Why it comes back — and what can reach it.
Maximal safe surgery and locally-delivered agents clear the tumor mass and the resection bed — but stop at the edge of what can be removed or physically reached.
Surgery and drugs clear the bulk. But the tumor has already sent single cells centimeters into healthy brain, and no agent large enough to act on them can travel that far. The one thing that can reach every one of them is the patient's own immune system.
Our mission is to improve the treatment available to people with grade 3–4 IDH-mutant astrocytoma — a brain cancer that most often strikes in a person's thirties or forties. Today's standard of care — the most surgery that is safely possible, plus radiation and chemotherapy — buys time, but it does not cure. The tumor comes back. Understanding why it comes back is the key to everything we do.
Why it always comes back
By the time this tumor is found, it has already sent single cells centimeters into normal-looking brain, far beyond anything a surgeon can see or remove. You cannot cut it all out, and you cannot physically reach those scattered cells with a drug delivered into the brain — anything large enough to act on a tumor cell is too large to travel that far through brain tissue. That fringe of hidden cells — the infiltrating margin — is where the tumor regrows.
The one thing that can reach the margin
There is a single agent in the body that already travels everywhere: the patient's own immune system. Immune cells go wherever blood and lymph go — including out to those scattered cells at the margin.
So our strategy is not to chase the margin with a drug. It is to ignite a controlled anti-tumor immune response right at the tumor, and let the immune system carry it outward to the cells surgery and drugs can't reach. This "ignite locally, let the immune system finish the job" idea is an active field of cancer research, and it has real precedent in brain tumors: some patients treated with related immune-igniting therapies have had durable responses, with the immune system visibly infiltrating the tumor.
How we do it today: a living vehicle
To ignite that response inside the tumor, we are engineering bacteria as a living, targeted vehicle. A bacterium is not a fixed dose — it is a small, programmable factory that can travel to the tumor, switch on there, and continuously release a payload that activates the immune system locally.
The tumor's own biology makes this possible. IDH-mutant tumors flood their surroundings with an oncometabolite called D-2-HG, and we use that one molecule three ways:
- Target — D-2-HG marks where the tumor is, so the therapy switches on there and not in healthy brain.
- Leash — the vehicle is built to depend on the tumor environment, so it thrives where the tumor is and falls back where it isn't. A built-in safety leash keeps it contained.
- Lift the immune brake — D-2-HG also paralyzes local immune cells. Consuming it in the ignition zone helps release that brake exactly where the immune response is being lit.
And because it is a programmable vehicle, immune ignition is the lead job but not the only one: the same platform can be built to destroy tumor tissue directly, or engineered to combine with other treatments.
Why this, and not the existing drug alone
A drug called vorasidenib blocks the mutant enzyme from making new D-2-HG. It is approved for grade 2 disease and is being studied in a Phase 2 trial in grade 3. It is an important advance — but grade 3–4 is a harder problem. At this stage, D-2-HG is only one of several ways the tumor hides from the immune system, so shutting down its production, on its own, is less of a complete solution. That harder disease — and that unreachable margin — is exactly what our mission is built to take on.
Where this stands
This is early-stage science — a testable hypothesis pursued rigorously, not a product available today. Nobody has cured grade 3–4 astrocytoma, and we do not claim to. Our honest ambition is meaningful, potentially durable responses in a subset of patients — a real advance over what today's standard of care can offer. The make-or-break question we are working to answer is a clear one: does an immune response ignited at the tumor actually reach and clear the infiltrating margin?
The IDH edge
One molecule, three jobs.
D-2-HG is both the tumor's immune-suppressing signal and a tumor-specific address. That lets one oncometabolite do three things at once for a therapy built around it.
Select each role.
D-2-HG marks where the tumor is. The therapy is designed to switch on where D-2-HG is present — concentrating its activity in the tumor, not in healthy brain.
A programmable vehicle
One living vehicle, several payloads.
A bacterium is a living, programmable factory — not a fixed dose. It can carry and continuously produce more than one payload, and the mix is engineered to the job.
The lead payload: activate a controlled anti-tumor immune response at the tumor, so the immune system can carry it to the margin.
A tumor-colonizing vehicle can also destroy tumor tissue directly at the sites it reaches.
Beyond immune ignition, the same vehicle can be engineered to carry additional payloads designed to combine with other treatments.
The disease pathway
How one mutation reshapes the tumor.
A single change in IDH turns a normal enzyme into a producer of an oncometabolite with two downstream effects — one of which the therapy is designed to exploit.
Hover or tap each node to explore.
The defining alteration of IDH-mutant astrocytoma. Instead of its normal reaction, the variant enzyme reduces α-KG into D-2-hydroxyglutarate.
The evidence, in detail.
This tier lays out the quantitative and mechanistic rationale behind the approach — the physics of reach, what is actually known about D-2-HG concentrations, the immune-suppression threshold, and the precedent. We keep our own engineering at the level of approach; the numbers below are drawn from the published literature.
The physics of reach — why the margin needs a relay
The infiltrating margin is a physical, not just a biological, problem. A bacterium is roughly 50–100× larger than the extracellular spaces of brain tissue; it swims in fluid but does not invade solid parenchyma, and every tumor-colonization precedent is vascular, not surface-crawl. A colonized or cavity-deposited focus therefore acts over a radius on the order of 1–3 mm. Against a tumor whose defining feature is single cells scattered centimeters out, no locally-acting agent — bacterial, viral, or small-molecule — can cover the margin by physical reach alone.
This is the entire reason the strategy is an immune relay rather than physical coverage: activated immune cells are the only agents mobile enough to traffic from an ignition focus to the dispersed margin. It is also why the program belongs to the intratumoral-immunotherapy class, whose glioma precedent (e.g. Delta-24 / DNX-2401 in recurrent disease) demonstrates that local ignition can produce systemic anti-tumor immunity with tumor T-cell infiltration in a subset of patients.
What is actually known about D-2-HG concentration — an honest calibration
The D-2-HG "immune brake" is a real lever, but a calibrated one, and we state it honestly.
- Bulk tissue D-2-HG in grade 3–4 IDH-mutant astrocytoma sits around 3–9 mM by modern quantitative measurement, reaching ~9–13 mM in high-cellularity cores (in-vivo MR spectroscopy: Choi et al. 2012, Nat Med, PMID 22281806; Suh et al. IPD meta-analysis 2018, Neuro-Oncol, PMID 30020513). The legacy "5–35 mM" figure (Dang et al. 2009) is a pooled, mixed-grade upper bound and runs ~10× above modern LC-MS/MS numbers; we do not treat it as the operative value.
- The compartment that matters for immunity is extracellular. Every bulk number is an upper bound on what a T cell actually sees. The one paired bulk-vs-interstitial measurement (mouse orthotopic model, microdialysis) put interstitial D-2-HG at only a few percent of bulk — implying a human extracellular level more likely in the ~0.1–1 mM range, with an uncertainty tail up to ~2.5 mM in high-D-2-HG cores. No human tumor-interstitial D-2-HG measurement exists; this is a genuine open quantity.
- The T-cell suppression threshold. On primary human T cells, extracellular D-2-HG suppresses proliferation by ~15–20% at 1 mM and ~50% at 5 mM, with cytokine/activation readouts needing ≥10–20 mM (Bunse et al. 2018, Nat Med, PMID 29988124).
Put together: local D-2-HG removal has conditional, likely-modest standalone immune-relief value at grade 3–4 — strongest in high-cellularity cores — and is best positioned as an enabling / combination lever alongside the ignited response, not as a standalone therapy. This is precisely why the therapeutic point of the program is immune ignition, with the brake-lift as a supporting role. It is also why grade 3–4 is the deliberate target: D-2-HG is only one of several immunosuppressive mechanisms there, so a durable effect requires actively igniting immunity, not only relieving one brake.
D-2-HG's triple role — the biochemistry
One reaction underlies all three jobs. Oxidizing D-2-HG back to α-KG lets the engineered vehicle use the oncometabolite as fuel; making survival depend on that D-2-HG-derived carbon creates an auxotrophic leash (the vehicle lives only where D-2-HG is); and consuming D-2-HG in the ~1–3 mm ignition zone locally lowers it, helping lift the immune brake on arriving T cells. The leash is partly redundant with — and reinforced by — hypoxia-gating in an anaerobe chassis, giving containment from two independent directions.
The immune-ignition cascade — the logic
Turning a cold, immunosuppressed tumor "hot" is a cascade, not a single signal: a controlled adjuvant provides the danger signal; the dominant, suppressive tumor macrophages are repolarized; chemokines recruit killer T cells into an otherwise excluded tumor; and locally-acting cytokines arm them, with IFN-γ amplifying the response into a self-reinforcing local response — while the D-2-HG brake-release keeps arriving T cells from being re-paralyzed. Payloads that are systemically toxic (e.g. certain cytokines) are confined to local, in-tumor production — one of the core advantages of a self-localizing living vehicle over a systemic drug. We describe this at the level of rationale; the specific payload set and release circuitry are protected.
Containment and safety — defense in depth
CNS deployment of a living therapeutic demands layered, independent containment. The approach combines physical/metabolic gating (hypoxia and/or D-2-HG dependence, keeping the vehicle in the tumor), endotoxin management appropriate to the chassis, orthogonal genetic kill switches, a population cap, and antibiotic sensitivity. No single layer is sufficient alone — genetic kill switches, for instance, have finite escape frequencies — so regulators expect several combined, which is the design intent here.
Positioning relative to vorasidenib
Vorasidenib blocks synthesis of new D-2-HG at the mutant enzyme. It is approved in grade 2 (INDIGO Phase 3, Mellinghoff et al. 2023, NEJM, PMID 37272516) and is being evaluated in a Phase 2 trial in grade 3. It is a genuine advance. But it is a production inhibitor addressing a single lever, and at grade 3–4 — where D-2-HG is only one of several immune-evasion mechanisms and the margin remains physically unreachable — blocking production alone is less of a complete solution. Our approach is complementary: it acts directly on the immune microenvironment and is designed to reach the margin through the immune relay.
Program stage
Early-stage science, pursued rigorously. The mechanistic and quantitative rationale is developed; the central hypothesis — that an immune response ignited at the tumor reaches and clears the infiltrating margin — is what the experimental program is built to test. Not a product available today; no cure is claimed.
Two ways in
Where each track goes to work.
Both tracks share one endgame — an immune response that reaches the infiltrating margin. They differ in where the living vehicle acts: Track A stays in the resection cavity at the tumor bed; Track B colonizes deeper, inside the tumor's hypoxic core, and damages it directly on the way.
After surgery, a safety-contained bacterium sits in the resection cavity and primes an anti-tumor immune response at the bed — the tissue closest to where recurrence begins.
Both feed the same immune relay out to the margin.
Track A — internal program
Contained cavity depot
A safety-contained probiotic bacterium, placed directly in the surgical resection cavity after tumor removal. It stays alive and active only where the tumor's own oncometabolite, D-2-HG, is present — the same molecule that suppresses the local immune system becomes the cue that keeps the vehicle localized.
There it secretes a payload built to prime an anti-tumor immune response at the resection bed — the tissue closest to where recurrence begins. It's the simpler, gentler approach, with the most straightforward safety story of the two tracks.
Track B — partnership
Tumor-colonizing collaboration
A partnership around a spore-forming bacterium built to germinate and colonize inside the tumor's hypoxic core, not just the resection cavity — using D-2-HG-based sensing to stay contained to tumor tissue.
There it destroys tumor tissue directly, where drugs can't penetrate — valuable on its own — and that destruction also helps ignite an anti-tumor immune response, as a programmable, multi-payload living factory rather than a single-trick construct. It goes deeper into the tumor than Track A: more potential potency, and correspondingly harder safety engineering.