On 21 September 2026 GenScript’s Bestzyme unit announced what it calls the first dedicated, automated sweet-protein line in the world, at Jinan in China. The release gives two numbers that matter: an annual capacity of “over 100 tons” of its Mellia brazzein, and a sweetness equivalent to “300,000 tons of sucrose annually”.
Divide one by the other and you get 3,000. Each ton of Mellia, on the company’s own arithmetic, sweetens like 3,000 tons of sugar.
That is not the number Bestzyme gave the US FDA. The notice that produced Mellia’s April 2025 “no questions” letter states the brazzein preparation is 24,000 times sweeter than sucrose. At that ratio, 100 tons would be worth 2.4 million tons of sucrose, not 300,000. The two figures for the same product differ eightfold, and a formulator sizing an inclusion rate, or a buyer pricing a kilogram, needs to know which one is real.
We covered the spread of sweetness claims across all six brazzein notices in our September analysis of the FDA’s brazzein letters. This piece adds a check on one of those six notifiers: Bestzyme’s filed potency can now be tested against its own capacity claim.
What Bestzyme said this month
The release, datelined Jinan on 21 September 2026 (also republished by BioSpace), says the new facility spans 5,000 square metres and integrates fermentation, purification and freeze-drying in one automated process. Capacity is “over 100 tons” a year and is expected to have “the potential to double within two years”. The output is described as “equivalent to the sweetness of 300,000 tons of sucrose annually”. The release also says Mellia “received a ‘No Questions Letter’ from the U.S. FDA in April 2025” and has “completed its ‘0 to 1’ commercial validation”, with a US launch through an unnamed sparkling water brand.
vegconomist and iGrowNews repeat the 100-ton and 300,000-ton figures. vegconomist adds that an application for market authorisation was filed with China’s National Health Commission in the second half of 2025, and that Bestzyme markets Mellia to formulators as a component of sweetener systems, blended with stevia and monk fruit. The release itself describes brazzein as “thousands of times sweeter than sucrose”, which is consistent with 3,000 and much less so with 24,000.
What Bestzyme told the FDA
GRN 1207 was filed by Nanjing Bestzyme Bio-Engineering Co. Ltd. The FDA’s inventory records the notice as closed on 10 April 2025, with a corrected response letter dated 23 May 2025. The substance is a brazzein preparation made by Aspergillus oryzae 90402 carrying the brazzein gene from Pentadiplandra brazzeana, specified at 95% or more brazzein on a dry basis. The intended use is as a general-purpose sweetener at levels consistent with good manufacturing practice.
The letter describes the preparation as “24,000 times sweeter than sucrose”. It then converts that into dietary exposure: between 0.011 and 0.028 mg/kg body weight a day for non-diabetic adults, and up to 0.038 mg/kg for diabetic children. As we set out in the earlier piece, every brazzein notifier used the same substitution method, so a higher sweetness claim mechanically produces a lower exposure estimate. Bestzyme’s is the lowest of the six letters by a wide margin.
The FDA’s own statement is narrow: it had no questions at that time about the notifier’s conclusion that the preparation is GRAS under its intended conditions of use. The agency was not asked to verify the potency.
Neither Bestzyme’s 26 April 2025 release nor ConfectioneryNews’s coverage of the clearance gave a sweetness multiple or a production capacity. They described a zero-calorie, general-purpose sweetener aimed at applications such as confectionery. The September release is therefore the first document we found that puts a capacity and a sweetness equivalent side by side.
| Figure | Value | Where it comes from |
|---|---|---|
| Relative sweetness in FDA notice | 24,000x sucrose | GRN 1207 letter, 23 May 2025 |
| Stated annual capacity | over 100 tons | Bestzyme release, 21 Sep 2026 |
| Stated sweetness equivalence of that output | 300,000 tons of sucrose | Same release |
| Implied relative sweetness | 3,000x (our arithmetic: 300,000 ÷ 100) | Derived |
| Sucrose equivalent of 100 tons at 24,000x | 2,400,000 tons (our arithmetic) | Derived |
| Ratio of the two | 8x (24,000 ÷ 3,000) | Derived |
Practical consequence: if you are asked to qualify Mellia, ask which figure applies to the product being shipped, and ask for the sucrose reference concentration behind it. To replace 1,000 kg of sucrose, the two numbers mean roughly 0.33 kg of protein or 0.04 kg. Whatever the price per kilogram, the cost of sweetening a product moves by the same eightfold factor.
Four ways both numbers could be honest
The mismatch does not show either number is wrong. We can name four explanations, none of which the company has stated.
A different reference concentration. Protein sweeteners lose relative potency as the sucrose they are compared against gets sweeter. A 24,000x figure measured against a very dilute sucrose solution and a 3,000x figure measured against beverage-strength sugar could both be accurate. No brazzein letter states its reference concentration, and the Jinan release does not either.
A different product. The FDA notice covers a preparation that is 95% or more brazzein. The line’s “over 100 tons” is described as Mellia sweet protein product. If what leaves the line is a standardised or diluted ingredient rather than near-pure protein, the sucrose equivalent per ton would be lower than the pure-protein figure. The release does not state the purity of the output.
Rounding and marketing convention. “Over 100 tons” and “300,000 tons” are round numbers. Even if capacity is 110 tons, the implied multiple stays under 3,000. Rounding does not close an eightfold gap.
A conservative commercial claim. A company might quote a lower potency to buyers than it used in a regulatory filing that benefits from a high figure, or the reverse. Nothing in the record shows which.
Only one of these, the reference concentration, can be tested from outside, and only by the company or by a buyer running a sensory panel.
Why it matters for a capacity buyer
Sweet proteins have the shortest route to revenue of any precision-fermented ingredient because the dose is tiny. That advantage rests wholly on the potency figure. At 3,000x, 100 tons a year supplies the sweetness of 300,000 tons of sugar, and at the stated doubling, 200 tons supplies 600,000. At 24,000x the same tonnage would supply eight times as much. The capacity needed to serve a given customer, the price per kilogram that is competitive with stevia and monk fruit, and the amount of product a customer must buy all scale with the number you choose.
We made a similar point about protein fermentation generally in our piece on basis mismatch in cost-parity claims, and about tonnes in the mycoprotein capacity comparison. Capacity in tons of what, at what potency or protein content, is the question that recurs.
There is also the contrast with actual demand. A line described as capable of 100 tons has been announced alongside one named commercial application, an unnamed sparkling water. The release does not disclose volumes shipped, customers or price. Compare the pattern in Calysseo’s 20,000-tonne plant, where announced capacity and booked revenue sat far apart. We are not suggesting the same here; we simply have no data on sales.
The counter-argument
Bestzyme’s defenders would say the release is a marketing document, not a filing, and that “thousands of times sweeter” is deliberately loose. Rounding a conservative sweetness equivalent for a general audience is ordinary practice. The regulatory number, by contrast, is a technical figure with a specified sample and method, and the FDA’s review turned on safety, not potency. On that reading there is no contradiction, only two documents written for two audiences.
That reading is plausible. It leaves the buyer where the earlier analysis left them: two figures from one company for one protein, no stated reference concentration for either, and a procurement decision that depends on the ratio. It also does not explain why the filing carries the highest potency figure among the six brazzein notifiers, which in turn yields the lowest exposure estimate.
What we could not establish
- The reference sucrose concentration for either the 24,000x or the 3,000x figure. Neither document states it.
- The purity or specification of the line’s output. The release says Mellia sweet protein product; the FDA notice specifies 95% or more brazzein. We have not confirmed they are the same specification.
- Whether the Jinan line and the notifier are the same legal entity. The FDA notifier is Nanjing Bestzyme Bio-Engineering Co. Ltd.; the release refers to Bestzyme, a GenScript subsidiary, and the April 2025 release describes Mellia as the product that received the letter. The release does not cite GRN 1207 by number. We have treated them as the same product on that basis.
- The 300,000-ton figure’s method. The release gives no calculation, so the 3,000x implied figure is our division, not a stated potency.
- Actual output. “Annual capacity” is a design figure. Nothing we read reports tonnes produced or sold.
- Price. No price per kilogram for Mellia is public in anything we found.
- The US customer. The sparkling water brand is unnamed.
- The status of the China filing. vegconomist reports an application to the National Health Commission in the second half of 2025. The Bestzyme release we read does not mention it, and we have found no decision.
What to watch
- A restated sweetness figure with a reference sucrose concentration from Bestzyme. If it publishes one, the eightfold gap either closes or becomes explicable.
- The named US customer and any disclosed volumes. A first commercial launch would turn capacity into demand.
- The National Health Commission decision on the filing.
- Whether the doubling to roughly 200 tons within two years is announced as built or as a target.
- Other Chinese brazzein notifiers still at FDA. Our earlier piece flagged a pending notice from another Chinese company, whose stated sweetness will show whether 24,000x is a regional convention or a single notifier’s choice.