Week 3: Designing a Cell-Free Reaction Composition

A. COMPARE CELL-FREE REACTION COMPOSITIONS

A1. Why does the concentration of potassium glutamate and magnesium glutamate differ between compositions? Which is more sensitive?

The concentrations differ because each formulation was optimized for its own conditions. The lysate strain and preparation method change how much Mg²⁺ and K⁺ the extract already carries over, and the energy system matters too. Jewett & Swartz (2004) had to lower Mg²⁺ when switching from PEP to pyruvate, because PEP binds Mg²⁺ more tightly (11.3 mM total in Cytomim vs about 23 mM in PANOx-SP). Olsen et al. (2026) likewise re-tuned Mg²⁺ and phosphate for each new lysate batch. The two ions also do different jobs. K⁺ is the main cation that sets ionic strength, so it is needed at roughly 80-362 mM across formulations. Mg²⁺ is a cofactor for ribosomes and polymerases, needed at about 8-10 mM.

Magnesium glutamate is the more sensitive of the two. We can observe this in Jewett & Swartz (2004), Table II, PANOx-SP gave maximum yield at 20 mM Mg-glutamate, but yield fell to 39% at 28 mM. So in simple terms, there is a sweet spot on how much magnesium glutamate you should use, use too little and you will not be rising to the full yield potential, too much and the yield will fall. In Cytomim, yield dropped to 67% at 12 mM and 41% at 16 mM, compared with the 8 mM optimum.

Potassium glutamate tolerated far larger changes in the experiments: PANOx-SP kept 92% of maximum yield at 225 mM against a 175 mM optimum, and Cytomim kept 95% at 200 mM against a 130 mM optimum. K⁺ is still required, since Cytomim yield fell to 20% with no added K-glutamate, but it has a broad plateau while Mg²⁺ has a narrow peak.

A2. Explain how a PEP+NTP reaction formulation is different from a Ribose+NMP reaction formulation.

A PEP+NTP formulation supplies energy directly. Phosphoenolpyruvate (PEP) is a phosphorylated energy substrate that transfers its phosphate to ADP to regenerate ATP, and the NTPs (ATP, GTP, CTP, UTP) are added already phosphorylated. In PANOx-SP (Jewett & Swartz, 2004), this means 33 mM PEP plus 1.2 mM ATP and 0.85 mM each of GTP, CTP and UTP.

A ribose+NMP formulation instead relies on the lysate’s endogenous metabolism. It provides sugars (glucose and ribose) and nucleoside monophosphates (AMP, CMP, GMP, UMP), and the lysate generates the energy and converts NMPs to NTPs. Olsen et al. show that this works: their optimized formulation (RFopt) reaches ATP levels comparable to PANOx-SP within 30 minutes despite starting with no added ATP.

A3. Choose two cell-free reaction compositions from the literature, compare and contrast the choice of reagents and the concentration of reagents.

I compared PANOx-SP (Jewett & Swartz, 2004) and the optimized formulation RFopt (Olsen et al.). Both use glutamate as the main anion (following the logic that glutamate better resembles the cell’s interior than acetate), both include all 20 amino acids and a crude E. coli lysate, and both depend on tuning magnesium glutamate carefully.

While PANOx-SP acquires energy in the form of PEP and nucleotides as NTPs. RFopt uses sugars and NMPs and lets the lysate’s metabolism do the conversion.PANOx-SP also includes expensive cofactors such as NAD, CoA, folinic acid and tRNA. RFopt drops all of them except nicotinamide. PANOx-SP also adds purified T7 polymerase, whereas RFopt omits it because a BL21 Star (DE3) strain induced with IPTG is used for the lysate (so we can assume it already comes with the polymerase).

In the case of RFopt, it uses about twice the K-glutamate but 2.5 times less Mg-glutamate, and 2.5 times more of each amino acid. Its lysate fraction is also higher (30% vs 24%). RFopt makes more than double the protein at roughly 1/70th of the cost per gram, by using fewer, cheaper components (12 vs about 20).

Use of AI

AI client: Claude (Anthropic), Claude Sonnet 5.5, web chat, October 2026. I used it to understand the papers and my design choices, and to draft and pressure-test answers to Section A and the sfGFP design. I checked its claims against the papers and the coopetition interface, and I can explain and justify the design.

Prompts used (paraphrased)

Readings

“Can you help me find these readings?” (six citations). Used to confirm the papers exist and find where to read them.

Understanding the papers

Uploaded Jewett 2004 and Smith 2026: “give me the keynote version, explain the key concepts like I’m 15.” Used to get an overview of CFPS, titer vs cost, Cytomim, and the GPT-5 lab. “Doesn’t that mean pyruvate didn’t work and it was the other stuff?” Checked the AI’s explanation against the paper’s own text (pyruvate only worked together with the new environment). Pasted the oxidizing/reducing passage: “explain please.” Used to understand why reteplase needs IAM, GSSG/GSH, DsbC and Δgor while sfGFP does not. Uploaded Olsen’s preprint: “which was the ideal formulation? what protein? help me understand Table 1.” Used to identify RFopt and read the table (cost per gram = $/L ÷ g/L).

Assignment scope and Section A

Pasted the instructions: “help me understand the questions first, should I find literature to validate them?” Used to separate Section A (answered from the papers) from Section B (design). “Help me with A1-A3,” then uploaded Jewett’s Table II: “what do the numbers mean” Used to verify the Mg vs K sensitivity claim against the table. “Explain A2 more clearly.” Used to understand PEP+NTP vs ribose+NMP. “Why is there an extra phosphate if PEP gives it up?” Used to understand how PEP’s phosphate becomes free phosphate (PEP → ATP → spent ATP). “So ribose+NMP is naturally metabolized vs PEP is introduced?” Used to check my understanding. The corrected version is that PEP delivers energy directly, while ribose+NMP makes the lysate generate it.

sfGFP design

Uploaded the reference CSV: “Wouldn’t two variables be fuzzier?” Used to decide to change one thing at a time. “Why not the optimal GFP recipe as base?” Used to compare RFopt with Smith’s recipe. “GFP Ribose has no spermidine, how can that be?” Used to find that Olsen saw no benefit while Smith found only a correlation, so I treat spermidine as a hypothesis. “Wouldn’t it be simpler to test first?” Used to plan a baseline comparison.

Fitting the design to the interface

Screenshots: “is this the RFopt to change?” and “there are two versions of RFopt, which one?” Used to identify the reducing version for sfGFP. The error “Total 22,375 nL exceeds the 20,000 nL reaction”: “should I adapt it so it adds up properly and then make the changes?” This showed the AI’s first mix didn’t fit in 20 µL, so I restarted from the sfGFP Ribose reference column, which does fit.

REFERENCES AND RESOURCES

  • Smith, A. A., et al. (2026). Using a GPT-5-driven autonomous lab to optimize the cost and titer of cell-free protein synthesis (Ginkgo Bioworks × OpenAI). bioRxiv preprint. DOI: 10.64898/2026.02.05.703998. https://www.biorxiv.org/content/10.64898/2026.02.05.703998v1

  • Olsen, M. L., Copeland, C. E., Sundberg, C. A., Aw, R., Shaver, Z. M., Rao, G., Swartz, J. R., Karim, A. S., &

  • Jewett, M. C. (2026). Design-driven optimization of low-cost reagent formulations for reproducible and high-yielding cell-free gene expression. Nature Communications. DOI: 10.1038/s41467-026-69605-8.

  • Jewett, M. C., & Swartz, J. R. (2004). Mimicking the Escherichia coli cytoplasmic environment activates long-lived and efficient cell-free protein synthesis. Biotechnology and Bioengineering, 86(1), 19–26. DOI: 10.1002/bit.20026.

  • Silverman, A. D., Karim, A. S., & Jewett, M. C. (2020). Cell-free gene expression: an expanded repertoire of applications. Nature Reviews Genetics, 21, 151–170. DOI: 10.1038/s41576-019-0186-3.

  • Rasor, B. J., Vögeli, B., Landwehr, G. M., Bogart, J. W., Karim, A. S., & Jewett, M. C. (2021). Toward sustainable, cell-free biomanufacturing. Current Opinion in Biotechnology, 69, 136–144. DOI: 10.1016/j.copbio.2020.12.012.