Plasmids That Go Anywhere

[SOFIA] Okay, here's a thing that sounds mundane but secretly controls whether half of biology is even possible: can you get a piece of DNA to stay put inside a cell that isn't E. coli or yeast?

[DANIEL] Hm. It does sound mundane until you try it.

[SOFIA] Right! Because the dream — the whole dream of engineering non-model organisms, a weird fungus, a probiotic bacterium, a diatom — is that you build a circuit, you put it on a plasmid, and the organism just carries it around and does what you want. And the reality is the DNA usually either falls out, gets chewed up, or won't express. So today's story is "plasmids that go anywhere," and I want to trace how the field has been chipping at that for thirty years.

[DANIEL] Let's define the problem for someone who lives in a model organism. A plasmid is a circle of DNA separate from the chromosome. In E. coli it's trivial — there's a well-characterized origin of replication, the cell copies the plasmid every generation, done. The trouble is that origin is species-specific. The machinery that recognizes it doesn't exist in your fungus or your diatom.

[SOFIA] So you've got two bad options. Either you integrate your DNA into the chromosome, which is permanent but slow and position-dependent, or you find something that replicates on its own — episomal, floating free — in that host.

[DANIEL] And "episomal" is the word to hold onto. Maintained outside the chromosome, ideally at some stable copy number, passed to daughter cells. That's the holy grail for a new organism because it makes genetics fast — transform, test, iterate.

[SOFIA] Which brings us to the root of the whole story, and it's a beautiful accident. 1991, Gems and colleagues, working in Aspergillus nidulans — a filamentous fungus. Filamentous fungi were notoriously miserable to transform. DNA went in, integrated rarely, low efficiency, painful.

[DANIEL] And they weren't hunting for a replicator. They were just re-isolating something from a gene bank.

[SOFIA] A 6.1-kilobase insert, pops out, and transformation efficiency jumps two hundred and fifty-fold. They call it AMA1. And the reason it works — the plasmid stops needing to integrate. AMA1 lets it sit episomally, ten to thirty copies per cell, free-form.

[DANIEL] Two hundred fifty-fold is the kind of number that makes me want to see the controls, but the mechanism they inferred is consistent — if the plasmid no longer has to find a chromosomal home, every cell that takes up DNA can keep it. That's where the efficiency comes from.

[SOFIA] And here's the part that made it a turning point and not a one-off: it worked across the genus. A. niger, A. oryzae — industrial workhorses. So AMA1 became this portable "stay resident" cassette for fungi.

[DANIEL] Which sets a template for the whole subject. Find a sequence that does one hard job — persistence — and reuse it.

[SOFIA] But persistence is only half the battle, and the 2010 paper from Collins and colleagues is where that bites. Coprinopsis cinerea — the inky cap mushroom, a model for mushroom-forming fungi. They build this lovely modular vector set, a multiple cloning site, an optional 5-prime intron to help expression.

[DANIEL] And they get a split result, which I appreciate them reporting honestly. Two fluorescent reporters, DsRed and mRFP — first time expressed in that organism. But GUS and LUC, the classic workhorse reporters, fail. Even with the intron added.

[SOFIA] Which is wild, because the intron is supposed to be the fix. In a lot of fungi, sticking an intron in boosts expression.

[DANIEL] So the barrier isn't splicing. It's downstream of splicing — something about the message or the protein that organism just won't produce. They couldn't fully pin it, and they said so.

[SOFIA] And that's the lesson the field keeps relearning. Getting DNA in and keeping it there doesn't guarantee it does anything. Expression is its own separate wall.

[DANIEL] It reframes "goes anywhere." Anywhere physically is not anywhere functionally.

[SOFIA] Okay, this is the good stuff — jump to 2018, totally different kingdom, and watch the field get systematic about that wall. Anik Debnath in the Church lab, engineering Lactobacillus — probiotic lactic acid bacteria — to secrete nanobodies. Little single-domain antibodies, VHHs.

[DANIEL] And the ambition here is therapeutic. Live biotherapeutics — a bug you swallow that manufactures drug in your gut. Anti-TNF-alpha for inflammatory bowel disease, anti-gp120 for HIV prevention.

[SOFIA] So they need the whole stack. A broad-host plasmid so it works in more than one strain. An inducible promoter — they bring an aTc-inducible system into LAB for the first time, so you can switch expression on. And crucially, signal peptides to get the protein out of the cell.

[DANIEL] And this is where I think they solved the Coprinopsis problem by brute force and smarts. They didn't guess at secretion tags. They did exoproteome analysis — looked at what each strain actually secretes — and built strain-specific consensus signal peptides from that.

[SOFIA] Metagenomics-guided plasmid design plus empirically-derived parts. And it works — over ten micrograms per mil of nanobody, out of both L. rhamnosus GG and L. gasseri, with neutralization within two percent of a conventionally made antibody.

[DANIEL] That two percent is the number that survives my skepticism. Potency essentially indistinguishable from the real thing. And they matched the parts to the host instead of forcing a universal part — which is the opposite philosophy from AMA1.

[SOFIA] Right, and I love that tension in the arc. AMA1 says "one portable module." Lactobacillus says "tune every part to the strain." Both are "go anywhere" strategies, they just disagree on how.

[DANIEL] Then 2020 pushes scale. Thalassiosira pseudonana — a diatom. They assemble the entire 44-kilobase mitochondrial genome in yeast from twelve PCR fragments, move it into E. coli, no growth burden.

[SOFIA] So now the "plasmid" is a whole organellar genome, riding through a host that isn't its home. But —

[DANIEL] But it's not stable. After about sixty generations in E. coli, five of thirty clones lost a diagnostic chunk to deletion. Versus zero of thirty for the higher-GC Phaeodactylum genome.

[SOFIA] Thirty percent G-plus-C content. The AT-rich diatom DNA is just fragile in E. coli — looks like deletions find those sequences.

[DANIEL] It's a real control — two genomes, different GC, same host, different outcome. That's a clean finger pointed at composition as the instability driver. Persistence again, biting back at the biggest scale yet.

[SOFIA] And then the newest paper flips to a question nobody asks until the product's on a truck — the 2023 LNP work. Lipid nanoparticles, the mRNA-vaccine delivery vehicle, also for DNA. How do you store the thing?

[DANIEL] Freeze it without sugar and the particles fuse into donut-shaped ghosts. Twelve percent sucrose prevents it. And DNA cargo tolerates storage much better than mRNA.

[SOFIA] So the "goes anywhere" problem now includes the freezer. Delivery and shelf life, not just replication.

[DANIEL] It's the full pipeline coming into view. Get in, stay, express, scale, and survive storage. Each paper owns one failure mode.

[SOFIA] And where it's heading, I think, is design-driven portability — exoproteome-style empirical parts, GC-aware assembly, formulation baked in from the start. Thirty years from an accident in Aspergillus to engineering every step on purpose.

[DANIEL] And still reporting the clones that fail. Which is why I believe the ones that work.

[SOFIA] Perfect place to leave it. More after the break.