In a recent survey, three-in-four respondents said they would prefer a once‑daily oral pill over a weekly injection of GLP-1s. So why aren't there more oral options?
I thought the use of cyanobacteria is a really interesting idea. As a plant synthetic biologist, I'm all for using biology and the power of photosynthesis to make high value products (aka, molecular farming). However, in this case, I'm curious: Why use spirulina or other non-model organisms over something like yeast which we eat everyday, is highly tractable genetically, and is already established in biomanufacturing?
I'm also curious how much research is still being done to make these peptides more stable in the gut (probably a lot). I've heard cyclizing peptides can be a way to stabilize them, so I wonder if that's being tried.
What fascinates me here is the engineering mindset behind GLP-1 drugs. The biology was known decades ago, but the real breakthrough came from chemical tweaks that stretched a molecule’s lifetime from minutes to days. Sometimes innovation is less about discovery and more about persistence in optimisation.
This is the kind of biotech story I love: the bottleneck isn’t the target, it’s delivery and manufacturing economics. The piece makes it clear why “just make it a pill” is a whole systems problem (gut degradation, absorption, dose, cost). Do you think small-molecule GLP-1 agonists ultimately outcompete oral peptides, or will delivery tech catch up?
The cost math in this piece is the part nobody talks about. Rybelsus needs ~1,500mg of active drug per month vs ~8mg for injectable Ozempic. Even if you solve the biology with spirulina or engineered cyanobacteria, you're still fighting a 200x dosing gap driven by sub-1% oral bioavailability. Orforglipron might actually be the more practical near-term path since it sidesteps the peptide degradation problem entirely by not being a peptide. The real question is whether Lilly can get orforglipron's efficacy close enough to injectable semaglutide that patients accept the tradeoff. The ACHIEVE trials suggest it's getting there but not quite matching Wegovy's 15%+ weight loss numbers yet.
>one must gain FDA approval by clinically demonstrating that the delivery of the microbe-encapsulated GLP-1 through this oral route leaves the drug molecule itself unchanged (with no post-translational modifications)
On the topic of post-translational modifications: how would the cyanobacteria be expected to put in the Aib and attach the lipid to the lysine? (The Aib might be particularly hard)
From the author, David Kim: Great question, short answer is that you will struggle to make semaglutide per se. But other GLP-1 drugs like lixisenatide or exenatide could be made.
Long answer:
So far, GLP-1 drugs that do not have non-canonical amino acids are far easier to turn into oral GLP-1 drugs. For example, exenatide (AstraZeneca) and lixisenatide (Sanofi) have only canonical amino acids and no lipidations. Therefore, as long as one can adequately demonstrate to the FDA that there are no new post-translational modification (PTMs) introduced by the expression host, the whole cell-encapsulated oral formulation of a GLP-1 peptide could be approved via the 505(b)2 pathway. This is conditional on having satisfactory safety, toxicology and PK/PD, but is only a viable approval route if the peptide is less than 41 amino acids in length under the current guidance.
To make semaglutide, quite right to point out that one has to incorporate non-canonical amino acids which currently, these cells cannot perform. But it is possible. This can be done through engineering orthogonal aminoacyl–tRNA synthetase (aaRS)/tRNA pairs to assign a new non-canonical amino acid, such as Aib, to an orthogonal codon (often by repurposing the UAG stop codon). 'Genetic code expansion' has been demonstrated in E coli by several groups using different approaches (Dunkelmann et al 2021, Costello et al 2024). Since it is a relatively recent field, so far, no one has published that they've replicated this in cyanobacteria, algae or any other edible organism — even though this could be a major missed opportunity if successful.
As for lipidations, it is harder to engineer specificity. Interestingly, cyanobacteria natively express enzymes called prenyltransferases. Prenylation, though only a subset of lipidation, can attach isoprenoid lipids (C15, C20) to a peptide's cysteine residues. This would not enable us to make semaglutide, but it opens doors to other peptide designs with lipid tails that can bind to albumin.
Great question, short answer is that you will struggle to make semaglutide per se. But other GLP-1 drugs like lixisenatide or exenatide could be made.
Long answer: So far, GLP-1 drugs that do not have non-canonical amino acids are far easier to turn into oral GLP-1 drugs. For example, exenatide and lixisenatide have only canonical amino acids and no lipidations. Therefore, as long as one can adequately demonstrate that there are no new post-translational modification (PTMs) introduced by the expression host, the whole cell-encapsulated oral formulation of a GLP-1 peptide could be approved, conditional on having shown satisfactory safety, toxicology and PK/PD clinically.
To make semaglutide, quite right to point out that one has to incorporate non-canonical amino acids which currently, these cells cannot perform. But it is possible. This can be done through engineering orthogonal aminoacyl–tRNA synthetase (aaRS)/tRNA pairs to assign a new non-canonical amino acid, such as Aib, to an orthogonal codon (often by repurposing the UAG stop codon). 'Genetic code expansion' has been demonstrated in E coli by several groups using different approaches (Dunkelmann et al 2021, Costello et al 2024). Since it is a relatively recent field, so far, no one has published that they've replicated this in cyanobacteria, algae or any other edible organism — even though this could be quite an interesting avenue to pursue.
Imho, the real hard question is the inconsistency of bioavailability of the oral GLP1a form (in currently existing formulations) not only on an individual basis but also with each administration. Basically, the oral form currently, it seems, delivers a "whatever" actual dosage. Thus, the potential long-term consequences of such inconsistency are in question, which might range from the broad swing of the "ordinary" side effects to the serious stuff that is more prevalent, it seems, with abrupt start/stops instead of gradual tapering of injected dosages.
I thought the use of cyanobacteria is a really interesting idea. As a plant synthetic biologist, I'm all for using biology and the power of photosynthesis to make high value products (aka, molecular farming). However, in this case, I'm curious: Why use spirulina or other non-model organisms over something like yeast which we eat everyday, is highly tractable genetically, and is already established in biomanufacturing?
I'm also curious how much research is still being done to make these peptides more stable in the gut (probably a lot). I've heard cyclizing peptides can be a way to stabilize them, so I wonder if that's being tried.
What fascinates me here is the engineering mindset behind GLP-1 drugs. The biology was known decades ago, but the real breakthrough came from chemical tweaks that stretched a molecule’s lifetime from minutes to days. Sometimes innovation is less about discovery and more about persistence in optimisation.
This is the kind of biotech story I love: the bottleneck isn’t the target, it’s delivery and manufacturing economics. The piece makes it clear why “just make it a pill” is a whole systems problem (gut degradation, absorption, dose, cost). Do you think small-molecule GLP-1 agonists ultimately outcompete oral peptides, or will delivery tech catch up?
The cost math in this piece is the part nobody talks about. Rybelsus needs ~1,500mg of active drug per month vs ~8mg for injectable Ozempic. Even if you solve the biology with spirulina or engineered cyanobacteria, you're still fighting a 200x dosing gap driven by sub-1% oral bioavailability. Orforglipron might actually be the more practical near-term path since it sidesteps the peptide degradation problem entirely by not being a peptide. The real question is whether Lilly can get orforglipron's efficacy close enough to injectable semaglutide that patients accept the tradeoff. The ACHIEVE trials suggest it's getting there but not quite matching Wegovy's 15%+ weight loss numbers yet.
>one must gain FDA approval by clinically demonstrating that the delivery of the microbe-encapsulated GLP-1 through this oral route leaves the drug molecule itself unchanged (with no post-translational modifications)
On the topic of post-translational modifications: how would the cyanobacteria be expected to put in the Aib and attach the lipid to the lysine? (The Aib might be particularly hard)
From the author, David Kim: Great question, short answer is that you will struggle to make semaglutide per se. But other GLP-1 drugs like lixisenatide or exenatide could be made.
Long answer:
So far, GLP-1 drugs that do not have non-canonical amino acids are far easier to turn into oral GLP-1 drugs. For example, exenatide (AstraZeneca) and lixisenatide (Sanofi) have only canonical amino acids and no lipidations. Therefore, as long as one can adequately demonstrate to the FDA that there are no new post-translational modification (PTMs) introduced by the expression host, the whole cell-encapsulated oral formulation of a GLP-1 peptide could be approved via the 505(b)2 pathway. This is conditional on having satisfactory safety, toxicology and PK/PD, but is only a viable approval route if the peptide is less than 41 amino acids in length under the current guidance.
To make semaglutide, quite right to point out that one has to incorporate non-canonical amino acids which currently, these cells cannot perform. But it is possible. This can be done through engineering orthogonal aminoacyl–tRNA synthetase (aaRS)/tRNA pairs to assign a new non-canonical amino acid, such as Aib, to an orthogonal codon (often by repurposing the UAG stop codon). 'Genetic code expansion' has been demonstrated in E coli by several groups using different approaches (Dunkelmann et al 2021, Costello et al 2024). Since it is a relatively recent field, so far, no one has published that they've replicated this in cyanobacteria, algae or any other edible organism — even though this could be a major missed opportunity if successful.
As for lipidations, it is harder to engineer specificity. Interestingly, cyanobacteria natively express enzymes called prenyltransferases. Prenylation, though only a subset of lipidation, can attach isoprenoid lipids (C15, C20) to a peptide's cysteine residues. This would not enable us to make semaglutide, but it opens doors to other peptide designs with lipid tails that can bind to albumin.
Great question, short answer is that you will struggle to make semaglutide per se. But other GLP-1 drugs like lixisenatide or exenatide could be made.
Long answer: So far, GLP-1 drugs that do not have non-canonical amino acids are far easier to turn into oral GLP-1 drugs. For example, exenatide and lixisenatide have only canonical amino acids and no lipidations. Therefore, as long as one can adequately demonstrate that there are no new post-translational modification (PTMs) introduced by the expression host, the whole cell-encapsulated oral formulation of a GLP-1 peptide could be approved, conditional on having shown satisfactory safety, toxicology and PK/PD clinically.
To make semaglutide, quite right to point out that one has to incorporate non-canonical amino acids which currently, these cells cannot perform. But it is possible. This can be done through engineering orthogonal aminoacyl–tRNA synthetase (aaRS)/tRNA pairs to assign a new non-canonical amino acid, such as Aib, to an orthogonal codon (often by repurposing the UAG stop codon). 'Genetic code expansion' has been demonstrated in E coli by several groups using different approaches (Dunkelmann et al 2021, Costello et al 2024). Since it is a relatively recent field, so far, no one has published that they've replicated this in cyanobacteria, algae or any other edible organism — even though this could be quite an interesting avenue to pursue.
This is a very well written and accessible explanation for what otherwise seems to be a very complex topic.
Imho, the real hard question is the inconsistency of bioavailability of the oral GLP1a form (in currently existing formulations) not only on an individual basis but also with each administration. Basically, the oral form currently, it seems, delivers a "whatever" actual dosage. Thus, the potential long-term consequences of such inconsistency are in question, which might range from the broad swing of the "ordinary" side effects to the serious stuff that is more prevalent, it seems, with abrupt start/stops instead of gradual tapering of injected dosages.