STRC Research

STRC ASO Phase2 STRCP1 Paralog Cross-Hybridization

8 min read

STRC ASO Phase 2 — STRCP1 paralog catches every Phase-1 splice-switch ASO

Question

Phase 1 (STRC ASO Phase1 Splice-Switch Design) generated 54 splice-switch ASOs across 4 high-priority cryptic exonic events (E1-E4) for the STRC pre-mRNA. Each was scored on Tm, GC%, hairpin, and self-fold (STRC ASO Phase1 Fold Check) — but none were scored against the STRCP1 paralog. The same paralog that just killed every PE pegRNA (STRC PE Phase4 STRCP1 Paralog Off-Target) shares ≥97 % sequence identity with STRC, including most exonic regions and splice sites. Phase 2 asks: how many of the 54 Phase-1 ASOs survive a STRCP1 cross-hybridization check?

Gate (per ASO):

  • 0 STRCP1 hits at ≤2 mismatches AND
  • ≥1 STRC hit at 0 mismatches

Method

aso_phase2_strcp1_specificity.py — pure-numpy sliding-window Hamming-distance scan against hg38 chr15 (~/STRC/genomes/hg38_chr15.fa, the same single-chromosome FASTA used by STRC PE Phase4 STRCP1 Paralog Off-Target). Runs in 96 s for 54 candidates × 2 strands × 102 Mb genome.

For each candidate, the target_sense sequence (the pre-mRNA region the ASO hybridises to) is searched on both strands of chr15 with ≤2-mismatch tolerance. Hits are binned by region:

Region labelhg38 windowmeaning
STRCchr15:43,580,000-43,640,000on-target locus
STRCP1chr15:43,680,000-43,720,000paralog (silenced pseudogene)
other_chr15rest of chr15random off-targets

(Region windows derived from the on-target chr15:43,600,548 + STRCP1 cluster chr15:43,700,346 observed in STRC PE Phase4 STRCP1 Paralog Off-Target.)

Whole-genome BLAST is not done in this Phase. STRCP1 is the dominant paralog risk because of the 97 % identity and same-chromosome proximity; the rest of hg38 is the next gate after fixing STRCP1.

Result — every ASO hits STRCP1, most perfectly

0 / 54 candidates pass the gate. All 54 fail on STRCP1 cross-hybridization.

STRCP1 best hitn candidates
0 mismatches (perfect)44 / 54
1 mismatch only10 / 54
2 mismatches only0 / 54

Per event:

Eventn candidatesn PASSn with STRCP1 0-mmn with STRCP1 1-mm
E1 (acceptor + donor)120120
E2 (acceptor + donor)14086
E3 (acceptor + donor)140122
E4 (acceptor + donor)140122

E2 has the only candidates that don’t hit STRCP1 with 0-mm — but they all hit at 1-mm, well within RNase-H1 cleavage tolerance (RNase-H1 tolerates 1-2 internal mismatches with 50-80 % cleavage efficiency, see Walder lab). Even the “least bad” candidates would cleave both transcripts.

The on-target STRC binding is intact in all 54 candidates (column “STRC[0]=1” appears for every row in the JSON).

Why this is worse than PE

PE Phase 4 also killed every candidate on STRCP1 — but the PE fix is straightforward in principle (STRC PE Phase4 STRCP1 Paralog Off-Target § Implications): redesign pegRNAs to include STRC-vs-STRCP1 discriminating bases in the seed region. The SpRY-NRN PAM permits enough design space to find ≥2 mismatch sites against STRCP1 in the seed.

For ASOs targeting splice sites, the redesign space is much narrower:

  1. Splice sites are the single most conserved sequence class in the genome. The 3’ acceptor (y(n)NCAG|G) and 5’ donor ((M)AG|GTRAGT) are invariant across all paralogs — the spliceosome doesn’t recognise variants. STRC and STRCP1 have identical acceptor and donor consensus motifs at every retained splice junction.
  2. The exon-internal 18-22 nt windows we scored carry the same identity bottleneck. ASOs targeting the exonic portion of the splice junction sit in regions where STRC and STRCP1 share 97-100 % identity over the relevant ~20 nt windows.
  3. Splice modulators have to be near the splice site to work (≤30 nt from the junction). Pushing the ASO 50-100 nt deep into the intron loses the splice-modulation activity — at that point you might as well design a gapmer for steady-state knockdown, which is a different hypothesis.

Net: ASO splice-switching for STRC is fundamentally compromised by the STRCP1 paralog. Unlike PE, there is no obvious “redesign with discriminating bases” path that preserves the splice-modulation mechanism.

Possible escape paths (pending follow-up)

None of these are free; each takes 1-2 weeks of design + computation work to evaluate:

  1. Verify STRCP1 expression in OHCs. If STRCP1 is transcriptionally silent in cochlear hair cells (the only target tissue), then the off-target ASO has no substrate — paralog risk becomes regulatory paperwork, not biology. Need: human OHC RNA-seq from a published atlas (Zheng-Quayle 2022 or Kolla 2020). If STRCP1 reads = 0 in OHC, downgrade severity from FAIL to YELLOW.
  2. Switch chemistry to morpholino or 2’-MOE blocker (no RNase-H). Steric-block ASOs bind both transcripts but don’t cleave. STRCP1 is a pseudogene → blocking its splicing has no protein-level consequence (it makes no protein anyway). Cost: morpholinos cost ~10× more per dose than gapmers and are harder to deliver to inner ear.
  3. Check 3’UTR exonic targets for STRCP1 divergence. Pseudogene 3’UTRs often diverge faster than coding regions. If STRC 3’UTR has 5+ % divergence from STRCP1, an ASO targeting 3’UTR could discriminate. But 3’UTR ASOs target steady-state knockdown, not splice modulation — switches the hypothesis from “rescue defective splicing” to “knockdown WT to clear room for AAV-delivered Mini-STRC”. That’s a different drug.
  4. Whole-genome BLAST + Tm-discriminating ASO design. Even at 1-mm, the Tm penalty against STRCP1 is ~5-7 °C. A carefully designed ASO at the upper Tm range may hybridise STRC fully and STRCP1 only marginally. Cost: requires re-design at higher Tm windows (24-26 nt) and biophysical validation.
  5. Allele-specific ASO targeting E1659A directly. Skip splice modulation entirely; design 18-22 nt gapmer around the c.4976A>C variant where STRC has C and STRCP1 has the WT genomic A. The 1-bp variant would give 1-2 °C ΔTm — borderline for RNase-H discrimination but possible in principle. This is essentially what PE does mechanically but with knockdown instead of editing.

Implications for the hypothesis

STRC ASO Exon Skipping: tier B → C.

  • Mechanism (5/5 → 4/5): RNase-H gapmer / 2’-MOE splice-switching chemistry is established (nusinersen, eteplirsen). What works for SMN2 and DMD also works mechanistically for STRC. -1 because the paralog contamination affects every Phase-1 design.
  • Misha-fit (3/5): unchanged. ASO works on the maternal E1659A allele’s mis-spliced transcript, which Misha has.
  • Delivery (1/5 → 1/5): unchanged. Inner-ear ASO delivery (intratympanic) is plausible-but-unbuilt for this indication. Same as before.
  • Evidence depth: +1 (Phase 2 STRCP1 cross-hybridization confirmed lethal across all 4 high-priority events).
  • Status: still active, but the candidate pool is zero until one of the 5 escape paths is evaluated.
  • Next step changed from “extend fold-check to all 54 candidates and triage by ΔG” → “verify STRCP1 OHC expression first (escape path 1, cheapest); if STRCP1 expressed, evaluate path 2 (morpholino) and path 5 (allele-specific gapmer at c.4976) as parallel sub-hypotheses”.

The downgrade B → C reflects the same severity-class jump as PE just took (A → B). Both hypotheses are now blocked on the same biological question: does Misha’s cochlea express STRCP1? If no, both unblock partially. If yes, both need substantial redesign at a higher phase.

Files

  • Driver: ~/STRC/models/aso_phase2_strcp1_specificity.py
  • Output JSON: ~/STRC/models/aso_phase2_strcp1_specificity.json
  • Phase 1 input: ~/STRC/models/aso_phase1_design.json (54 candidates)
  • Genome: ~/STRC/genomes/hg38_chr15.fa (shared with PE Phase 4)

Limitations

  • chr15-only scan. Distal off-targets on other chromosomes not assessed. STRCP1 is the dominant paralog so the per-candidate verdict is robust, but the 16 candidates that hit STRCP1 with 1-mm could in principle have additional off-target loci elsewhere — would need full-genome BLAST.
  • No structural / accessibility weighting. Pre-mRNA secondary structure could in principle render a target site inaccessible in STRCP1 specifically, sparing it. Requires NUPACK or RNAfold of the full STRCP1 transcript ± ASO docking — substantial extra computation.
  • Hamming distance only. No bulges, no insertions, no chemistry-specific affinity scoring. Real RNase-H1 cleavage tolerates 1-2 internal mismatches with reduced (50-80 %) but non-zero efficiency.
  • STRCP1 transcript model assumed. We assume STRCP1 mRNA exists with the same exonic boundaries as STRC. If STRCP1 splices differently (or doesn’t splice at all due to mutated splice sites), the per-event ASO hits may be irrelevant. RNA-seq splice junction analysis would resolve this.
  • No OHC expression filter (this is the central limitation — see escape path 1).

Ranking delta

  • STRC ASO Exon Skipping: Tier B → C. Mechanism 5/5 → 4/5, Misha-fit 3/5 unchanged, Delivery 1/5 unchanged. Evidence depth +1. Status: active, but candidate pool = zero pending STRCP1 OHC expression check + chemistry pivot. Next step changed from “extend fold-check + triage by ΔG” → “verify STRCP1 OHC expression (escape path 1, cheapest); evaluate morpholino (path 2) and allele-specific gapmer at c.4976 (path 5) as parallel sub-hypotheses if STRCP1 expressed”. Mirrors the PE A→B downgrade for the same biological reason.
  • Prime Editing for STRC: no further change beyond STRC PE Phase4 STRCP1 Paralog Off-Target’s A → B. The two hypotheses now share a common gate (“STRCP1 OHC expression check”) — clearing it unblocks both partially.
  • STRC Mini-STRC Single-Vector Hypothesis: no change. AAV protein delivery is paralog-orthogonal.
  • STRC mRNA-LNP Strategy B Full-Length: no change. Exogenous mRNA delivery doesn’t depend on endogenous transcript discrimination.
  • STRC Pharmacochaperone Virtual Screen E1659A: no change.
  • All other S/A/B/C tier hypotheses: no change.

New cross-cutting next step: pull human OHC RNA-seq (Zheng-Quayle 2022 or Kolla 2020 atlases) and quantify STRCP1 expression. Result resolves the gate for both PE and ASO simultaneously.

Connections

Graph View

Backlinks