Hypothesis

Autophagy does not merely clear misfolded proteins generated by aberrant splicing; it selectively degrades key splicing regulators (e.g., SRSF1, SRSF2) to re‑balance the spliceosome and keep cryptic splicing below a toxic threshold. This creates a tunable buffering system: as mutant splicing factor burden rises, autophagy is up‑regulated and preferentially consumes SR proteins, dampening their activity and reducing the production of deleterious splice isoforms. When the autophagic capacity is saturated, SR protein levels rebound, splicing fidelity collapses, and the cell shifts from adaptive survival to lethal proteotoxic stress.

Mechanistic Rationale

1. Selective cargo recognition – Recent work shows that autophagy can be substrate‑specific under stress (1). We propose that p62/SQSTM1 binds phosphorylated SR proteins via their RS domains, targeting them for autophagic clearance when spliceosome‑derived misfolded proteins accumulate.
2. Feedback on splicing – Depletion of SR proteins reduces exon inclusion of cryptic splice sites, directly lowering the load of toxic isoforms that drive autophagy induction (2, 3).
3. Threshold behavior – The heterozygous advantage of SF3B1 mutations in MDS (4) reflects a sweet spot where autophagy‑mediated SR protein turnover sufficiently buffers splicing noise. Beyond this point, autophagy flux plateaus (measured by LC3B‑II turnover) while SR protein synthesis outpaces degradation, leading to a sudden rise in cryptic splicing and p62 accumulation.

Testable Predictions

• Prediction 1: In cells expressing graded levels of mutant SF3B1, pharmacological activation of autophagy (e.g., rapamycin) will decrease SR protein abundance and cryptic splicing events, whereas autophagy inhibition (e.g., BafA1 or ATG5 KO) will increase them, but only up to a mutant burden threshold.
• Prediction 2: Measuring autophagic flux (LC3B‑II turnover) alongside SR protein levels will reveal an inverse correlation that breaks down at high mutant SF3B1 expression, coinciding with a spike in intron retention and loss of viability.
• Prediction 3: Rescue of SR protein levels via expression of autophagy‑resistant mutants (e.g., lysine‑to‑arginine RS domain mutants) will abolish the protective effect of autophagy activation, shifting the toxicity curve leftward.

Experimental Design

• Generate isogenic human cell lines (e.g., RPE‑1) with inducible SF3B1^K700E at low, medium, high doxycycline doses.
• Treat with rapamycin, Torin1, or vehicle; parallel sets with ATG5 CRISPR knockout or BafA1.
• Quantify: (a) LC3B‑II/LC3B‑I and p62 by Western blot; (b) SR protein (SRSF1, SRSF2) levels; (c) cryptic splicing via RNA‑seq (percent spliced‑in of intronic cryptic exons); (d) cell viability/apoptosis.
• Perform epistasis: overexpress autophagy‑resistant SRSF1 (RS domain mutant) in WT background to see if it mimics high burden phenotype.
• Use mathematical modeling to fit a sigmoidal dose‑response curve; the inflection point predicts the autophagy capacity threshold.

Potential Outcomes and Falsifiability

• If autophagy manipulation shifts the viability curve as predicted, the hypothesis is supported.
• If altering autophagy flux does not change SR protein levels or cryptic splicing across any mutant burden, or if autophagy‑resistant SR proteins fail to modify the toxicity threshold, the hypothesis is falsified.
• Additionally, observing that autophagy inhibition exacerbates toxicity only at low mutant burden (contrary to prediction) would refute the selective cargo model.

This framework extends the ‘siege’ analogy by positing that autophagy is not just a generic rationing system but a specific, adjustable valve that controls the spliceosome’s own regulators, defining the point at which cellular cannibalism can no longer keep pace with splicing‑derived entropy.