Truncation as a Rewiring Event: Interfaces, Degrons, and Condensates Behind NMD Escape

Question: How do disease-causing truncating variants that escape NMD produce gain-of-function or dominant-negative effects through altered protein-protein interaction interfaces or aberrant biomolecular condensation?

Answer

Across the retrieved papers, the answer is that NMD escape is a permissive step, not the mechanism: it licenses accumulation of a truncated polypeptide whose pathogenic activity is then determined by which interface, degron, and low-complexity segment the truncation retains, deletes, or appends. Escape alleles are quantitatively graded rather than a terminal-exon binary, being set by PTC position, the distance to the last exon-exon junction, the amino acid preceding the stop, and alternative polyadenylation [4] [1] [9]. Critically, the genes carrying these alleles are not distinguishable from controls by dosage sensitivity, interaction-network centrality, or tissue specificity, so the gain-of-function/dominant-negative behaviour cannot be predicted at the gene level and must be read off the protein [5]. Two mechanisms recur: stop-gain alleles that delete interaction-prone, disordered, positively charged C-terminal sequence and leave a truncated protein that still engages partners (dominant-negative poisoning, as with LDB1–LHX2 and BICC1–ANKS3/ANKS6), and frameshift alleles whose divergent tails are disorder-rich and phase-separation-prone, producing aberrant condensation or aggregation (RARB, desmin R150X, ASXL3 accumulation) [13] [11] [5] [10]. The strongest cases are those where the truncation also removes a degradation signal, making protein accumulation itself the first hit of a two-hit gain-of-function [7].

Escape is a sequence-encoded continuum, not a terminal-exon binary

Saturation genome editing of LMNA measured NMD activity for 722 PTCs and found that NMD is completely absent over the first 21 codons, then activates sharply over the next four, and that in the penultimate exon the position effect extends beyond the canonical 50–55 nucleotide rule as a quantitative function of PTC–exon junction distance [4]. The identity of the amino acid immediately preceding the PTC changes NMD efficiency substantially, with glycine codons promoting robust decay and peptide release rate acting as the primary determinant in a "window of opportunity" model [1]. Escape can also be engineered post-transcriptionally: poison exons are more often followed by an active intronic polyadenylation site than cassette exons, and blocking those sites converts NMD-escape isoforms into NMD targets in VRK3, NFX1, TM2D3 and DDX31 [9]. Even mRNA abundance is an unreliable proxy for escape, since downstream PTCs in a fission yeast reporter produced modest reductions or outright increases in mRNA without changes in turnover [3].

Gene-level dosage properties do not predict the mechanism; truncation position within the protein does

In 134 autosomal dominant genes enriched for NMD-escape frameshift or stop-gain variants, those genes were indistinguishable from length-matched controls in dosage sensitivity, interaction-network centrality and tissue specificity, despite being over-represented among neurodevelopmental-disorder risk genes [5]. Within KAT6A, early truncations (exons 1–15) are degraded by NMD and cause loss-of-function, whereas late truncations (exons 16–17) escape NMD and produce gain-of-function effects on a KAT6A–FOS–RBFOX–PTBP axis with inverse neuronal metabolic states [2]. In LDB1, N-terminal variants that disrupt homodimerization behave as loss-of-function, while NMD-escaping C-terminal variants that spare the dimerization domain instead impair binding to LHX2 in a dominant-negative manner, and C-terminal variants exacerbated toxicity in Drosophila whereas N-terminal ones did not [13]. A proteome-scale comparison of 5,168 canonical-versus-truncated homodimer pairs predicts complete interface ablation in 26.1%, partial destabilization in 17.5%, and truncation-associated emergence of a new high-confidence interface in 4.0% [12].

Stop-gain alleles poison through retained or exposed interfaces

Pathogenic stop-gain NMD-escape regions overlapped protein-protein interaction interfaces more often than expected (log odds ratio +1.22), and this enrichment survived adjustment for truncation position, alongside enrichment for Pfam domains [5]. Such alleles remove longer, more disordered and interaction-prone C-terminal sequence but leave a truncated protein enriched in arginine and positive charge — a composition expected to favour nucleic-acid and acidic-surface engagement rather than simple loss of function [5]. BICC1 truncations exemplify the dominant-negative route: the mutant products localize to cytoplasmic bodies, self-associate and bind ANKS3 and ANKS6 to varying degrees, and all truncations increased expression of a BICC1 target reporter despite wild-type BICC1 acting as a repressor, which the authors interpret as dominant-negative or gain-of-function [11]. In population-scale MODY data, only the NMD-escape class of INS loss-of-function variants was enriched in cases (OR 181), with protein modelling implicating aberrant proinsulin molecules rather than simple haploinsufficiency, and NEUROD1 and PDX1 were likewise enriched only for NMD-escape variants [6].

Frameshift-encoded tails and exposed disorder drive aberrant condensation and aggregation

Pathogenic frameshift alleles generated novel C-terminal tails that were less positively charged and had a higher phase-separating disordered-region fraction in their divergent segment than benign alleles in the same genes, and an NMD-escape truncating variant in RARB formed aberrant condensates in cells [5]. The Arg150Stop desmin variant, which removes the C-terminal tail, produced persistent aggregate-like structures in both overexpression and CRISPR-engineered homozygous cardiomyocytes, structures that were proteasome-resistant, partially autophagy-cleared, and disruptive to actin filament organization — interpreted as a toxic gain-of-function rather than a null state [10]. Predicted phase-separating disordered regions carry a three-fold elevated pathogenic substitution rate relative to non-phase-separating disordered regions, with the highest pathogenicity where a predicted phase-separating region also contains a short linear interaction motif, indicating that condensation propensity and interaction motifs are jointly constrained by disease selection [8].

Degron loss makes accumulation the first hit of a two-hit gain-of-function

ASXL3 patient truncations cause Bainbridge-Ropers syndrome by escaping both NMD and Cullin 4-dependent degradation, producing aberrant protein accumulation, widespread transcriptional dysregulation and altered chromatin accessibility; CRISPR knock-in of patient versus population truncations excluded simple haploinsufficiency, forced full-length ASXL3 failed to rescue the disease expression signature, and antisense oligonucleotides normalized it [7]. Deletion mapping identified a broad C-terminal destabilizing region, explaining why benign truncations cluster at the 3′ end of the gene [7]. This places a degradation signal, rather than an interaction interface or a condensate, as the decisive element for at least one chromatinopathy, and it reframes the "gain" as failure to clear a partially functional protein.

Conflicts and gaps

The retrieved papers disagree on how sharply dominant-negative and gain-of-function can be separated. BICC1 is explicitly left as "dominant-negative- or gain-of-function" without discrimination [11], and ASXL3 gain-of-function rests on exclusion of haploinsufficiency by failed full-length rescue, which excludes dosage but does not establish a positively neomorphic activity [7]. KAT6A gain-of-function is inferred from transcriptional and metabolic changes running opposite to the NMD-degraded early-truncation class rather than from an assay of a new biochemical activity [2]. The two proposed mechanisms assort by variant class in the same dataset — interface overlap for stop-gain, disordered tail composition for frameshift — but the interface analysis is enrichment statistics on ClinVar annotations and the condensate evidence is a single RARB allele, so the mechanistic split is a hypothesis rather than a demonstrated partition [5]. The claim that truncation can create new high-confidence interfaces is entirely computational, derived from AlphaFold homodimer models without experimental validation [12]. Whether condensates are causal or an epiphenomenon of exposed aggregation-prone sequence is unresolved: the desmin aggregates are operationally defined by persistence and protease resistance rather than by measured phase behaviour or saturation concentration [10]. Finally, prediction of escape itself remains incomplete, with LMNA SGE identifying 5′ and penultimate-exon patterns that current NMD models do not explain [4].

What would change the answer

Separation-of-function alleles at endogenous loci would settle the dominant-negative versus neomorphic question directly: knock-in of the same PTC with and without the divergent frameshift tail, or with point mutations that abolish a candidate interface while preserving protein abundance, would test whether the interface or the condensate is required. Allele-specific knockdown of the mutant transcript — already demonstrated for ASXL3 [7] — is the cleanest available discriminator, because rescue by removing the mutant protein, but not by adding wild-type protein, defines a toxic gain. Isogenic interaction proteomics (proximity labelling or affinity purification) comparing full-length, NMD-triggering, and NMD-escaping truncations would convert the interface-overlap enrichments into measured neomorphic interactions [5]. Optical and biophysical measurements — saturation concentration, FRAP recovery, 1,6-hexanediol sensitivity, and partition coefficients of patient versus benign truncations — are needed to establish that condensation, rather than generic aggregation, drives the phenotype [10]. Experimentally validated truncation libraries, rather than AlphaFold-only models, are required before the predicted emergence of new interfaces in 4% of human homodimers can be treated as a disease mechanism [12]. Finally, saturation genome editing with protein-level readouts combined with NMD inhibition would decouple transcript loss from protein gain at every truncation position, as LMNA SGE does for mRNA alone [4].

References

[1] Kolakada, D., Fu, R., Biziaev, N., Shuvalov, A., Lore, M., Campbell, A. E., Cortazar, M. A., Sajek, M. P., Hesselberth, J. R., Mukherjee, N., Alkalaeva, E., Akdemir, Z. H. C., Jagannathan, S.. "Systematic analysis of nonsense variants uncovers peptide release rate as a novel modifier of nonsense-mediated mRNA decay". bioRxiv. 2024-11-10T00:00:00.000Z. URL: https://doi.org/10.1101/2024.01.10.575080

[2] Nava, A. A., Perez-Rodriguez, Y., Hsieh, T.-C., Byrne, A. S., Krall, A. S., Freudenberg, J., Mansooralavi, N., Pandey, V., Stiles, L., Beninca, C., Li, J.-M., Choufani, S., Singh, M., Moosa, S., Valenzuela, I., Tizzano, E. F., Piton, A., Lacombe, D., Perrin, L., Marquez, J., Ortigoza-Escobar, J. D., Ahmadyar, S., Pimentel, H., Wohlschlegel, J. A., de la Torre-Ubieta, L., Christofk, H. R., Weksberg, R., Lowry, W. E., Arboleda, V. A.. "Dominant truncating variants in KAT6A cause two neurodevelopmental disorders with opposite gene regulatory and metabolic changes". medRxiv. 2026-08-17T00:00:00.000Z. URL: https://doi.org/10.64898/2026.08.11.26358095

[3] Owuamalam, P., Hossain, M. N., Brogna, S.. "Nonsense mutations can increase mRNA levels". bioRxiv. 2025-12-16T00:00:00.000Z. URL: https://doi.org/10.1101/2025.11.16.688500

[4] Cortazar, M. A., Schmidt, J., Egab, I., Coban Akdemir, Z. H., Jagannathan, S.. "Genomic stop codon scanning reveals quantitative principles of nonsense-mediated mRNA decay". bioRxiv. 2025-12-23T00:00:00.000Z. URL: https://doi.org/10.64898/2025.12.20.695734

[5] Xu, J., Schmidt, J., Bozkurt-Yozgatli, T., Egab, I., Mestroni, L., Taylor, M., Posey, J. E., Gibbs, R. A., Boerwinkle, E., de Vries, P. S., Morrison, A., Yi, S., Shaw, C., Carvalho, C. M. B., Lupski, J. R., Jagannathan, S., Timucin, E., Coban-Akdemir, Z.. "Phase Separation Contributes to Pathogenicity for Nonsense Mediated Decay-Escaping Variant Alleles". medRxiv. 2026-10-19T00:00:00.000Z. URL: https://doi.org/10.64898/2025.12.01.25341387

[6] Laver, T. W., Sriram, A., Wakeling, M. N., Siklar, Z., Kalyon, O., Hattersley, A., Weedon, M. N., Flanagan, S., De Franco, E., Colclough, K., Patel, K. A.. "Systematic analysis of loss-of-function variants across MODY genes demonstrates gene- and NMD-specific effects and identifies NMD-escape INS variants as a novel cause of MODY". medRxiv. 2025-09-08T00:00:00.000Z. URL: https://doi.org/10.1101/2025.09.07.25334427

[7] Nakamura, Y., Nguyen, T., Mor, N., Dominissini, D., Tang, I., Torio, C. J., Thulaseedharan, H., Zhou, R. Y., Zhang, W., Skourti-Stathaki, K., Douville, J., Mignon, L., Dung, A., Freier, S., Watt, A., Jagannathan, S., Crooke, S. T., Gleeson, J. G.. "ASXL3 truncating patient variants mediate transcriptional gain-of-function and are antisense oligonucleotide-responsive". medRxiv. 2026-07-22T00:00:00.000Z. URL: https://doi.org/10.64898/2026.07.20.26358515

[8] Kipp, O. L., Lewis, K. A., Hough, L. E., Whitten, S. T.. "Missense mutations in intrinsically disordered protein regions link pathogenicity and phase separation". bioRxiv. 2025-08-12T00:00:00.000Z. URL: https://doi.org/10.1101/2025.02.19.639140

[9] Vlasenok, M., Kuznetsova, A., Skvortsov, D. A., Dontsova, O. A., Pervouchine, D. D.. "Novel examples of NMD escape through alternative intronic polyadenylation". bioRxiv. 2026-03-27T00:00:00.000Z. URL: https://doi.org/10.1101/2025.08.27.672552

[10] Mitra, S., Ghosh, T., Mishra, A. K., Sanga, S., Maulik, M., Acharya, M.. "Novel truncating Desmin mutation Arg150Stop disrupts structural integrity and cellular homeostasis by formation of persistent aggregate-like structure". bioRxiv. 2025-10-15T00:00:00.000Z. URL: https://doi.org/10.1101/2025.10.14.682287

[11] Eylath, N. S., Kidd, K. O., Alyea-Herman, P., Meyersiek, J., Colombo, D. A., Rennke, H. G., Guleserian, A. J., Adams, V. W., Bianchi, G., Maillard, A., Faguer, S., Izzi, C., Bergmann, C., Lecker, S. H., Astley, M., Taylor, A., Martin, L. M., Means, S., Sanchez, A., Weller, N., Hodanova, K., Kmochova, T., Stranecky, V., Hartmannova, H., Svojsova, K., Sikora, J., Pavlovicova, L., Yang, H., Harris, P. C., Kmoch, S., Bleyer, A. J., Zivna, M., Czarnecki, P. G.. "Heterozygous truncating variants in BICC1 are a novel cause of autosomal-dominant tubulointerstitial kidney disease". medRxiv. 2026-08-23T00:00:00.000Z. URL: https://doi.org/10.64898/2026.08.20.26360556

[12] Karagöl, T., Karagöl, A.. "A structural grammar of truncation across the human homodimer landscape". bioRxiv. 2026-05-09T00:00:00.000Z. URL: https://doi.org/10.64898/2026.05.06.723091

[13] Fluri, R., Coll-Tane, M., Brunet, T., Cogne, B., Conrad, S., Nizon, M., Nicita, F., Travaglini, L., Novelli, A., Glissmeyer, M., Peterson, A., Buchan, J. G., Serber, D., Meier, K., Gaertner, J., Diegmann, S., Pingault, V., Attie-Bitach, T., Courtin, T., Schneider, M. C., Hung, W., Sahai, I., OGrady, L., Steindl, K., Mehta, S. G., Depienne, C., Heron, D., Keren, B., Heide, S., McKee, S., Laccone, F., Dyer, L. M., Melver, C., Motter, C., Jones, W., Wilson, Z. T., Vats, D., Huss, K., Zweier, C., Sticht, H., Gregor, A.. "De novo variants in LDB1 are linked to distinct neurodevelopmental phenotypes determined by variant location and differing pathomechanisms". medRxiv. 2026-02-28T00:00:00.000Z. URL: https://doi.org/10.64898/2026.02.26.26347174

How this survey was made

This survey answers one question: “How do disease-causing truncating variants that escape NMD produce gain-of-function or dominant-negative effects through altered protein-protein interaction interfaces or aberrant biomolecular condensation?”

We keep our own collection of recent preprints and refresh it every day from bioRxiv, medRxiv, arXiv, PsyArXiv, SocArXiv and ChemRxiv. For each survey, we search that collection for recent work and keep the strongest matches: papers that use the same terms as the question, and papers on the same topic in different words.

The text was written from the titles and abstracts of those papers; the full texts were not read. Statements that draw on a particular paper are marked with a number in brackets. The reference list just above is the 13 papers whose work this text draws on, chosen from the 213 papers those searches returned; the rest of the shortlist is not printed. It is not a complete review of everything published on the question.

Treat this as a starting point and a reading list, not a peer-reviewed review. The searches behind this survey ran on October 21, 2026 and covered work from the previous two years.