---

Version of record: 10.5281/zenodo.21799864, published 5 August 2026. That identifier is the citable address for this paper and it resolves at https://doi.org/10.5281/zenodo.21799864. It is a version identifier; Zenodo minted a second one that resolves to all versions, and the version identifier is the one to cite.

What this file is. This project's authoritative copy of the manuscript, SUBMIT_THESE/papers/PUBLISH_7_DOMINANT_NEGATIVE_DEGENERACY.md, which is the file the deposited PDF was built from. Synced 7 August 2026 by scripts/sync-manuscripts.mjs, which copies the source byte for byte and prepends this note. Nothing in the manuscript below has been rewritten for the website.

What was here before, because nothing on this site is deleted quietly. Until 6 August 2026 this file was a copy taken before the corrections of 4 August 2026 evening and never resynced, so it served the retired 34.1 percent rescaling of the measured current. Earlier on 6 August 2026 it was given a banner reading "superseded revision, do not cite any figure in it". That banner was an accurate description of a stale file and a poor thing to serve on a research site, so the stale file has been replaced with the authoritative text rather than annotated. The retired rescaling divided the measured 68.3 percent by two; O'Neill 2022's own two-allele control reads 218.4 percent of a single allele rather than 200, so the divisor is 2.184, the baseline is 31.3 percent and the comparator for simple loss of one allele is 45.8 percent.

Warning, and it points at the published record rather than at this page. The version deposited on 5 August 2026 says this paper's derived tables are in the shared data archive and names two: the per-study numbers used in the power calculation, and the dimer-arithmetic solutions for x. The archive at 10.5281/zenodo.21799234 contained neither, nor any other file belonging to this paper. That is a statement a reader cannot check except by downloading the archive and finding nothing there. It was found in an audit of all eleven papers' data availability statements on 6 August 2026, which found the same class of defect in six of them.

No saved output of either calculation exists anywhere, because the tables were assembled to write the paper and never written to disk. Both were regenerated from first principles rather than reconstructed from the paper, which is possible here in a way it is not for most papers, because everything in them is closed-form arithmetic over published summary statistics and the method is fully stated in the Methods. p7_power_and_dimer.py also recomputes the three structural distances from the public 8VYJ chain A coordinates. The regeneration reproduces 36 of the 38 quantities this paper prints, at the precision printed, including every solved x, both cross-study z-tests and all four structural measurements. The two that do not reproduce exactly are rounding propagations, and both are named in the paper rather than quietly adjusted: a minimum detectable difference printed as 37.3 percentage points that recomputes as 37.245 from the unrounded standard error, and a z of 4.25 that recomputes as 4.258. Neither changes an argument.

One thing the regeneration sharpened, and it cuts this paper's own way. The post-hoc power of about 27 percent for Wang's endoplasmic-reticulum null is a normal approximation, reproducing as 26.5 percent. The exact non-central t at the group sizes this paper itself assumes gives 21.7 to 24.7 percent. The printed figure is therefore the generous one, so the argument it supports, that Wang's null is a failure to resolve rather than evidence of absence, is if anything understated by it.

Separately, this copy is also ahead of the version of record, and that addition is dated. On 6 August 2026, after the deposit, Tano et al. 2026 (PMID 41582807) was folded in at six places: one new sentence in the abstract, a provenance note in the methods, a new results section on the Tano family variants p.G833R and p.T1396P, a sixth experimental arm that the abstract's arm count deliberately still does not include, a new limitation, and reference 10. The record at the identifier above contains none of it. (Until 7 August 2026 this header described the paper as additive only, meaning ahead of its record with nothing corrected. That was true when written and was falsified by the data availability audit, so it is corrected here rather than left standing.)

No conclusion changes, and the new limitation cuts against this paper rather than for it. It is repeated here rather than left buried: Tano's own claim is that phenotypic severity is independent of dominant-negative and coupled-gating status, which if true means resolving the mechanism buys less clinical information than this paper's framing implies.

If a figure on this page disagrees with the same figure at the identifier above, this page is the corrected one. The full divergence, and what a version-2 deposit would have to include, is recorded in SUBMIT_THESE/ZENODO_DIVERGENCE_20260806.md. (This paragraph used to open by asserting that no version 2 had been deposited and nothing had been uploaded. That was true when written on 6 August 2026 and is not a claim a generated page can keep true, because it would turn false the moment anything is deposited and nothing here would notice. The sentence is removed rather than updated: to find out what is deposited, resolve the identifier, which is the only source that cannot go stale.)

None of this is peer reviewed, and none of it has been through a wet lab. No cell has been edited and no current has been recorded for this variant by this project. Every therapeutic statement in the manuscript below is a prediction.

---

Peak current alone cannot separate trafficking arrest from coupled gating in SCN5A dominant-negative variants

Ethan Bradley

Independent researcher, no institutional affiliation

ORCID: 0009-0008-8925-7975

Abstract

Co-expression peak-current measurements alone cannot distinguish trafficking arrest from coupled gating for SCN5A variants under the stated dimer model. Both mechanisms reduce to I/I_null = 1 minus f times x, where f is the mutant fraction of the assembly-competent pool and x represents retention or functional silencing of partnered wild-type channels. A measured current constrains the product, not its biological source. Comparing reports on R104W reveals different assay scopes: mutant-only surface measurements cannot resolve wild-type retention during co-expression. For Wang 2020's co-expression endoplasmic-reticulum comparison, the reported summary values yield a conditional approximate 95 percent interval of -8.36 to 43.76 percentage points for increased co-localisation. This is compatible with both no change and substantial retention; observed-effect power adds no independent evidence. R121W shows a cross-study numerical discrepancy that survives numerator/reference covariance sensitivity under a first-order normal model, but differences in experimental systems and normalization remain possible explanations. L96P is a reported loss-of-function and surface-expression lead, not yet a source-verified allele-resolved co-expression discriminator: the accessible abstract does not establish the necessary comparator and assay details. Tano 2026 reports p.G833R association with wild type without a dominant-negative or coupled-gating effect, showing that association alone is insufficient. A five-arm, three-readout design for R104Q can constrain abundance and functional contributions. Unique mechanistic attribution still depends on qualified measurements, explicit assumptions and possible coexistence of mechanisms.

A key to the terms used here

• SCN5A is the gene for the heart's main sodium channel; Nav1.5 is the protein. R104Q, R104W, R121W and L96P are variants, named as original amino acid, position, replacement.
• Peak current is the largest sodium current measured from a cell, the standard readout in these experiments and the quantity this paper argues cannot settle the question on its own.
• Trafficking is the cell's delivery of a finished channel to the surface membrane. Trafficking arrest means the channel is built but never delivered, so it cannot pass current: stock in the stockroom rather than on the shelf.
• Coupled gating means two channel molecules influence each other's opening and closing, so a defective one drags down a healthy partner. Both mechanisms lower peak current, which is why the measurement alone cannot separate them.
• Degenerate here is used in its mathematical sense: two different mechanisms produce the same number, so the number has more than one explanation and cannot identify which is true.
• Dominant-negative means the variant copy interferes with the working copy, making the loss worse than losing one copy of two.
• Co-expression means putting the variant and the normal version in the same cell, as happens in a carrier, rather than testing the variant alone.
• Heterozygous means carrying one variant copy and one normal copy.
• A dimer is a pair of molecules acting as a unit. The dimer arithmetic here is the algebra of how many pairs form and what each contributes.
• Statistical power is the chance an experiment would detect a real effect of a given size. An underpowered experiment can miss a real effect, so its negative result does not establish absence.
• Patch clamp is the electrical measurement of current through channels in a single cell.
• PDB 8VYJ is the public 3D structural model of the channel used here.

Why a fifteen-year disagreement has not resolved

SCN5A dominant-negative variants reduce sodium current when co-expressed with the wild-type channel, and Brugada syndrome follows. Two mechanisms compete to explain that reduction. Under trafficking arrest, the mutant subunit is retained in the endoplasmic reticulum and drags its wild-type partner back with it, so the deficit is a protein-delivery failure. Under coupled gating, both subunits reach the membrane, dimerise there, and the mutant suppresses its partner's opening, so the deficit is a functional failure at normal protein levels. The distinction is not academic. A trafficking perturbation may alter both WT and mutant abundance and activity. Its direction of effect depends on which populations change and by how much, not solely on the mechanism label. Competing intervention responses therefore need measurement.

The published literature on this question looks, at first reading, like a straightforward contradiction between two well-regarded groups. It is not. Section 3 below shows why. Section 4 shows something stronger: no amount of additional co-expression current recording, however carefully controlled, can settle it, because the two mechanisms are algebraically indistinguishable at the level of current alone.

Methods

Literature search. I ran eighteen PubMed queries against PubMed E-utilities, covering every N-terminal-domain SCN5A missense variant with published functional data that I could identify (R104W, R104Q, R121W, Y87C, A124D, R27H, K126E, G35S, E17K, L96P, and N-terminal deletion constructs) crossed against the assay classes capable of bearing on the trafficking-versus-gating question: surface biotinylation, glycosylation-state analysis, endoplasmic reticulum co-localisation, single-channel and gating-current recording, temperature and chemical-chaperone rescue, proximity ligation, FRET or BiFC, iPSC-derived cardiomyocyte recording, and dimerisation assays. This retrieved 305 unique records. I screened 303 abstracts and classified 80 as containing a direct measurement bearing on the fork; eight of those concern an N-terminal-domain variant specifically (PMIDs 19632629, 22529811, 22739120, 26786162, 32815768, 34122134, 42082654, 42492110). Two of the eight bear on the domain but not on this fork and one concerns a compound-heterozygous frameshift; I added O'Neill 2022 (PMID 35305865), which a keyword screen classes as a general dominant-negative survey but which in fact reports R104W and R121W individually. Six papers plus O'Neill yield eleven variant-level measurements. Every PMID cited in this paper was checked against PubMed E-utilities on 2026-07-25; author, journal, year, volume and pages are as returned by that service. The date the original search queries were run is not separately recorded.

One paper cited below did not come from those queries. Tano 2026 (PMID 41582807) was flagged screened-relevant by this project's own standing literature screen, which runs a broader query set than the eighteen above, and was folded into this manuscript at revision after the version of record was deposited. Its full text was read from PubMed Central (PMC12911491) rather than from an abstract, and the numbers attributed to it below are from that full text except where the paper's own abstract is the source.

This is a bounded search over named queries on a stated date, not an exhaustive one. Clatot 2012's full text was not obtainable: its PubMed Central record is abstract-only and the article itself is paywalled, so statements attributed to it below come from the abstract. Mercier 2015, Pujolas 2026, and a second 2026 paper by Iamshanova and colleagues (PMID 42082654) were likewise read from their abstracts and whatever metadata PubMed E-utilities returns; I did not have full-text access.

Structural measurements. Distances were measured on PDB accession 8VYJ, chain A, a cryo-EM structure solved to 3.6 Å resolution (deposited 8 February 2024, released 12 February 2025), using Biopython's Bio.PDB module with the ShrakeRupley algorithm (module version not recorded), after stripping hydrogens and heteroatoms. These are my own computations, not a value taken from a published table. The Cβ to Cβ distance between residue 96 and residue 104 is 11.0 Å. The dimer interface mapped by Clatot 2017 (PMID 29233994, residues 493-517, first intracellular loop) is unresolved in 8VYJ: the model has a 247-residue gap spanning residues 433 to 681. Measured to the nearest resolved flanking residues, the centroid-to-centroid distance from R104 is 23.4 Å to residue 433 and 32.0 Å to residue 681, but a disordered 247-residue linker could place the true interface almost anywhere relative to R104. Structure, at the current resolution of the available model, cannot place the N-terminal domain at or away from the dimer interface.

Statistics. Cross-study summaries use conditional normal comparisons of reported means and SEMs, assuming independent study estimates and compatible quantities. Wang's current ratio uses first-order error propagation; numerator/reference correlation is also varied from -1 to +1 as a sensitivity analysis, not estimated from data. The endoplasmic-reticulum contrast uses independent-group SEM propagation and an approximate 95% normal interval. Actual experimental units, pairing, clustering and error definitions require source-level verification. Historical observed-effect power is retained as arithmetic provenance only: it is a transformation of the observed test statistic, not independent evidence about power for a prespecified biological effect.

The algebra: why current alone cannot tell the mechanisms apart

Nav1.5 α-subunits assemble and gate as dimers, and the interaction site maps to residues 493-517 in the first intracellular loop (Clatot 2017, PMID 29233994). Let f be the mutant fraction of the assembly-competent subunit pool. Under binomial pairing, the fraction of wild-type subunits partnered with a mutant subunit is exactly f.

Under coupled gating, a partnered wild-type subunit retains residual activity a, so co-expression current follows I/I_null = 1 minus f times (1 minus a). Under co-retention, a partnered wild-type subunit is pulled from the surface with probability c, so current follows I/I_null = 1 minus f times c.

Both expressions have the same form: I/I_null = 1 minus f times x, with x bounded between 0 and 1. A measured co-expression current constrains only the product f times x. It cannot separate f from x, and it cannot identify which biological process, retention or silencing, supplies x. This is the central negative result of this paper. It is not a limitation of the available studies. It is a property of the observable itself, and no re-analysis of published current percentages, however careful, escapes it.

There is a further correction that any re-analysis must apply before comparing numbers across studies. In a matched-total-DNA co-transfection, the heterozygous arm receives half the wild-type plasmid of a wild-type-alone arm, so the no-dominant-negative null is close to 50 percent of wild-type-alone current, not 100 percent. Wang 2020 measured this directly in their own system: their wild-type plus empty-vector arm reaches 53.9 percent of wild-type alone in one table and 44.3 percent in another, a mean of 49.1 percent. Studies differ in how they handle this. Wang 2020 included a wild-type plus empty-vector reference arm and needs no correction. O'Neill 2022 used a landing-pad system with stable dual integration, in which the wild-type allele is present at equivalent levels in both arms; there the no-effect null genuinely is 100 percent, and their measured two-allele reference reaches 218.4 ± 7.7 percent. Clatot 2018 used a full-dose wild-type-alone reference, so its headline 75 percent reduction, read against the correct roughly 50 percent null, corresponds to near-complete interference rather than an implausibly large effect.

With that correction in hand, the arithmetic still constrains something useful. At f = 0.5, the largest dominant-negative effect a 1:1 dimer can produce is 50 percent of null. Solving for the x each measured variant requires: R104W needs x = 0.61 (O'Neill, 69.6 ± 7.3 percent of null) or x = 0.80 (Wang Table 2, 60.2 ± 13.1 percent); R121W needs x = 0.95 (O'Neill, 52.7 ± 8.4 percent); Y87C needs x = 0.98 (Wang Table 1, 50.9 ± 15.9 percent); L325R requires x = 1.00 under these assumptions. L96P is excluded from this inference until its dose-matched comparator is verified; an abstract-level roughly 50 percent reduction does not establish I/I_null. R104Q's magnitude sits at the mild end of this set: O'Neill 2022 reports 68.3 ± 6.1 percent of null (n = 34), which requires only x ≈ 0.63. Whatever the mechanism turns out to be, its size requires nothing unusual.

Clatot 2012 and Wang 2020 were never measuring the same thing

The apparent contradiction between these two papers dissolves once the methods are read side by side.

Clatot 2012 (PMID 22739120) assayed localisation by immunocytochemistry in rat neonatal cardiomyocytes under co-expression, mutant and wild-type together, and reported that mutant subunits were mostly retained in the endoplasmic reticulum and that co-expression led to wild-type channel retention. This is the correct design for a dominant-negative question, but it is reported qualitatively: no percentage, no co-localisation coefficient, no stated n.

Wang 2020 (PMID 32815768) ran two separate protein experiments with different designs. Their surface biotinylation used 768 ng of a single construct, wild-type or variant alone, with no co-expression arm at all. That measures the mutant's own fate in isolation. It cannot, by construction, speak to whether the mutant affects wild-type trafficking, which is the dominant-negative question, so the widely quoted finding of reduced fully-glycosylated Nav1.5 at the surface describes the mutant alone, not the interference mechanism. Their endoplasmic reticulum co-localisation assay, by contrast, used a genuine co-expression design matching Clatot's cell type: GFP-wild-type plus FLAG-variant plus a calreticulin marker. This is the only Wang measurement that addresses the question Clatot addressed, and it is the sole point of overlap between the two papers.

The overlap leaves substantial uncertainty. Wang reports wild-type endoplasmic-reticulum co-localisation of 57.4 ± 9.3 percent for wild-type plus wild-type and 75.1 ± 9.5 percent for wild-type plus R104W. Treating the reported errors as independent-group SEMs gives a difference of 17.7 percentage points, SE 13.2944, z = 1.3314 and p = 0.1831. The approximate 95% normal interval is -8.36 to 43.76 percentage points: both no increase and substantial retention remain compatible with this calculation. The point estimate is in the direction of Clatot's qualitative report, but neither agreement nor absence of retention is established. Historical observed-effect power of approximately 27 percent is a deterministic transformation of the same test statistic and adds no independent evidence. A prospective power claim requires a prespecified effect and variance/design assumptions. These summaries do not establish whether a biologically meaningful retention effect was adequately tested.

Two smaller inconsistencies in Wang 2020 are worth flagging for anyone quoting its prose rather than its tables. The text states a roughly 55 percent current decrease for wild-type plus R104W relative to wild-type plus empty vector; Table 2 gives 23.1 versus 38.4 pA/pF, a 39.8 percent reduction. The text also reports Y87C's decrease as 25.1 ± 11.0 percent, but 51.1 minus 26.0 equals 25.1 pA/pF, a current-density difference reported as if it were a percentage; the paper's own ratio-based statement elsewhere, roughly 50 percent less, matches a recomputed 49.1 percent. Neither error changes a conclusion, but both propagate if the prose is cited uncritically.

R121W has a cross-study numerical discrepancy; R104W remains uncertain

For R104W, O'Neill 2022 reports 69.6 ± 7.3 percent of a single-wild-type-allele reference and Wang 2020's Table 2 reports 60.2 ± 13.1 percent of its own reference. The conditional comparison gives z = 0.63, p = 0.53. Failure to detect a difference does not establish equivalence or agreement within a biologically useful margin; no such margin was prespecified here.

For R121W, Clatot 2012 qualitatively reports a strong dominant-negative effect, whereas O'Neill 2022 reports 52.7 ± 8.4 percent of the single-wild-type-allele reference. Wang 2020's Table 2 reports 46.9 ± 3.4 versus 38.4 ± 3.4 pA/pF for wild-type plus empty vector. With zero numerator/reference covariance, first-order propagation gives 122.135 ± 13.976 percent and a conditional comparison with O'Neill of z = 4.258, p approximately 2.1×10⁻⁵. Varying that correlation from -1 to +1 changes the ratio SE from 19.668 to 1.960 and the comparison z from 3.247 to 8.050; the largest corresponding two-sided normal p is approximately 0.00117. Thus covariance alone does not erase the numerical discrepancy within this model. This does not rule out normalization, comparator, expression-system or other laboratory differences, and a qualitative third report cannot be treated as another quantitative estimate.

The practical lesson is to specify the variant, comparator and observable. R104W co-expression current reduction and wild-type retention are different questions. R121W summary currents differ across studies under the calculations above, but their cause is not identified by a cross-study test or by counting reports on either side.

L96P: a surface-expression lead awaiting methodological qualification

Pujolas and colleagues (PMID 42492110, Forensic Sci Int Genet 2026;86:103587), a molecular-autopsy study of SCN5A c.287T>C, p.Leu96Pro, report no measurable current when the variant is expressed alone, a roughly 50 percent reduction of peak current on co-expression with wild-type, cell-surface biotinylation showing preserved total and membrane Nav1.5, and patient-derived iPSC cardiomyocytes with reduced current but no reduction in SCN5A transcript.

L96 sits 11.0 Å (Cβ to Cβ) from R104 in the same domain in my measurement on 8VYJ chain A. This proximity motivates follow-up; it does not transfer a mechanism. The accessible abstract does not establish the WT dose and total-DNA comparator behind the co-expression reduction, whether surface measurements used that same co-expression condition, or whether WT surface protein was distinguished from mutant signal. A roughly 50 percent reduction can have different interpretations under different reference designs. L96P cannot yet be used here as a verified surface-preserved dominant-negative example selecting coupled gating.

Full methods and figure-level evidence are needed before drawing that inference. Relevant checks include the no-interference reference, condition matching, allele specificity, independent preparations and uncertainty against a justified preservation margin. Glycoform resolution is an additional issue: the Mercier 2015 abstract (PMID 25721215) reports surface-localized core-glycosylated channels without current, so pooled surface signal need not represent a functional population. The reported additional familial SCN5A splice-site deletion also complicates genotype-phenotype attribution. These are limits on the evidence accessible for this review, not claims that the investigators omitted appropriate controls.

Set against this, the older Clatot 2012 data point is instructive by contrast: a construct deleting the entire N-terminal domain abolished current yet was still addressed to the membrane, and an isolated N-terminal-domain-only construct doubled current. Losing the domain, on that evidence, is not intrinsically a trafficking lesion.

At the strength accessible here, Clatot's qualitative co-expression retention report, the surface-resident deletion construct and the L96P abstract motivate distinct tests. The deletion is not itself an allele-resolved dominant-negative experiment, and L96P lacks the verified comparator and surface-assay details needed for the proposed inference. These observations do not justify ranking coupled gating above retention for the N-terminal domain; both remain live hypotheses.

Tano 2026: dimer formation and dominant-negative effect come apart

A dimer can form and do nothing. Tano and colleagues (PMID 41582807, Circ Arrhythm Electrophysiol 2026) genotyped a family presenting with sick sinus syndrome and sudden cardiac death, and found a proband carrying two SCN5A variants, p.T1396P from her mother and p.G833R from her father. Expressed alone, p.T1396P generated no sodium current. Co-expressed with wild type it reduced peak sodium current by 37 percent, prolonged current decay, and shifted steady-state inactivation 5.6 mV rightward: a dominant-negative effect with a coupled-gating signature. p.G833R expressed alone gave current comparable to wild type, and in co-expression it produced no dominant-negative effect and no coupled-gating effect at all. Co-immunoprecipitation and proximity ligation nonetheless returned positive signals for every pairwise combination of wild-type and variant plasmids tested, p.G833R included, and the paper states the point in its own words: the signals appear irrespective of the electrophysiological characteristics.

That is an empirical version of this paper's algebraic result, arriving from the other side. The negative result in the algebra section is that a co-expression current cannot say which mechanism produced it. Tano's is that a physical association assay cannot say whether any mechanism is operating. Co-immunoprecipitation and proximity ligation are the assays most often offered as mechanistic corroboration when a co-expression current comes in low, and here they are positive for a variant that does nothing to its partner. Between the two results, both of the cheap routes to a mechanism are closed: current alone cannot separate co-retention from coupled gating, and dimerisation alone cannot establish that either is happening.

Their surface measurement has a different scope from the co-expression current assay. Tano reports no significant surface-biotinylation difference between separately expressed wild-type and variant channels. This does not establish equivalence, and it does not directly measure allele-resolved WT surface abundance in the co-expression condition exhibiting interference. It therefore cannot, by itself, establish preserved WT surface protein during the dominant-negative phenotype. Compatible co-expression measurements and uncertainty against a justified preservation margin are needed for that inference.

Difopein supplies a perturbation to investigate, not a mechanism-specific verdict. Tano reports reversal of the T1396P/WT gating changes with difopein. The source table gives peak-current magnitudes of 100.6 without and 253.6 pA/pF with difopein: a treated/untreated ratio of 2.52, or a 152% increase relative to the untreated mean. Figure 3G marks the WT/T1396P treatment comparison significant, but this is not a treatment-by-background interaction test. A WT-only inactivation response also makes background-matched controls important. Difopein responsiveness would support a perturbation-sensitive process; attribution specifically to coupled gating requires additional evidence addressing relevant abundance, trafficking and direct functional effects. Insensitivity would not diagnose retention. None of these findings establishes R104Q responsiveness or a therapy.

The experiment that would break the degeneracy

Additional readouts can test specific explanations that peak current alone cannot separate. The proposed design can constrain abundance and functional contributions, but unique attribution depends on qualified measurements, the model and whether mechanisms coexist. Published components do not establish that their combined implementation will be decisive for every variant.

Clatot 2018 (PMID 30118344) ran it on L325R, a domain I to II linker variant, using matched total DNA throughout. Co-expression gave a 75 percent current reduction against a wild-type-alone reference; since the correct no-effect null under matched DNA is close to 50 percent, this corresponds to near-complete interference rather than an implausible super-maximal effect. Surface biotinylation showed L325R itself reaches the surface at 40 percent when expressed alone, and in co-expression, total surface Nav protein was not significantly altered relative to wild-type alone: the trafficking deficit could not account for the current loss. Single-channel recording then supplied the deciding evidence: open probability fell 54 percent, accounting for only about half the whole-cell loss, while double-level, coupled openings fell 85 percent and coupled closings fell 92 percent. Co-immunoprecipitation, DSS crosslinking and blue-native gels confirmed the dimer itself stayed intact throughout. Surface protein preserved, plus open probability and coupling destroyed, together identify coupled gating in a way current alone cannot.

The design needs two negative controls to be interpretable, and both already exist in the published record. R878C (PMID 18616619) carries no current, persists at the membrane, and exerts no dominant-negative effect on co-expressed wild-type: a dead subunit at the surface does not automatically interfere. W822X (PMID 16239976) gives roughly 50 percent current reduction on heterozygous co-expression with no dominant-negative effect at all, the empirical haploinsufficiency benchmark against which any claimed interference should be measured.

The design I would run for R104Q, or for any single variant, has five conditions at matched total DNA, with the wild-type dose held constant across every co-expression arm: wild-type plus empty vector as the reference arm (not wild-type alone at full dose); the variant alone, to measure its own residual current; wild-type plus wild-type, to control for DNA dose and establish the two-allele ceiling; wild-type plus the variant, the question itself; and wild-type plus a loss-of-function variant that is not dominant-negative, W822X or R878C, to make the result interpretable against a known null. Three readouts are needed. Allele-resolved surface protein, with wild-type and mutant differentially tagged and quantified separately, and with glycosylation state resolved by EndoH or PNGase F digestion, since total surface signal alone cannot distinguish a genuinely functional population from a core-glycosylated, silent one. Single-channel recording for open probability and coupled-gating fraction, the Clatot 2018 measurement. And whole-cell current across all five arms, to anchor the new data to the existing literature.

Under the stated pure-mechanism assumptions and negligible mutant current, a current of 68.3 percent of the matched no-interference reference would correspond to about 68 percent WT surface abundance in a pure-retention explanation, versus unchanged WT surface abundance in a pure activity-loss explanation. This roughly 32-point model separation is not an assay-resolution or power estimate. Reporter specificity, matched references, preparation variance, uncertainty and possible mixed mechanisms must be qualified before interpreting a real result. An inconclusive result must remain possible; a nonsignificant surface comparison alone does not establish preservation.

The five arms above preserve the original design. A difopein follow-up requires matched perturbation and control conditions in both WT/reference and WT-plus-variant backgrounds, with delivery and reporter burden addressed. Estimate the treatment-by-background contrast rather than infer it from significance in one background and nonsignificance in another. This is not merely one additional condition and does not, alone, convert the experiment into a uniquely coupled-gating-specific causal test.

What would falsify this

The binomial-dimer parameterization fails if its dimer or random-pairing assumptions fail in the system under study. That would invalidate the specific formula, not establish that peak current uniquely identifies the biological mechanism. Ambiguity between abundance and effective activity can persist outside that parameterization. The model assumptions and the broader identifiability claim therefore require distinct tests.

The Wang endoplasmic-reticulum uncertainty calculation assumes the printed errors are SEMs and the groups are independent. A paired or clustered design would require different propagation. At fixed reported SEMs, increasing assumed equal group sizes changes degrees of freedom, not precision: the historical noncentral-t observed-effect power rises from 21.7–24.7 percent at n = 5–12 toward 26.5 percent, rather than supplying evidence that a larger unknown n resolves the effect. Prospective sample-size planning is a separate question requiring biological effect and variance assumptions.

A mechanistic use of L96P first requires verification of its no-interference comparator and condition-matched, preferably allele-resolved surface assay. Glycoform identity and preservation uncertainty then constrain whether unchanged measured surface signal can exclude a delivery or maturation contribution. None of these missing details can be inferred from the abstract alone.

A qualified co-expression study measuring WT retention together with surface abundance and function could change the relative support for these mechanisms. The direction of the existing R104W point estimate motivates that measurement; it does not predict its result or establish which mechanism dominates this domain.

The most serious challenge to this paper is not to its algebra but to why the algebra matters, and it comes from the same paper cited above in its support. Tano 2026's title is a claim and not a hedge: phenotypic severity in their family was independent of dominant-negative and coupled-gating status. The variant that carries the dominant-negative effect and the coupled-gating signature, p.T1396P, is the one the mother carries alone, and she was the milder of the two. She developed sinus bradycardia in her forties, lost consciousness once under antiarrhythmic medication and needed temporary pacing, and at seventy had only right bundle branch block and first-degree atrioventricular block, with no permanent device. The proband carries that same variant plus p.G833R, which produces neither effect, and she presented at thirty-two with severe sinus bradycardia and recurrent ventricular tachycardia. The authors read this as evidence that losing coupled gating is itself harmful, which inverts the usual framing rather than merely weakening it. If severity is set by something other than dominant-negative status, then resolving the dominant-negative mechanism for a variant, which is the whole object of this paper, buys less clinical information than the framing around it implies.

I do not think that conclusion follows from one family, and the reasons are worth stating rather than implying. The phenotype is sick sinus syndrome and bradycardia, not Brugada syndrome. It is a single pedigree, and the severity comparison is two individuals of different ages and different genotypes, each an n of one. The proband is a compound heterozygote and her mother is not, so straightforward allelic burden is an explanation the study cannot separate from its coupled-gating interpretation, although the authors' own finding that p.G833R alone gives current comparable to wild type argues against the simplest form of it. Both variants are classified as variants of uncertain significance under ACMG. Every functional measurement is in HEK293 cells, which the authors flag themselves as their first limitation. The deceased sister, the only other affected member, was never genotyped. But the direction of the challenge is real and it is the direction this work should want tested. A mechanism paper that cannot be cashed out in phenotype is a mechanism paper and not a clinical one, and this paper does not claim otherwise.

Two further limits apply specifically to R104Q. No protein-level or mechanistic co-expression data exist for R104Q at all; everything above reasons from R104W, R121W, Y87C and L96P by structural and positional analogy within the same domain, and R104Q is a different substitution from R104W at the same position, which need not share its mechanism. And structure cannot adjudicate this question either: the dimer interface mapped by Clatot 2017 is unresolved in the best available cryo-EM model of Nav1.5, so the N-terminal domain can be neither placed at, nor excluded from, that interface on structural grounds at present.

Correction, 6 August 2026: the data availability statement was false, and the tables it named have been regenerated

Interpretation update, 7 September 2026. The dated record below preserves the historical arithmetic and deposit correction. Its statements that observed-effect power strengthens the biological argument are superseded by the present uncertainty analysis. Reproducing a calculation does not validate that interpretation. The historical L96P dimer row also remains a conditional arithmetic record, not a verified mechanistic estimate.

Version 1 of this paper said its derived tables were deposited in the shared data archive. They were not. The statement named two — the per-study numbers used in the power calculation, and the dimer-arithmetic solutions for x — and the archive at 10.5281/zenodo.21799234 contained neither, nor any other file belonging to this paper. That is a statement a reader cannot check except by downloading the archive and finding nothing there. It was found in an audit of all eleven papers' data availability statements on 6 August 2026, which found the same class of defect in six of them.

No saved output of either calculation exists anywhere: the tables were assembled to write the paper and never written to disk. Both have been regenerated from first principles rather than reconstructed from the paper, which is possible here in a way it is not for most papers, because everything in them is closed-form arithmetic over published summary statistics and the method is fully stated in the Methods above. The regenerating script, p7_power_and_dimer.py, also recomputes the three structural distances from the public 8VYJ chain A coordinates.

The regeneration reproduces 37 of the 39 quantities checked, at the precision printed, including every solved x, every predicted surface-protein separation, both cross-study z-tests, both arithmetic slips this paper flags in Wang 2020, and all four structural measurements. The two that do not reproduce exactly are recorded here rather than quietly adjusted, and both are rounding propagations:

Neither changes an argument. The printed values are kept, and their provenance is now stated and deposited so that a reader can compute it either way.

One thing the regeneration sharpened. The post-hoc power of approximately 27 percent for Wang's endoplasmic-reticulum null is a normal approximation, reproducing as 26.5 percent; the Methods now say so. The exact non-central t at the group sizes this paper itself assumes for those panels, n = 5 to 12 per group, gives 21.7 to 24.7 percent. The printed figure is therefore the generous one, and the argument it supports — that Wang's null is a failure to resolve rather than evidence of absence — is if anything understated by it.

This section is not in the record deposited at 10.5281/zenodo.21799865 on 5 August 2026, which carries the false data availability statement and, separately, no mention of Tano 2026.

Data availability

The 7 September statistical interpretation audit is saved locally as check_p7_statistical_scope.py, P7_STATISTICAL_SCOPE_RESULTS.json and P7_STATISTICAL_SCOPE_REVIEW.md. These new audit artifacts are not claimed to be included in the existing public archive. The older deposited tables below preserve the historical calculations; their existence does not validate the superseded observed-power interpretation or the L96P comparator assumption.

This paper draws exclusively on published data retrieved through PubMed E-utilities (accessed 2026-07-25) and on structural coordinates from PDB accession 8VYJ, chain A. No new experimental data were generated.

The derived tables are deposited in the data archive whose identifier is recorded in DATA_DOI.txt alongside this manuscript, and they are these four historical files, named individually so that a reader can check this statement against the archive rather than take it on trust:

• P7_POWER_INPUTS.csv — the per-study numbers used in the power calculation and in the two cross-study z-tests: O'Neill 2022's four arms, Wang 2020's two current tables, its endoplasmic-reticulum co-localisation arms and its two wild-type-plus-empty-vector references, Clatot 2018's headline reduction and Pujolas 2026's L96P reduction, each with its dispersion, dispersion type, n, unit, PMID and the source table it was read from.
• P7_DIMER_ARITHMETIC.csv — the dimer-arithmetic solutions for x, one row per variant, with I/I_null, f, the solved x, the surface-protein level each mechanism predicts and the separation between them.
• p7_power_and_dimer.py — the script that writes both tables from those inputs and checks them against the historical quantities checked in the August regeneration, including the three structural distances, which it recomputes from the public 8VYJ chain A coordinates.
• P7_VERIFICATION_OUTPUT.txt — that script's output. 37 of 39 checked quantities reproduce at the precision this paper prints them to; the two that do not are named in the correction section below.

Every number in these tables is a reformulation of numbers already published under the PMIDs cited throughout, and P7_POWER_INPUTS.csv records which publication and which table each came from. The four rows of O'Neill 2022 Supplementary Table 1 that this paper uses are quoted inside p7_power_and_dimer.py so that it runs standalone; that table is not redeposited in full, because it is a third party's supplementary material and this paper prints the four rows it needs in its own text.

Two things this paper reports are not reproducible from anything deposited, and this statement says so rather than implying otherwise. The eighteen PubMed E-utilities query strings behind the literature search were not recorded verbatim, and the date they were first run is not recorded either, as the Methods already state; the retrieval counts that follow from them — 305 unique records, 303 abstracts screened, 80 classified as bearing on the fork, eight on an N-terminal-domain variant, eleven variant-level measurements — therefore cannot be re-derived. And every figure attributed to Clatot 2012, Mercier 2015, Iamshanova 2026 and Pujolas 2026 is transcribed from an abstract, because those full texts were not obtainable. Tano 2026 measurements instead come from the full text at PMC12911491, as stated in Methods.

Archive versioning. The concept DOI 10.5281/zenodo.21799233 always resolves to the current version of the data archive and is the identifier to follow for access. The version current at the time of this revision is version 2, 10.5281/zenodo.21840036. Version DOIs cited elsewhere in this manuscript name the specific version read and are deliberately not rewritten.

Competing interests

The author is a heterozygous carrier of the SCN5A variant discussed as the motivating example in this paper, p.Arg104Gln.

Use of AI tools

This work was carried out with AI coding and research assistants (Anthropic Claude, via Claude Code; OpenAI assistants, via Codex, for subsequent analysis and revisions). That use is disclosed here rather than left to inference.

Analysis code. The great majority of the analysis code in this project -- parsers, genome scans, regeneration scripts and verification scripts -- was written by an AI assistant working to my specification. I set what each script had to compute, chose the thresholds and the decision rules, and checked the output against the claims it is used to support.

Manuscript text. The prose of this manuscript was drafted by an AI assistant. I directed the drafting and revised the result, and I am responsible for every claim it makes.

Scientific decisions. The questions asked, the thresholds set, what was allowed to count as a refutation, and what was published were mine.

Verification, which does not depend on any of the above. Where a claim in this manuscript is regenerable from deposited inputs, the script that regenerates it and that script's own output are in the data deposit. Reproduction does not require trusting any account of who wrote what.

What no AI system did. No AI system generated, altered or selected any experimental measurement; this project contains no wet-lab data of any kind. All primary literature cited was retrieved from PubMed, PMC and publisher sources. Every reference in this manuscript has been machine-resolved against its own record, including a check that each PMID's first author and year match the author and year printed beside it in the text.

References

1. Clatot J et al. Dominant-negative effect of SCN5A N-terminal mutations through the interaction of Nav1.5 α-subunits. Cardiovasc Res 2012;96(1):53-63. PMID 22739120.
2. Wang Z et al. Calmodulin binds to the N-terminal domain of the cardiac sodium channel Nav1.5. Channels (Austin) 2020;14(1):268-286. PMID 32815768.
3. O'Neill MJ et al. Dominant negative effects of SCN5A missense variants. Genet Med 2022;24(6):1238-1248. PMID 35305865.
4. Clatot J et al. Voltage-gated sodium channels assemble and gate as dimers. Nat Commun 2017;8(1):2077. PMID 29233994.
5. Clatot J et al. Mutant voltage-gated Na+ channels can exert a dominant negative effect through coupled gating. Am J Physiol Heart Circ Physiol 2018;315(5):H1250-H1257. PMID 30118344.
6. Pujolas AR et al. Molecular autopsy identifies the NaV1.5 p.Leu96Pro variant causing sodium current loss-of-function in unexplained sudden cardiac death. Forensic Sci Int Genet 2026;86:103587. PMID 42492110.
7. Mercier A et al. Nav1.5 channels can reach the plasma membrane through distinct N-glycosylation states. Biochim Biophys Acta 2015;1850(6):1215-1223. PMID 25721215.
8. Zhang Y et al. Correlations between clinical and physiological consequences of the novel mutation R878C in a highly conserved pore residue. Acta Physiol (Oxf) 2008;194(4):311-323. PMID 18616619.
9. Keller DI et al. A novel nonsense mutation in the SCN5A gene leads to Brugada syndrome and a silent gene mutation carrier state. Can J Cardiol 2005;21(11):925-931. PMID 16239976.
10. Tano A, Kato K, Yamauchi K, Jinzai H, Iguchi T, Toyoda F, Baba Y, Kubo T, Ohno S, Makiyama T, Nakagawa Y, Horie M. Phenotypic severity of SCN5A-related bradycardia is independent of dominant-negative and coupled gating effects. Circ Arrhythm Electrophysiol 2026. doi:10.1161/CIRCEP.125.014270. PMID 41582807, PMC12911491.

Back to all ten papers