Week 1 HW: Za'atar One — Biomanufacturing Carvacrol

Za’atar One

زعتر — choosing which isomer the za’atar pathway makes.

Thymol and carvacrol are the same molecule with the hydroxyl in a different place. One cytochrome P450 decides which. Nobody has learned to tell it what to do and perhaps Origanum syriacum, the za’atar of the Levant, already has the answer.

1. What I want to biomanufacture, and why

Carvacrol - made selectively, not as a mixture.

Carvacrol (C₁₀H₁₄O, CAS 499-75-2) is the phenolic monoterpene that gives Origanum syriacum (AKA za’atar) its smell, its bite and its antimicrobial activity. Its regioisomer thymol has the same formula and the same biosynthetic route, differing only in where the hydroxyl sits relative to the isopropyl group.

Carvacrol has previously been produced in microbes, back in 2025, in Yarrowia lipolytica. I want to make try to poduce one of the isomers, e.g. carvacrol rather than thymol, since there has never been a report controlling the ratio.

Why it matters to me. Za’atar in Lebanon is a spice we eat mainly at breakfast. Origanum syriacum is part of the Levantine heritage, and it can represent us lebanses more than any other plant product from our country, including the cedar on our flag. Wild za’atar is under harvest pressure across the Levant, and the commercial molecule that bears its chemistry has nothing to do with the plant.

Why it matters industrially. Carvacrol is a veterinary antibiotic replacement. Antibiotic growth promoters are banned in the EU and US and since 2020 in China, and phytogenic feed additives filled the gap, standardised on carvacrol and thymol as marker compounds. An additive tested against the antibiotic avilamycin in Cobb 500 broilers was formulated at 10% carvacrol + 10% cinnamaldehyde, dosed at 100 g per tonne of feed (PMID 40342754). And the biggest painful fact is that carvacrol sold today is petrochemical, made from p-cymene by sulfonation and alkali fusion.

Who it is for: feed formulators needing standardised, traceable, natural-label carvacrol without depending on wild-harvested oregano — and anyone who needs to steer a plant P450’s regiochemistry.

What is known, and what is left

Row three is my project.

2. What living system I would use

Two stages, because the enzyme used and the product are different problems.

Stage A — the system that makes the catalyst. E. coli cell-free lysate on a digital-microfluidic cartridge for screening; E. coli or Y. lipolytica to scale up whichever constructs win.

• In: synthetic linear DNA per variant, lysate, energy mix, and — for the P450 — heme, NADPH, O₂ and a cytochrome P450 reductase partner.
• Out: soluble full-length enzyme, reported by a split-GFP tag that only fluoresces on full-length soluble protein.

Stage B(i) — whole-cell fermentation, the manufacturing platform. Y. lipolytica, following the published route.

• In: glucose (or a glycerol/lipid feed, exploiting Yarrowia’s oleaginous metabolism), minerals, O₂.
• Process: glucose → acetyl-CoA → mevalonate → GPP. TvTPS2 cyclises GPP to γ-terpinene, releasing pyrophosphate. TvCYP71D507 hydroxylates it at C-3 (→ thymol) or C-6 (→ carvacrol), consuming NADPH and O₂, giving an unstable cyclohexadienol. TvSDR1 oxidises that to a dienone, which aromatises spontaneously to the phenol.
• Out: carvacrol + thymol (a mixture — see §5), plus p-cymene, a dead-end byproduct formed when the dienol dehydrates before the SDR reaches it.

3. How I would use automation

Primary — programmable digital microfluidics with cell-free expression. Synthetic DNA with no cloning, with a split-GFP soluble-yield readout and overnight scale-up of hits to µg–mg. Its membrane-protein workflow adds cofactor and redox screening, which matters because CYP71D needs a reductase partner and NADPH.

Secondary — Opentrons liquid handling for the assays. Terpene synthases release pyrophosphate, detectable colorimetrically (PMID 29438678, 24681801). The selectivity assay needs each sample split into a pH 9 well and a pH 12 well with a shared incubation clock — exactly what a liquid handler removes variance from.

4. How much, and how long

How much. Carvacrol’s largest market by volume is poultry feed additives, so I scale to one year of Lebanon’s broiler flock: 130,000 t meat × 1.6 kg feed/kg meat × 10 g carvacrol/t feed = 2.1 t/yr (dose from PMID 40342754; meat figure from a 2008 review, so pre-crisis and dated). Small enough that this is a specialty problem about consistency and traceability, not tonnage.

How long. At the published 61.31 mg/L a 200 m³ fermenter makes 12.3 kg per 7-day batch, so 2.1 t is 170 batches — 3.3 years in one vessel, or ~10 months in four running in parallel. At the ~200 mg/L target it is 52 batches: one year, one vessel — which is why 3.3× is the number worth engineering for.

And the prototyping platform is a different machine on a different clock. The cartridge answers which isomer across 192 conditions in ~24 h, while the microreactor at the only measured immobilised-P450 turnover (20.3 fmol/min, PMID 30073515) makes 1.6 µg per reactor-year — twelve orders of magnitude short of manufacturing. It produces a number; only the fermenter produces tonnes.

5. What I don’t know yet

Q1 — Does the P450 sequence set the isomer ratio, or does the expression environment? Known: C-3 → thymol, C-6 → carvacrol, and D180/D181 track the carvacrol chemotype — but the isoforms leak the other product, the deciding residues are unstated, and D507 changed regiochemistry between yeast and tobacco (PMID 34930840). Measure: carvacrol:thymol across D179 / D180 / D181 / D507 / an I363F revert, by the two-pH gold assay — 550 nm, carvacrol at pH 9, thymol by difference (PMID 29968089). Gate: NADPH also reduces Au(III), so the extract matrix must validate first.

Q2 — Does the physical distance between the P450 and the SDR control the p-cymene byproduct? The intermediate dehydrates to p-cymene unless the SDR intercepts it, and the authors hypothesise the ratio depends on interenzyme distance — untestable in a cell (PMID 34930840). Measure: phenol (A550, pH 12) ÷ γ-terpinene made (PPi) across CYP–SDR fusions of varied linker length. p-Cymene has no hydroxyl, so a phenol assay is blind to it by construction.

Q3 — What does NADPH cost once the enzymes leave the cell? In a cell regeneration is free; in a reactor the P450 consumes it stoichiometrically, and the 2025 authors blame their own plateau on cofactor supply. Unverified: if the SDR runs oxidatively, part of the demand may balance internally. Measure: moles NADPH per mole carvacrol, then a regeneration module (PMID 39908731) — precedent being a cell-free plant monoterpene cascade on orthogonal regeneration (PMID 35355836).

Why both platforms coexist: monoterpenes poison the host that makes them — 1.052 g/L of 1,8-cineole but only 0.605 g/L limonene on product toxicity (PMID 36397160), and phenolic monoterpenes are worse. In an immobilised system that constraint is not mitigated, it is absent.

6. The plan

Timeline, with two hard gates. Weeks 1–2 design and freeze the 24-construct panel; week 3 order synthesis — slipping either gate by two weeks costs the activity work. Weeks 4–5 Opentrons method development including the matrix-blank ladder that decides whether the colorimetric screen is viable at all. Week 6 cartridge run. Week 7 scale up and purify. Week 8 NADPH-coupled turnover. Weeks 9–10 selectivity screen, HPLC confirmation, fusion constructs. Weeks 11–12 write-up.

AI use disclosure

Clients: Claude Science (Anthropic), model Claude Opus 5, session of 22 September 2026, for the literature retrieval, calculations and drafting. Pippit AI for the cover banner, from a prompt written in that session describing dried Levantine oregano beside a microplate holding one wine-red and one colourless well.

Form of this disclosure. The session was iterative and my individual messages were short, because each rested on reading I had already done outside the conversation. Quoting them verbatim would hide that prior research and make one-line follow-ups look like the whole specification. So what follows is the consolidated brief — the instructions actually in force.

Help me write HTBAA Homework #1. Lebanese background. Platform: custom DNA
synthesis, cell-free expression on a digital-microfluidic cartridge (~24
constructs x 8 conditions, split-GFP soluble-yield readout, ug-mg overnight
scale-up), an Opentrons, a plate reader. ONE cartridge run; due December 2026.

Answer all five assignment questions in detail: what to make and why (cultural
motivation is allowed); the chassis with inputs, process and outputs, my
assumptions named as assumptions; the automation and which step it automates; a
back-of-envelope calculation with units and sources for how much and how long,
treating prototyping and manufacturing platforms separately; and the 2-3 biggest
unknowns with what I would measure.

Find a target anchored in the history of present-day Lebanon. Score candidates on
an explicit rubric. A classmate took alpha-pinene and the Egyptian-cedar framing -
avoid both. Criteria in order: a named enzyme with a published failure mode; a
plate-reader readout; eleven weeks, staged so it cannot fully fail; real cultural
authenticity; still industrially wanted for a non-nostalgic reason.

Method rules, which matter more than the conclusions:
1. Verify by live retrieval - no PMID, accession, titer or pKa from memory.
2. Check prior work academically AND commercially; say what framing I must drop.
3. Tell me plainly when my premise is wrong.
4. Quote the key paper's own stated future work and bottleneck, and map my
   project against it item by item. An author-stated gap beats one I assert.
5. Do not assume instrumentation - if the obvious method is GC-MS, find
   plate-reader, colorimetric and electrochemical alternatives.
6. Run the arithmetic as code, flag dated inputs, then invert it to give me the
   engineering target that would change the answer.
7. Treat anything I paste from another AI as claims to check, not sources.
8. Separate what is known from what is open, even if it makes me less novel.

The redirections that were mine. I raised that a Boston company already produces an earlier candidate molecule, forcing the commercial-precedent check; rejected two candidate targets after seeing the evidence, which is how the search reached za’atar; asked why both isomers are co-produced, which prompted retrieving the PNAS full text and yielded the C-3/C-6 mechanism and the p-cymene problem; and pointed out that GC-MS is outside this course’s scope, supplying the colorimetric reference myself — which is why the primary screen is now a plate reader. I did factcheck and modify all the texts that were AI-generated.

The image generation prompt was the following:

Make a wide cinematic banner, 16:9, editorial science photography. Left third: a close-up
of dried Levantine oregano — sage-green leaves, crushed stems, a scatter of dark
red sumac and pale sesame — in warm olive, terracotta and sand, lit by soft raking
daylight, shallow depth of field so the fine leaf hairs catch the light. Right
two-thirds: a black anodised microplate seen at a low three-quarter angle, two
adjacent wells in sharp focus — the left well a deep translucent wine red, the
right well almost colourless pale straw — the remaining wells falling out of focus
into darkness. A single clear droplet hangs from a fine glass pipette tip above the
wine-coloured well. The herb's dusty texture dissolves into the plate's grid across
the middle of the frame. Palette: olive green, terracotta, sand, deep wine, one
cool steel highlight on the pipette. Mood: quiet, precise, restrained — a journal
cover, not an advertisement. Photographic realism, natural materials, hard shadows.
No text, no lettering, no numbers, no logos, no molecular diagrams, no glowing
helices, no blue-teal laboratory cliché.

References

1. (2021) The biosynthesis of thymol, carvacrol, and thymohydroquinone in Lamiaceae proceeds via cytochrome P450s and a short-chain dehydrogenase. PNAS. PMID 34930840. — the pathway, and the C-3/C-6 regiospecificity this project turns on
2. (2025) Engineering Yarrowia lipolytica for co-production of the phenolic monoterpenes thymol and carvacrol. Microbial Cell Factories. PMID 41024186. — the only microbial production of these compounds, and the source of the 61.31 mg/L baseline
3. (2018) Extraction-free colorimetric determination of thymol and carvacrol isomers in essential oils by pH-dependent formation of gold nanoparticles. Microchimica Acta. PMID 29968089. — the two-pH plate assay that replaces GC-MS.
4. (2018) Digital microfluidic immobilized cytochrome P450 reactors with integrated inkjet-printed microheaters for droplet-based drug metabolism research. Anal Bioanal Chem. PMID 30073515. — immobilised cytochrome P450 in digital microfluidics, and the source of the 20.3 fmol/min turnover used in §4
5. (2026) Membrane composition influences expression yield of plant cytochrome P450s in E. coli lysate. ACS Synth Biol. PMID 42081616. — evidence that the membrane is a pipettable variable in cell-free expression
6. Karim, A. S. & Jewett, M. C. (2021) Toward sustainable, cell-free biomanufacturing. Curr Opin Biotechnol 69, 136–144. doi:10.1016/j.copbio.2020.12.012 — course reading
7. Petzold, C. J. & Mukhopadhyay, A. (2025) From bench to biofactory: high-throughput technologies and automated workflows to accelerate biomanufacturing. Curr Opin Biotechnol 94, 103320. doi:10.1016/j.copbio.2025.103320 — course reading

Week 2 HW: DNA Design

Week 3 HW: Cell-Free Protocols

Week 4 LAB: Cell-Free Manufacturing

Final Project