Bacterial DNA Defense And Modification

[SOFIA] Okay, so if you're a bacterium, your whole life is a knife fight. Phages are trying to inject their DNA into you, other bacteria are shooting DNA at you through conjugation, and your genome is basically a fortress with a sign-in sheet. And the question underneath everything we're talking about today is: how does a cell tell its own DNA apart from somebody else's?

[DANIEL] Right, and that's not a philosophical question for them, it's survival. If you can't distinguish self from invader, you either let a phage kill you or you chop up your own chromosome. So bacteria evolved these marking systems — chemical tags on the DNA that say "this is mine."

[SOFIA] And the classic one everybody learns is restriction-modification. Let me set it up for someone who doesn't live in this world. You've got a methyltransferase — an enzyme that sticks a methyl group onto a specific DNA sequence, like G-A-T-C. And then you've got a restriction enzyme that cuts that same sequence, but only if it's unmethylated.

[DANIEL] So your own DNA is methylated at every GATC, protected. Incoming phage DNA shows up naked, no methyl marks, and the restriction enzyme shreds it. It's an innate immune system written in chemistry.

[SOFIA] And it's an arms race, which is the through-line for this whole segment. Because once hosts have restriction-modification, phages evolve ways around it, hosts evolve new marks, and the whole thing spirals. That's the story — the escalation.

[DANIEL] I'd add one term before we go further, because it comes up: the nucleoid. Bacteria don't have a nucleus, but their chromosome isn't just floating loose either — it's condensed, organized, coated with proteins. Accessibility of that DNA matters for everything we're going to discuss.

[SOFIA] Which is actually where our oldest paper lands, and it's a weird one — 2009, Rebelo. They take bacterial DNA methyltransferases and import them into human mitochondria.

[DANIEL] Hm. Which sounds like it has nothing to do with bacterial immunity.

[SOFIA] Right, except the mitochondrion is a domesticated bacterium, and its little genome, the mtDNA, is organized into a nucleoid just like a bacterial one. So they used these bacterial methyltransferases as a footprinting probe — the enzyme can only tag DNA that's physically exposed.

[DANIEL] That's the clever part. Wherever a protein is bound to the DNA, the methyltransferase can't reach, so you get a shadow. They found one site, the mTERF1 binding site, almost permanently occupied, and promoter regions partially protected. And accessibility shifted with the level of TFAM — the packaging protein — and with oxidative stress.

[SOFIA] So the deep idea here: these self-recognition enzymes aren't just immune tools, they're reporters. They tell you how open or shut the genome is. That's a thread that pays off later when people want to edit chromosomes in situ.

[DANIEL] It also quietly establishes that these bacterial methylation systems are modular enough to drop into a completely foreign compartment and still work. That portability matters.

[SOFIA] Okay, then 2012, An and colleagues, and this is where the chemistry gets wild. Phosphorothioation. Instead of adding a methyl group to a base, you swap one of the oxygens in the DNA backbone — the phosphate — for a sulfur atom.

[DANIEL] A sulfur in the backbone. That's a completely different kind of mark from methylation. It's on the phosphate, not the base, and it's installed by the Dnd system — dnd for "DNA degradation," because mutants degrade their own DNA during electrophoresis, which is how it was found.

[SOFIA] And the beautiful finding here — where does the sulfur come from? There's an enzyme, DndA, a cysteine desulfurase, that normally pulls sulfur off the amino acid cysteine. But An's group showed that in E. coli, 19 of 31 known phosphorothioation gene clusters don't even carry their own DndA.

[DANIEL] They parasitize the host's enzyme. IscS — the cell's own multipurpose desulfurase — physically contacts the Dpt proteins through its active-site cysteine and hands over sulfur. So the immune system is wired straight into central metabolism.

[SOFIA] Which is such an evolutionary move, right? Why carry your own sulfur-donor when the host already has one running? Steal the supply chain.

[DANIEL] And it tells you these systems are deeply integrated, not bolt-on cassettes. That constrains how they can move between organisms.

[SOFIA] So 2020 is my turning point. Salmonella, phosphorothioate epigenetics, and this gorgeous idea of epigenetic competition. The Dnd proteins — DndCDE — want to modify certain sequences. But some of those sites are already occupied by Dam methylation, that same GATC mark we started with.

[DANIEL] So two self-marking systems collide on the same DNA.

[SOFIA] Exactly! The Dnd complex gets "parked" at Dam-methylated GATC sites it can't actually modify — it recognizes them but can't act. So the phosphorothioate marks get pushed off to other motifs, GAAC and GTTC, chosen by DNA shape rather than strict sequence.

[DANIEL] And here's the number that got my attention — genome-wide, the density of phosphorothioate marks stayed clamped at around 1,500 per million nucleotides, regardless of which sites were used.

[SOFIA] Density homeostasis. The cell cares how much marking there is, not exactly where. That's a regulatory principle nobody would've predicted from the 2012 picture.

[DANIEL] And it connects back to Rebelo in a way I like — both papers are really about occupancy and accessibility. Who's bound where, and what that excludes. Different systems, same logic of competition for the same stretch of DNA.

[SOFIA] Now here's the pivot in the arc, and this is the good stuff — all this knowledge about how bacteria defend their DNA becomes the obstacle when we want to put DNA in on purpose. 2021, Rubin and colleagues. They're trying to edit bacteria inside a real community — synthetic soil, infant gut microbiota — without isolating anybody first.

[DANIEL] Which runs straight into restriction-modification. The community members chew up your incoming DNA with exactly the immune systems we've been describing.

[SOFIA] So they built ET-Seq, which measures which organisms in a mixed community will actually accept foreign DNA — basically mapping who's tractable. Then VcDART, a single vector using a CRISPR-associated transposase, to drop targeted insertions into specific species at specific loci, in situ.

[DANIEL] A CRISPR-guided transposase — so you get RNA-programmed targeting without needing to make a double-strand break and rely on the host's repair. In a community where you can't culture half the members, that's the enabling trick.

[SOFIA] And 2023, Liu's review, extends the same thinking into probiotics — cataloging which CRISPR-Cas types different probiotic genera already carry endogenously, and charting therapeutic uses, antimicrobial resistance, IBD, even oncology, with things like E. coli Nissle.

[DANIEL] The point being you can sometimes repurpose the organism's own CRISPR system rather than fighting it.

[SOFIA] Which brings us to 2024, Vajente, and Cupriavidus necator — a non-model bacterium people want for bio-based chemicals. Its restriction-modification system was wrecking transformation. So instead of fighting it, they matched it: a plasmid backbone methylated to look like native DNA, built by Golden Gate assembly, plus a deletion strain with the restriction system removed.

[DANIEL] Seventy-fold improvement in delivery. And notice the move — they used the host's self-recognition rules as the design spec. Disguise your DNA as self, or delete the gatekeeper entirely.

[SOFIA] That's the whole arc in one sentence. We spent fifteen years figuring out how bacteria mark self versus non-self, and now we're forging the ID badge to walk our DNA right through the front door.

[DANIEL] And where it's heading is predictability — knowing a genome's methylation and restriction profile well enough to design delivery for any new organism up front.

[SOFIA] Fortress biology, turned into an engineering manual. Daniel, great one. Back after this.