Skip to content

C1-INH (SERPING1) Protease Stability + Glycosylation Feasibility in EcN LBP: Computational Analysis (comp-037)

Frozen analysis archived to ./etc/experiments/comp-037-c1-inh-protease-stability-ecn/ (full pipeline: analyze.py + inputs/ + outputs/ + README.md). This wiki page is the interpretive summary; the long content + reproducible code lives next to the experiment.

Status: Complete — 2026-05-17 Experiment folder: etc/experiments/comp-037-c1-inh-protease-stability-ecn/ Evidence level: Mechanistic Extrapolation — AlphaFold pLDDT-based structural inference; no wet-lab confirmation.

Sister analyses: comp-006 (DAF/CD55 ectodomain, koji, HIGH) · comp-012 (DAF/CD55 SCR1-4 truncated, koji, LOW) · comp-024 (complestatin BGC on LBP, RED; C1-INH-on-EcN comparator GREEN-provisional 0.774)

Question

Would human C1-INH (SERPING1, UniProt P05155) survive the colonic-luminal protease environment if expressed as a secreted LBP-luminal payload in engineered E. coli Nissle 1917 (EcN)? And does the heavy native glycosylation (~26 kDa glycan on ~52 kDa polypeptide) — which EcN cannot reproduce — preclude functional inhibitor activity?

This is the computational gate that decides whether C1-INH joins DAF SCR1-4 as the second CP0 engineering candidate, completing the two-chassis architecture surfaced 2026-05-16 in complement-c5a-gout.md §9.8: C1-INH at classical/lectin entry (LBP chassis) + DAF SCR1-4 at surface convertase decay (koji chassis). See engineered-lbp-chassis.md for the chassis peer track and comp-024 for the upstream feasibility prior that flagged C1-INH as the right next computational gate.

Headline verdict

MODERATE — kinetic-competition gated.

Three independent verdicts decompose to:

Scope / axis Verdict Driver
Serpin core, strictly-degradative (non-RCL, aa 123-451 + 468-500) LOW (0.1) Once mucin domain (aa 23-119) truncated, serpin body is protease-resistant; only buried OmpT di-basic sites remain
Serpin core including RCL (aa 123-500) RED (0.8) DegP and elastase have 9 + 11 exposed sites in the RCL (aa 452-467) — but RCL exposure is by design (serpin suicide-substrate mechanism)
Glycosylation feasibility, serpin core GREEN EcN cannot N-glycosylate; for luminal-secreted topology plasma half-life concern is moot; serpin catalytic mechanism is encoded in polypeptide, not glycan

The combined verdict is MODERATE, not RED, because:

  1. Strictly-degradative protease risk on the folded serpin body is LOW after mucin truncation.
  2. Glycosylation feasibility is GREEN for luminal topology.
  3. The remaining risk lives in the RCL kinetic competition — does C1s / C1r / MASP-2 engage the reactive bond R466-T467 productively before DegP / elastase cleaves it unproductively? That is a wet-lab kinetic assay question, not an in silico question.

A naive read of the "RED 0.8 serpin core" verdict would mis-attribute the risk axis. The serpin body folds and persists; the question is whether the inhibitor mechanism wins the kinetic race at the RCL.

What the analysis establishes (and what it doesn't)

Establishes:

  • The mucin-like N-terminal domain (aa 23-119) is the dominant protease liability in an unglycosylated EcN expression context. Truncating it eliminates 55 elastase exposed sites, 18 DegP sites, 7 chymotrypsin sites, and 4 trypsin sites. This is directly analogous to the CD55 stalk truncation in comp-006 → comp-012.
  • The serpin body (aa 123-450, anchored by the two disulfides C123-C428 and C130-C205) is well-folded (AlphaFold pLDDT 80-95) and has no exposed protease sites outside the RCL after mucin truncation.
  • The RCL (aa 452-467) is intrinsically exposed — and must be, for the serpin suicide-substrate mechanism. The "RED 0.8" score reflects this by-design exposure, not degradative vulnerability of the body.
  • N-glycans are not strictly required for inhibitor mechanism in the luminal-secreted format. The catalytic activity (acyl-enzyme covalent trap) is encoded in the polypeptide; the plasma-half-life function of N-glycans is moot for a gut-luminal therapeutic.

Does NOT establish:

  • Whether the truncated serpin-core construct retains full inhibitor activity against C1r, C1s, and MASP-2. Native C1-INH includes the N-terminal mucin-like domain; whether that domain contributes to substrate recognition beyond plasma half-life is not fully resolved in primary literature. Wet-lab kinetic assay against C1r/C1s/MASP-2 substrates is the gate.
  • Whether EcN can fold the serpin-core construct with both disulfides correctly formed. EcN periplasmic DsbA/DsbC can in principle catalyze both C123-C428 and C130-C205 bonds, but serpin-fold quality in a bacterial chassis is a known wet-lab risk class (broader serpin-bacterial-expression literature is mixed).
  • The k_C1s_engagement vs k_DegP_RCL_cleavage kinetic ratio at the RCL. This is the load-bearing wet-lab measurement.
  • Whether luminal C1-INH reaches the CP0 priming sites (submucosal macrophages, MSU-crystal-interface convertases) at functionally relevant concentrations. This is the same H05-class clinical-translation question facing the DAF SCR1-4 koji track.

Engineering implications

If the wet-lab gates clear, the recommended C1-INH-on-EcN construct is:

  • Boundary: aa 123-500 (serpin core, mucin truncated, retains RCL R466-T467).
  • N-terminal anchor: truncation starts at C123 (first canonical disulfide cysteine) to preserve the disulfide-anchored fold.
  • Signal peptide: human aa 1-22 replaced with EcN-native secretion signal (OmpA, PelB, or YebF for extracellular release — choice TBD by expression-system optimization).
  • Disulfide partner support: EcN periplasmic DsbA/DsbC; potentially DsbC-overexpression strain if folding-quality issues emerge.
  • No N-glycosylation system needed — for luminal-format activity, the catalytic mechanism does not strictly require N-glycans. C. jejuni pgl-pathway heterologous transplant (Wacker 2002) is an unnecessary engineering burden for this topology.

Architectural read: the two-chassis CP0 architecture

complement-c5a-gout.md §9.8 (2026-05-16) proposed pairing C1-INH on EcN with DAF SCR1-4 on koji as a two-engineered-payload, two-independent-chassis, two-cascade-node CP0 coverage architecture. This comp-037 computational gate substantiates the C1-INH-on-EcN side:

Property DAF SCR1-4 on koji (comp-012, H05) C1-INH on EcN (comp-037)
Cascade node Surface convertase decay (already-assembled C4b2a, C3bBb) Classical / lectin entry (C1r, C1s, MASP-2) — upstream of convertase formation
Mechanism logic Disassembles convertases that form Prevents convertases from forming
Chassis A. oryzae (koji, food-grade, transit) E. coli Nissle 1917 (LBP, regulated drug, resident)
Distribution Home-fermentable condiment Pharmacy-distributed pharmaceutical
Computational verdict LOW (post-stalk-truncation) MODERATE (kinetic-competition gated)
Wet-lab gate §1.25 validation-experiments.md (queued) RCL kinetic competition assay (to be specified)

Together: two independent mechanisms at two distinct cascade points, delivered via two completely independent chassis, with two completely independent regulatory paths. Neither competes for the other's production infrastructure; neither mechanism is mechanistically redundant.

What this does NOT claim:

  • C1-INH on EcN-LBP has NOT been wet-lab validated for luminal stability or complement-regulatory activity in the gut environment. The MODERATE verdict here is a feasibility prior, not an empirical result.
  • The C1-INH + DAF composition has NOT been wet-lab tested as a combined intervention. The mechanism-level coverage claim is a testable hypothesis, not an asserted synergy.
  • No clinical effect-size prediction. Both arms must validate individually before any composition question can be tested.

Files

Pre-commit verification (CLAUDE.md Rule 4)

Every load-bearing number in this analysis was grep-verified against primary source before commit:

  • Sequence length 500 aa, signal peptide aa 1-22, mature aa 23-500: UniProt P05155 SV=2 (fetch 2026-05-17, direct grep against text-format flatfile).
  • Reactive bond R466-T467: UniProt FT SITE 465..466 ("Reactive bond for chymotrypsin") + SITE 466..467 ("Cleavage; by C1S" + "Reactive bond"); asserted programmatically in analyze.py main() (seq[465]=='R' and seq[466]=='T').
  • Disulfide bonds C123-C428 and C130-C205 — exactly 2 total: UniProt FT DISULFID (exactly two entries; counted with grep -c). Asserted programmatically in analyze.py main() for both Cys positions. Note: casual literature sometimes quotes higher counts; the canonical SV=2 UniProt entry has 2.
  • N-glycan sites 25, 69, 81, 238, 253, 272-variant, 352; O-glycan sites 47, 48, 64, 71, 83, 88, 92, 96: UniProt FT CARBOHYD direct grep; N-sequon (N-X-S/T) verified by direct sequence inspection for each N-site (see inputs/provenance.md).
  • ~26 kDa glycan on ~52 kDa polypeptide: Bos 1998 PMID 9799502 + Stavenhagen 2018 PMID 29381136 (cited PMID-anchored in provenance.md).
  • DegP P1 specificity V/I/L/F/Y/A: Krojer 2008 PMID 18261546.
  • OmpT di-basic P1-P1' specificity: Dekker 2001 PMID 11226160; Hwang 2007 PMID 17263510.
  • Colonic pH 6-7: Fallingborg 1999 PMID 10204470.
  • Trypsin / chymotrypsin / elastase MEROPS IDs (S01.151, S01.001, S01.153) and P1 specificities: MEROPS release 12.4 + Schechter & Berger 1967 + Hedstrom 2002 PMID 12475195 + Largman 1976 PMID 988044.

The DAF SCR1-4 disulfide-count incident (2026-05-06, documented in CLAUDE.md Rule 4) is the canonical case this verification gate exists to prevent. C1-INH disulfide count was the first place that gate was applied here — 2 bonds, grep-verified, not the inflated counts sometimes appearing in secondary literature.

Limitations

  • Disulfides not modeled by the scoring algorithm. The C123-C428 and C130-C205 bonds reduce backbone flexibility in the serpin core; this analysis likely overestimates risk in the body. C1-INH has only 2 disulfides — far less than CD55 SCR1-4 (8) — so the overestimate is modest.
  • O-glycosylation of the mucin domain not modeled in the native baseline. The scoring treats the mucin domain as the unglycosylated polypeptide that EcN would produce; this is the right model for the EcN-expression context but means the score is NOT comparable to in vivo (heavily glycosylated) protease stability.
  • pLDDT ≠ solvent accessibility. Surface-exposed loops on a well-folded serpin core may have high pLDDT but still be protease-accessible. SASA refinement would tighten the per-residue score.
  • P1/P1' rules only. Extended subsite specificity (P2-P4) not modeled. Trypsin K/R-P bonds resist cleavage; DegP and OmpT have non-trivial P2 preferences. Site counts likely over-estimated for some proteases.
  • Commensal-bacterial protease load not modeled. The five-protease panel covers the dominant identifiable risks (pancreatic + EcN-native); the diffuse colonic-microbiome protease landscape is not enzyme-level characterized for this scope.
  • Bile-acid effects not modeled. Bile acids in the upper colon may partially unfold proteins and enhance protease access.
  • EcN secretion topology not directly modeled. Type I (HlyA-like), Type II (secretome), Sec/YebF (extracellular), or signal-peptide-driven release affects exposure to OmpT and DegP. This analysis pools both risks without conditioning on a specific export pathway.
  • RCL kinetic-competition treatment is in-scope only as a flagged wet-lab question, not in silico modeled. k_C1s_engagement vs. k_DegP_RCL_cleavage on the recombinant construct is the load-bearing wet-lab measurement that decides whether C1-INH is a viable luminal CP0 candidate.

Cross-references