STRC RBM24 Exon Mapping to Human Protein
All four RBM24-regulated STRC exon-skipping events (Sun et al. 2026 SD03) fall inside the mini-STRC therapeutic window (aa 700–1775). The strongest event (E1, 56 aa in-frame, dPSI −0.54) lies at human aa 1047–1102. The TMEM145 interaction interface (from AF3 Job 2 CIF, atom-atom < 5 Å) is at aa 1603–1770 — the C-terminal ARM repeats. None of the RBM24-regulated exons directly overlap the TMEM145 interface; nearest (E2/E4) is 227 aa upstream.
Mapping result
Ensembl canonical transcripts: mouse ENSMUST00000038389 (1809 aa) → human ENST00000450892 (1775 aa). Cross-species aa positions via BLOSUM62 global pairwise alignment (Biopython).
| Event | mm10 coords | nt | ÷3 | Mouse aa | Human aa | In mini-STRC (700–1775) | Distance to TMEM145 interface |
|---|---|---|---|---|---|---|---|
| E1 | chr2:121203399–121203567 | 168 | 56.0 ✓ | 1083–1138 | 1047–1102 | ✅ | 501 aa (nearest IF res: 1603) |
| E2 | chr2:121198816–121198894 | 78 | 26.0 ✓ | 1345–1410 | 1311–1376 | ✅ | 227 aa |
| E3 | chr2:121206192–121206262 | 70 | 23.33 ✗ | 750–770 | 712–732 | ✅ (boundary) | 871 aa |
| E4 | chr2:121198816–121199013 | 197 | 65.67 ✗ | 1345–1410 | 1311–1376 | ✅ | 227 aa |
E2 and E4 share the same mouse exon (ENSMUSE00000293748) but differ in splice-site choice — E2 is a short 78 nt variant (in-frame), E4 uses the full 197 nt canonical exon (frameshift). RBM24 appears to regulate both splice-site selection and exon inclusion at this locus.
Actual TMEM145 interface (from AF3 Job 2 CIF, 2026-03-16)
Reading job2-mini-complex.cif (Mini-STRC 594–1775 + TMEM145, ipTM 0.43) with Biopython and computing all residue pairs with min atom-atom distance < 5 Å gives 32 STRC interface residues clustered in 6 segments, all in the C-terminal ARM repeats region:
| Interface cluster | STRC aa | Residues |
|---|---|---|
| I1 | 1603–1607 | 5 |
| I2 | 1630–1638 | 9 |
| I3 | 1648–1651 | 4 |
| I4 | 1669–1680 | 12 |
| I5 | 1692–1707 | 16 |
| I6 | 1770 (C-terminus) | 1 |
This matches Derstroff et al. 2026: “STRC binds TMEM145 via C-terminal armadillo repeats.” Job 1 (Full STRC + TMEM145) showed 60 N-terminal contacts — likely artifactual sliding of the intrinsically disordered N-terminal region onto TMEM145.
What this means
The naïve mechanistic link (“RBM24 loss → E1 deletion → TMEM145 anchor broken”) is not supported by direct interface overlap. E1 is ~500 aa upstream of the nearest TMEM145 contact residue.
Three remaining possibilities for how RBM24 regulation could still matter for TMEM145 anchoring:
- Allosteric / foldability — E1 deletion may destabilize the ARM-repeat fold globally, indirectly affecting TMEM145 binding. Testable with AF3 ΔE1 solo (pTM) and ΔE1 × TMEM145 (ipTM).
- Scaffold geometry — a 56 aa shortening changes distance from the membrane-embedded GPI anchor (C-terminus) to the TMEM145 interface. Geometry could misregister.
- Independent contribution — E1 may serve a separate function (protein interaction with another partner, glycosylation patch, disulfide bridge — C1053/C1081 are conserved in 6/6 mammals). TMEM145 anchoring is not the mechanism by which RBM24 loss affects STRC function.
The weaker version of the link still holds: all four RBM24-regulated exons are inside the mini-STRC construct, so mini-STRC naturally preserves the sites RBM24 normally protects from skipping. But this is a “the construct covers it” statement, not “the construct’s key binding site is what RBM24 regulates.”
Therapeutic implications (revised)
- Mini-STRC (700–1775) covers all RBM24-regulated segments. The construct remains therapeutically sound.
- Hypomorphic patients with intact RBM24 may benefit from boosting RBM24 — if E1/E2 inclusion matters for STRC function (still an open question given the TMEM145 interface is downstream). See STRC RBM24 Regulatory Hypothesis.
- ASO-mediated E1 inclusion remains a targetable intervention if E1 function is confirmed (conservation is very high, see below).
- NMD from E3/E4 reduces total STRC protein — this is an RBM24-loss mechanism independent of the TMEM145 hypothesis.
Follow-up (2026-04-20)
E1 region is extremely conserved (functional signal, independent of TMEM145)
The 56 aa E1 segment (human STRC 1047–1102) is almost invariant across mammals:
| Species | Identity to human |
|---|---|
| Chimp | 100 % (56/56) |
| Rabbit | 98 % |
| Dog | 95 % |
| Pig | 95 % |
| Rat | 93 % |
| Cow | 89 % |
Human sequence: LSLEELCSLHLLLPGLSPQTLQAIPRRVLVGACSCLAPELSRLSACQTAALLQTFR. The two cysteines (C1053, C1081) are conserved in all 6 species — consistent with a disulfide-anchored fold or interaction surface. 100 % conservation over ~90 Myr of mammalian evolution is a strong functional signal, even if the function is not TMEM145 binding.
E3 and E4 trigger NMD
NMD prediction per the 50-nt rule (mouse Strc CDS = 5430 nt):
- E3 (70 nt skip, +1 frameshift) → PTC at codon 773, 3111 nt upstream of the original stop → NMD triggered
- E4 (197 nt skip, +2 frameshift) → PTC at codon 1353, 1371 nt upstream of the original stop → NMD triggered
Consequence: RBM24 loss reduces total STRC mRNA levels via NMD on these two events, independent of any interface argument.
Method and caveats
- rMATS coordinates treated as 0-based half-open (168 nt span ÷ 3 = 56 aa — the in-frame check confirms this convention).
- Cross-species aa mapping via pairwise BLOSUM62 alignment (gap open −11, gap extend −1). Mouse and human STRC are ~90 % identical in the C-terminal ARM region; positional error ≤ 2 aa.
- TMEM145 interface defined empirically from AF3 Job 2 CIF file at contact cutoff 5 Å. Job 2 ipTM = 0.43 (low confidence), so the interface position is suggestive, not definitive — Derstroff-style GOLD-only pruning job (pending) would give a higher-confidence interface map.
- Prior estimate of “~950–1200” for TMEM145 interface in earlier notes was incorrect and has been superseded by this analysis.
Remaining computational steps
- AF3: mini-STRC × TMEM145 GOLD-only — Derstroff-style pruning for cleaner ipTM (predicted >0.7); confirm interface location is 1600–1770.
- AF3: mini-STRC ΔE1 solo — does deletion destabilize the ARM-repeat fold? Tests the allosteric hypothesis. pTM drop from 0.86 = destabilization signal.
- AF3: mini-STRC ΔE1 × TMEM145 — does deletion drop ipTM vs baseline mini-STRC × TMEM145? Tests the foldability/allosteric hypothesis.
- AF3: mini-STRC (700–1775) × TMEM145 full — tighter version of Job 2 (which used 594–1775). Replaces the old data for the revised construct.
Job JSONs ready at ~/STRC/models/af3_jobs_2026-04-20/.
Files
- Script (mapping):
/tmp/strc_exon_mapping.py - Script (local analyses):
/tmp/local_analyses.py - Script (interface extraction):
/tmp/extract_interface.py - Mapping result:
~/STRC/models/strc_exon_protein_mapping.json - Conservation + NMD + motif scan:
~/STRC/models/strc_e1_conservation_nmd_motifs.json - TMEM145 interface from CIF:
~/STRC/models/strc_tmem145_interface_from_cif.json - Source splicing data:
~/STRC/models/rbm24_sd03_splicing_analysis.json - Source AF3 CIF:
~/Sites/site-strc-egor-lol/public/models/job2-mini-complex.cif - Pending AF3 jobs:
~/STRC/models/af3_jobs_2026-04-20/
Connections
[part-of]STRC RBM24 Regulatory Hypothesis — this refines the open question “do RBM24-regulated exons fall in the therapeutic window”[supports]STRC Mini-STRC Single-Vector Hypothesis — confirms 700–1775 covers all RBM24-regulated segments[contradicts]STRC TMEM145 GOLD Domain Interaction — earlier claim that E1 overlaps TMEM145 interface is not supported; actual interface is aa 1603–1770[see-also]STRC RBM24 Exon Splicing Quantification — upstream quantitative data[see-also]STRC AlphaFold3 Computational Experiments — source for Job 2 CIF file[see-also]STRC ASO Exon Skipping — ASO targeting E1 is the translational handle, independent of TMEM145 story[source]2026-04-17-sun-rbm24-strc-splicing — SD03 rMATS dataset[source]2026-04-17-derstroff-tmem145-ohc-stereocilia — TMEM145 interaction paper (C-terminal ARM repeats)[about]Jeffrey Holt — both source papers share Holt as editor / co-author[about]Misha