Writing Genomes From Scratch

[THEO] Okay, picture this. You want to build a house. You could do what everyone's always done — go to the one lumber yard that happens to stock the exact beams you need, hope they take your order, hope the delivery truck fits down your street. Or you could learn to make the beams, machine the joints, and deliver to any address you want. That second world — where biology isn't limited to a couple of cooperative organisms — that's what today's story is about. Writing genomes from scratch, in whatever bug you choose.

[DR. MARA] And the reason that's hard comes down to a small set of model organisms doing all the heavy lifting. E. coli, a handful of others — we have a century of tricks for getting DNA into them, moving pieces around, selecting what we want. Step outside that club and almost nothing works the same. The DNA won't go in. If it goes in, the cell chews it up. If it survives, it won't replicate.

[THEO] So when we say "non-model organism," we mean the ninety-nine percent of microbial life where the toolbox just... doesn't fit.

[DR. MARA] Right. And "writing a genome" isn't one act. It's a stack of problems. You need a way to deliver DNA. A way to assemble large constructs accurately before you deliver them. A way to edit the chromosome once it's inside. And ideally a way to read out, at scale, what all those edits actually did. Each paper in this arc chips at a different layer of that stack.

[THEO] Let's talk key terms first, because a few of these come up over and over. Conjugation?

[DR. MARA] Bacterial sex, loosely. One cell builds a physical bridge to another and pushes DNA across. It's a natural process, and crucially it moves DNA as a protected handoff rather than dumping naked DNA into a hostile environment.

[THEO] Which matters because naked DNA gets eaten.

[DR. MARA] By restriction-modification systems — the cell's way of recognizing foreign sequence and cutting it. The microbial immune memory. A lot of why transformation fails in weird organisms is those defense systems.

[THEO] And recombineering, integrases, transposases — those are all ways of getting DNA to land where you want in the chromosome?

[DR. MARA] Different mechanisms, same goal. Recombineering uses homology — matching sequence — to swap something in. Integrases and transposases are enzymes that insert DNA at specific sites. Hold those three, and most of this arc makes sense.

[THEO] So let's start at the root. 2022, Deinococcus radiodurans. The bug famous for surviving radiation that would turn you and me into a cautionary tale.

[DR. MARA] A gorgeous, difficult organism. High G+C genome, aggressive restriction-modification, and chemical transformation just fails for anything large. The 2022 work sidesteps all of that by using conjugation from E. coli — the protected handoff — to deliver both replicating and nonreplicating plasmids into Deinococcus.

[THEO] And nonreplicating is the clever bit, right? You send in DNA that can't copy itself, so it has to integrate or die?

[DR. MARA] Exactly. If it can't replicate on its own, the only way it persists is by recombining into the chromosome. They used that to do sequential gene deletions — knocking out four restriction-modification genes by replacing each with a resistance marker. Disarming the immune system, one gene at a time.

[THEO] But here's the part that made my jaw drop. They ran it backwards.

[DR. MARA] They did. A nonreplicating plasmid carrying just one kilobase of homology to a gene called McrC, and they captured the entire 178-kilobase megaplasmid — the whole thing, about 62% G+C — as a roughly 190-kilobase clone, pulled back into E. coli and verified by MinION sequencing.

[THEO] So conjugation isn't only a delivery truck. It's a tow truck. You can haul a native megaplasmid out of an exotic organism and park it in E. coli where all your tools work.

[DR. MARA] That's the conceptual turning point. Delivery and capture as two directions of one process. It reframes the exotic genome as something you can borrow.

[THEO] Okay, so 2022 cracks delivery into a hard organism. The next beat, 2023, is about precision inside the chromosome. ORBIT.

[DR. MARA] This one's in E. coli, but it's solving a problem that haunts every organism: recombineering with long payloads is miserably inefficient. Lambda Red, the standard tool, drops off fast as your insert gets bigger. ORBIT splits the job. A short oligo carries a landing pad — a Bxb1 integrase attachment site — and lambda Red only has to install that tiny site. Then the Bxb1 integrase, a serine integrase, delivers the kilobase payload to that pad.

[THEO] So instead of asking the weak enzyme to do the heavy lifting, you have it tack up a doorframe, and then the strong enzyme walks the furniture through the door.

[DR. MARA] A thousand-fold over lambda Red, and it scales to libraries of thirty thousand members. That last part matters — it's not just one clean edit, it's edits at population scale.

[THEO] And now the tools start branching. 2024 gets busy. In- and Out-Cloning, the SapI work.

[DR. MARA] This is the assembly layer. Before you deliver anything, you have to build it, and ideally without scars — leftover junk sequence at every junction. SapI is a type IIS enzyme that leaves three-nucleotide overhangs, which lines up with codons, so you can stitch transcription units together scar-free. And "Out-Cloning" builds the acceptor plasmids on demand, for any organism. That's the quiet theme of the whole arc, actually — "for any organism."

[THEO] Then CAST, same year, which feels like the heir to that 2022 integration idea but programmable.

[DR. MARA] CRISPR-associated transposases. You get RNA-guided insertion of large payloads at a programmed genomic site — and critically, without double-strand breaks. No cut, no reliance on the cell's repair machinery, which varies wildly in non-model organisms. They built genome-scale knockout, overexpression, and fusion libraries in bacteria with it.

[THEO] And the complementation paper — Coaux-Seq — that's the read-out layer finally catching up?

[DR. MARA] That's the one I'd underline. They barcode roughly three-kilobase genomic fragments from eleven diverse bacteria, express them in E. coli across twenty different auxotrophic knockout backgrounds, and use cheap barcode sequencing to see which fragment rescues which defect. First experimental validation for fifty-three protein functions. Including — and this is the lovely detail — the first sulfate importer in the TauE family, which everybody had filed as an exporter family.

[THEO] So you don't just predict function from sequence, you test it, by the thousands, in a living cell.

[DR. MARA] Gain-of-function at scale. That closes a loop. Deliver, assemble, edit, and now measure.

[THEO] And the last paper zooms all the way out — bacteria as delivery vehicles into mammalian cells. Vaccines, gene therapy, microbiome editing.

[DR. MARA] Five strategies — attenuated strains, lysis circuits, T4SS conjugation, outer membrane vesicles, nanoparticle coatings. The same conjugation machinery from the Deinococcus story, pointed at human cells. Large cargo, mucosal access that viral and lipid-nanoparticle vectors can't reach.

[THEO] So the arc goes from "can we even get DNA into a radiation-proof bug" to "bacteria as programmable couriers for medicine." The through-line being — stop waiting for the organism to cooperate.

[DR. MARA] Build the delivery, the assembly, the editing, and the measurement so they travel with you. The genome stops being something you inherit and becomes something you write. We're not at whole genomes from scratch in arbitrary organisms yet — but every layer of that stack now has a real tool under it.

[THEO] Which is a good place to leave it. After the break, the mailbag — and someone wrote in about tardigrades, so Mara, warm up.

[DR. MARA] I'm bracing.