Case 2 · Reports 04 Quality and 05 Biosafety
Leghemoglobin is the soybean protein that makes the Impossible Burger look and taste like meat. The company makes it in engineered yeast, Pichia pastoris. Its first FDA filing stalled. The second one is the clearest public example of a specification and safety dossier for a protein made by a microbe.
Figure 1 · What leghemoglobin is, and why anyone wants it
A soybean protein that carries a heme. The heme is red, and heme is a large part of what makes cooked meat taste of meat.
Where it comes from
Soybean roots grow nodules that house nitrogen-fixing bacteria. The plant makes leghemoglobin there to control oxygen around them. Cut a working nodule open and it is pink, which is the protein.
What it is
A globin: a small protein folded into helices around one heme group. The same architecture as the haemoglobin in blood and the myoglobin in muscle, which is why it behaves like a meat protein in a pan.
The part that matters
Heme b · C₃₄H₃₂FeN₄O₄. An iron atom held in a flat ring. Extracting enough of it from plants is impractical, so the protein is made in yeast instead: 16 copies of the soybean gene, plus 8 of the yeast's own heme-pathway genes turned up.
FDA accepted the second filing, GRN 737, in July 2018. These four numbers carry most of its argument.
Minimum share of total protein that must be leghemoglobin. Five lots measured 71–86%.
Report 04 · GRN 737Independent production runs tested against every line of the specification.
Report 04 · GRN 737Highest dose with no adverse effect in rats, divided by the 90th-percentile human intake.
Report 05 · Fraser 2018Viable production strain in the product: not detected. No antibiotic-resistance genes in the strain.
Report 05 · GRN 737The protein is made inside the yeast. To get it out, the cells are broken open, so everything else the yeast makes comes along. The first filing described leghemoglobin well and said little about the rest.
The 2015 FDA memos are known only through press reports of documents released under freedom-of-information requests. Every other date is from FDA and court documents.
Figure 2 · Why the host proteins were the problem
The route from fermenter to ingredient, with the step that creates the whole regulatory question marked.
An excerpt of Table 1 in GRN 737. Each row has four parts: what is measured, the limit, how it was measured, and what five real lots gave. This is the layout to copy.
| Property | Limit | Method | Result, 5 lots |
|---|---|---|---|
| Leghemoglobin, % w/w | 6–9% | UPLC | 6.28–6.95% |
| Purity, % of total protein | ≥65% | SDS-PAGE, Coomassie, densitometry | 71, 77, 82, 85, 86% |
| Solids | ≤24% | — | 12.6–18.4% |
| pH | 6.5–8.5 | — | 6.77–7.38 |
| Lead (Pb) | <0.4 ppm | — | all <0.01 |
| Arsenic (As) | <0.05 ppm | — | ≤0.01 |
| Aerobic plate count | <10⁴ CFU/g | AOAC 990.12 | <10 |
| Salmonella | absent | AOAC 2011.03 | absent |
| Production strain, viable | not detected | — | not detected |
Values read from the scanned GRN 737 notice (pp. 8–9) and checked against FDA's response letter, which confirms the 6–9%, ≥65% and lead limits. A dash means our reading of the notice found no method named on that row; check the scanned page before quoting. The release rule underneath the table: any lot that fails any line is discarded and the cause is investigated.
Figure 3 · Five lots against one specification
A specification is only believable with real batch results beside it. This is the layout to copy.
LeghemoglobinResidual host protein
| Lot | Leghemoglobin, % of total protein | Specification |
|---|---|---|
| Lot 1 | 71% | ≥65% |
| Lot 2 | 77% | ≥65% |
| Lot 3 | 82% | ≥65% |
| Lot 4 | 85% | ≥65% |
| Lot 5 | 86% | ≥65% |
"Highly pure" has no limit and no method. "Free of contaminants" is untestable, because nobody can test for every contaminant.
Range, n, limit, method, and a plain statement of what the rest of the material is.
The safety argument is one chain of arithmetic. Anyone can redo it from the notice, which is why it persuades.
25 g of burger a day × 0.8% leghemoglobin = 200 mg/day.
90th percentile taken as twice the mean: 400 mg/day ÷ 60 kg = 6.67 mg/kg/day.
28-day feeding, no adverse effect at the top dose: 750 mg/kg/day leghemoglobin.
750 ÷ 100 safety factor = 7.5 mg/kg/day allowed, above the 6.67 eaten.
From GRN 737 Tables 2–3 and Fraser et al. 2018. A later 90-day study (Reyes et al. 2023) found no adverse effect at the highest dose either. The 2019 colour additive petition used newer intake data with higher numbers; do not mix the two sets.
Figure 4 · The safety argument in one line
Three numbers on a log scale. Anyone can redo the arithmetic from the notice, which is why it persuades.
| Quantity | Value | Where it comes from |
|---|---|---|
| Estimated intake, 90th percentile | 6.67 mg/kg/day | 25 g burger per day at 0.8%, doubled, over 60 kg |
| Acceptable daily intake | 7.5 mg/kg/day | NOAEL ÷ 100 |
| No-adverse-effect level, rats | 750 mg/kg/day | 28-day dietary study, top dose |
Host strain from a known safe lineage. Eight of the yeast's own heme-pathway genes boosted, plus 16 copies of the soybean gene. Whole genome sequenced; stable for 150–200 generations without selection; no antibiotic-resistance genes.
Cells are broken by bead mill, then removed by centrifugation and microfiltration; ultrafiltration concentrates the protein. Viable strain: not detected. Residual yeast DNA: about 0.2 mg/L, measured and stated.
"Cannot" is a promise. A reviewer will ask what was tested, how often, and what the detection limit was.
A test, a result, a count, and the leftovers stated as a number.
Our product, ACC deaminase, is secreted through a 0.2 µm membrane, so we never break cells open. That helps with host proteins. It does not prove containment: small vesicles and debris pass a 0.2 µm pore, and hollow fibres can develop pinholes.
| Property | Limit | Method | Our result |
|---|---|---|---|
| Identity | band at 37.6 kDa (mature); 41.9 kDa means unprocessed | anti-His Western, shell and lumen | from the blot record |
| Concentration | to be set | ELISA (planned) | not yet measured |
| Enzyme activity | to be set | ACC deaminase assay, α-ketobutyrate | fill in or "not yet measured" |
| Viable cells in shell fluid | no growth | 100 µL plated on LB, every sample point | x of y plates clean |
| Antibiotic-resistance genes | state them | plasmid map | name the marker on pSTK-Exp1 |
This is the starting table for the Product specification section. The Bioreactor Report builder has the same rows.