UTI Drug Discovery
Recurrent UTIs aren't reinfections — they're reactivations. E. coli hides inside bladder cells, survives antibiotics, and re-emerges weeks later. GSK's Phase 2b FimH inhibitor trial failed because it targeted the wrong stage. We designed the protocol that should have come first.
8
compounds tested
literature-backed selection
18
page protocol
v1.3, delivered to lab
$3,300
total budget
24-well format, 5637 cells
6
week timeline
POC sprint to go/no-go
The problem
About 25% of women who get a UTI will get another one. The standard explanation is reinfection — new bacteria, bad luck, take more antibiotics. That explanation is wrong.
What actually happens: uropathogenic E. coli (UPEC) invades bladder epithelial cells and forms quiescent intracellular reservoirs (QIRs). These are dormant bacterial communities sitting inside Lamp1+ compartments — essentially hiding in the host cell's own recycling machinery. Antibiotics can't reach them. The immune system can't see them. They wait.
Then Gardnerella vaginalis shows up (a vaginal microbiome disruption, often triggered by sex or antibiotics themselves), secretes vaginolysin (VLY), damages the bladder epithelium, and the dormant UPEC reactivate. New infection. Same bacteria.
The key papers: Mysorekar & Hultgren 2006 (PNAS) discovered QIRs. Gilbert et al. 2017 (PLoS Pathog) identified the Gardnerella/VLY reactivation trigger. Blango & Mulvey 2010 (AAC) showed antibiotics fail against intracellular UPEC. The biology has been known for years. The therapeutics haven't caught up.
Why GSK's trial failed
GSK3882347 was a FimH inhibitor — a drug that blocks the adhesin E. coli uses to stick to bladder cells. Smart target. Good chemistry. They took it through Phase 2b (VOLCANO-2). It didn't work.
The reason is straightforward once you know the QIR biology: anti-adhesion monotherapy can't reach bacteria that are already inside cells. FimH inhibitors prevent new attachment. But if the reservoir is intracellular, preventing attachment only blocks one half of the recurrence mechanism. The bacteria are already inside. They don't need to attach — they need to escape and re-colonize.
Anti-adhesion alone
Blocks new attachment. Doesn't clear intracellular reservoirs. Bacteria re-emerge, reattach when drug wears off. That's VOLCANO-2.
Flush + block
Force bacteria out of intracellular reservoirs first. Then block reattachment. Clear the source, not just the symptom.
The protocol: flush and block
The dual-agent concept: a QIR clearance compound (the flush) paired with a FimH inhibitor (the block). Neither works alone for the reasons above. Together, they address both stages of the recurrence cycle.
The flush agent forces UPEC out of their intracellular hiding spots. The block agent prevents them from reattaching to the bladder epithelium once they're expelled. You need both, in sequence.
The practical upshot: if the POC sprint identifies even one compound that achieves ≥1 log₁₀ CFU reduction in intracellular bacteria with ≥70% host cell viability, that's a GO for the next phase — combination testing, VLY emergence models, 3D organoids, and eventually Klebsiella QIR characterization.
The autophagy paradox
This is the part I'm proudest of, because it's where we caught ourselves being wrong.
The initial compound panel included metformin and rapamycin — autophagy inducers. The logic seemed clean: autophagy clears intracellular debris, UPEC is intracellular debris, induce autophagy, clear UPEC. Three-step syllogism.
Except UPEC has read that syllogism too.
Zhang et al. 2019 showed that UPEC actively exploits autophagic machinery for intracellular survival. The bacteria hijack the autophagosome for nutrients and protection. Inducing autophagy doesn't clear them — it feeds them. And Yen et al. 2022 (N = 40,774 patients) found zero clinical UTI benefit from metformin, confirming the mechanism matters more than the intuition.
We removed both compounds. The panel got smaller and the protocol got better.
The lesson: biological plausibility isn't enough. "Autophagy clears intracellular pathogens" is a true general statement that happens to be false for this specific pathogen. Literature review that doesn't interrogate its own assumptions is just confirmation bias with citations.
The compound panel
Eight compounds, each with a distinct mechanism of action. The lead is Cytosporone B — a Nur77 agonist with published data showing active UPEC expulsion from 5637 cells.
| Compound | Mechanism | Role |
|---|---|---|
| Cytosporone B | Nur77 agonist | Lead — published UPEC expulsion data |
| Dasatinib | EPHA2 inhibitor | Blocks receptor-mediated endocytosis |
| AKB-4924 | HIF-1α stabilizer | Metabolic remodeling |
| Niclosamide | Autophagy modulator | With caveats — see paradox above |
| Deferiprone | Iron chelation | Starves UPEC of Fe²⁺ |
| Nitazoxanide | Intracellular antibiotic | Broad-spectrum, intracellular penetration |
| Ciprofloxacin | Fluoroquinolone | Positive control |
| Fosfomycin | Cell-wall synthesis inhibitor | Negative control (poor intracellular penetration) |
The protocol
Format
24-well plates, 5637 bladder cells
Readout
Intracellular CFU + host cell viability
Go/No-Go
≥1 log₁₀ CFU reduction + ≥70% viability
Timeline
6 weeks, POC sprint
Budget
$3,300 all-in
Status
Awaiting lab kickoff
If GO: expand to VLY emergence model (does Gardnerella vaginolysin trigger reactivation in the assay?), combination testing (flush + block together), 3D bladder organoids, and Klebsiella QIR characterization. The $3,300 sprint decides whether a multi-month program is worth running.
Literature foundation
This project required synthesizing across urology, microbiology, cell biology, and clinical trial analysis. The supporting documentation runs 30+ papers deep:
- 1
FimH structural biology
How the FimH adhesin protein grips bladder cells at the molecular level — the specific amino acid (Tyr48) that acts as a gate. We estimated the energy barrier for this gating mechanism and identified errors in a published crystal structure.
- 2
Who else is working on this
20+ clinical programs surveyed. GSK historical data analyzed. The result: no one is directly targeting the intracellular reservoir where bacteria hide. The entire therapeutic space is focused on preventing attachment — not clearing the source.
- 3
QIR biology & therapeutics
30+ papers on intracellular persistence, VLY-triggered reactivation, and the failure of extracellular-only approaches. This is the evidence base for the flush-and-block design.
- 4
GSK failure analysis
VOLCANO-2 Phase 2b post-mortem. Why anti-adhesion monotherapy was insufficient. Strategic analysis of what the field missed and what to do next.
Collaborators
-
Pisces
AI scientist
Literature synthesis, competitive landscape analysis, compound selection, FimH structural biology, protocol design.
-
Lab researcher
BSL-2 cell culture specialist
Compound testing, flow cytometry, wet lab execution. The hands that make it real.
-
Alex Andonian
Project architect
Strategic direction, lab coordination.