Brazzein is a 53- or 54-amino-acid protein from the fruit of a West African vine. It is sweet, heat-stable and, since March 2024, the subject of six FDA “no questions” letters from six different notifiers, three of them issued this summer. A seventh notice is pending. That makes it one of the most-notified precision-fermented food ingredients in the GRAS inventory, and a reasonable proxy for how the notification system handles a molecule several companies make at once.

We read every letter. They agree on the safety case, the toxicology study and the exposure method. They do not agree on how sweet the thing is. Oobli told the FDA its preparation is 330 times sweeter than sucrose, later revised to 281 times for its major-isoform product. Perfect Day said 500. Sweegen said 1,125. GreenLab said 1,200. Nanjing Bestzyme said 24,000.

Because every notifier used the same substitution method, in which the sweetener’s estimated intake is sucrose-equivalent intake divided by relative sweetness, the exposure estimates scale inversely with that number. Oobli’s 90th-percentile adult exposure is 2.05 mg/kg bw/d. Bestzyme’s is 0.028 mg/kg bw/d. That is a 73-fold difference for the same protein, at the same intended use, in the same regulatory inventory.

For a toxicologist this is a curiosity; the safety margins are large under every assumption. For a formulator it is the whole question. Relative sweetness is the inclusion rate, and the inclusion rate multiplied by the price per kilogram is the cost of sweetening a product. Six letters give six different answers to that, and none of the trade coverage of the summer’s “approvals” has noticed.

The inventory, assembled

The GRAS inventory now lists fifteen sweet-protein notices across four proteins. Eleven have closed with no questions; four are pending.

GRN Protein Production system Notifier Received → filed → closed Outcome
738 Thaumatin (plant expression; host not stated in inventory) Nomad Bioscience 18 Oct 2017 → 8 Nov 2017 → 18 Apr 2018 No questions; sweetener, 5–400 mg/kg by category
910 Thaumatin II (Nomad, “non-edible host”) Nomad Bioscience 4 Feb 2020 → 12 Mar 2020 → 9 Sep 2020 No questions; sweetener, 5–400 ppm
920 Thaumatin II (as 910) Nomad Bioscience 23 Mar 2020 → 18 May 2020 → 24 Nov 2020 No questions; flavour modifier, 1–150 ppm
1142 Brazzein K. phaffii P-BRZ-013 Oobli 18 Apr 2023 → 21 Aug 2023 → 11 Mar 2024 No questions; general-purpose sweetener at GMP. Supplement 2 Sep 2025 adds strain P-BRZ-336 (major isoform)
1167 Brazzein K. phaffii Perfect Day 23 Oct 2023 → 16 Jan 2024 → 8 Oct 2024 No questions; general-purpose sweetener
1183 Monellin (modified) K. phaffii P-MON-040 Oobli 28 Feb 2024 → 30 May 2024 → 22 Nov 2024 No questions; general-purpose sweetener
1207 Brazzein Aspergillus oryzae 90402 Nanjing Bestzyme 2 Aug 2024 → 25 Sep 2024 → 10 Apr 2025 (corrected 23 May 2025) No questions; general-purpose sweetener
1269 Monellin (single-chain MNEI) K. phaffii CBS 150005 Amai Proteins 6 May 2025 → 7 Aug 2025 → 4 Feb 2026 (corrected 5 Feb) No questions; general-purpose sweetener
1286 Brazzein Zea mays (corn) GreenLab 11 Jun 2025 → 1 Dec 2025 → 30 Jun 2026 No questions; general-purpose sweetener
1293 Brazzein K. phaffii RA-b Sweegen 23 Jul 2025 → 27 Feb 2026 → 23 Jun 2026 No questions; general-purpose sweetener
1295 Brazzein K. phaffii GS115 MicroFarmtory closed 23 Jun 2026 No questions (letter not retrieved; see below)
1305 Brazzein K. phaffii GDMCC 70101 Inner Mongolia Kingdomway no closure date; letter pending Pending
1319 Thaumatin II K. phaffii P-THM-354 Oobli no closure date; letter pending Pending; sweetener and flavour enhancer
1320 Thaumatin II (Nomad) Nomad Bioscience no closure date; letter pending Pending; sweetener up to 10 mg/kg by category
1330 Mycodulcein K. phaffii DSM 35637 MycoTechnology no closure date; letter pending Pending; general-purpose sweetener

Three structural points fall out of the table before the sweetness question.

First, the notifier list is not the one the trade press covers. Oobli, Sweegen and Amai issue press releases. Perfect Day — better known for recombinant whey — holds a brazzein letter that predates Sweegen’s by twenty months and has said little about it. Nanjing Bestzyme, GreenLab and MicroFarmtory hold letters and have, as far as we can find, issued nothing. Inner Mongolia Kingdomway Pharmaceutical has a notice pending. As with HMOs, the notification record and the marketing record describe different industries.

Second, filing-to-closure has stretched. Oobli’s first brazzein notice took 203 days from filing to letter. Sweegen’s took 116 days from filing, but 335 days from receipt, because the notice sat 219 days between receipt and filing. GreenLab’s sat 173 days before filing. The FDA does not explain pre-filing intervals in its letters.

Third, the intended use converged. Oobli’s original 2023 notice, as the inventory still records it, listed maximum levels “from 2 to 99 mg/100 g” across specific categories. The letter that closed it, after amendments in December 2023 and February 2024, describes “general-purpose sweetener in food at levels determined by current good manufacturing practices”. Every brazzein notice since has used the general-purpose formulation. The category-by-category dosing that Nomad’s thaumatin notices still carry, and that its pending GRN 1320 carries at “up to 10 mg/kg”, has left the brazzein record entirely. A formulator looking to the GRAS letter for a use level will not find one.

Six sweetness numbers for one protein

Here is what each notifier told the FDA, as recorded in the agency’s letters, and the resulting exposure estimates. Every notifier used Renwick (2008), a method that takes measured intakes of intense sweeteners, assumes the new sweetener substitutes for them, and converts sucrose-equivalents to milligrams by dividing by relative sweetness.

GRN / notifier Stated relative sweetness Preparation purity stated Adult exposure, mean / 90th pct (mg/kg bw/d) Child (diabetic), mean / high
1142 Oobli (2024) 330x (preparation) ~80% protein, ~40% of it brazzein 0.77 / 2.05 2.04 / 2.75
1142 supplement, Oobli (2025) major isoform 550x; preparation 281x (51% major isoform) as above 0.91 / 2.41 (major isoform) 2.40 / 3.24
1167 Perfect Day 500x ≥90% protein, ≥90% of it brazzein 0.51 / 1.35 1.34 / 1.82
1207 Nanjing Bestzyme 24,000x ≥95% brazzein, dry basis 0.011 / 0.028 0.028 / 0.038
1286 GreenLab 1,200x ≥80% brazzein, dry basis 0.21 / 0.56 0.56 / 0.76
1293 Sweegen 1,125x (brazzein), ~92% content ≥90% protein, ≥90% of it brazzein 0.21 / 0.55 0.55 / 0.74

Some of the spread is composition. Oobli’s 330x is for a preparation in which brazzein is roughly a third of the mass; its own supplement puts the pure major isoform at 550x, and Sweegen’s 1,125x is stated for brazzein itself before adjusting for 92% content. Read that way, the “pure protein” figures cluster: 550, 1,125, 1,200 — a two-fold band, which is what one would expect from different isoforms, different reference concentrations and different panels.

Bestzyme’s 24,000x does not sit in that band. It is 20 to 44 times the other pure-protein figures, for a preparation the letter specifies at ≥95% brazzein. The letter records the number without comment, as it records the others; the FDA’s review is of the safety conclusion, and a higher sweetness claim produces a lower exposure estimate and therefore a larger margin. Nothing in the process penalises optimism about potency.

We are not in a position to say which figure is right. Relative sweetness is concentration-dependent — a protein sweetener’s potency relative to sucrose falls as the sucrose reference rises — and Amai’s monellin letter makes the point explicitly, stating its figure as “sweeter than a 6% sucrose solution”. No brazzein letter states the reference concentration. A 24,000x figure measured against a threshold-level sucrose solution and a 330x figure measured against a beverage-strength one could both be honest.

Practical consequence: do not take a sweetness figure from a supplier’s regulatory summary. Ask for the reference sucrose concentration and the panel method, and run the substitution at your own product’s sugar level. At a cola’s roughly 10% sugar equivalent, the difference between a 500x and a 1,200x ingredient is 2.4 times the inclusion rate, and the same multiple on cost.

What the letters share

The safety architecture is identical across the six brazzein letters, which is why the sweetness divergence stands out.

Each cites prior human consumption of Pentadiplandra brazzeana fruit “in endemic regions of Africa”. Each cites a 90-day subchronic oral toxicity study — Oobli, Perfect Day and Bestzyme “a published” one; Sweegen its own repeated-dose study with a NOAEL of “2,000 mg/kg bw/d, the highest dose tested”. Each K. phaffii notifier incorporates the earlier GRAS conclusions for proteins from that host, and the letters cross-reference each other: Bestzyme “states that the subject brazzein preparation is similar to the subject of GRN 001142”; GreenLab “incorporates relevant safety information from GRN 001142”; Sweegen incorporates GRNs 001142 and 001167. The first notifier carried the evidentiary load. The five that followed leaned on it.

That is how the system is designed to work, and it is efficient. It also means the FDA has never had to adjudicate the sweetness question, because no notifier’s safety case depends on it.

The other three proteins

Thaumatin’s record is older and more specific. Nomad’s GRN 910 letter records thaumatin II as “2000-3000 times sweeter than sucrose” and an intended use with per-category limits of 5 to 400 ppm; its GRN 920, closed two months later, covers the same protein as a flavour modifier at 1 to 150 ppm. Two notices, one protein, two functions, two dose ranges — the split between sweetener and flavour that sweet-protein marketing now leans on is visible in the 2020 record. Oobli’s pending GRN 1319 asks for both functions in a single notice.

Monellin has two letters. Oobli’s (GRN 1183) states 2,700x and 63% monellin content; Amai’s (GRN 1269) states “approximately 3,500 times sweeter than a 6% sucrose solution” at “approximately 83% monellin protein”, and is the only sweet-protein letter we read that specifies a reference concentration. Amai’s exposure estimate — 0.09 / 0.23 mg/kg bw/d for adults — is the lowest in the set after Bestzyme’s.

Mycodulcein, a protein from the fungus Mattirolomyces terfezioides, is MycoTechnology’s GRN 1330, pending, general-purpose sweetener. The inventory does not yet link a notice document.

What this means for a fermentation-capacity buyer

Sweet proteins are the precision-fermentation product with the shortest route to revenue: use levels are tens of milligrams per kilogram of food, so a single 50,000-litre campaign covers a great deal of product. That arithmetic depends entirely on the sweetness figure. At 24,000x, a tonne of brazzein replaces 24,000 tonnes of sugar-equivalent; at 500x, 500 tonnes. The capacity a company needs to book, and the price it can charge per kilogram, differ by the same 48-fold factor.

We have argued before that the sector’s figures are incomparable because nobody states the basis. The GRAS record is the cleanest demonstration yet. Six companies, one molecule, one regulator, one method, and a 73-fold spread in the number that drives the economics.

Counter-argument

A reasonable reading is that the spread is noise the market will remove. Buyers will run their own panels, the suppliers with inflated figures will lose the bake-off, and the GRAS letters will remain what they are meant to be: safety documents. Relative sweetness is not a regulatory quantity and it was never meant to be checked by the FDA.

There is force in that, with two caveats. First, procurement in mid-sized food companies does read the GRAS letter, because it is the one public document about the ingredient, and it is where the “24,000x” figure is now permanently recorded with the agency’s letterhead above it. Second, the same figure feeds the exposure estimate, and the exposure estimate is the number a competent authority in another jurisdiction will compare against an ADI if one is ever set. A sweetness overstatement that lowers estimated exposure is not neutral in every regulatory system, even if it is in this one.

What we could not establish

  • MicroFarmtory’s GRN 1295 letter. The inventory lists it as closed with no questions on 23 June 2026, the same day as Sweegen’s. The response-letter link returned an error on every attempt we made, with and without a browser identity. We could not read its sweetness figure or exposure estimate, so the table has six brazzein letters and five sets of numbers.
  • The reference sucrose concentration behind any brazzein sweetness figure. No brazzein letter states it.
  • Whether Bestzyme’s 24,000x is a transcription of a literature value for a purified isoform at threshold concentration, or its own measurement. The letter attributes the figure to the notifier and gives no method.
  • Why Sweegen’s press release is dated 25 June when the FDA letter closed the notice on 23 June. The release was posted to Sweegen’s site on 6 August. A two-day gap is unremarkable; we note it because trade reports carry the 25 June date as the “approval” date.
  • The identity of GRN 738’s expression host. The inventory entry does not name it; the GRN 910 letter describes a “non-edible host”. We have left the cell blank rather than guess.
  • Commercial prices for any of the six brazzein preparations. None is public.

What to watch

  1. GRN 1305, Kingdomway. The second brazzein notice from a Chinese notifier. Its stated sweetness figure, when the letter issues, will show whether 24,000x was one notifier’s choice or a regional convention.
  2. GRN 1320, Nomad. The only pending brazzein-era notice that still asks for per-category limits, at up to 10 mg/kg. If the FDA closes it on those terms, the category-dosed route remains open alongside general-purpose.
  3. Any notifier restating relative sweetness with a reference concentration. Amai has done it for monellin. The first brazzein notifier to do so sets the comparable the rest will be held to.
  4. The FDA’s proposed rule on mandatory GRAS notification, which we covered on 20 August. If self-affirmed routes close, every sweet-protein launch will pass through this inventory, and the sweetness figures in it will become the industry’s reference set by default.