Glial mitochondrial transfer

In adults with type 2 diabetes diagnosed within five years and no clinical DPN, the normal supply of intact mitochondria from lumbar satellite glial cells (SGCs) to their connected peptidergic sensory neurons becomes insufficient in amount or abnormal in the functional properties defined below before terminal injury. Correcting that incoming supply would avert at least 20% of five-year net distal peptidergic C-fiber terminal loss, in a baseline group covering at least 10% of reference-care incident DPN.

This is a claim about organelle supply at the sensory-neuron soma and its eventual structural consequences at distal endings. It does not assume that every transferred organelle must itself travel to the skin: somatic metabolism or signaling could protect the axon. Nor does it require fusion with the recipient mitochondrial network as the sole mechanism. Uptake, persistence, disposal, local effects and onward transport remain possible downstream processes.

Reference care continues without organelle-supply correction. Incident DPN means new bilateral length-dependent clinical signs with objective small-fiber loss or abnormal nerve conduction, excluding another sufficient cause. The endpoint is actual distal-leg/foot terminals with stable peptidergic identity, not pain, a variable peptide stain, cultured axon length or aggregate DRG respiration.

For each person, L = (terminal disappearances − successful new terminals)/baseline terminal number over five years. A new terminal subsequently lost contributes to both counts. Average territories within people; independent people are the inference unit. Define B = (mean L_reference − mean L_Q)/mean L_reference, requiring positive reference loss, and C = P(group | incident DPN under reference care).

The fixed candidate baseline groups are the whole eligible population, HbA1c ≥7.5%, or an abnormal organelle-supply phenotype measured during a 90-day baseline window using the criteria below. Membership precedes treatment and future structural outcomes. Overlapping groups are tested separately, not combined after identifying responders.

What counts as transfer, quantity and quality

An intact transferred mitochondrion has an enclosing mitochondrial membrane system, matrix and mitochondrial genome, with its SGC origin distinguished from the recipient's original organelles. Its location must be inside the recipient cell boundary, not in an apposed SGC process or on the neuronal surface. Endocytosed intact organelles count as entry; their subsequent escape, assimilation or degradation does not automatically follow. Dye fluorescence alone cannot establish identity or transfer.

Use transferred organelle volume, not organelle number, because fission can increase number without adding mitochondrial material. Let J be completed SGC-to-neuron mitochondrial volume transfer per baseline neuron per day, averaged over 90 days. Normal values match age, sex, ganglion, stable neuronal subtype and activity state. R1 requires this route in normal human physiology: over a year, ≥10% of the mapped target neurons in the matched nondiabetic reference population receive at least one directly identified intact SGC mitochondrion. This physiological comparator is distinct from the early-diabetic group followed in R2–R5.

The two abnormal-supply alternatives are:

• A — amount: J is <50% of the corresponding normal mean, with positive normal mean supply.
• F — organelle properties: ≥20% of incoming mitochondrial volume has either ATP-synthesis capacity below the matched normal fifth percentile or hydrogen-peroxide release above its 95th percentile, and that fraction is at least twice its normal reference fraction.

ATP-synthesis capacity is measured per mitochondrial volume under a specified common challenge: 37°C, pH 7.2, oxygen 100 μM, ADP 1 mM, phosphate 5 mM, pyruvate 5 mM, malate 2 mM and succinate 10 mM. Measure ATP output and peroxide release over the same ten-minute interval, without uncoupler or respiratory inhibitor. These chosen assay conditions define the comparison; they are not assumed neuronal concentrations. An ideal nondestructive equivalent measures the organelles immediately before entry. “Quality” means these measured properties, not their presumed benefit to a neuron.

The R2 abnormality must occur in ≥10% of the mapped baseline target neurons, averaging fractions first within people and then across the group, during a 90-day year-one interval before those terminals disappear. A transient dye difference after established axon loss does not satisfy it. ATP capacity, peroxide release, the 50%/20% cutoffs and duration are definition choices. Other forms of harmful mitochondrial cargo are broader hypotheses.

The intervention and its alternatives

There are three fixed corrections: amount alone, organelle properties alone, or both. Q denotes whichever specified correction is being assessed.

Amount correction restores deficient J to the normal mean while preserving the current distribution of incoming organelle properties. It can therefore increase delivery of dysfunctional organelles and be harmful. Properties correction replaces the excess out-of-range incoming organelle volume with equal-volume, patient-compatible SGC mitochondria sampled from the normal reference functional distribution; it preserves J. Combined correction does both. In-range supply is left alone.

These are ideal substitutions at the source-to-recipient boundary, with explicit organelle mass and energetic costs, rather than an undefined instruction to provide “beneficial” mitochondria. They retain physiological delivery timing and do not deplete the source SGC's own mitochondrial pool. Replacement organelles are compatible with the patient's mitochondrial genotype; an unrelated immune mismatch is not the proposed mechanism. Manipulation does not independently change TNT abundance, SGC–neuron spacing, glial soluble factors, neuronal firing or neuronal biogenesis. All endogenous downstream responses—including clearance, fusion, compensation, inflammation and axonal transport—may change.

Matched handling controls, restoration of the original incoming supply, and a second way of producing the same organelle-volume/property change distinguish the intended intervention from handling artifacts. Q corrects the received supply; it does not claim that MYO10 overexpression, an actin inhibitor or injection of an unstandardized mitochondrial fraction implements it.

Source assessment

Best judgment: about 8% (7.90% by multiplication); interpretation sensitivity: 0.09–52%. These are subjective probabilities of the proposition, not treatment effects or confidence intervals. Reviewed 14 September 2026. Probability method · Collection index.

Reasoning through the chain

Transferred fluorescence can reflect dye exchange; organelle uptake can terminate in degradation; added respiratory material can have little effect if intrinsic biogenesis already meets demand. Conversely, an organelle could influence signaling without supplying much ATP. None of these downstream possibilities should be imposed as an additional mandatory mediator.

Low MYO10 can reflect altered glial geometry or established degeneration. A lower organelle yield from a fixed number of cultured donor cells can produce an apparent “quality” difference even if each organelle functions normally. Donor disease, age, sex, culture history, total mitochondrial volume and per-volume function must therefore be separated.

Physical contact has many consequences. Cytochalasin B alters actin and TNTs; MYO10 perturbation changes SGC–neuron spacing. Effects of those interventions do not isolate mitochondrial cargo. A normal-source mitochondrial fraction outperforming a diabetic-source fraction supports a source-dependent effect, but alone cannot distinguish quantity, organelle function, nonmitochondrial contaminants or soluble signaling.

Probabilities of the necessary claims

Each row is conditional on all preceding rows, retaining the same eligible diabetic group and correction alternatives from R2 onward. Their conjunction expresses the complete proposition.

See the claims and probabilities table above.

R1

Mouse genetic labeling and ultrastructure support the route; human structures and dye assays do not yet quantify intact transfer into the specified peptidergic subtype.

R2

Human MYO10 differences are real measurements but the donor comparisons are small, the full roster is unmatched, and dye signal is not organelle flux.

R3

Mouse manipulation and source-dependent structural rescue support causality, with substantial uncertainty about selective organelle mediation, compensation and the human target.

R4

A small mouse endpoint contrast and analgesia do not determine material five-year human terminal preservation.

R5

The commonness of the specified pre-loss phenotype and its incident-case coverage are unmeasured.

Merging R1 and R2 gives 33.75%, leaving the joint price unchanged. Compensatory biogenesis and soma-to-terminal consequences are included in R3's net structural comparison; they are not extra automatic discounts. R3 and R4 differ in magnitude: B=0.10 can pass R3 and fail R4. A group covering 3% of incident cases can pass both structural rows and fail R5. Amount and organelle-property defects are OR alternatives; no product assumes both must occur.

Evidence bearing on the judgment

The principal direct study is Xu et al., Nature 2026. SGC-directed genetic MitoTag labeling, mitochondrial-marker colocalization, directionality controls and microscopy support transfer in mouse DRG. The genetic-label experiments generally used 3–4 mice per condition. Human microscopy showed intercellular structures, and snRNA-seq supplied cell context from two donors, not thousands of independent humans. MYO10 perturbation and actin/contact disruption support a protective SGC relationship but are not mitochondrial-cargo-specific interventions.

Human evidence: donor counts and an unmatched reference group

Recalculation of Fig. 4 source data gives mean MYO10 signal 29.59 versus 15.16 after averaging images within 4 nondiabetic and 5 diabetic donors. The diabetic/nondiabetic ratio is 0.513, with a difference of −14.42 source units and a donor-level Welch interval of −25.67 to −3.17. This is a measured expression difference, not a transfer rate.

The complete donor roster contains eight people per group. Mean ages are 32.75 versus 55.25 years; female counts are 2/8 versus 6/8. Age alone discriminates the roster's diabetes labels with AUROC 0.9375. Diabetic disease duration averages 15 years, range 4–26; one control is 17. The exact mapping of this roster to each assay subset was not recovered, so an adjusted MYO10 estimate cannot be constructed by assuming row order. These imbalances provide competing explanations, not proof that age caused the signal. The roster does not establish baseline absence of DPN or subsequent incident disease.

The functional Fig. 4m comparison contains six versus nine neurons from only two versus three source experiments. After source aggregation, normalized MitoTracker means are 0.671 versus 0.0849, Welch p=0.073 and exact allocation p=0.10. The interval for the difference is wide, −1.36 to +0.189. Correct aggregation repairs the inference unit, but not assay identity: the outcome remains dye fluorescence.

Hole et al. (2026) dual-labeled astrocyte mitochondria with MitoTracker and genetically encoded fluorophores. Dye entered or closely associated with neurons without corresponding genetically labeled organelle transfer, including after conditioned-medium exposure; longer 48-hour cocultures did not restore agreement. Imaging used three biological repeats for the contact comparison and four for the conditioned-medium comparison. This is a different glial system, so it does not refute mouse SGC genetic-label evidence. It does prevent treating MitoTracker-only human intensity as a calibrated mitochondrial-transfer rate. Normalizing neuronal dye to donor-cell dye does not solve that problem.

Structural rescue: a positive direction with a limited comparator

In Fig. 5j source data, db/db mice received fractions derived from diabetic or nondiabetic human SGCs. Averaging three sections per mouse gives 12.24 versus 16.21 on the published PGP9.5 epidermal-branch scale, 7 recipient mice per group. The difference is +3.97, a ratio of 1.324. A Welch interval is −0.02 to +7.95, p=0.051; exact allocation p=0.059. The published equal-variance test gives p=0.0475. These small changes around 0.05 do not make the directional observation disappear or establish a precisely known benefit.

The figure labels the outcome “Number of branches”; the source values are retained without inventing a fibers/mm conversion. This structural panel has no vehicle or nondiabetic-mouse reference arm. A 32.4% difference between mitochondrial-source groups is therefore not 32.4% prevention of diabetic terminal loss. It is also not seven independent human donor comparisons.

The study dosed the mitochondrial fraction obtained from 10,000 source SGCs, rather than a reported matched mitochondrial volume or ATP-generating capacity. Source-dependent yield can therefore explain part of the difference. The methods also describe myxothiazol pretreatment of SGCs before isolation, although the results summarize damaged mitochondria. That comparison and the differing analgesic response support source-function relevance without cleanly holding mitochondrial quantity and other fraction components fixed.

Contact, source condition and the distal bridge

The damaged-SGC culture source data show axon outgrowth falling from 312.6 to 145.1 with paclitaxel-exposed SGCs, then rising to 204.2 with CytoB: 35.3% of that outgrowth deficit was recovered. There are 15 plotted neurons per condition from three experiments, without recoverable experiment identifiers for those observations. The result supports a source-condition-dependent, contact-sensitive effect. CytoB does not identify harmful mitochondrial cargo specifically; other contact-mediated material or signaling can change with it.

The same source contains PGP9.5 epidermal-fiber loss after CytoB or antimycin, six mice per condition, and after paclitaxel, five per condition. These interventions have direct nontransfer effects. They establish a distal structural phenotype after broad perturbation, not isolated loss of physiological SGC mitochondrial supply.

The proposed preference for medium/large recipients also needs restraint. Published total-neuron and labeled-neuron size distributions are not linked denominators; some nominally corresponding bins contain more labeled neurons than total neurons. Size and contact opportunity can explain much of the distribution shift. Near-disjoint MitoTag and activity-indicator labeling after injury does not establish activity-directed delivery to each needy neuron. Neither observation proves that human peptidergic neurons receive no support.

The combined evidence supports mouse SGC organelle transfer and source-dependent neuroglial effects. It supplies weaker evidence for the exact early-human amount/property deficit and its material structural contribution. Chemotherapy injury, established db/db neuropathy and short-lived analgesia inform mechanism or repair, not incident-human-DPN prevention.

Ideal experiments that would resolve each claim

The definitions above fix population, molecular identity, units and interventions. Each experiment uses independent people, or the stated matched human physiological comparator, with stable connected-cell and terminal identities. Confidence intervals or equivalent precision must fall wholly on one side of the specified boundary. A nonsignificant result alone is not NO. For an OR, a precise NO requires every retained group/intervention alternative to fail; use simultaneous precision across those alternatives.

R1 — normal human organelle supply; no predecessors. Observe matched nondiabetic adults without neuropathy for a year, mapping lumbar SGCs to peptidergic neurons supplying the target skin. Label SGC mitochondrial membranes, matrix and genome independently without labeling recipient organelles or altering mitochondrial function. Combine continuous trajectories with ultrastructural confirmation of completed entry; exclude dye diffusion, label synthesis in neurons, adjacent glial processes and surface-bound material. No supply manipulation is needed, and routine physiological inputs vary normally. YES requires ≥10% of mapped target neurons, averaged within and across people, to receive at least one intact identified SGC mitochondrion. NO requires a precisely smaller fraction over the whole year with adequate detection sensitivity. Labels that can move without organelles or inadequate observation time leave it unresolved. This resolves the human physiological route, not diabetic causation or useful recipient function.

R2 — pre-loss diabetic amount/property defect; assume R1. Follow the specified early-T2D groups in parallel with matched normal references, measuring J and the defined per-volume ATP/peroxide properties of incoming organelles in the same connected neurons during a 90-day year-one pre-disappearance interval. Keep measurement conditions identical, account for age, sex, activity, cell identity and culture-independent source handling, and measure actual transferred volume rather than MYO10 or fluorescence intensity. This is descriptive; no intervention is required and ordinary disease exposures may evolve. YES requires at least one group with ≥10% person-weighted mapped neurons satisfying A or F's exact boundary before loss. NO requires all groups to fail both alternatives. An amount reduction with normal organelles passes A; excess abnormal-property volume with normal J can pass F. Uncertain source identity, post-loss-only measurements, an interval crossing a cutoff or unmeasured delivered volume leaves the claim unresolved.

R3 — net causal terminal preservation; assume R1–R2. Randomize the three defined supply corrections versus matched handling control before structural loss and sustain them for five years in retained groups. Verify completed incoming volume and property distributions, source-cell viability, genotype compatibility and absence of directly imposed changes in nonmitochondrial cargo, glial contact or neuronal survival machinery. Keep glucose, lipids, insulin, oxygen supply, medications and initiating activity inputs comparable; permit all downstream biogenesis, clearance, fusion, inflammation and transport responses. Restore the original incoming supply in a separate control and use an independent manipulation of the same input to exclude handling artifacts. Track actual terminal disappearances and successful replacements with stable subtype identity. YES requires at least one retained B≥0.05; NO requires all retained B<0.05 with positive reference loss and faithful correction. Increased amount can worsen outcomes when source properties are abnormal, a valid negative result. Uncertain terminal identity, negligible reference loss, incomplete input correction or an interval spanning 0.05 is unresolved. Somatic ATP or analgesia without the terminal endpoint cannot resolve this claim.

R4 — material effect; assume R1–R3. Use the same five-year randomized interventions, input matching, fidelity checks and structural measurements with enough people to estimate B against 0.20. YES requires at least one retained B≥0.20; NO requires all retained B<0.20. A precise B=0.12 passes R3 but fails R4. An interval crossing 0.20 is unresolved. Direct distal transport of the donor organelle is not additionally required if the defined somatic supply correction preserves distal structure through another endogenous route.

R5 — reference-case coverage; assume R1–R4. In a representative reference-care cohort from the same early-T2D population, assign the fixed baseline groups without future-response information and ascertain incident DPN for five years. Estimate C for every group retained through R4, counting independent people. No Q is used in the reference-incidence denominator. YES requires at least one C≥0.10; NO requires every retained C<0.10. Do not pool overlapping groups or redefine membership after benefit is observed. Incomplete clinical ascertainment, uncertain baseline classification or an interval crossing 0.10 is unresolved. Coverage is separate from effect magnitude.

Uncertainty and sensitivity

Conditional vectors [40,15,25,25,25]%, [75,45,65,60,60]% and [95,80,90,85,90]% give 0.09375%, 7.8975% and 52.326%. The skeptical interpretation combines uncommon human target-cell supply with confounded donor differences and nonspecific manipulation effects. The favorable interpretation accepts a common early supply defect and effective physiological restoration. These are coordinated interpretations, not independently sampled probabilities or a fitted posterior.

A human-confounding scenario, [60,25,40,40,40]%, gives 0.96%; an early common-deficit scenario, [90,70,85,80,80]%, gives 34.3%. No evidence assigns probability mass to these scenarios. Resolving R2 much more favorably, from 45% to 80%, raises the whole price to 14.0% with other rows unchanged; lowering it to 15% gives 2.6%. A donor-replicated, dye-independent pre-loss result would be more informative than another recipient-mouse analgesia comparison.

The amount and property cutoffs, 90-day window, physiological assay and normal-human route requirement define this hypothesis. Broadening it to every form of intercellular mitochondrial cargo or removing the 20% effect boundary changes the proposition and requires re-elicitation. A five-percent structural proposition omits R4. Existing data do not identify separate numerical probabilities for quantity and quality mechanisms or a continuous “good mitochondria” threshold.

Overlap and implication

This soma-level route differs from Remak metabolic output. It can affect energy supply, mitochondrial dynamics, transport or oxidative injury. Those processes share evidence and are not independent contributions to add.

The organelle-transfer route is credible. The central unresolved question is whether its physiological amount or incoming properties are abnormal early enough, in the human target neurons, for selective correction to preserve a material fraction of their distal terminals.