletion. Increased synthesis can compensate, and a depleted NAD pool can originate from reduced production or another consumer. A low pool can also coexist with a low absolute SARM1 flux because little substrate remains.

NMN and NAD participate in regulation as well as substrate availability. Published human-enzyme kinetics are not determined by NMN/NAD ratio alone: the same ratio at different absolute concentrations need not produce the same cleavage rate. Reconstructed fitted kinetics even permit reduced NAD to raise or lower flux depending on NMN. This is biochemical context, not a calibrated rate law for a diabetic human axon. [Angeletti et al., 2022](https://pmc.ncbi.nlm.nih.gov/articles/PMC8844822/).

The causal structure is:

`several initiating stresses → SARM1 catalysis → nucleotide loss and/or product signaling → terminal destruction minus replacement.`

The intermediate alternatives are not extra independent probability discounts. Irreversible NAD/ATP collapse is not obligatory: some activated neurons can recover, and catalysis may influence calcium signaling before a catastrophic pool loss. What matters is the five-year structural difference when catalysis is selectively suppressed.

A hypothetical delay in fragmentation is not automatically durable prevention. Upstream metabolic damage can remain, another destruction route can take over, or dysfunctional terminals may persist without effective replacement. Conversely, an intervention can satisfy the structural proposition without normalizing pain or conduction. Those endpoints must be reported separately rather than used to redefine the claim.

## Evidence that moves the judgment

**A direct diabetic structural effect is partial and measured at small animal n.** Cheng's STZ study used constitutive Sarm1 deletion in male mice and assessed skin after 25 weeks of diabetes. Figure 4 has **4–6 mice per arm**. Approximate native-image means are 23.9/8.0 fibers/mm for control/diabetic WT and 22.4/13.1 for control/diabetic knockout. This is about **+5.0 fibers/mm**, 32% closure of the raw WT deficit, or **37% reduction of the deficit normalized to each genotype's control**. Two-pixel reading variation gives approximately 32–42% for the latter, not a sampling interval. Glucose at the endpoint was similar. WT controls were commercially sourced rather than stated littermates; global lifelong deletion and genetic background limit selective adult attribution. [Cheng et al., 2019](https://pmc.ncbi.nlm.nih.gov/articles/PMC6804630/).

Sciatic NAD and structural findings support the same study's mechanism, not additional independent replications. SARM1 protein actually decreased in diabetic sciatic nerve, while mRNA did not change. Expression therefore cannot substitute for enzyme activity or refute activity-dependent injury.

**The earlier HFD result has a metabolic confound.** After eight weeks of 60%-fat feeding, Sarm1-null mice retained epidermal fibers, but only WT mice developed the reported fasting hyperglycemia. Structural n was four per WT arm and eight per knockout arm; both genotypes gained weight. The knockout may act through systemic metabolism as well as nerves. Paclitaxel skin protection in the same paper supports broader axon preservation, with a dissociation from large-fiber electrical outcomes. [Turkiew et al., 2017](https://pmc.ncbi.nlm.nih.gov/articles/PMC5585053/).

A later STZ study supplies an additional functional replication under similar severe glycemia: Sarm1 deletion protected male electrophysiology over 15 weeks, with **4–7 mice per group**. Females did not develop the measured dysfunction, so their result is not evidence of failed treatment. Its separate initial cohort showed no gross cutaneous structural deficit at 14 weeks; the paper does not supply a replicated diabetic skin-rescue effect merely by describing prevention of DPN. [Brazill et al., 2024](https://insight.jci.org/articles/view/175159).

**The 2026 mutant experiment supports activity-related injury without proving the proposed acetylation switch.** Neonatal Sarm1-null mice received AAV expressing GFP, wild-type human SARM1 or K641Q before HFD/STZ. At the skin endpoint, **n=8 per group**, exact source means were 28.55 fibers/mm for nondiabetic GFP, 18.67 for diabetic GFP, 8.87 for diabetic SARM1-WT and 18.08 for diabetic K641Q. K641Q minus WT was **+9.20 fibers/mm, Welch interval 5.97–12.43**. K641Q minus diabetic GFP was **−0.60, interval −4.31 to +3.12**. Null diabetic controls retained a **34.6% deficit** relative to null nondiabetic controls. The mutation chiefly avoids the extra injury associated with WT re-expression; equivalence to GFP is not established by its nonsignificant contrast. [Chen et al., 2026](https://pmc.ncbi.nlm.nih.gov/articles/PMC12799148/), [Figure 3 source data](https://figshare.com/articles/dataset/30958283).

K641Q is an activity-deficient amino-acid substitution, not authentic acetyl-lysine. The fluorescence assay uses 2 mM NAD, 500 µM NMN and PC6, whose product reports base exchange rather than an absolute hydrolytic turnover. Neither that assay nor the neonatal experiment establishes adult human axonal K641 occupancy or the causal effect of changing it.

SIRT3 is also not an obligatory upstream gate. In the same paper, SARM1-WT expression causes additional skin injury in Sirt3-null diabetic mice; source values give a 35.1% IENFD reduction versus GFP. Opposite-sign SIRT3 interventions in a separate rat study therefore do not overturn SARM1's catalytic capability or resolve its human trigger. [Chen et al., Figure 5](https://figshare.com/articles/dataset/30960032), [Yang et al., 2024](https://pmc.ncbi.nlm.nih.gov/articles/PMC10993345/).

**Human-neuron evidence establishes conserved machinery, not population prevention.** A 2021 study generated two independent knockout clones from **one BJ fibroblast-derived iPSC line**. Knockout protected sensory axons after transection and vincristine; SARM1 re-expression restored injury susceptibility, and dominant-negative SARM1 protected WT cells. Axon-only metabolite measurements linked SARM1 to NAD depletion and cADPR production. Multiple clones, fields and assay samples do not increase the independent human donor count. These are acute cultured injuries, not early diabetes in mature human skin. [Chen et al., 2021](https://pmc.ncbi.nlm.nih.gov/articles/PMC8171232/).

A later human dopamine-neuron study directly challenges inevitable commitment: after four hours of vacor exposure with altered NAD and mitochondrial function, removing vacor and adding nicotinamide preserved axons for subsequent weeks. Adding nicotinamide while vacor remained chiefly delayed injury. That comparison changes the initiating exposure as well as metabolism, so it does not establish protection under continued diabetes. It does show why activated SARM1 cannot be defined as an already irreversible state. The study used five donor lines overall, with at least three for most non-knockout experiments. [Reversible human-neuron study, 2025](https://pmc.ncbi.nlm.nih.gov/articles/PMC7617922/).

**Catalytic inhibition and deleting the whole protein differ.** Constitutive sensory-neuron knockout can increase collateral branching without increasing epidermal crossing density, with effects depending on age and substrate. Its legacy allele and breeding background leave some morphological attribution unresolved. Independent CRISPR alleles retain injury protection, so background concerns do not erase the degeneration mechanism. Neither result proves that catalytic suppression reproduces every growth phenotype. [Ketschek et al., 2022](https://pmc.ncbi.nlm.nih.gov/articles/PMC8899016/), [Uccellini et al., 2020](https://pmc.ncbi.nlm.nih.gov/articles/PMC7226674/).

**An inhibitor can fail to implement Q.** A 2026 study found that base-exchange inhibitors accelerated degeneration at low concentrations while inhibiting it at higher concentrations in rodent and human-neuron assays. Longer exposure also distinguished delayed protection from sustained preservation. This restricts particular compounds and dose trajectories; it is not a negative experiment on uniformly inactive SARM1. [Lundbäck et al., 2026](https://www.nature.com/articles/s42004-026-02074-8).

TNV108 entered a healthy-volunteer phase 1 study in April 2026. The accessed sponsor updates provide no human DPN structural efficacy result. Development activity adds no measured prevention effect. [Sponsor announcement](https://tenvie.com/news/tenvie-announces-dosing-of-tnv108/).

## Probabilities of the logical claims

Each row after R1 assumes **all preceding rows are true**. They are conditional judgments about one same group and intervention, not independent marginal probabilities.

| ID | Additional proposition | Skeptical / best / favorable | Best cumulative |
|---|---|---:|---:|
| R1 | The specified excess axonal SARM1 flux occurs before sustained terminal deficit in a qualifying group. | 35 / **75** / 95% | 75% |
| R2 | With that condition, selective catalytic suppression produces five-year structural benefit B≥5%. | 45 / **80** / 95% | 60% |
| R3 | Conditional on R2, benefit reaches B≥20%. | 45 / **80** / 95% | 48% |
| R4 | A same group passing R1–R3 accounts for C≥10% of reference-care incident DPN. | 35 / **75** / 90% | 36% |

R1 is plausible because related diabetic perturbations affect structure and the catalytic machinery operates in human sensory neurons. It remains an inference: early axonal flux has not been measured in the intended people. Human nerve expression results neither supply nor falsify that flux.

R2 is relatively high conditional on actual pre-deficit activation, supported by structural knockout and re-expression comparisons. It remains below certainty because global/developmental interventions differ from Q, and continued metabolic injury can bypass catalysis. R3 credits the sizeable, though partial, animal structural effects; translating their magnitude and duration is uncertain. R4 reflects a potentially shared executor with no measured early-human case coverage.

A B of 10% satisfies R2 and fails R3. A group with B=30% but C=3% satisfies R3 and fails R4. These separate boundaries make the conjunction sufficient for the full defined proposition without repeating an undefined claim that activation is harmful.

## Ideal experiments that would resolve each claim

These are hypothetical resolving experiments, not a proposed intervention program. Assume continuous noninjurious human axonal flux and structure measurements, durable subtype identification, and selective adult SARM1 catalytic control. Use independent people as the inference units. Missing activity specificity, inadequate Q fidelity or imprecise boundaries leave the relevant claim unresolved.

**R1 — activity before physical loss.** Follow representative eligible adults from entry through year one, with matched nondiabetic reference measurements. In mapped peptidergic peripheral axons, quantify SARM1-specific NAD cleavage, absolute substrate/product pools and other input/output fluxes. Establish the ≥2-fold and >95th-percentile rate, ≥60-minute cumulative duration, ≥10% neuron fraction and ordering before the sustained 5% net-deficit landmark. Use rapid selective suppression or equivalent enzyme-specific reaction identification to separate SARM1 from CD38/PARP activity without mistaking pool changes for rates. YES requires at least one complete group-level condition; NO means no candidate satisfies all boundaries. Activity found only in blood or after terminal destruction is NO for this claim.

**R2 — a structural causal effect.** Assuming R1, randomize members of the same baseline groups to Q or reference care for five years. Verify ≥95% catalytic suppression across relevant neuronal compartments and the endogenous activation/substrate range, without changing SARM1's noncatalytic functions. Confirm the result using an independent catalytic implementation and restoration of active, intervention-resistant SARM1 at endogenous abundance; a catalytically inactive restoration control separates catalysis from protein presence. Keep systemic glucose, lipids, blood pressure, medications and other initiating exposures comparable; permit downstream NAD, calcium, energy, disposal and repair to respond. Count physical terminal loss and replacement, including late bypass injury. YES requires positive mean reference loss and B≥5%; NO is B<5% in every qualifying group. Pain relief alone or short-lived metabolic rescue without five-year structure does not resolve R2.

**R3 — material magnitude.** Assuming R1–R2, estimate the same randomized, person-weighted five-year contrast with enough precision to locate B relative to 20%. Do not switch to acute transection, select future responders or change the loss denominator. YES is B≥20%; NO is 5%≤B<20% for every group still passing R2. A brief delay that leaves the final net loss almost unchanged fails the magnitude criterion.

**R4 — incident-case coverage.** Assuming at least one group passes R1–R3, follow a representative reference-care cohort spanning the full eligible population. Ascertain incident DPN uniformly and calculate the fraction of cases belonging to each qualifying baseline group, resolving informative dropout through continued outcome ascertainment. YES requires C≥10% for a group already passing R1–R3; NO means all such groups have C<10%. The frequency of SARM1 expression, a rare activating variant, or response among treated volunteers cannot substitute for this denominator.

## Uncertainty, sensitivity and overlap

Best inputs **[75,80,80,75]%** give **36%**. Skeptical **[35,45,45,35]%** and favorable **[95,95,95,90]%** give **2.480625% and 77.16375%**. These are coordinated interpretations of incomplete evidence, not a confidence interval or an empirically calibrated posterior.

A bypass-dominated interpretation, [50,60,55,50]%, gives **8.25%**. A common-executor interpretation, [90,90,90,85]%, gives **61.97%**. No weights or population frequencies have been estimated for these scenarios. Variation in early activity, intervention specificity, sustained effects and case coverage must remain correlated.

Holding other judgments fixed, evidence moving R1 to 95% raises the whole probability to **45.6%**; moving it to 30% lowers it to **14.4%**. Establishing R1 with certainty would leave **48%**. These are hypothetical evidence updates. Merging R1 and R2 must preserve their **60% joint probability**; adding an unnecessary compulsory NAD-collapse row must not lower the same biological belief.

Relaxing the activity, duration, B or C thresholds expands the qualifying proposition, but current data do not quantify the probability increment. A broader claim about any SARM1 protein function, or repair after established DPN, needs a separate estimate.

SARM1 can lie downstream of [transport](26_axonal_transport.md), [energy insufficiency](07_bioenergetic_insufficiency.md), [lipid exposure](08_fatty_acid_delivery.md) and [calcium dysregulation](31_hyperexcitability_calcium.md). [PARP](28_parp_nad_depletion.md) is a different nucleotide-consuming route. These overlapping probabilities cannot be added as independent portions of avoidable nerve loss.