Skip to content

Uricase Protease Stability in Shio-Koji — Computational Analysis

Question: Will A. flavus uricase (Q00511) retain meaningful activity after 7–14 days in a shio-koji ferment (15–20% NaCl, pH 4.5–5.0, ~22°C)?

Verdict: LOW risk (provisional). Computational analysis finds no exposed protease recognition sites in the uricase structure, combined with strong salt-mediated suppression of the dominant koji protease. Three assumptions in this analysis compound toward safety (pLDDT as burial proxy, point-estimate salt values, monomer structure); a SASA-based reanalysis on the D2 tetrameric assembly would remove "provisional." Wet-lab confirmation (validation-experiments.md §1.10) remains necessary but this shifts §1.10 from a feasibility gate to a confirmation experiment.

Reproducible artifact: experiments/comp-001-uricase-shio-koji-protease-stability/ — script, inputs, and full outputs committed to the repo.

Why this matters

The shio-koji format is strategically important: it is the only koji product format that unifies home-fermentation simplicity, long shelf life (~6 months refrigerated), and palatability (used as a marinade, seasoning, condiment). If uricase survives in this format, the platform's gut-lumen-sink intervention can be delivered as a condiment rather than a capsule or powder — a meaningful accessibility advantage.

Shio-koji contains active A. oryzae proteases, which are the same enzymes the organism uses to break down protein in food fermentation. The concern: those same proteases might degrade the engineered uricase payload before it's consumed. This analysis evaluates whether that concern is structurally supported.

Method summary

Three independent protective factors were evaluated:

  1. Structural burial — AlphaFold v6 per-residue confidence (pLDDT) as a proxy for fold quality and burial. Residues with pLDDT ≥ 80 are in well-folded regions unlikely to be protease-accessible; < 65 are disordered/exposed.

  2. Recognition site mapping — P1/P1' cleavage-site prediction for each of the three major A. oryzae koji proteases (ALP/alkaline subtilisin, NPr/neutral metalloprotease, acid protease/aspergillopepsin), using specificity rules from MEROPS.

  3. Shio-koji condition corrections — Salt inhibition (17.5% NaCl midpoint) and pH activity factors applied per protease, based on primary literature characterization of each enzyme.

Full methodology and limitations in experiments/comp-001-uricase-shio-koji-protease-stability/analyze.py.

Key results

Structural confidence is exceptional

Metric Value
Sequence length 302 aa
Mean pLDDT (AlphaFold confidence) 97.1 / 100
Minimum pLDDT (most flexible residue) 80.5 / 100
% residues pLDDT > 80 (well-folded) 100%
% residues pLDDT > 90 (core-folded) 97%

A mean pLDDT of 97.1 is exceptional even among well-characterized enzymes. Crucially, the minimum is 80.5 — there are no disordered loops, no flexible termini, no exposed unstructured regions. Proteases typically initiate cleavage at disordered regions and N/C-terminal tails; uricase offers none of these entry points.

Every recognition site is buried

Protease Recognition sites Exposed Partially exposed Buried Effective activity at shio-koji conditions Max risk score (0–1)
ALP (alkaline subtilisin) 215 0 0 215 18.8% (salt-suppressed) 0.019
NPr (neutral metalloprotease) 97 0 0 97 38.8% (moderately salt-suppressed) 0.039
Acid protease (aspergillopepsin) 44 0 0 44 19.5% (salt + sub-optimal pH) 0.020

All 356 recognition sites across all three proteases map to buried regions. Risk scores well below the 0.15 LOW threshold.

Two independent protective mechanisms

The low-risk verdict rests on two orthogonal factors, either of which alone would be partially protective:

Factor 1 — Structural compactness. Zero exposed cleavage sites means there is no structural vulnerability even if the proteases were fully active. This is the primary protective factor.

Factor 2 — Salt inhibition of ALP. The dominant koji protease (ALP, alkaline subtilisin) retains only ~19% of its normal activity at 17.5% NaCl. This is why salted ferments preserve protein foods — the salt partially disables the same proteases we're worried about.

The acid protease is the least salt-inhibited (retains ~65% activity relative to unsalted) and operates closest to shio-koji's pH range — it is the residual risk case. But all its recognition sites are in buried regions, limiting practical risk.

Factor 3 (not modelled) — Tetramer burial. Uricase functions as a homotetramer in vivo. This analysis modelled the monomer only. The tetramer interface buries additional surface area beyond the monomer fold — the real-world structure is more protease-resistant than this analysis captures. The monomer analysis is therefore conservative.

Limitations

  • pLDDT ≠ solvent-accessibility. Some high-pLDDT surface loops may be accessible in reality — pLDDT measures local structural confidence, not burial. High-pLDDT surface helices (scored 85–90) can still be fully solvent-exposed. A SASA calculation on the AlphaFold PDB is the correct tool and would sharpen the analysis materially. See comp-002 (planned) for this follow-up if §1.10 wet-lab results are unexpected.
  • P1/P1' specificity only. Real proteases use extended subsites (P2–P4, P2'–P4') that are not modelled here. This may over-count recognition sites (making the analysis conservative — more hits than a full subsite model would predict). ALP's 215 sites against a 302 aa sequence reflects its genuinely broad P1 specificity, rendering the raw site count uninformative; SASA-filtered exposed-site count would be more meaningful.
  • Monomer structure only. Uricase is a D2-symmetric homotetramer. See Factor 3 above. Beyond mere "conservatism," the active-site tunnel is located at subunit interfaces — residues that appear surface-exposed in the monomer analysis are buried at tetramer interfaces and not accessible to protease. The correct analysis should use the biological assembly PDB (D2 tetramer), not the monomer chain.
  • NPr salt inhibition value has higher uncertainty than stated. The 39% residual at 17.5% NaCl is a point estimate; thermolysin-family metalloproteases show variable NaCl sensitivity depending on Ca²⁺ availability (NaCl can competitively displace Ca²⁺ coordination). The true value may be lower (20–30% residual), which would reduce NPr's max risk score below the 0.039 reported. Treat as ±0.15 uncertainty range.
  • Secreted protease concentration not quantified. In vitro inhibition values describe specific activity, not [E]. A. oryzae may secrete substantially less protease under 15–20% NaCl osmotic stress. If secretion is reduced, effective enzyme exposure is lower than the model captures. This is an additional suppressive factor not in the risk formula.
  • Temperature suppression not modelled. All three proteases have activity optima at 37–50°C; the shio-koji fermentation temperature (~22°C) suppresses activity by an additional 2–5× relative to in vitro reference conditions at optimum. This is a conservative omission — real-world risk is lower than the model states.
  • pH range treated as a point value. pH 4.5 vs. 5.0 represents approximately a 2-fold difference in aspergillopepsin I activity. The pH_factor of 0.30 is a midpoint estimate; at pH 5.0 the value may be closer to 0.15–0.20. The range [0.15, 0.30] is more accurate than the single value.
  • No fermentation dynamics. Proteases are most active before the salt environment is established (during initial koji growth). Shio-koji by definition starts after koji is harvested — the peak-activity window has already passed. This analysis models the shio-koji-phase conditions, not the pre-salt koji phase.
  • pH unfolding not modelled. At pH 4.5–5.0, some proteins partially unfold, exposing buried residues. AlphaFold models standard folding conditions; acid-induced partial unfolding is not captured. If §1.10 shows unexpected degradation, cooperative unfolding at pH 4.5 is the first mechanism to investigate.

Impact on experimental priorities

Before this analysis: validation-experiments.md §1.10 was positioned as a feasibility gate — "does this format even work?" with the outcome potentially disqualifying shio-koji entirely.

After this analysis: §1.10 is a confirmation experiment — "we expect this to work; the experiment confirms and quantifies it." This changes the risk framing for the platform's shio-koji delivery thesis from "unknown" to "structurally supported, pending empirical confirmation."

The planned comp-002 (thermal/pH stability MD simulation) is low priority unless §1.10 produces unexpected degradation results.

Cross-references