Jamie Dixon — HTBAA Fall 2026
About me
Colorado Sweet Potato King
Contact info
Homework, Labs, and Projects
See the nav bar on the left to look through my work for How To Biomanufacture Almost Anything (Fall 2026)!
Subsections of Jamie Dixon — HTBAA Fall 2026
Week 1 HW: Final Project Ideas
HTBAA Homework #1 Due: Tuesday September 22, by 2:30p ET
For your first homework, propose and describe one idea for something you would like to biomanufacture. This could be medicine, material, food, energy, an art, design, or architectural artifact, or beyond. Draw from your own background and interests for inspiration.
(1) What you would like to biomanufacture and why? Describe the target product or output you would like to make.
I want to create a Coffee Cherry Borer resistant tree.
Why are you interested in making it? Your motivation might be utilitarian—for example, addressing a problem in health, sustainability, food, energy, or manufacturing—but it does not need to be. Your project might instead have a social, cultural, artistic, or personal motivation.
I own a coffee farm in the Kona coffee belt on Hawaii. One of the bigger problems I face is the pest called coffee cherry borer
https://www.ctahr.hawaii.edu/site/cbbmanage.aspx
I would love to accelerate Mendal and create a cherry that does not taste good to the beatle.
(2) What living system would you use? Propose at least one living-system manufacturing chassis that could produce your target—for example, a microbial, mammalian, or cell-free system.
The chasis would be a tree itself
At a high level, describe how the system would make your target product. What goes into the system? What biological process or processes occur? What comes out?
I would edit the genome in tissue culture (CRISPER-CAS9?), regenerate one engineered plantlet, then clonally propagate that exact tree via grafting/cuttings.
(3) How would you use automation? Identify at least one automation technology that you would combine with your living system.
I would use liquid-handling robots preparing CRISPR reagents and applying them to many callus/explant samples in parallel, then automated PCR-based genotyping to screen which calli actually took up the edit correctly
I could use robotic arms with fine scalpels or laser dissection for explant transfer, automated subculturing systems (e.g. NuPlant’s SmartClone), and machine-vision quality control that sorts healthy vs. struggling plantlets
If available I could use a commercial robotic grafting system built specifically for woody plants: it 3D-scans the rootstock and scion, calculates the optimal cut, and assembles the graft with controlled pressure
(4) How much would you need to make, and how long would it take?
Perform a back-of-the-envelope calculation estimating what it would take to manufacture your target at a meaningful scale.
(a) How much?
I would create enough viables shoots to test in 1/4 acre of my farm - appx 200 trees
(b) How long?
No idea - I would need the TAs help in estimating that
- What don’t you know yet? After completing your analysis, identify the 2–3 biggest technical unknowns, bottlenecks, or assumptions in your proposed biomanufacturing system.
If a viable DNA strand can be identified and created
If the CRSPER-CAS9 technique will work when intorducing it into the tree
If the social factors about bio-engineered crops will allow testing on my farm
AI
AI Prompts (Claude Sonnet 5.0)
I want to create a Cherry Beatle Borer resistant tree. what kind of living system would that be? create a seed whose DNA is altered?
how would I use automation to edit the genome in tissue culture, regenerate one engineered plantlet, then clonally propagate that exact tree via grafting/cuttings
Paused Start a new session Sonnet 5’s safeguards flagged this message. This sometimes happens with safe, normal conversations. Switching models won’t resume this session, so continue in a new one. Send feedback or learn more.
how would I use automation to edit the genome in tissue culture, regenerate one engineered plantlet, then clonally propagate that exact tree via grafting/cuttings
Paused Start a new session Sonnet 5’s safeguards flagged this message. This sometimes happens with safe, normal conversations. Switching models won’t resume this session, so continue in a new one. Send feedback or learn more.
Papers
https://www.nature.com/articles/s41597-025-05658-6
https://academic.oup.com/g3journal/article/15/1/jkae262/7900928
https://coffee-genome-hub.southgreen.fr
https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1215012/
Week 2 HW: DNA Design
Overview
The steps to create a CBB resistant coffee tree Does not work: altering the genome of the tree using CRISPR of the tree b/c no one know how to do it - what site expresses the cherries for example
What will works: altering the genome of the tree using a protein that destroys the gut of the CBB 1) Get the cry10Aa sequence and design the full expression cassette. 2) Order the synthesized cassette from Twist, cloned into your binary vector. 3) Amplify the plasmid in E. coli and verify it. 4) Transform Agrobacterium with the purified plasmid. 5) Co-culture with coffee somatic embryos, then select on antibiotic. 6) Regenerate plants, confirm the insert (PCR or Southern blot), and bioassay against the borer.
Why the cry10Aa sequeence works 1) The insect eats the protein. A borer chewing into the berry ingests the toxin along with plant tissue. 2) The gut activates it. Insect midguts are typically alkaline and full of proteases, which convert the protein into its active form. 3) It binds gut receptors. The active toxin attaches to specific receptors on the midgut lining cells. The article notes that Cry proteins work this way, and that it's why organisms lacking the right receptors aren't affected. 4) The cells rupture. Binding leads to pores forming in the membrane, and the gut cells burst. 5) The insect dies. The damaged gut lets gut contents and bacteria leak into the body cavity, causing septicemia.
What molecule am I expressing?
Cry10Aa is the one protein with demonstrated CBB resistance in coffee. Valencia-Lozano et al. (2021)
expressed it in C. arabica and got 85–100% larval mortality within 10 days, with seed
damage under 9% versus 100% in controls. I considered two alternatives:
• α-amylase inhibitor-1 (α-AI1) from Phaseolus vulgaris (UniProt P02873). CBB larvae depend on gut α-amylase to digest the bean's storage carbohydrate. Barbosa et al. (2010) expressed α-AI1 in C. arabica and got up to 88% inhibition of H. hampei α-amylase in seed extracts. It is a plant protein rather than a bacterial one, which is an easier regulatory and public-perception story — a real consideration for a Kona-branded farm. But the published endpoint is in vitro enzyme inhibition, not dead beetles.
• cry1Ac (Leroy et al. 2000). Works in coffee, but it is a lepidopteran-active toxin aimed at leaf miner. CBB is a beetle, so I need a coleopteran-active Cry — which is exactly why Cry10Aa.
Decision: Cry10Aa as the primary, with α-AI1 named as the second gene for a future pyramided stack (two modes of action slows resistance evolution — see Part 6).
• α-amylase inhibitor-1 (α-AI1) from Phaseolus vulgaris (UniProt P02873). CBB larvae depend on gut α-amylase to digest the bean's storage carbohydrate. Barbosa et al. (2010) expressed α-AI1 in C. arabica and got up to 88% inhibition of H. hampei α-amylase in seed extracts. It is a plant protein rather than a bacterial one, which is an easier regulatory and public-perception story — a real consideration for a Kona-branded farm. But the published endpoint is in vitro enzyme inhibition, not dead beetles.
• cry1Ac (Leroy et al. 2000). Works in coffee, but it is a lepidopteran-active toxin aimed at leaf miner. CBB is a beetle, so I need a coleopteran-active Cry — which is exactly why Cry10Aa.
Decision: Cry10Aa as the primary, with α-AI1 named as the second gene for a future pyramided stack (two modes of action slows resistance evolution — see Part 6).
What is the sequence?
Cry10Aa, 675 aa, 77.8 kDa. UniProt P09662 (CR10A_BACTI); nucleotide GenBank M12662
(protein AAA22614.1). Import into Kernel by accession rather than pasting — a single
transcription error in a 2 kb CDS is a frameshift.
Native CDS = 675 × 3 + stop = 2,028 bp.
Native CDS = 675 × 3 + stop = 2,028 bp.
What host or chassis?
Coffea arabica L. var. 'Kona Typica' — the tree itself, as in HW1. This is a
tetraploid perennial plant (2n = 4x = 44), which is neither column of the assignment's
prokaryote/eukaryote table, so I extended it in Part 2. Transformation is into embryogenic callus /
somatic embryos, then regeneration and clonal propagation.
Tetraploidy matters: I need the transgene to work as a single dominant insert, which it does since I am adding a gene rather than knocking four allele copies out. This is another argument against my original splice-site plan — editing a trait out of a tetraploid means hitting all four copies.
Tetraploidy matters: I need the transgene to work as a single dominant insert, which it does since I am adding a gene rather than knocking four allele copies out. This is another argument against my original splice-site plan — editing a trait out of a tetraploid means hitting all four copies.
Where does the product need to go?
Cytosol of the developing endosperm (the bean). No signal peptide and no targeting sequence.
Cry toxins are not active in the plant — they are solubilized and proteolytically activated in the
insect midgut after ingestion, then bind midgut receptors. So the protein only has to accumulate,
intact, where the larva chews. That is the endosperm.
This is the one place I depart from the published construct, and it is my main design decision — see the promoter choice in Part 2.
This is the one place I depart from the published construct, and it is my main design decision — see the promoter choice in Part 2.
Part 2: Design the genetic parts
The assignment’s table has a prokaryote column and a eukaryote column. A plant needs a third, so here is my version with the plant-specific layers added (T-DNA borders, plant selection marker):
| Layer | Part type | Short name | Source | Why this part |
|---|---|---|---|---|
| T-DNA left border | Misc / feature | LB | pCAMBIA-class binary vector | Defines the start of what integrates into the coffee genome |
| Transcription start | Promoter | pCaSSP1 (csp1, ~998 bp) | C. arabica 11S seed storage protein gene; Marraccini et al. 1999. Re-derive the actual Kona Typica allele from the assembly (see below) | Seed/grain-specific. Puts the toxin only in the bean, where CBB feeds |
| Translation start | 5′ UTR + Kozak | 5UTR-csp1-kozak | Native csp1 5′ UTR + plant Kozak consensus aAcaATGGC | Efficient initiation in a plant ribosome; plant Kozak differs from mammalian |
| Coding sequence | CDS | cry10Aa-CaOpt | UniProt P09662 / GenBank M12662, codon-optimized for C. arabica | The insecticidal protein; optimization is not optional for a Bt gene (Part 4) |
| Targeting | — | none | — | Cry toxin activates in the insect midgut, not in planta. Cytosolic accumulation is correct |
| Transcription stop | Terminator + poly(A) | tNOS (~253 bp) | Nopaline synthase terminator, Agrobacterium Ti plasmid | Standard, compact, well-behaved plant terminator with its own poly(A) signal |
| Selection cassette | Promoter + CDS + term | p2X35S :: hptII :: tNOS | pMDC85 / pCAMBIA1300 | Hygromycin selection of transformed callus — what Valencia-Lozano et al. used in coffee |
| T-DNA right border | Misc / feature | RB | pCAMBIA-class binary vector | Defines the end of the integrating segment |
| Vector context | Backbone | pCAMBIA1300-class binary | Existing part — reuse, do not design | Provides pVS1 ori (Agrobacterium), pBR322 ori (E. coli), bacterial aadA, and the T-DNA borders |
Promoter
Valencia-Lozano et al. drove cry10Aa with a 2X CaMV 35S promoter — constitutive, on in
every tissue, all the time. It works, but for a perennial tree I think it is the wrong choice, and
I am deliberately swapping it for the coffee-native seed-storage-protein promoter:
1. Metabolic burden. A coffee tree is productive for 20–30 years. Making a 78 kDa bacterial protein in every leaf for three decades is a yield tax I am paying year-round for a pest that attacks one tissue.
2. Non-target exposure. 35S expression puts Cry10Aa in flowers and leaves. Kona farms depend on pollinators. Confining the toxin to the endosperm means a bee visiting the blossom is not exposed.
3. It is where the pest actually is. CBB is the only insect that completes its life cycle inside the bean. The toxin only needs to be where the larva feeds.
4. Social acceptance — which I named as an unknown in HW1. "The bean makes a protein that only affects the beetle that eats it" is a far easier conversation in Kona than "the whole tree is transgenic in every cell."
Why csp1 specifically. Marraccini et al. (1999) cloned the complete C. arabica 11S seed storage protein gene and showed that ~998 bp of 5′ upstream sequence drove reporter expression in tobacco grain and in no other tissue tested. 11S globulins are the dominant storage protein of the coffee endosperm, so the promoter is both native and strongly expressed in exactly my target tissue.
Why not CaSMP. I also looked at the seed maturation protein promoter (Quintero et al. 2018, ~1.4 kb). I rejected it: its activity is confined to the ovule of immature fruit, with no activity in the seeds of ripening fruit. CBB attacks the developing and maturing cherry, so CaSMP would switch off right when I need the toxin present.
This is where the Kona Typica genome assembly earns its place in this assignment. Rather than using the 1999 sequence from a different cultivar, I will locate the 11S storage protein gene models in the 'Kona Typica' annotation (65,458 protein-coding genes across 22 chromosomes; Lyu et al. 2025) and pull the true upstream region from my own variety. That guarantees the promoter is an exact match to the genetic background it is going back into, and it gives me the coffee codon usage table I need in Part 4.
1. Metabolic burden. A coffee tree is productive for 20–30 years. Making a 78 kDa bacterial protein in every leaf for three decades is a yield tax I am paying year-round for a pest that attacks one tissue.
2. Non-target exposure. 35S expression puts Cry10Aa in flowers and leaves. Kona farms depend on pollinators. Confining the toxin to the endosperm means a bee visiting the blossom is not exposed.
3. It is where the pest actually is. CBB is the only insect that completes its life cycle inside the bean. The toxin only needs to be where the larva feeds.
4. Social acceptance — which I named as an unknown in HW1. "The bean makes a protein that only affects the beetle that eats it" is a far easier conversation in Kona than "the whole tree is transgenic in every cell."
Why csp1 specifically. Marraccini et al. (1999) cloned the complete C. arabica 11S seed storage protein gene and showed that ~998 bp of 5′ upstream sequence drove reporter expression in tobacco grain and in no other tissue tested. 11S globulins are the dominant storage protein of the coffee endosperm, so the promoter is both native and strongly expressed in exactly my target tissue.
Why not CaSMP. I also looked at the seed maturation protein promoter (Quintero et al. 2018, ~1.4 kb). I rejected it: its activity is confined to the ovule of immature fruit, with no activity in the seeds of ripening fruit. CBB attacks the developing and maturing cherry, so CaSMP would switch off right when I need the toxin present.
This is where the Kona Typica genome assembly earns its place in this assignment. Rather than using the 1999 sequence from a different cultivar, I will locate the 11S storage protein gene models in the 'Kona Typica' annotation (65,458 protein-coding genes across 22 chromosomes; Lyu et al. 2025) and pull the true upstream region from my own variety. That guarantees the promoter is an exact match to the genetic background it is going back into, and it gives me the coffee codon usage table I need in Part 4.
Part 3: Assembling construct
5′→3′ order of the expression cassette inside the T-DNA:
Approximate sizes: expression cassette ≈ 3.4 kb; selection cassette ≈ 2.1 kb; total T-DNA ≈ 5.5 kb; whole binary plasmid ≈ 12–14 kb.
Checks to run in Kernel and record here:
- Single ATG at the intended start, in frame with the Kozak
- Exactly one stop codon, at position 2,026–2,028 of the CDS
- No internal in-frame stops (this is the check codon optimization can silently break)
- No spurious ATGs upstream of the real start in the 5′ UTR
- Parts sit in the intended order and orientation on the map — the selection cassette is often cloned in reverse; confirm which way Kernel drew it
- T-DNA borders intact and flanking everything that should transfer
TODO — export the Kernel construct map and drop the image in here.
Part 4: Optimize and quality-check
Codon optimization
TODO — run it in Kernel, then report the before/after GC% and % codons changed.
What to expect and why it matters: Bacillus thuringiensis is AT-rich, and native cry genes are the textbook case of a bacterial gene that fails in plants. The AT-rich sequence contains cryptic plant polyadenylation signals (
Valencia-Lozano et al. (2019) changed 73.6% of the codons of cry10Aa against coffee codon usage and reached 13.88 µg/g fresh weight in somatic embryos. I expect a similar magnitude. Constraints I will apply: raise GC toward the coffee average, eliminate
What to expect and why it matters: Bacillus thuringiensis is AT-rich, and native cry genes are the textbook case of a bacterial gene that fails in plants. The AT-rich sequence contains cryptic plant polyadenylation signals (
AATAAA) that truncate the transcript early, and
ATTTA mRNA instability motifs. This is not a subtle tuning problem — unoptimized
cry genes can express at near-zero in planta.Valencia-Lozano et al. (2019) changed 73.6% of the codons of cry10Aa against coffee codon usage and reached 13.88 µg/g fresh weight in somatic embryos. I expect a similar magnitude. Constraints I will apply: raise GC toward the coffee average, eliminate
AATAAA and
ATTTA, remove internal BsaI / BsmBI and the restriction sites I need for
the Gibson junctions, avoid runs >6 of one base, and keep no repeat longer than ~20 bp so the
fragment is synthesizable (Part 5).Sequence search / BLAST
TODO — run it and paste the top hits.
Expected result, which is itself the interesting part: the protein should come back ~100% identical to Cry10Aa / CryIVc from B. thuringiensis subsp. israelensis, while the nucleotide sequence should show only partial identity to GenBank M12662 — because I changed ~3/4 of the codons. Same protein, novel DNA. If the nucleotide identity comes back high, my optimization did not actually run. If the protein identity is not ~100%, I have introduced a mutation and need to go back.
Expected result, which is itself the interesting part: the protein should come back ~100% identical to Cry10Aa / CryIVc from B. thuringiensis subsp. israelensis, while the nucleotide sequence should show only partial identity to GenBank M12662 — because I changed ~3/4 of the codons. Same protein, novel DNA. If the nucleotide identity comes back high, my optimization did not actually run. If the protein identity is not ~100%, I have introduced a mutation and need to go back.
Safety screen
TODO — run it and summarize what it reported.
I expect this to flag, and I should be able to defend it rather than be surprised. Screens of this kind check a construct against sequences of concern — toxins, virulence factors, select agents. An insecticidal crystal protein is a toxin by any reasonable definition, so a hit is the correct behavior of the tool, not a problem with my design.
My response: Bt Cry proteins are the active ingredient in the most widely used biopesticide in the world, including in certified organic agriculture; Bti is applied to standing water for mosquito control. Cry10Aa's specificity comes from requiring solubilization at insect midgut pH and binding insect-specific midgut receptors — the mammalian gut has neither. It is not active on contact and not active in the plant. The honest caveat is that "no mammalian receptor" is an argument from mechanism plus a long deployment record, and for a food crop I would still expect to do the standard allergenicity and digestibility work before anything left the greenhouse.
I expect this to flag, and I should be able to defend it rather than be surprised. Screens of this kind check a construct against sequences of concern — toxins, virulence factors, select agents. An insecticidal crystal protein is a toxin by any reasonable definition, so a hit is the correct behavior of the tool, not a problem with my design.
My response: Bt Cry proteins are the active ingredient in the most widely used biopesticide in the world, including in certified organic agriculture; Bti is applied to standing water for mosquito control. Cry10Aa's specificity comes from requiring solubilization at insect midgut pH and binding insect-specific midgut receptors — the mammalian gut has neither. It is not active on contact and not active in the plant. The honest caveat is that "no mammalian receptor" is an argument from mechanism plus a long deployment record, and for a food crop I would still expect to do the standard allergenicity and digestibility work before anything left the greenhouse.
Part 5: From design to build
How would I build it?
A mixed approach, chosen fragment by fragment:
• cry10Aa-CaOpt (2.0 kb): order as synthetic DNA. Three quarters of the codons are changed, so there is no template in nature to amplify from — it has to be written, not copied. 2 kb is routine commercial gene synthesis.
• pCaSSP1 promoter (~1.0 kb): PCR from Kona Typica genomic DNA. I have the trees. It is cheaper than synthesis and, more importantly, it captures the real allele from the exact genetic background the construct is going back into — which is the whole point of using my own variety's genome.
• Backbone and selection cassette: reuse an existing binary vector. No reason to build 10 kb of well-characterized replication machinery.
• Join them with Gibson assembly. Gibson et al. (2009) showed one-step isothermal assembly of multiple overlapping fragments, up to hundreds of kilobases. I have 3–4 fragments totalling ~14 kb, which is comfortably inside that envelope, and it is scarless — no restriction sites left between my promoter and my start codon, which matters when the spacing there controls translation initiation.
• cry10Aa-CaOpt (2.0 kb): order as synthetic DNA. Three quarters of the codons are changed, so there is no template in nature to amplify from — it has to be written, not copied. 2 kb is routine commercial gene synthesis.
• pCaSSP1 promoter (~1.0 kb): PCR from Kona Typica genomic DNA. I have the trees. It is cheaper than synthesis and, more importantly, it captures the real allele from the exact genetic background the construct is going back into — which is the whole point of using my own variety's genome.
• Backbone and selection cassette: reuse an existing binary vector. No reason to build 10 kb of well-characterized replication machinery.
• Join them with Gibson assembly. Gibson et al. (2009) showed one-step isothermal assembly of multiple overlapping fragments, up to hundreds of kilobases. I have 3–4 fragments totalling ~14 kb, which is comfortably inside that envelope, and it is scarless — no restriction sites left between my promoter and my start codon, which matters when the spacing there controls translation initiation.
What does synthesis constrain?
Two things in my design are synthesis-relevant. First, the AT-richness of the native Bt
sequence — happily, codon-optimizing for coffee raises GC and makes the fragment easier
to synthesize, so the optimization I need for expression and the optimization I need for
manufacturability point the same direction. Second, error rate over 2 kb. Chip-synthesized
oligos carry roughly 1 error per 300–500 bp, so a 2 kb gene assembled naively from a pool is
overwhelmingly likely to carry at least one error — and in a CDS most errors are frameshifts or
premature stops that kill the protein outright.
This is exactly the problem Carr et al. (2004) addressed: they used a mismatch-binding protein (MutS) to remove error-containing heteroduplexes from synthetic DNA pools, cutting errors ~15-fold. Kosuri et al. (2010) attacked the same bottleneck from the other side, selectively amplifying subpools from high-fidelity microchips so that cheap chip oligos become usable gene-length constructs. Practically, either route means I do not have to sequence-verify hundreds of clones — but for a single 2 kb gene I would just buy the sequence-verified fragment and let the vendor absorb the error correction. The oligo-pool economics only start to win when I want many variants.
This is exactly the problem Carr et al. (2004) addressed: they used a mismatch-binding protein (MutS) to remove error-containing heteroduplexes from synthetic DNA pools, cutting errors ~15-fold. Kosuri et al. (2010) attacked the same bottleneck from the other side, selectively amplifying subpools from high-fidelity microchips so that cheap chip oligos become usable gene-length constructs. Practically, either route means I do not have to sequence-verify hundreds of clones — but for a single 2 kb gene I would just buy the sequence-verified fragment and let the vendor absorb the error correction. The oligo-pool economics only start to win when I want many variants.
Prototype vs. scale
This is the part that surprised me, and it mirrors HW1's scale question directly.
Prototype: one binary plasmid, micrograms of DNA, maybe $500 of synthesis. But I would not build one construct — I would build a small library: csp1 promoter vs. 2X35S vs. a promoter stack, crossed with 2–3 codon-optimization variants. Six to twelve plasmids, assembled in parallel by Gibson, is the honest prototype, because Part 6 says I do not know which promoter works.
Scale: I never need more DNA. This is the key difference from a fermentation product. The 200 trees I costed in HW1 are not 200 transformation events — they are one validated event, clonally propagated by grafting and cuttings. DNA is a one-time input of a few micrograms consumed in a single biolistic or Agrobacterium experiment. What scales is tissue culture and grafting throughput, which is precisely where the automation from HW1 goes (liquid handlers for the reagent prep and genotyping, machine-vision sorting of plantlets, robotic grafting). The DNA synthesis question is a prototype question only.
So the bottleneck is not making DNA. It is event selection — generating enough independent transformation events to find one with a clean single-copy insert, the right expression level, and no agronomic penalty — and then the 3–4 years before a Typica tree bears cherry you can actually run a bioassay on.
Prototype: one binary plasmid, micrograms of DNA, maybe $500 of synthesis. But I would not build one construct — I would build a small library: csp1 promoter vs. 2X35S vs. a promoter stack, crossed with 2–3 codon-optimization variants. Six to twelve plasmids, assembled in parallel by Gibson, is the honest prototype, because Part 6 says I do not know which promoter works.
Scale: I never need more DNA. This is the key difference from a fermentation product. The 200 trees I costed in HW1 are not 200 transformation events — they are one validated event, clonally propagated by grafting and cuttings. DNA is a one-time input of a few micrograms consumed in a single biolistic or Agrobacterium experiment. What scales is tissue culture and grafting throughput, which is precisely where the automation from HW1 goes (liquid handlers for the reagent prep and genotyping, machine-vision sorting of plantlets, robotic grafting). The DNA synthesis question is a prototype question only.
So the bottleneck is not making DNA. It is event selection — generating enough independent transformation events to find one with a clean single-copy insert, the right expression level, and no agronomic penalty — and then the 3–4 years before a Typica tree bears cherry you can actually run a bioassay on.
Part 6: What don’t you know yet?
1. Will csp1 fire in coffee endosperm, early enough and hard enough?
This is my biggest bet, because it is the one place I departed from a published working construct. The csp1 promoter was validated in tobacco grain, not in coffee, and "seed-specific" does not tell me the developmental timing. CBB bores into the cherry during endosperm development — if the promoter switches on late, the larva is already inside. And Cry10Aa worked at 13.88 µg/g under a doubled 35S; csp1 may deliver less.
How I would find out: before committing to the transgene, clone pCaSSP1::GUS and pCaSSP1::GFP and transform coffee somatic embryos, then score reporter activity across fruit development stages against the known CBB infestation window. Cheap, fast, and it de-risks the whole design. Fallback variants to build alongside: 2X35S (proven, accept the burden), or a dual cassette with a pericarp-active promoter added.
2. Does seed-confined expression actually stop the damage, or just the beetle?
A CBB female bores through the pericarp before reaching the endosperm. If the toxin lives only in the bean, she may do her boring damage and lay eggs before the larvae ingest anything lethal. Valencia-Lozano et al. got seed damage under 9% — but with a constitutive promoter, so toxin was in the pericarp too. My seed-only design may not reproduce that number.
How I would find out: whole-fruit bioassay scoring penetration depth, egg deposition, and larval mortality separately — not just "is the seed damaged." If boring damage alone is economically significant, I add a pericarp-expressed second cassette, which is also where α-AI1 would go.
3. Will 'Kona Typica' regenerate at all?
Every published coffee transformation I found used other cultivars — Catimor, Catuaí, elite Brazilian lines. Typica is an old, narrow, inbred lineage, and somatic embryogenesis efficiency is highly genotype-dependent in coffee. I could have a perfect plasmid and no plant.
How I would find out: run a somatic embryogenesis and regeneration trial on Kona Typica explants with no transgene at all, purely to measure callus induction and regeneration frequency, before spending anything on constructs. If Typica regenerates poorly, the fallback is transforming a compatible cultivar and using it as a rootstock, or grafting Typica scion onto it — which does not deliver the trait to the bean, so more likely it means a longer optimization of the tissue-culture media.
4. Resistance evolution, and whether I am allowed to plant it.
Single-gene Bt resistance has been broken in the field repeatedly. With one mode of action and a perennial crop I cannot rotate out of, CBB has decades to adapt — which is the argument for pyramiding α-AI1 (a completely different mechanism: starch digestion, not midgut receptors) into a second cassette from the start. And the non-technical unknown from HW1 has not gone away: a transgenic tree in the Kona appellation is a brand question before it is a regulatory one. How I would find out: for resistance, lab selection experiments on CBB colonies under sublethal Cry10Aa; for the brand, talk to the Kona Coffee Farmers Association before I talk to APHIS.
This is my biggest bet, because it is the one place I departed from a published working construct. The csp1 promoter was validated in tobacco grain, not in coffee, and "seed-specific" does not tell me the developmental timing. CBB bores into the cherry during endosperm development — if the promoter switches on late, the larva is already inside. And Cry10Aa worked at 13.88 µg/g under a doubled 35S; csp1 may deliver less.
How I would find out: before committing to the transgene, clone pCaSSP1::GUS and pCaSSP1::GFP and transform coffee somatic embryos, then score reporter activity across fruit development stages against the known CBB infestation window. Cheap, fast, and it de-risks the whole design. Fallback variants to build alongside: 2X35S (proven, accept the burden), or a dual cassette with a pericarp-active promoter added.
2. Does seed-confined expression actually stop the damage, or just the beetle?
A CBB female bores through the pericarp before reaching the endosperm. If the toxin lives only in the bean, she may do her boring damage and lay eggs before the larvae ingest anything lethal. Valencia-Lozano et al. got seed damage under 9% — but with a constitutive promoter, so toxin was in the pericarp too. My seed-only design may not reproduce that number.
How I would find out: whole-fruit bioassay scoring penetration depth, egg deposition, and larval mortality separately — not just "is the seed damaged." If boring damage alone is economically significant, I add a pericarp-expressed second cassette, which is also where α-AI1 would go.
3. Will 'Kona Typica' regenerate at all?
Every published coffee transformation I found used other cultivars — Catimor, Catuaí, elite Brazilian lines. Typica is an old, narrow, inbred lineage, and somatic embryogenesis efficiency is highly genotype-dependent in coffee. I could have a perfect plasmid and no plant.
How I would find out: run a somatic embryogenesis and regeneration trial on Kona Typica explants with no transgene at all, purely to measure callus induction and regeneration frequency, before spending anything on constructs. If Typica regenerates poorly, the fallback is transforming a compatible cultivar and using it as a rootstock, or grafting Typica scion onto it — which does not deliver the trait to the bean, so more likely it means a longer optimization of the tissue-culture media.
4. Resistance evolution, and whether I am allowed to plant it.
Single-gene Bt resistance has been broken in the field repeatedly. With one mode of action and a perennial crop I cannot rotate out of, CBB has decades to adapt — which is the argument for pyramiding α-AI1 (a completely different mechanism: starch digestion, not midgut receptors) into a second cassette from the start. And the non-technical unknown from HW1 has not gone away: a transgenic tree in the Kona appellation is a brand question before it is a regulatory one. How I would find out: for resistance, lab selection experiments on CBB colonies under sublethal Cry10Aa; for the brand, talk to the Kona Coffee Farmers Association before I talk to APHIS.
AI
Claude Opus 5 (Claude Code), September 2026.
Prompts: asked it to review my HW1 submission and the HW2 assignment, and said I assumed I needed to find a splice site in the Kona Typica reference genome to alter expression of the cherries. It pushed back on the splice-site framing — that HW2 wants an expression cassette rather than a genome edit, that there is no characterized resistance locus in the assembly to target, and that CBB tolerates caffeine — and pointed me to the α-AI1 and Cry10Aa literature. I then had it pull the accessions, the published construct details, and candidate coffee seed promoters, and draft this page. The promoter swap away from 2X35S, and the reasoning about non-target exposure and the Kona brand, came out of that back-and-forth. Kernel work in Parts 0, 3 and 4 is my own.
Prompts: asked it to review my HW1 submission and the HW2 assignment, and said I assumed I needed to find a splice site in the Kona Typica reference genome to alter expression of the cherries. It pushed back on the splice-site framing — that HW2 wants an expression cassette rather than a genome edit, that there is no characterized resistance locus in the assembly to target, and that CBB tolerates caffeine — and pointed me to the α-AI1 and Cry10Aa literature. I then had it pull the accessions, the published construct details, and candidate coffee seed promoters, and draft this page. The promoter swap away from 2X35S, and the reasoning about non-target exposure and the Kona brand, came out of that back-and-forth. Kernel work in Parts 0, 3 and 4 is my own.
References
- Lyu, H. et al. A chromosome-level genome assembly of Coffea arabica L. var. ‘Kona Typica’. Scientific Data 12, 1314 (2025). https://doi.org/10.1038/s41597-025-05658-6
- Valencia-Lozano, E., Cabrera-Ponce, J.L., Noa-Carrazana, J.C. & Ibarra, J.E. Coffea arabica L. resistant to coffee berry borer (Hypothenemus hampei) mediated by expression of the Bacillus thuringiensis Cry10Aa protein. Frontiers in Plant Science 12, 765292 (2021). https://doi.org/10.3389/fpls.2021.765292
- Valencia-Lozano, E., Cabrera-Ponce, J.L., Gómez-Lim, M.A. & Ibarra, J.E. Development of an efficient protocol to obtain transgenic coffee, Coffea arabica L., expressing the Cry10Aa toxin of Bacillus thuringiensis. IJMS 20, 5334 (2019). https://doi.org/10.3390/ijms20215334
- Barbosa, A.E.A.D. et al. α-Amylase inhibitor-1 gene from Phaseolus vulgaris expressed in Coffea arabica plants inhibits α-amylases from the coffee berry borer pest. BMC Biotechnology 10, 44 (2010). https://doi.org/10.1186/1472-6750-10-44
- Ceja-Navarro, J.A. et al. Gut microbiota mediate caffeine detoxification in the primary insect pest of coffee. Nature Communications 6, 7618 (2015). https://doi.org/10.1038/ncomms8618
- Marraccini, P., Deshayes, A., Pétiard, V. & Rogers, W.J. Molecular cloning of the complete 11S seed storage protein gene of Coffea arabica and promoter analysis in transgenic tobacco plants. Plant Physiology and Biochemistry 37, 273–282 (1999). https://doi.org/10.1016/S0981-9428(99)80025-4
- Quintero, F.O. et al. Identification of a seed maturation protein gene from Coffea arabica (CaSMP) and analysis of its promoter activity in tomato. Plant Cell Reports 37, 1257–1268 (2018). https://doi.org/10.1007/s00299-018-2310-9
- Leroy, T. et al. Genetically modified coffee plants expressing the Bacillus thuringiensis cry1Ac gene for resistance to leaf miner. Plant Cell Reports 19, 382–389 (2000). https://doi.org/10.1007/s002990050744
Assigned readings cited (Part 5):
- Gibson, D.G. et al. Enzymatic assembly of DNA molecules up to several hundred kilobases. Nature Methods 6, 343–345 (2009). https://doi.org/10.1038/nmeth.1318
- Carr, P.A. et al. Protein-mediated error correction for de novo DNA synthesis. Nucleic Acids Research 32, e162 (2004). https://doi.org/10.1093/nar/gnh160
- Kosuri, S. et al. Scalable gene synthesis by selective amplification of DNA pools from high-fidelity microchips. Nature Biotechnology 28, 1295–1299 (2010). https://doi.org/10.1038/nbt.1716
Final Project
Week 3 HW: Cell-Free Reaction Design (Coopetition)
Week 3 Homework: Designing a Cell-Free Reaction Composition
Due: Tuesday, October 6, 2026
Coopetition website: coopetition.htgaa.org
Goal: Using the collective intelligence of the HTBAA community and accessible AI systems, design novel cell-free protein synthesis (CFPS) reaction compositions for each of the three target proteins (sfGFP, PETase, and Reteplase) to reduce the cost per gram of functional protein. Each design should be compatible with the experimental framework and reflect the distinct expression and folding requirements of that protein.
A. Compare Cell-Free Reaction Compositions
A1. Potassium glutamate vs. magnesium glutamate
Question: Why does the concentration of potassium glutamate and magnesium glutamate differ between compositions? Which is more sensitive?
My answer:
Based on on this paper: https://pmc.ncbi.nlm.nih.gov/articles/PMC3161345/ the concentrations differ beacause each expierment requires a different level of patassium and magnesium to work - why would you assume it is static? magnesium is much more senitive
A2. PEP+NTP vs. Ribose+NMP
Question: Explain how a PEP+NTP reaction formulation is different from a Ribose+NMP reaction formulation.
My answer:
Based on this paper: https://www.nature.com/articles/s41467-026-69605-8
Both the (1) PEP+NTP and the (2) Ribose+NMP are ways of providing the cell-free system enegry for the reaction. (1) contains pre-built energy and the pre-defined nucleotides so it is faster compared to (2) which contains the sugar that needs to be turned into engergy and the 1/2 built nucleotide strain that needs to be completed. I am not sure why anyone would use (2) over 1 since it is slower.. unless it is more costly? I don’t see that in the article.
A3. Compare two published compositions
Question: Choose two cell-free reaction compositions from the literature, compare and contrast the choice of reagents and the concentration of reagents.
Suggested starting points: PANOx-SP (Jewett & Swartz 2004), the cost-optimized formulation from Olsen et al. (2026), and/or the AI-designed formulation from Smith et al. (2026).
Composition 1:
This is a high-performance reaction - it uses PEP+NTP
Composition 2:
This is a low-cost reaction - it uses Ribose+NMP - which answers my question from the prior question :-)
Comparison of reagents and concentrations:
For energy ANOx‑SP uses PEP - high energy but expensive Olsen replaces PEP with Ribose - lower energy but cheaper
For RNA building blocks ANOx‑SP uses NTP - preassembled nucluotides Olsen uses NMP - sysetm turns into NTP - slower and cheaper
Other Factors (NAD, CoA, tRNA, etc.) ANOx‑SP includes all possible cofactors Olsen includes a lot less - paper argues most are not needed
Salts ANOx‑SP - PEP/NTP uses higher magnesium and potassium levels Olsen adjusts based on cheaper chemstry (I think?)
Amino Acids ANOx‑SP - standard 20‑amino‑acid mix at ~2 mM each Olson - less to lower costs
B. Design One Reaction Composition for Each of the Three Proteins
Rules: total reaction volume is 20 µL (20,000 nL). Fixed components (lysate 5,000 nL, DNA template 2,000 nL, base buffer 2,000 nL) take 9,000 nL. I choose up to 11,000 nL from the reagent menu; water fills the rest. All volumes must be multiples of 25 nL.
B1. Publish my first compositions on coopetition.htgaa.org
Question: Sign in to coopetition.htgaa.org and publish your first unique 20 µL CFPS reaction composition for sfGFP, PETase, and Reteplase. For each composition, specify the reagents you would adjust from the relevant categories, including salts, energy sources, building blocks, cofactors, and redox components.
sfGFP composition
My answer (reagents, volumes, and reasoning):
I took the default - just figuring out the user interface
PETase composition
My answer (reagents, volumes, and reasoning):
I took the default - just figuring out the user interface
Reteplase composition
My answer (reagents, volumes, and reasoning):
I took the default - just figuring out the user interface
B2. Community Reagents suggestion
Question: Contribute (or upvote) at least one reagent to the “Community Reagents” section of the website and provide a short rationale. Include how you decided on that specific reagent and provide links to any relevant research.
Reagent I suggested or upvoted:
FkpA - it was the first one listed
How I decided and my rationale:
FkpA - it was the first one listed
Links to relevant research:
N/A
B3. cfps-coopetition forum
Question: Visit the cfps-coopetition forum page to ensure you will be able to take part in the discussion. Bonus: send a note in the channel about a CFPS optimization idea that you have.
My answer (confirmation, plus my optimization idea if I posted one):
So the link just goes to
https://forum.htgaa.org/#narrow/channel/43-cfps-coopetition
which I joined and posted a test message
C. Asimov Kernel (Optional)
C1. Build T7-compatible plasmids
Question: Build T7-compatible plasmids for cell-free expression using the amino acid coding sequences provided for sfGFP, PETase, and Reteplase.
NA
Type your answer here.
C2. Bonus: T7 helper plasmid for Reteplase
Question: Design a T7 helper plasmid that you could add to the cell-free reaction to improve the yield of functional Reteplase (general chaperones, disulfide-bond machinery, etc.).
My answer:
NA
Use of AI
If I used AI, I need to cite the AI client and the prompts I used, and be able to explain and justify my design reasoning in recitation.
AI tools used:
Claude (Anthropic), model Claude Opus 5.5, used through the Claude desktop app with access to my homework folder. I used it to summarize the assignment in plain language and to set up the structure of this homework page. It did not write my answers or design my reaction compositions.
Prompts used:
To understand the assignment:
review the two documents found here: ‘/Users/jamiedixon/Documents/MIT - HTBMAA/class 03’ tell me what I am supposed to be doing as if I am in 6th grade
To set up this page:
OK I created the _index page for week 3 here update the mark down to sections for my homework that I have to answer. Include the question and then leave a block where I can type in my answer
To give Claude the location of the page:
sorry, here is the file ‘/Users/jamiedixon/Documents/MIT - HTBMAA/webpages-2026b/content/week-03-HW-coopetition/_index.md’
To document AI use:
Fill in the AI tools used section and the prompts used so far
References
- 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. doi:10.64898/2026.02.05.703998
- 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