FRONTIER SIGNAL | DEEP DIVE | BIOTECH & INVESTING | 8 MIN READ | MONDAY, 14 SEPTEMBER 2026
KEY TAKEAWAYS
Engineered Saccharomyces cerevisiae produced galantamine from supplied 4OMe-norbelladine using three daffodil enzymes.
Matching the plant P450 enzyme with a better reductase and cytochrome b5 partner raised output to 7.0 milligrams per litre.
The process achieved a 7.6 percent molar yield and is far below commercial manufacturing readiness.
The reusable frontier is P450 and redox-partner engineering, which can unlock other difficult plant-derived molecules.
Antheia and Codexis are category read-throughs, not disclosed participants or direct beneficiaries of this result.
Galantamine still begins in a field. The Alzheimer medicine is extracted from plants such as snowdrops and daffodils because its tightly folded molecular structure is difficult to make economically through conventional chemistry. Crops grow slowly, concentrations are low, and supply moves with harvests.
A research team has now made galantamine inside baker's yeast. The cells completed the reaction that had blocked earlier microbial efforts and produced 173 times more of the drug than the starting strain.
There is one important correction to the easy headline. The yeast was given a specialised precursor. It did not brew galantamine from sugar.
THE NUMBERS
7.0 mg/L reported titre · 173x gain from starting strain · 7.6 percent molar yield
Posted to bioRxiv in September 2026. The work has not been peer reviewed and used a supplied precursor.
Here is what happened
A team led from the Technical University of Denmark reconstructed the last three steps of galantamine biosynthesis in Saccharomyces cerevisiae, inserting three genes from Narcissus cv. Tete-a-Tete.
Three borrowed enzymes. NtCYP96T6 folds the precursor into the galantamine scaffold, NtNMT1 methylates an intermediate, and NtAKR1 performs the final reduction.
The first strain made nothing. It produced intermediates but no detectable galantamine. Dropping the culture from 30 degrees Celsius to 25, swapping glucose for trehalose and raising medium pH from 5.5 to 7.0 produced about 41 micrograms per litre.
The real gain came from the wiring, not the enzyme. Screening redox partners found a reductase from Narcissus pseudonarcissus that raised pathway flux to 7.0 plus or minus 0.6 milligrams per litre. It is the same move we watched in September when a team stopped trying to fix a broken gene and fixed the machinery that reads it: leave the hard target alone and rebuild what supports it.
The reaction that held the pathway back
Galantamine has a compact, stereochemically demanding ring system. Plants assemble it through an oxidative phenol-coupling reaction that forms the core scaffold in the correct orientation. Synthetic chemists can make the molecule, but doing so through many controlled steps is costly.
P450s are versatile because they are complicated. They sit in a membrane, bind a substrate and rely on partner proteins to shuttle electrons from NADPH.
Moved into yeast, they misbehave. A plant P450 may express badly, receive electrons inefficiently, or send the precursor into unwanted side reactions.
So the transferable result is not the 7 milligrams. It is the evidence that redox-partner selection can unlock a plant P450 inside a microbial factory, which matters for many alkaloids whose pathways stall at similarly awkward chemistry.
What the yeast did and did not do
The yeast made galantamine. It did not make the process independent of plants or chemical synthesis.
The cells were fed. Researchers supplied 4OMe-norbelladine, the shared precursor from which plants build hundreds of Amaryllidaceae alkaloids. The yeast performed only the downstream conversion.
The paper says so itself. It describes the process as biocatalytic production and identifies de novo production as future work.
That distinction changes the commercial story. A full fermentation route could replace fields of bulbs with tanks of microbes, in the way engineered bacteria sealed inside the body replace a pharmacy with a colony. A fed-precursor route still depends on the cost, purity and availability of its feedstock.
THE OTHER WAY TO SOLVE THIS
This team borrowed three enzymes nature had already evolved and spent its effort on making them work somewhere else. The opposite bet is to stop borrowing: we wrote about a delivery shell drawn from scratch by a generative model, a structure with no natural counterpart at all. Borrowing gets you chemistry evolution already solved, and inherits every reason that chemistry is fussy. Designing from scratch avoids the inheritance and gives up the head start.
The second product may matter more
The team also fed a fluorinated version of the precursor and detected a product consistent with 7-fluorogalantamine. The assignment rests on accurate mass and fragmentation rather than full pharmacological characterisation, so it remains putative.
A pathway that accepts modified precursors is a scaffold factory. Enzymes build the hard ring system, chemists alter the feedstock, and the combination reaches chemical space either method covers poorly alone.
The existing drug is not the ceiling. Galantamine offers moderate symptomatic benefit, and dose escalation is limited by cholinergic side effects.
Discovery value can arrive before manufacturing value, if the platform makes structures that were previously inaccessible. New analogues would still need medicinal chemistry, toxicology and trials.
Who captures the value
No pharmaceutical company has announced a galantamine programme based on this preprint. The result is too early and too low-yield to change the economics of the current medicine.
Antheia is the clearest private read-through. It is building yeast-based routes to plant-derived pharmaceutical ingredients, on the thesis that contained fermentation reduces dependence on crops. This work validates the category problem, not Antheia's pipeline.
Codexis is the nearer listed comparison. Its business includes engineering enzymes for pharmaceutical biocatalysis from gram to ton scale, which is exactly the bottleneck the 173-fold gain addressed. There is no disclosed Codexis involvement, so this is thematic, not a revenue forecast.
The winner may own the whole chain. Precursor biosynthesis, enzyme optimisation, scale-up, purification and regulatory-grade consistency. That is the lesson from Moderna, which did not beat its manufacturing bottleneck so much as pay to go around it.
The useful diligence questions are concrete. Can the precursor be made from sugar inside the same organism? Does titre rise without new impurities? Can recovery work at scale? Can each batch meet pharmaceutical specifications? Those decide cost per gram, not the 173-fold headline.
The honest catch
Seven milligrams per litre is a proof of chemistry, not a commercial titre. The process converted only 7.6 percent of the supplied precursor on a molar basis.
Intermediates piled up. Narwedine in particular, which means the final reduction step is another bottleneck.
It is a preprint. Not peer reviewed. The experiments are detailed and the molecule was compared with analytical standards, but replication and scale-up data do not exist yet.
The precursor still comes from outside the yeast. Until the upstream pathway is rebuilt from a cheap carbon source, claims that galantamine has been brewed like beer are misleading. The accurate description is narrower and still meaningful: yeast has completed the plant-derived final assembly line.
OUR VIEW
This is a real synthetic-biology advance and an easy story to overstate. The team crossed the step that turns a useful precursor into galantamine and improved output by roughly two orders of magnitude through enzyme-system engineering.
It did not replace daffodil farming. The yeast still needs a specialised starter molecule, the yield is low and the work is unreviewed. The frontier lies in taming plant P450 chemistry inside microbes, and using that capability to make both known medicines and new analogues.
For investors the signal favours full-stack biomanufacturing over single-pathway demonstrations. It is the same bet as moving drug manufacture into orbit: the interesting question is never the molecule, it is whether the new place to make it survives contact with cost, purity and scale.
THE ONE QUESTION
What does one gram of purified galantamine cost when the precursor, fermentation losses and downstream purification are all included?
Titre alone can improve while the process stays uneconomic. The commercial threshold is set by total cost per gram, batch consistency, impurity control and the ability to operate under pharmaceutical manufacturing standards.
Five signals to watch next
Peer review and replication. Independent groups need to confirm the 7 milligram per litre result and the product identity.
De novo precursor production. The decisive milestone is galantamine made from a simple carbon source with no externally supplied 4OMe-norbelladine.
Titre and molar yield. Commercial work needs large gains beyond 7 milligrams per litre and 7.6 percent conversion.
Downstream recovery. Purity, solvent use and recovery losses decide whether fermentation beats extraction or synthesis.
Analogue validation. Putative 7-fluorogalantamine needs structural confirmation and pharmacological testing before it is a discovery asset.
Quick questions
Did yeast make galantamine from sugar?
No. The engineered yeast converted externally supplied 4OMe-norbelladine into galantamine. A complete de novo route from a simple carbon source has not been demonstrated.
How much galantamine did the yeast produce?
The best reported strain produced 7.0 plus or minus 0.6 milligrams per litre, a 7.6 percent molar yield from 250 micromolar supplied precursor.
Why was the 173 fold improvement possible?
The team optimised culture conditions and screened redox partners for the plant cytochrome P450. A reductase mined from Narcissus pseudonarcissus supported the highest pathway flux.
Is the process ready to replace plant extraction?
No. The titre and conversion yield are low, the precursor is still supplied, downstream economics are unknown and the work is a preprint.
What is the investment angle?
The result supports the broader case for engineered enzymes and yeast-based manufacturing of complex plant medicines. Antheia and Codexis illustrate the relevant platform categories, but neither is connected to this study.
READ MORE: BIOTECH SIGNALS
Four pieces on making medicine somewhere it has never been made before.
AI Designed a Protein Shell Nature Never Tried 8 Sep. Designing from scratch instead of borrowing from biology.
Living Bacteria Sealed in the Body to Make Drugs on Demand 18 Aug. Microbes as the factory, moved inside the patient.
Moderna Didn't Beat the Bottleneck. It Paid It. 21 Aug. What manufacturing actually costs when the science is finished.
Scientists Stopped Trying to Fix the Gene. They Fixed the Reader 1 Sep. Leaving the hard target alone and rebuilding what supports it.
KEEP READING
Three deep dives a week, Monday, Wednesday and Friday.
Sources
Galantamine yeast preprint, bioRxiv: every figure quoted above.
University of Texas precursor study: the 2024 E. coli route to 4OMe-norbelladine.
ResearchGate full-text mirror, the Antheia biosynthesis platform and the Codexis biocatalysis platform. Cited by name only: this desk could not verify their URLs, and we do not ship a guessed one.
This article is general information, not investment advice. Frontier technologies carry technical, commercial, regulatory and financing risks. Company statements and project forecasts may not be achieved. Verify current disclosures and conduct independent due diligence before making investment decisions.