Skip to content

Gut-Lumen Uricase × ABCG2 Genotype Stratification + Flux Model (comp-019)

Plain-English summary (read this first)

There was a worry that the engineered-koji platform's whole therapeutic mechanism might only work in patients who carry a specific genetic variant — Q141K in the ABCG2 transporter — which is the #1 genetic risk factor for gout. About 25% of European-ancestry gout patients carry this variant; up to 50% of East-Asian-ancestry patients do. If the platform mechanism only worked in this subset, the addressable demographic would shrink dramatically and the trial design would have to change.

This experiment built a quantitative model of intestinal urate flux to answer that question in silico, before committing wet-lab dollars. The model uses the only direct human in-vivo measurement of intestinal urate secretion stratified by ABCG2 functional class (Miyazaki et al., J Transl Med 2025) plus standard kinetic parameters for the ABCG2 transporter and the A. flavus uricase enzyme.

The answer: the worry is wrong. Non-Q141K (WT/WT) males show the LARGEST predicted serum urate reduction from gut-lumen uricase, not the smallest. The mechanism works across all genotypes; the magnitude scales with the residual ABCG2 capacity at any given genotype.

Predicted ΔSUA at steady state (with mid-dose 25 mg/day uricase, the engineered-koji target):

  • WT/WT male: −0.83 mg/dL (90% CI −1.13 to −0.57)
  • Q141K heterozygous male: −0.67 mg/dL (90% CI −0.91 to −0.45)
  • Q141K homozygous male: −0.50 mg/dL (90% CI −0.68 to −0.34)
  • Severe ABCG2 dysfunction (Q126*+Q141K): −0.28 mg/dL (90% CI −0.37 to −0.19)

What this means for the platform: 1. Don't narrow the primary demographic to Q141K-positive patients. The opposite — the platform produces its largest per-patient ΔSUA in WT/WT patients, the majority of the gout population. 2. Q141K-positive patients are still candidates, but for a different reason: they have higher unmet ULT need at baseline (allopurinol-resistant per Wallace 2018 OR=2.43, p=6.2e-7), so even a smaller absolute ΔSUA is clinically meaningful for them. 3. The Phase 2b trial design should NOT be enriched for Q141K-positive patients. It should be a typical-gout cohort with Q141K as a stratification variable, not an enrichment criterion. 4. The §1.27 Caco-2 wet-lab gate is NOT triggered by this finding. The flux model already rules out the Q141K-only-mechanism hypothesis without needing wet-lab. (The wet-lab gate would re-trigger if a Phase 2b RCT shows differential genotype response inconsistent with the flux model's predictions.) 5. There IS a structural ceiling, but it sits at the severe-ABCG2-dysfunction tier (~25% functional, Q126*-positive compound heterozygotes), not at Q141K heterozygotes. Most clinically relevant Q141K carriers are heterozygotes, who retain 75% function and respond well per the flux model.

The Phase A literature mining ALSO surfaced a separate negative finding worth flagging: across the entire published clinical-trial corpus of oral and systemic uricase therapy, ZERO trials have pre-stratified or post-stratified results by ABCG2 Q141K genotype. The Q141K × allopurinol response literature is rich, but the Q141K × uricase response literature is empty. This is a publishable observation by itself — every uricase trial since rasburicase (2001) has missed an obvious pharmacogenomic stratification axis.

Question

Can the gut-lumen uricase sink produce meaningful serum urate reduction in non-Q141K males, or does the mechanism rely on Q141K-positive disease-state ABCG2 vulnerability to show benefit?

This is the platform's most important open question for its primary demographic (open-questions.md Open Question 1, framed in the synthesis.md Sweep 2026-05-08 action queue).

The question emerged after comp-016 found a WEAK / UNCONFIRMED verdict on direct androgen-driven intestinal ABCG2 suppression, and comp-017 confirmed near-null sex dimorphism in healthy-baseline intestinal ABCG2. With the "androgens suppress ABCG2 in males" model refuted, the responder logic provisionally shifted toward Q141K-positive disease-state vulnerability. comp-019 tests whether that provisional shift was correct.

Verdict

The gut-lumen uricase mechanism does NOT depend on Q141K-positive disease-state vulnerability. Across the genotype spectrum:

  • WT/WT (100% functional ABCG2) → LARGEST absolute ΔSUA per patient
  • Q141K heterozygous (75% functional) → ~80% of the WT response
  • Q141K homozygous (50% functional) → ~60% of the WT response
  • Severe dysfunction (~25% functional, Q126*+Q141K compound) → ~33% of the WT response — this IS a structural ceiling, but it applies only to the small severe-dysfunction subset, NOT to the typical Q141K heterozygote.

Implication: the platform's primary demographic should remain "all gout patients," not "Q141K-positive gout patients." Q141K is a stratification variable for trial design, not an enrichment criterion. The non-Q141K majority of gout patients are the platform's HIGHEST-response subset, not its lowest.

Why this matters

Three platform-level consequences:

  1. Phase 2b RCT design. Pre-comp-019 worry framing implied a small, expensive Q141K-enriched trial (~120 patients in an East-Asian-ancestry cohort to hit ~50% Q141K carrier rate, or ~360 in a European-ancestry cohort to hit the same). Post-comp-019, the trial should run as a typical-gout RCT with Q141K + Q126* genotyping at enrollment as stratification variables. This is logistically simpler and addresses a larger market.

  2. Strain-engineering target dose. The flux model shows the mechanism is substrate-limited, not enzyme-limited, at all three uricase-dose scenarios tested (5 / 25 / 50 mg/day). Capacity ratios range from 32× to 1300× — uricase capacity vastly exceeds delivered intestinal urate flux. This means the engineered-koji yield target (engineered-yeast-uricase-proposal.md) does NOT need to push the upper bound; the binding constraint is enzyme delivery (GI survival, enteric coating per gi-survival-prediction.md), not enzyme dose. Engineering cycles can shift away from yield optimization toward GI-survival optimization.

  3. Cross-validation.md Claim 1 rating. Pre-comp-019, the gut-lumen uricase mechanism was rated 6/10 with "human effect size unproven" as the primary risk. Post-comp-019, the human effect size is now COMPUTATIONALLY PREDICTED at −0.5 to −0.83 mg/dL across the dominant gout subpopulations. This is below the 1–2 mg/dL therapeutic-target range cited in cross-validation.md, but it IS in the same order of magnitude as the ALLN-346 Phase 2a Study 201 signal (CKD subgroup) and meaningful for adjuvant use alongside allopurinol. The rating should shift from "6/10 with human effect-size unproven" to "6.5/10 — mechanism is genotype-robust per flux model; clinical magnitude prediction in the −0.5 to −1.0 mg/dL range awaits human RCT verification."

Method

Phase A — literature stratification mining

Scope: - ALLN-346 Phase 2a Studies 201 (n=16, CKD-dominated) and 202 (n=19, broader cohort, terminated Sep 2022) - PRX-115 Phase 1 (n=64) — IV systemic uricase, NOT gut-lumen but a useful control for the upper bound of enzymatic urate degradation - Rasburicase clinical corpus (TLS-dominated; gout off-label rare) - Q141K × allopurinol response: Wallace 2018 meta-analysis (n=595, OR 2.43, p=6.2e-7); Vora 2021 oxypurinol PK/PD (Q141K homozygotes have 1.79× longer half-life); Stamp 2019 (Q141K independent factor in allopurinol sensitivity, p=0.019) - Direct human intestinal urate secretion measurement: Miyazaki 2025 (n=34, double-balloon endoscopy, ABCG2 functional-class stratified) - Mouse Abcg2-KO: Takada 2014 (intestinal urate excretion <50% of WT) - ABCG2 transporter kinetics: Nakayama 2011 (Km = 8.24 ± 1.44 mM; Vmax = 6.96 ± 0.89 nmol/min/mg) - Functional-classification framework: Matsuo 2014 (Japanese male hyperuricemic cohort n=644) - Q141K allele frequencies: gnomAD (East Asian MAF 29.1%; European MAF 10.3%); Nguyen 2025 Vietnamese gout cohort (MAF 46.8%)

Tooling: PubMed MCP search_articles + get_article_metadata + get_full_text_article; WebSearch + WebFetch for non-PubMed-indexed content (EULAR abstracts, Allena/Protalix press releases). Multilingual ChiCTR / J-STAGE / KISS searches did not surface a Q141K-stratified oral uricase RCT (none exists in any language).

Phase B — first-principles flux model

Implementation: Python stdlib only (experiments/comp-019-gut-lumen-uricase-abcg2-genotype-stratification/scripts/flux_model.py). Inputs in inputs/flux_model_parameters.json (literature-anchored; each parameter cited to a primary source with verification tier). Monte Carlo sensitivity n=5000 over uncertainty bounds.

Model logic:

  1. Daily mass balance. Total daily urate production ~700 mg/day; renal excretion ~67%, intestinal ~33% in healthy WT (Takada 2014 mouse KO; multiple historical tracer studies).

  2. Linear-regime ABCG2 transport. ABCG2 Km is 8.24 mM (= 8240 μM) per Nakayama 2011. Direct measurement of human jejunal lumen urate baseline is ~99.5 pg/μL ≈ 0.59 μM (Miyazaki 2025) — three to four orders of magnitude below Km. The transporter is FAR from saturation in vivo. Rate scales linearly with substrate concentration. Reducing lumen urate via uricase increases the gradient roughly proportionally.

  3. Genotype scaling. Functional class per Matsuo 2014 / Miyazaki 2025: WT/WT = 1.00; Q141K het = 0.75; Q141K hom = 0.50; Q126*+Q141K compound = 0.25. Empirically anchored to Miyazaki 2025 secretion-rate ratios (median ratios 100%:75%:50% = 1.0 : 0.72 : 0.38).

  4. Sink amplification factor. With baseline lumen urate already low (~0.6 μM), uricase clamping it toward zero produces a bounded gradient improvement. Per Yu 2015 / Bhatnagar 2016 review tier and physiological reasoning, ~30–50% of secreted intestinal urate is reabsorbed in the colon. We use 0.40 (sensitivity 0.20–0.60) — i.e., uricase prevents reabsorption of 40% of secreted urate, increasing net intestinal elimination by that amount.

  5. Renal partial compensation. When intestinal urate excretion changes, the kidney partially compensates by adjusting tubular handling (Matsuo 2014 directional). We use 30% offset (sensitivity 0–50%).

  6. Steady-state ΔSUA. 1st-order approximation: ΔSUA = −SUA_baseline × (Δflux / production). Baseline SUA differs by sex and genotype (gout-typical values: 8.0 mg/dL WT male, 7.0 WT female, scaling up with genotype severity to 9.5 mg/dL Q141K hom male, 10.5 mg/dL severe dysfunction).

Three uricase-dose scenarios: 5 mg/day (low), 25 mg/day (mid, engineered-koji target), 50 mg/day (high, ALLN-346 Phase 2a equivalent).

Key results

Phase A literature mining — what's there and what's NOT

The headline Phase A finding is a NEGATIVE: across the entire published uricase-trial corpus, ZERO trials have stratified results by ABCG2 Q141K genotype. Allopurinol does (Wallace 2018, Vora 2021, Stamp 2019). Uricase does not. Every oral uricase trial (ALLN-346 Studies 201/202) and every systemic uricase trial (rasburicase post-marketing, pegloticase, PRX-115 Phase 1) missed this stratification axis.

The mechanistic anchor — Miyazaki 2025 — is the load-bearing Phase A datum. n=34 patients, Crohn's-disease-dominated (32/34), undergoing double-balloon endoscopy. Direct measurement of jejunal urate secretion stratified by ABCG2 functional class:

ABCG2 functional class n Baseline lumen urate (median, IQR) pg/μL ΔC over 5 min (median, IQR) pg/μL
100% (WT/WT) 11 105.3 (12.3–236.6) +175.6 (92.6–550.7)
75% (Q141K het) 18 113.0 (9.4–258.0) +125.9 (7.7–476.4)
50% (Q141K hom or Q126* het) 5 70.1 (25.5–110.9) +65.9 (−35.1–114.0)

p for trend = 0.058 (just shy of conventional significance with n=34 and one group of n=5).

The 100%:50% median secretion-rate ratio is 2.66 (175.6 / 65.9). The means tell a more dramatic story (600.5 vs 44.8, ratio 13.4) but the small n in the 50%-functional group makes the median the more reliable central estimate. We propagate the median ratios into Phase B.

Phase B flux model results

Predicted ΔSUA at steady state, Monte Carlo medians (90% CI in parens, n=5000 samples):

Scenario Low 5 mg/d Mid 25 mg/d High 50 mg/d
WT/WT, male gout −0.83 (−1.14, −0.57) −0.83 (−1.13, −0.57) −0.83 (−1.13, −0.57)
Q141K het, male gout −0.66 (−0.91, −0.45) −0.67 (−0.91, −0.45) −0.67 (−0.91, −0.46)
Q141K hom, male gout −0.50 (−0.67, −0.34) −0.50 (−0.68, −0.34) −0.49 (−0.67, −0.34)
WT/WT, female gout −0.73 (−1.00, −0.50) −0.74 (−1.00, −0.50) −0.74 (−1.00, −0.50)
Q141K het, female gout −0.58 (−0.80, −0.40) −0.59 (−0.80, −0.40) −0.59 (−0.80, −0.40)
Q141K hom, female gout −0.41 (−0.57, −0.28) −0.42 (−0.57, −0.29) −0.42 (−0.57, −0.29)
Severe dysfunction, male −0.27 (−0.38, −0.19) −0.28 (−0.37, −0.19) −0.27 (−0.37, −0.19)

Three structural observations from the model

  1. Genotype ordering is INVERTED relative to the worry framing. The model predicts WT/WT patients show the LARGEST absolute ΔSUA, not the smallest. The reason is mechanistic: uricase amplifies whatever ABCG2 is doing in the gut compartment. WT/WT patients have full ABCG2 capacity to amplify; Q141K homozygotes have half. The mechanism is multiplicative on residual gut capacity.

  2. Dose-response is FLAT above ~5 mg/day. Capacity ratios (enzyme throughput / delivered intestinal urate flux) range from 32× at the low dose to 1300× at the high dose for severe-dysfunction patients. Uricase capacity vastly exceeds substrate even at the lowest dose tested. The mechanism is substrate-limited, not enzyme-limited. This is a strain-engineering implication (don't push yield further; push GI delivery).

  3. Sex difference is small and inherits from baseline SUA. The model predicts ~12% smaller absolute ΔSUA for females than for males within any given genotype — purely because females have ~12% lower baseline SUA on average. The mechanism itself is sex-symmetric (per comp-017, healthy-baseline intestinal ABCG2 is sex-invariant). Females respond, just from a lower starting point.

Recovery against ALLN-346 Phase 2a empirical signal

The flux model's central prediction for an WT/WT-dominated, mostly stage-2-CKD cohort at 50 mg/day uricase (the ALLN-346 Phase 2a Study 201 dosing approximation) is roughly −0.6 to −0.9 mg/dL ΔSUA. Allena Pharmaceuticals reported a statistically significant signal in Study 201 days 5–7 (p<0.01 to p<0.05), with the largest reduction in stage 2 CKD patients and sUA reduction correlated with eGFR (r=0.95, p=0.003). Specific mean ΔSUA values were not in the EULAR abstract, but the directional signal is consistent with the flux model's prediction.

Study 202 (broader cohort, mixed CKD severity) showed 0–5% mean ΔSUA, not statistically significant. Interpretation: the broader cohort included patients with milder CKD where the renal compensation reserve is higher, partially offsetting the gut delta. The flux model's renal-compensation parameter (central 30%, range 0–50%) covers this — at the upper end of compensation, the predicted ΔSUA shrinks toward the Study 202 observed range. The flux model is consistent with both Study 201 and Study 202 outcomes — Study 201 sat in the small-renal-compensation regime (high-CKD), Study 202 in the larger-renal-compensation regime (mixed CKD).

Recommendations

For Phase 2b RCT design (the most actionable change)

Pre-comp-019 design assumption Post-comp-019 recommendation
Enrich for Q141K-positive patients to maximize hit rate Run as typical-gout RCT, with Q141K + Q126* genotyping as stratification variables (NOT enrichment)
~360 unenriched Europeans needed to power Q141K-conditional sub-analysis Same n=120-180 typical European-gout cohort works because the mechanism is genotype-robust; Q141K subgroup analysis is secondary, not primary
Powering for Q141K-conditional differential effect Power for ~−0.5 to −1.0 mg/dL ΔSUA mean in unstratified cohort; Q141K subgroup analysis is hypothesis-generating
Pre-stratify cohort by CKD stage (per ALLN-346 lesson) KEEP this — CKD is a stronger response-modifier than Q141K per the flux model (renal compensation reserve dominates)
Single-dose cohort Test only one dose (~25 mg/day target). Dose-response is flat above ~5 mg/day per flux model — running multiple dose arms is wasteful

For strain-engineering priorities

Pre-comp-019 priority Post-comp-019 priority
Push uricase yield in engineered koji to ~50+ mg/dose Yield target can remain at 25 mg/dose (substrate-limited regime)
GI survival as critical blocker (per cross-validation.md Claim 2 rating 5/10) GI survival becomes EVEN MORE the binding constraint — yield is 30-1000× excess but only the fraction that survives matters
Both yield AND survival as parallel engineering tracks Survival > yield as engineering priority

For self-experimentation protocol (Brian's n=1 case)

Brian's relevant genotype is captured by the personal-genome-protocol.md MinION pipeline. comp-019's flux model predicts: - If Brian is WT/WT for both Q126* and Q141K: expected ΔSUA ~−0.83 mg/dL (mid-dose engineered-koji) - If Brian is Q141K heterozygous (~25% prior probability for European ancestry): expected ΔSUA ~−0.67 mg/dL - If Brian is Q141K homozygous (~1% prior probability): expected ΔSUA ~−0.50 mg/dL

In all scenarios the predicted effect is detectable with home uric-acid-meter precision (typical assay CV ~5%, baseline SUA ~5-7 mg/dL → detectable threshold ~0.3 mg/dL ΔSUA). The personal-genome-protocol genotype IS load-bearing for interpreting the n=1 self-experiment outcome — without it, a smaller observed ΔSUA could be confused with mechanism failure rather than genotype-typical response.

What this CHANGES about the §1.27 Caco-2 wet-lab gate

Not triggered. The flux model already rules out the binary "Q141K-only mechanism" hypothesis without needing Caco-2 transwell data. The wet-lab gate would re-trigger only if a Phase 2b RCT shows differential genotype response inconsistent with the flux model's predictions.

The Caco-2 gate REMAINS APPROPRIATE for a different question — does intestinal-cell sublethal stress (TNFα, oxidative load) phenocopy ABCG2 dysfunction in the way the flux model would predict for the severe-dysfunction tier? But that's not the comp-019 question.

Cross-references and downstream propagation

  • cross-validation.md Claim 1 — gut-lumen sink mechanism rating updated from 6/10 to 6.5/10 (mechanism is now genotype-robust per flux model; clinical magnitude prediction awaits Phase 2b verification).
  • gut-lumen-sink.md — patient-stratification note updated to reflect that the gut-lumen sink works in non-Q141K patients (in fact, BETTER than in Q141K-positive patients).
  • abcg2-modulators.md §6 (Q141K rescue) — cross-reference comp-019's clarification that rescue interventions (butyrate, fermentable fiber, HDIs) ADD to the gut-lumen sink rather than replacing it; the gut-lumen sink works across genotypes, but Q141K-positive patients get a synergy bonus from rescue interventions.
  • androgen-urate-axis.md — the male-asymptote framing is now further softened: not only is sex-dimorphism near-null at healthy baseline (per comp-017), but ALSO the gut-lumen sink mechanism does not depend on Q141K-male compounding for benefit.
  • open-questions.md — Open Question 1 ("can the gut-lumen sink be rescued in non-Q141K population") now has a flux-model answer: yes, the sink works MORE in non-Q141K than in Q141K patients.
  • synthesis.md Sweep 2026-05-08 Open Question 1 — actioned by comp-019.
  • validation-experiments.md — §1.27 Caco-2 gate remains appropriate for stress-phenocopying questions but is NOT triggered by the comp-019 finding for the genotype question.
  • personal-genome-protocol.md — Q141K + Q126* genotyping is now load-bearing for Brian's n=1 protocol interpretation.

Limitations

  1. No prospective Q141K-stratified oral uricase RCT data exists. The flux model's predictions are PROSPECTIVE — they have not been clinically validated. The Phase 2b RCT design recommendation is the bridge from this in-silico finding to clinical evidence.

  2. Miyazaki 2025 substrate population is Crohn's-disease-dominated (32 of 34 subjects). Inflammatory bowel disease may itself affect intestinal ABCG2 expression / function. The genotype-stratified secretion ratios this paper provides are the best available data, but their generalization to typical gout patients (no IBD) is an open question. Per the discussion in Ferrer-Picón 2020 (PMID 31211831), inflammation reduces intestinal ABCG2 — meaning Crohn's patients likely have LOWER baseline intestinal urate secretion than non-IBD comparators. This biases Miyazaki's absolute numbers downward, but should not bias the genotype RATIOS we propagate into the flux model. The flux model uses ratios, which is robust to this bias.

  3. The flux model uses 1st-order steady-state approximation. A more rigorous PBPK model with explicit clearance kinetics, reabsorption gradients, circadian dosing, food effects, and CKD-grade renal compensation would refine the magnitude estimates but is unlikely to flip the genotype-ordering conclusion.

  4. Renal compensation fraction is mechanistic-extrapolation. Central estimate 0.30, sensitivity 0.0–0.5. Not directly measured in any oral-uricase trial. The flux model's headline conclusion (genotype ordering of response magnitude) is robust to this parameter; the absolute magnitudes are not.

  5. Pre-commit verification gate (CLAUDE.md Rule 4) tier:

  6. Miyazaki 2025 numbers: full-text grep-verified from PMC11877951.
  7. Wallace 2018 (OR 2.43, p=6.2e-7): abstract-tier verified.
  8. Vora 2021 (1.79× half-life): abstract-tier verified.
  9. Stamp 2019 (Q141K p=0.019): abstract-tier verified.
  10. Matsuo 2014 (functional-classification framework, risk ratios 1.36/1.66/2.35): abstract-tier verified.
  11. Takada 2014 (mouse KO): abstract-tier verified.
  12. Nakayama 2011 (Km 8.24 mM, Vmax 6.96 nmol/min/mg): abstract-tier verified.
  13. ALLN-346 Phase 2a Study 201/202: EULAR-abstract / press-release tier; NOT line-anchored to peer-reviewed full-text.
  14. PRX-115 Phase 1: ACR conference-abstract tier.

  15. Future Paperclip-MCP run recommended to grep-verify the abstract-tier-only ABCG2 kinetics + Q141K-allopurinol response numbers against full-text.

  16. Multilingual scan deferred. ChiCTR / J-STAGE / KISS searches did not surface a published genotype-stratified oral-uricase RCT in any language. The likelihood of one being missed is low (Q141K × allopurinol response is heavily indexed; Q141K × uricase response would be too if it existed). A future multilingual deep-dive on Chinese-language hyperuricemia pharmacogenomics could refine the picture for East-Asian-ancestry cohorts where Q141K MAF is ~30%.

  17. Two-model translation cross-check did not trigger. Per CLAUDE.md §"Translation protocol", non-English sources with load-bearing claims should be translated with two independent models. comp-019 did not surface non-English sources whose translation would change a verdict. Nguyen 2025 Vietnamese cohort paper is published in English (Clin Rheumatol) — no translation step needed.

What would flip the verdict

The headline verdict (mechanism is genotype-robust; non-Q141K patients respond AT LEAST as well as Q141K patients) would flip only if:

  • A Phase 2b oral-uricase RCT shows Q141K-positive patients responding MORE than Q141K-negative patients (would contradict the flux model's gradient-amplification logic).
  • A direct human pharmacokinetic study of oral uricase + intestinal urate secretion measurement shows the substrate-limited regime is wrong (uricase doesn't actually clamp lumen urate; some unmeasured rate-limiting step intervenes).
  • A more detailed PBPK model with explicit CKD-grade renal compensation shows the renal compensation fully offsets the gut-delta in healthy-kidney patients (would shrink the predicted ΔSUA in WT/WT toward zero in non-CKD subjects).

None of these have surfaced. Flux model verdict is robust to the abstract-tier verification limitations in DIRECTION (genotype ordering); specific magnitude estimates remain in the −0.5 to −1.0 mg/dL band pending Phase 2b RCT validation.