Persistent chromatin regulation

In adults with T1D or T2D who have improved sustained hyperglycemia, a history-dependent chromatin state in peripheral neurons or supporting cells persists and changes gene output. Replacing a specified set of chromatin modifications with a matched lower-exposure profile would preserve at least 20% of subsequent five-year net physical peptidergic C-fiber terminal loss, in a baseline group containing at least 10% of reference-care incident or progressive DPN cases.

Population exception. Entry follows HbA1c ≥9% for at least 12 months and then six months at 6–7%, with continued comparable glucose management during follow-up. Include people with and without established DPN and any diabetes duration. This assesses further injury after improvement; it does not establish prevention in recently diagnosed T2D. The six-month interval also separates the claim from the acute syndrome following rapid glucose correction.

Chromatin is DNA and its associated proteins. The finite state panel is DNA 5-methylcytosine, histone H3 lysine 4, 9, 27 and 36 mono-, di- and trimethylation, and H3K9, H3K27 and H4K16 acetylation. Histone names specify the protein and amino-acid position. DNA sequence, methylated RNA and independent noncoding-RNA feedback are outside the direct intervention. Their downstream responses to chromatin changes can contribute.

This is a bounded chromatin-dependent version of persistent gene regulation. It does not assign a probability to every possible form of molecular memory. A circular RNA can sustain altered regulation without this panel carrying the effect.

Q — replacement of the chromatin profile. In mapped sensory neurons, connected Schwann and satellite glial cells, and keratinocytes, fibroblasts and microvascular cells serving the distal skin territories, replace the panel's joint, locus-specific modification profile with the profile following the lower glycemic history defined below. Apply the corresponding cell-type profile from entry through five years, updating it daily. Preserve its joint relationships across loci and marks rather than independently selecting favorable values.

The reference history has HbA1c 6–7% during the preceding high-exposure year and thereafter. Compare the same genetic background and cell identity, with age, current extracellular glucose, lipids, oxygen, hormones, medications and entry nerve structural damage matched. This ideal reference isolates a regulatory difference attributable to history; it is not a young healthy cell replacing an injured one.

Q changes the chemical marks, not DNA sequence, cell number or existing axonal damage. It does not select marks because they predict a favorable outcome. Transcription, cell identity maintenance, inflammation, metabolism and repair respond naturally. All costs of removing useful regulatory adaptations count. Broad methylation or histone-enzyme drugs are incomplete tests because they do not implement this specified profile.

Occurrence and persistence. At least one panel modification must show an exposure-history effect before entry in units serving ≥10% of mapped terminals in ≥10% of a candidate group. A locus qualifies if modification occupancy differs by ≥10 percentage points from the matched lower-history profile and lies outside its central 95% distribution. Occupancy means the fraction of corresponding DNA sites or nucleosomes carrying the specified mark, within the same cell lineage.

After six months of improved glucose, that deviation must remain for ≥90 cumulative days in follow-up year one, before an additional baseline-normalized net terminal deficit of ≥5% lasts 30 days. Ordinary turnover is not the deficit. The state can participate in a feedback loop; it need not be the sole storage site of glycemic history. However, altered cell proportions or a record without a modifiable downstream effect cannot complete the claim.

Gene output. In the same persistent units, Q must move at least one linked transcript and its protein abundance or biochemical activity by ≥20% toward the lower-history profile, for ≥30 cumulative days. This requirement establishes a functional regulatory consequence, without presuming that the output is harmful. A compensated metabolic shift can satisfy it and provide no structural benefit.

Outcome and groups. Track stable-identity peptidergic terminals in fixed bilateral distal-leg and dorsal-foot territories. Count physical disappearance minus newly established connected terminals, including later loss of replacements; peptide expression alone is insufficient. Normalize to entry terminal amount, average territories within each person and then people equally. For reference and Q losses L₀ and L_Q, require L₀>0 and define B=(L₀−L_Q)/L₀.

Candidate baseline groups are all eligible people, T1D, T2D, established DPN, no DPN, and age ≥60 years. No additional intersections or future responders qualify. C is the fraction of reference-care people developing objective incident or progressive DPN who belong to a qualifying group. Incident DPN requires new bilateral length-dependent neuropathy with objective small-fiber loss or abnormal conduction; progression requires new objective deterioration in established disease, not increased pain alone.

Each claim requires at least one group meeting every preceding requirement. Different groups cannot separately supply the pieces of the conjunction. These thresholds define this question; they are not measured biological cutoffs or a clinical recommendation.

Source assessment

Best judgment: about 7%; uncertainty range (sensitivity): 0.17–50%. Past hyperglycemia can leave persistent regulatory changes. Whether resetting those changes preserves a material amount of human nerve structure remains unestablished. These numbers express belief in the complete proposition below, not an expected treatment effect. Reviewed 14 September 2026. Probability method · Collection index.

Why the causal distinction matters

Clinical “metabolic memory” means that past exposure predicts later outcomes despite improved current control. Several mechanisms can generate that pattern: previously lost axons, damaged matrix, reduced reserve, persistent inflammation or regulatory changes. Only the last category directly tests this chromatin intervention.

A methylation association is therefore insufficient. Persistent occupancy must belong to the same cells or lineages, affect gene output and permit net structural preservation when changed. Conversely, chromatin need not cause every downstream event by itself. It could sustain a Schwann-cell metabolic change or alter a repair response while other injuries continue.

Improvement after treatment given during high glucose demonstrates prevention of installation, not reversal after improvement. Better nerve function also need not mean more terminals. These distinctions make the peripheral washout studies informative without treating them as completed human structural tests.

Primary evidence

Schwann cells show direct persistence, with limited donor and endpoint support. Kim and colleagues cultured the immortalized mouse IMS32 line at 25 mM glucose for >8 weeks, then 5.6 mM. After one week, Hk2 expression remained 1.943-fold normal versus 1.803 during high glucose; Fbp1 remained 0.037-fold versus 0.049. Pfkl retained about 79% of its initial deviation from normal, calculated from 2.026- and 1.809-fold expression.

Selected expression changes persisted for four weeks; Fbp1/Fbp2 methylation differences persisted for one week. Assay n=3–8 is not independent animal or human donor n. ATP remained stable, and the study did not measure axon loss. One week of 10 μM pioglitazone after the chronic exposure did not reverse selected expression changes; treatment during exposure prevented some. Chromatin marks were associated with persistence, not selectively erased to establish mediation. Primary human Schwann cells were studied only over shorter exposure. Kim et al., 2013.

A sensory-neuron study separates maintenance from installation. After seven days of high-sugar diet, fly sweet-taste defects persisted through up to 20 days on control food. Inhibiting PRC2 or reducing its targeting factor Pcl during recovery restored taste, supporting continuing chromatin-dependent regulation after exposure withdrawal.

After seven recovery days, 310/658 previously altered transcripts, 47.1%, remained differential at q<0.1. Expression used two pooled replicates per recovery condition, with 300 heads per replicate; 658 genes are not 658 independent tests of the whole mechanism. Taste behavior involved tens of flies per group. Pcl overexpression did not reduce sensory-neuron number. This is adult neural regulatory plasticity, not evidence of diabetic terminal destruction or five-year structural rescue. Vaziri et al., 2020.

Endothelial experiments support chromatin causation in an adjacent compartment. Sixteen hours at 30 mM glucose produced approximately 2.5-fold H3K4 monomethylation at the NF-κB p65 promoter, persisting after six days at 5 mM. SET7 knockdown before exposure prevented the response. Histone acetylation increased with trichostatin A without reproducing p65 induction, showing that different marks are not interchangeable.

A six-hour mouse glucose clamp followed by normal glucose supported persistence in aortic endothelium, with very small molecular assay samples: figure 4c reports n=2 for chromatin and n=3 for transcript measurements. These experiments support installation and output, not delayed reset of peripheral nerve chromatin or structural preservation. El-Osta et al., 2008.

Human glycemic-history associations remain confounded by how participants were selected. Miao studied 30 conventional-treatment participants with high historical HbA1c and renal/retinal progression versus 30 intensive-treatment participants with lower HbA1c and no progression. At blood collection, HbA1c still differed, 8.0% versus 7.4%.

Monocytes had 6,470 versus 6,105 H3K9-acetylated regions, P=0.0096, adjusted P=0.048 across five tests. The selected 39-region list had an estimated 33% false-discovery rate. The result supports blood regulatory associations with exposure and selected complications, but neither matched current exposure nor prospective peripheral-cell mediation. It is not a 60-person randomized chromatin intervention. Miao et al., 2014.

A prospective neuropathy methylation marker is not equivalent to a causal memory carrier. Zhou followed 282 initially DPN-free people with T1D; 154 developed DPN over 28 years. Five percentage points more baseline blood methylation at PKNOX1 cg18945945 predicted HR 0.370, approximately 0.255–0.538 from the reported log-HR and standard error; risk-factor adjustment gave 0.361.

However, the one-sample genetic estimate −1.44±0.76 gives two-sided P≈0.058, agreeing with the Results' P=0.06 rather than the abstract's P=0.01. The paper's independent two-sample estimate was also inconclusive and pointed oppositely, +0.14±0.20, reported P=0.47. Blood prediction does not establish a target-cell state, its persistence after normalization, or an effect of Q. Zhou et al., 2025.

Human sural multi-omics cannot supply the missing longitudinal state contrast. The public study has 77 expression and 53 methylation samples, including 35 paired molecular samples in degeneration/regeneration tails. Nerve sampling is not a repeated same-cell chromatin trajectory after normalized glucose. In the available paired analysis, no tested promoter–expression association survived q<0.05, and no corrected methylation or expression association established the measured fiber-density-change mechanism.

That restricts what this small dataset supports. It does not exclude an unmeasured cell-specific mark, histone mechanism or a response shared by both groups. The parent study's molecularly defined contrasts and a direct association with structural change answer different questions. Public sural study; quantitative support (background note outside this collection).

New peripheral RNA evidence broadens the alternatives. A 2025 rat Schwann-cell study reported persistent circ_0000756/miR-7b/NRF2 changes and improvement after circRNA knockdown. Metabolomics had n=3 per condition, without an established independent donor count. The Methods specify 48 hours high glucose followed by 24 hours normal glucose; the metabolomics Results instead say one week.

Its animal experiment reports six rats per group, pain responses, inflammatory infiltration and TUNEL staining. It does not demonstrate in-vivo glucose normalization or count retained peptidergic terminals. Methods and Results also differ in diet/induction and group enumeration. This is suggestive peripheral regulatory evidence with unresolved reporting defects, not a demonstrated chromatin-mediated structural rescue. Chen et al., 2025.

Conditional probabilities

Every value after R1 is conditional on the preceding requirements being true, including a surviving qualifying group. The lower/best/upper numbers form coherent sensitivity scenarios; they are not independently estimated frequencies.

See the claims and probabilities table above.

R1

Glucose-induced chromatin changes occur experimentally; human peripheral dose, cell type and prevalence remain uncertain.

R2

Direct washout evidence exists, but days or weeks in models do not settle persistence after six improved months in humans.

R3

Mark-specific perturbation and sensory plasticity favor functional regulation; correlational methylation and unrelated drug effects remain alternatives.

R4

No human delayed chromatin reset demonstrates retained terminals; beneficial adaptation can offset relief of injury.

R5

A persistent amplifier could be material, but pre-existing loss, matrix damage and parallel injury may dominate.

R6

Severe historical exposure is shared within this exception population; the distribution of structural benefit is unknown.

R3 does not label the altered output harmful. R4 establishes a net structural effect; R5 asks whether it is material. A small, rare positive can therefore fail the complete claim. Installation without persistence, or persistence without gene output, also fails it.

Ideal experiments that would resolve each claim

These are definitions of decisive evidence, not proposed laboratory work. Assume exact state and output measurements, the specified selective Q, matched exposure-history counterfactuals and nondestructive connected-terminal tracking. Independent people are the inference unit. Follow complete structural and vital-status trajectories, retaining initial population membership. Finite intervals crossing a threshold, incomplete engagement or loss of decisive follow-up leave a claim unresolved.

R1 — history installs the state. Compare the specified high and lower glycemic histories in otherwise matched peripheral units before entry; measure the full finite panel and cell-lineage identity. YES: at least one group meets the ≥10-percentage-point, reference-distribution and ≥10%-of-people/units boundaries, with a causal effect of history. NO: complete measurement excludes the conjunction in every candidate. Blood differences, age differences or different surviving cell proportions do not establish YES. Kim and El-Osta supply model installation evidence; the human association studies do not supply this comparison.

R2 — persistence after improvement. Follow the same R1-qualified lineages through six months of matched current glucose and the next year, measuring new structural loss concurrently. YES: a surviving group retains the defined deviation for ≥90 cumulative days before the additional sustained 5% deficit. NO: every candidate fails duration, occupancy, coverage or order under complete observation. Continued low-amplitude deviation below the explicit boundary is NO for this claim, even if biologically interesting. Short culture washout and cross-sectional cell mixtures leave the human requirement unresolved.

R3 — the persistent state changes output. Randomize verified Q versus reference care within the surviving groups. Measure locus-linked transcripts and their corresponding protein abundance or biochemical activity in the same units. YES: a group meets the specified ≥20% movement toward the lower-history profile for ≥30 days. NO: no surviving group does despite complete state replacement and output measurement. Drug benefit without the specified mark changes does not resolve it; neither does gene expression alone. Adult fly recovery supports this type of maintenance comparison, with a different species and output.

R4 — net structural causation. Use the same randomized groups and Q for five years, allowing downstream inflammatory, metabolic and repair responses. Track physical connected terminals independently of peptide abundance. YES: at least one group satisfying R1–R3 has L₀>0 and B≥5%. NO: all have L₀≤0 or B<5%, with complete structural accounting. Better pain behavior, metabolic activity or TUNEL staining alone leaves the claim unresolved. Existing peripheral studies do not implement this test.

R5 — material structural preservation. Retain the same intervention, denominator and qualifying groups. YES: a group satisfying R1–R4 has B≥20%. NO: every remaining group has B<20%. An imprecise estimate compatible with both 5% and 30% remains unresolved; statistical nonsignificance does not decide it. None of the cited studies bounds this human five-year ratio.

R6 — coverage within the stated population. Follow the full eligible reference-care population and count objectively incident or progressive cases by baseline group. YES: at least one group satisfying R1–R5 supplies ≥10% of cases. NO: all such groups supply less. Do not add overlapping groups, select future responders, or substitute the prevalence of a blood mark. If some candidate groups remain unmeasured, the complete claim remains unresolved.

Calibration and sensitivity

The best joint product is 0.80×0.60×0.75×0.50×0.60×0.65 = 7.02%. There is no empirical reference class that converts these studies into a validated human probability. The judgment favors occurrence and some persistence more than useful structural reversal.

The lower scenario [45,30,40,25,35,35]% → 0.165% treats much human chromatin variation as a record, cell-selection effect or useful compensation, with little net benefit after existing damage. The upper [95,90,95,85,85,85]% → 49.88% treats persistent local regulation as a common, reversible amplifier and the missing human structural experiments as the main limitation. Both interpretations are compatible with current direct evidence. The approximately 49.7-percentage-point width measures interpretive sensitivity, not sampling variance or a confidence interval.

A favorable hypothetical update would demonstrate persistent human peripheral marks after matched current exposure, output response to selective reset and substantial retained structural benefit across common baseline groups. Coordinated [90,85,90,80,80,80]% would give 35.25%. A fully engaged human reset that changes output but yields little physical preservation would instead support [75,50,60,20,35,40]% → 0.63%. These are explicit reassessments, not likelihood ratios estimated from the papers.

A distinct alternative is that RNA feedback stores memory while this chromatin panel is dispensable. That could strengthen a broader gene-regulation hypothesis while driving this claim's R3 or R4 toward NO. Another is that a reset promotes initial sprouting but destabilizes mature identity; long-term net preservation then falls despite early benefit. Neither alternative is captured by merely widening assay error bars.

Merging R5 and R6 gives a conditional 39%, leaving the joint 7.02% unchanged. The number of written steps carries no penalty. Lowering B or C weakens the conjunction and cannot lower its probability; changing the modification panel or Q can alter benefit in either direction because useful regulation also changes.

The unresolved question is modifiable structural mediation after improvement, not whether glycemic history matters. Persistent matrix modification, senescent secretions, inflammation and Schwann-cell support can share downstream paths. Their probabilities must not be added as independent explanations.