Sensory-neuron cell-body apoptosis

In adults with type 2 diabetes diagnosed within five years and no clinical DPN at entry, classical apoptosis destroys peptidergic sensory cell bodies before substantial degeneration of their peripheral arbors. Interrupting the defined somal execution machinery from entry would reduce cell-body death and 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 target is the physical skin terminals of adult peptidergic unmyelinated neurons and their connected dorsal root ganglion (DRG) cell bodies. A DRG is a collection of sensory cell bodies outside the spinal cord. Support-cell apoptosis, spinal neuronal death, large myelinated sensory-neuron loss and apoptosis confined to an axon are different propositions.

Here a completed classical apoptotic event requires a sequence in the same identified soma: activation of executioner caspases 3, 6 or 7, nuclear condensation/fragmentation and cellular disassembly, followed by irreversible loss of the viable cell body. Persistent absence must be confirmed for 30 days. A protein-abundance change, DNA-break stain or temporary loss of an identity marker alone does not qualify.

Timing and occurrence. An early event must precede that neuron's first ≥5% net physical terminal deficit maintained for 30 days. Over five years, the person-weighted proportion of entry-mapped somata completing such events must exceed that in age-, sex- and site-matched nondiabetic reference people by at least one percentage point. This is R1. Normal terminal turnover does not count as the sustained deficit.

These timing and frequency boundaries make the proposition finite; they are chosen definitions, not established biological thresholds. A soma dying after extensive distal loss can matter for later repair without meeting this early causal claim.

Q — selective interruption. From entry for five years, Q suppresses at least 95% of BAX/BAK-mediated mitochondrial outer-membrane pore formation and catalytic activity of caspases 3, 6 and 7 within the target somata. BAX and BAK are proteins that can assemble mitochondrial pores; executioner caspases cleave cellular substrates during apoptosis. The combined intervention covers mitochondrial initiation and direct executioner activation, without requiring one upstream stress pathway.

Q preserves the proteins' direct functions outside these specified activities and has no direct action in axons, glia, blood vessels or other organs. It does not directly correct glucose, insulin, lipid delivery, trophic supply or oxygen delivery. It is an ideal molecular intervention, not a description of an available drug. Downstream effects on mitochondria, signaling, inflammation, transport, alternative death pathways and repair are allowed to change and count in the result.

Preventing an apoptotic appearance while the same soma dies by another process is insufficient: R2 requires Q to reduce all-cause five-year somal death by at least 0.2 percentage points, measured across all entry-mapped target somata. R1 identifies the early apoptotic branch; R2 tests whether blocking its machinery actually keeps additional somata alive.

The structural comparison measures Q's total effect, including changes before final somal disassembly. It does not claim that every saved terminal was rescued solely by preventing a completed death. Conversely, terminal protection accompanied by no additional soma survival fails this proposition.

Groups and endpoint. The finite candidate groups are the whole eligible population, males, females, baseline BMI ≥30 kg/m², baseline HbA1c ≥7.5%, and a baseline somal-caspase-active group. The last requires at least 10% of mapped target somata to show executioner-caspase activity ≥2 times the positive matched nondiabetic median and above its 95th percentile for at least ten cumulative minutes during a 30-day entry observation. It does not require or predict future death. No additional intersections or groups defined by subsequent response qualify.

Identify target neurons by combined anatomy and stable subtype classification. Follow physical arbors independently of changing CGRP or PGP9.5 staining. For each human, net terminal loss equals destruction minus successful replacement, divided by baseline terminal amount. Average humans equally and define:

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

The reference mean must be positive. Reference care is contemporary diabetes care without Q. 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). All five claims concern the same group and connected neurons.

Source assessment

Best judgment: about 8%; uncertainty range (sensitivity): 0.1–52%. The main uncertainty is whether completed apoptosis of peptidergic sensory cell bodies occurs before substantial loss of their terminals in early human diabetes. These are subjective probabilities of the proposition below, not treatment effects or statistical confidence intervals. Reviewed 14 September 2026. Probability method · Collection index.

Reasoning from the biology

A permanently absent soma cannot indefinitely supply its axon with newly synthesized proteins, membranes and regulatory signals. That makes soma loss a credible cause of eventual axon loss. It does not establish the converse: an axon can fail while its soma remains alive.

Two trajectories must be distinguished:

somal apoptotic execution → cell-body loss → withdrawal of axonal support → terminal loss

distal maintenance/replacement failure → terminal loss → later somal stress or death.

The second trajectory can produce apoptosis-related transcripts, activated caspases and empty neuronal sites at a late autopsy without establishing the first trajectory in early diabetes.

Apoptotic signaling is also not an irreversible binary switch at its first detectable marker. Antiapoptotic proteins, repair and limited mitochondrial permeabilization can permit survival. Blocking execution can preserve a soma that still cannot maintain its long axon. Alternatively, other surviving neurons may replace a lost arbor, reducing the effect of soma survival on net terminal amount.

The magnitude therefore depends on the terminal territory of the somata saved, the timing of their deaths, concurrent axonal damage and replacement. For example, with 30% reference net terminal loss, a 20% relative benefit means saving six percentage points of baseline terminal amount. Saving 1% of equally sized arbors cannot supply that benefit by itself. Cell counts must be connected to the amount of distal arbor supported, rather than assuming one percentage point of somata equals one percentage point of all five-year loss.

Evidence that moves the judgment

Human ganglia show substantial established-disease degeneration. Shiers examined histology from 90 organ donors. Nageotte nodules—glial structures occupying sites of lost neurons—were more abundant with diabetes and neuropathy. In the publicly recoverable ordinal scores, the diabetes/no-prescribed-analgesic group had mean score 2.33, n=15, versus 1.29, n=48 without diabetes; rank discrimination was AUROC 0.810. That is evidence against treating somal pathology as confined to the most advanced recorded DPN.

However, “no recorded DPN” was not a prospectively verified normal nerve examination. Diabetes duration and the temporal sequence of skin and soma loss were not established. Nodules do not identify the dead neuron's subtype or death mechanism, and non-diabetic neuropathic-pain donors also had nodules. Sprouting peptidergic axons within a nodule can originate from surviving neighbors. Shiers et al., 2025.

The human apoptosis signal is not a death count. The 2026 preprint analyzed 30 single-nucleus samples from 28 donors. Its apoptosis-high label selects the top decile of a composite combining apoptosis, p53, DNA repair and unfolded-protein-response scores minus a survival score. The most prominent susceptible populations were A-fiber subtypes. C-peptidergic populations near nodules were discussed as possible sources of sprouting.

The accompanying public bulk-protein comparison supplies four definite DPN versus nine diabetes-without-DPN donors. Only five of an eleven-gene apoptosis-related panel were complete in all thirteen. A score contrasting BAD/BAX/CASP7 with BIRC2/BCL2L2 gave Hedges g=0.92, with donor-bootstrap interval −0.42 to 3.53 and synchronized multiple-score permutation p=0.185. Simply averaging all five gave g=−0.18. Neither the favorable point estimate nor the flat score resolves apoptosis. These proteins are not measurements of completed caspase-dependent death. Human DRG preprint, 2026, public data.

Foundational diabetic-rat studies genuinely disagree. Schmeichel used male streptozotocin (STZ) rats, 8 diabetic and 8 control at one and three months, 9 and 9 at twelve months. TUNEL-positive nuclei with chromatin clumping were 8%, 7% and 11%, versus 2%, 1% and 1% in controls. About 100–125 neurons were examined per rat. This supports an apoptotic interpretation more directly than an expression panel.

Cheng and Zochodne instead found increased cleaved caspase-3 but no neuronal TUNEL, apoptotic nuclear morphology or electron-microscopic evidence at twelve months. Those histological comparisons had three rats per group, not the 26/27 animals in the parent cohort; electron microscopy sampled 60–80 neurons per rat. Different STZ exposures—Schmeichel's reported concentration and volume imply 85.8 mg/kg, versus Cheng's 65 mg/kg—and assay methods leave several explanations for the disagreement. Neither study follows an individual soma to completed death and then follows its skin arbor. Cheng's negative snapshot cannot exclude rare events, while Schmeichel's marker-positive percentage cannot be converted directly into an annual death rate. Schmeichel et al., 2003, Cheng and Zochodne, 2003.

Actual late neuron loss need not have classical apoptotic morphology. In ten-month diabetic BB/Wor rats, Kamiya estimated 9,968 versus 13,582 neurons per ganglion, a 26.6% deficit, with three rats per group for morphology. Estimated CGRP-positive counts were 2,338 versus 4,488. Nuclear-diameter-corrected profile counts and peptide staining have limitations, but this is more than a pathway score.

Despite increased active caspase-3 and Bax, the investigators found no apoptotic nuclear changes or membrane blebbing among roughly 937 examined profiles per DRG. They reported progressive Golgi vacuolation instead. This supports late somal degeneration in type 1 diabetes while weakening the inference that any diabetes-associated soma loss is classical apoptosis. A terminal snapshot still cannot prove that no apoptotic events occurred earlier. Kamiya et al., 2006.

A diabetic intervention preserves counted somata, without isolating apoptosis. Shi used stereological counts in db/db mice at 32–33 weeks, 5–6 mice per group. CoQ10 at 1 g/kg food, started at seven weeks for about six months, changed diabetic mean L5 counts from 6,495 to 9,628; corresponding control means were 9,643 and 11,180. The diabetes-associated deficit therefore fell from 32.6% to 13.9%. Glucose and body weight did not show a treatment effect.

The raw diabetic treatment difference is +3,133 neurons; subtracting the control treatment difference leaves +1,596. Using the reported SDs and every allocation of n=5–6 gives an approximate 95% interval envelope of −2,562 to 5,754 for that interaction. Thus “significant loss untreated, nonsignificant loss treated” does not establish selective normalization. CoQ10 alters redox chemistry, energetics and other pathways; the study did not identify early peptidergic apoptotic events or measure skin-terminal preservation. Shi et al., 2013.

The newer db/db result measures composition. Filfilan studied eight male db/db and eight lean mice at 32 weeks. N200-positive profiles fell from 41.2% to 28.8% of sampled neuronal profiles. The advertised 30% loss is the proportional change in that fraction, not a whole-ganglion neuron count. Peripherin-positive profiles increased in proportion. Neither finding establishes the absolute number of peptidergic neurons or completed apoptosis. The samples contained 4,042 versus 2,975 profiles, but those are sampled profiles, not independent mice or exhaustive counts. Filfilan and Nassar, 2025.

Selective execution inhibition can preserve neurons in other settings. An August 2026 study identified covalent BAX inhibitors with enantiomer, target-engagement and BAX-dependence controls. Rat DRG experiments used E15 embryonic neurons, trophic factors reduced on day three, 1.25–10 μM WX-02-16 before 10 μM ABT-737, and metabolic viability/caspase assays at 24/12 hours. Technical triplicates were repeated with independent cultures, but an exact independent embryo count was not stated.

CBI-3 also preserved counted motor neurons derived from one healthy donor challenged with ABT-737 and one ALS donor undergoing spontaneous culture attrition. The human figures used technical triplicates/quadruplicates. This supplies a much cleaner execution-intervention anchor than a botanical rescue with changed Bax expression; it does not establish adult diabetic sensory exposure, sustained arbor function or population coverage. P. Shi et al., 2026.

Adult DRG cultures provide the competing phenotype: Gumy reported 20–30% slower neurite growth with hyperglycemia and negligible apoptosis, using adult mouse preparations including Bax-null neurons. The accessible report does not supply all independent sample counts for those comparisons. This informs impaired regeneration, not a five-year prevention effect. Gumy et al., 2008.

The reviewed human evidence contains no selective somal-execution intervention with early-diabetes structural follow-up. CoQ10 clinical symptom or conduction results cannot fill that comparison. The public molecular data also do not link each future dying soma to its baseline skin arbor. Quantitative support (background note outside this collection).

Probabilities of the necessary claims

Each conditional assumes all preceding claims hold in the same qualifying group. The complete hypothesis is their conjunction. The ranges are sensitivity to named, coordinated interpretations; they are not independent distributions.

See the claims and probabilities table above.

R1

Human nodules favor a somal component, but mechanism, subtype and early order are unresolved; adult and chronic-rat evidence permits predominantly distal-first or nonapoptotic loss.

R2

Defined execution blockade can preserve neurons, but persistent metabolic injury may cause alternative death; acute or embryonic protection need not last five years.

R3

A saved soma can continue supporting an arbor, but small numbers saved, independent distal injury and replacement can dilute structural benefit.

R4

Assumes a real smaller structural effect; remaining uncertainty concerns how much terminal territory depends on the affected somata.

R5

Somal pathology occurs across diabetic donors, but available groups are selected and established; the early peptidergic branch could be uncommon.

Central inputs [35,80,70,65,65]% give:

0.35 × 0.80 × 0.70 × 0.65 × 0.65 = 0.08281, or 8.281%.

The independently reasoned whole-claim judgment is that a preventable somal-execution contribution is credible, while its early timing in the target subtype is more likely to fail than hold. Existing disease counts and selective neuronal rescue make a near-zero central belief too pessimistic. They do not make this common early-human route probable. The product fits that low-but-substantial neighborhood; its decimal precision is arithmetic.

Uncertainty and consequential updates

A distal-first, nonapoptotic interpretation uses [10,50,35,30,25]%, giving 0.13125%. It treats most early caspase signals as survivable stress, rare somal deaths as difficult to prevent durably, and preserved somata as poor guarantees of distal maintenance.

An early somal-subtype interpretation uses [75,95,95,90,85]%, giving 51.78094%. It transfers the genuine somal pathology and execution-rescue evidence to a common early peptidergic subgroup whose arbors remain serviceable when apoptosis is interrupted. Existing cross-sections cannot exclude that interpretation. The resulting sensitivity width is about 52 percentage points.

Other defensible interpretations move the rows differently. Rare but highly responsive early apoptosis, [45,85,80,75,25]%, gives 5.74%. Frequent preventable soma death with persistent distal injury, [60,90,50,70,75]%, gives 14.18%. These expose coverage and soma-to-arbor coupling as separate uncertainties. The scenarios have no empirical weights; no variance or probability interval over these judgments is asserted.

Hypothetically, direct target-subtype tracking demonstrating the R1 timing and excess frequency could raise R1 from 35% to 70%, yielding 16.56% if later conditionals were unchanged. Representative tracking showing rare or predominantly late events could lower R1 to 10%, yielding 2.37%. Settling R1 as true would leave 23.66%, not certainty about terminal rescue or case coverage. Evidence that changes several links requires a coordinated update.

The one-percentage-point occurrence boundary, 5% contribution boundary and 20%/10% materiality definitions are separate sensitivity choices. Relaxing early timing would answer a broader established-injury question; lowering 20% to 5% would remove R4. Neither change is new biological evidence.

Ideal experiments that would resolve each claim

Shared setup. Conceptually enroll a representative population of the defined early-T2D adults and matched nondiabetic reference people. Map target somata and connected skin arbors before intervention, classify the finite groups from entry information, and track continuously enough to order molecular execution, irreversible somal loss, terminal destruction and replacement. These require capabilities unavailable in living humans today.

Randomize Q versus reference care from entry for five years. Verify ≥95% suppression of the specified somal pore formation and caspase activities, with no direct axonal or systemic action. Use an independent molecular implementation and restoration of the targeted activities to identify manipulation artifacts. Other initiating exposures remain comparable; downstream effects and alternative deaths remain free. Analyze independent humans, with cells and arbors nested within them. For each numerical claim, an interval wholly on its YES side supports YES; one wholly below its boundary supports NO. A boundary-crossing interval, incomplete engagement or missing structural tracking is unresolved.

R1 — completed apoptosis with early order and excess frequency; no predecessors. In reference-care diabetic participants and matched nondiabetic people, directly follow the defined sequence in every sampled target soma through disassembly and persistent absence. Measure the person-weighted five-year frequency of events occurring before the connected arbor's sustained ≥5% deficit. YES means diabetic minus matched nondiabetic frequency ≥1 percentage point; NO means less than one point. For example, a precisely resolved excess of 2 points qualifies and 0.3 points does not. Isolated caspase activation, TUNEL without completed death, unresolved subtype or unknown temporal order remains unresolved. Existing human nodules establish prior neuronal loss at selected sites; no available dataset supplies this longitudinal comparison. R1 alone establishes neither Q efficacy nor terminal benefit.

R2 — durable prevention of somal death; assume R1. In the randomized human comparison, count all irreversible target-soma deaths, whatever their morphology, through five years and divide by all entry-mapped target somata, averaging humans equally. YES requires reference minus Q death frequency ≥0.2 percentage points; NO is a smaller difference. A one-point reduction in apoptotic-looking events with unchanged all-cause death is NO. An apparent reduction caused by preserving a marker in nonviable somata is unresolved. Verify the specified molecular engagement separately from survival. Selective BAX experiments support intervention causality in other contexts, but supply neither this population nor duration. R2 does not establish an arbor benefit.

R3 — smaller structural contribution; assume R1–R2. Measure actual five-year net terminal loss in both randomized arms and estimate B over all enrolled members of the qualifying group. Track successful replacement as well as disappearance; do not restrict analysis to somata that survive treatment. YES is B≥0.05; NO is B<0.05, with a positive reference denominator. A precisely estimated B=0.08 satisfies R3 but leaves R4 false. Preserved soma counts with no change in physical terminals can resolve NO; pain relief or altered staining cannot. Existing soma-preservation studies do not supply this target-terminal comparison.

R4 — material terminal effect; assume R1–R3. Use the same human structural experiment to estimate B against 0.20, including the costs of altered nonlethal signaling and alternative injury. YES is B≥0.20; NO is B<0.20. Reference loss of 30% and Q loss of 23% gives B=23.3% and YES; Q loss of 27% gives B=10% and NO despite satisfying R3. Precision crossing 20% remains unresolved. The CoQ10 soma counts concern different anatomy and a broader intervention, so they cannot settle this magnitude.

R5 — population coverage; assume R1–R4 in the same entry-defined group. Use a population-representative reference-care cohort with complete five-year incident-DPN ascertainment. Measure the fraction of all incident cases belonging at entry to the group already shown to meet R1–R4. YES is C≥0.10; NO is C<0.10. For the entire eligible population, C is 1 by definition once the preceding claims hold there. A trial selected for high caspase activity cannot by itself estimate how many incident cases that group covers. Existing organ-donor series and rare neuronal disease models do not establish this denominator.

Overlap and boundary

SARM1-dependent axon destruction, transport failure and impaired replacement can occur with living somata. Ferroptotic membrane injury can destroy an axonal domain without classical soma apoptosis. Those hypotheses do not imply this one, nor does rescuing a soma imply their absence. Broad inhibition of inflammatory or stress pathways may affect several routes simultaneously; its benefits cannot be assigned wholly to somal apoptosis.