The Bacterial Immune Arms Race

[SOFIA] Okay, so picture a war that's been going on for billions of years, with no winner, and the weapons are molecular. That's what I want to talk about today — the bacterial immune arms race. Bacteria get attacked constantly, mostly by phages, by viruses, and by foreign DNA floating in from other microbes. And they've evolved this whole arsenal to tell self from non-self and shred anything that doesn't belong.

[DANIEL] And it matters to us for a very practical reason, which is — that same arsenal is exactly what fights back when we try to engineer these organisms. Every time you want to put a plasmid into a new bacterium, you're walking into the defenses.

[SOFIA] Right! The immune system doesn't know you're trying to help. So let me set up the pieces, because this spans a few different mechanisms. The classic one everybody knows is restriction-modification. Restriction enzymes cut DNA at specific sequences, and the cell protects its own genome by methylating those same sequences — sticking a little chemical tag on so the enzyme reads it as "mine, don't cut."

[DANIEL] Methylation as a passport, basically. Your DNA carries the stamp, incoming DNA doesn't, so incoming DNA gets destroyed.

[SOFIA] Exactly. Then there's a weirder, less famous one — phosphorothioate modification. Instead of tagging a base, the cell swaps one of the oxygen atoms in the DNA backbone, in the phosphate, for a sulfur. It's a backbone modification. The system that does it is called the Dnd system — Dnd, as in "DNA degradation," because when you lose certain genes the DNA literally degrades during electrophoresis.

[DANIEL] And sulfur insertion is chemically non-trivial. You need a source of reactive sulfur, and you need to hand it off without poisoning the cell. Hold that thought, because it becomes one of the turning points.

[SOFIA] And then CRISPR, which everyone knows as a tool now, but in the bacterium it's an adaptive immune memory — it keeps little snippets of past invaders and uses them to target the DNA again. So: restriction-modification, phosphorothioate, CRISPR. Three flavors of self/non-self.

[DANIEL] And the arc of these papers, I think, is us going from watching these systems, to understanding their chemistry, to understanding how they're regulated, and finally to outsmarting them to do engineering.

[SOFIA] That's the through-line. So let's go to the roots. The oldest paper here is 2009, Rebelo, and it's a little bit of a sideways entry point — they took bacterial DNA methyltransferases and imported them into the mitochondria of human cells.

[DANIEL] Which is clever, actually. The methyltransferase only marks DNA that's physically accessible. So wherever a protein is sitting on the DNA, that stretch is shielded from methylation. You read out the methylation pattern afterward and you've got an in vivo footprint of what's bound where.

[SOFIA] A bacterial immune enzyme repurposed as a molecular ruler. They mapped the mitochondrial nucleoid and found the mTERF1 site was almost permanently occupied, and that accessibility shifted with things like TFAM levels and oxidative stress.

[DANIEL] And for our arc, the lesson isn't the mitochondrial biology per se. It's that a modification enzyme reports on DNA state and DNA accessibility. These enzymes aren't blunt instruments — where and when they act is informative.

[SOFIA] Then 2012, An and colleagues, and this is the chemistry turning point. The phosphorothioate question — where does the sulfur come from? They showed that in E. coli, the Dnd proteins — DptC and DptE — physically grab onto the host's own cysteine desulfurase, IscS.

[DANIEL] IscS is a housekeeping enzyme. Its day job is making iron-sulfur clusters. It pulls sulfur off cysteine and parks it as a persulfide on an active-site cysteine — residue 328 here. And the finding is that the Dnd system taps that persulfide for its own sulfur.

[SOFIA] And here's the part I love — 19 of 31 known PT gene clusters have dropped their own dedicated sulfur enzyme, DndA, entirely. They just freeload off the host IscS.

[DANIEL] Which is evolutionarily tidy. Why carry your own sulfur machinery if the cell already has one running? It's parasitizing the metabolism. From a rigor standpoint, the thing that sells it is the direct protein-protein contact and the dependence on that specific cysteine — that's a mechanistic claim you can break by mutating one residue.

[SOFIA] So now we know the chemistry. 2020 moves to regulation, and this one genuinely surprised me. Salmonella, the Dnd proteins — DndCDE — and the question is how the cell decides where to put these phosphorothioate marks across the whole genome.

[DANIEL] And it turns out the two immune systems interfere with each other. Salmonella also runs Dam methylation, which tags GATC sites. The Dnd proteins get stuck — "parked" — at those Dam-methylated sites. They bind but they can't modify there.

[SOFIA] So the restriction-modification world and the phosphorothioate world are literally competing for the same DNA. The Dnd machinery gets sequestered at sites it can't use, and that forces the phosphorothioate marks to redistribute to other motifs — GAAC, GTTC — which it finds by DNA shape, not just sequence.

[DANIEL] And the number that makes it a real finding: genome-wide phosphorothioate density stays clamped at around fifteen hundred per million nucleotides. The locations move around, but the total is held constant.

[SOFIA] Density homeostasis! The cell cares how much modification there is, not exactly where. That's a regulatory logic nobody would've guessed from the chemistry alone.

[DANIEL] It reframes the whole thing. These aren't rigid sequence-reading systems. They're plastic, they compete, and they're tuned to a set point.

[SOFIA] Okay, and this is where the arc turns from understanding to using. Because once you understand these defenses, you can engineer around them — and that's the last three papers. 2021, Rubin, is ET-Seq and VcDART. In a mixed microbial community — soil, infant gut — they first measure which bacteria will actually take up foreign DNA.

[DANIEL] Which is the honest first question. Not all of them will. ET-Seq quantifies the tractable members before you waste effort.

[SOFIA] And then VcDART uses a CRISPR-associated transposase — so, CRISPR's targeting repurposed as a tool — to drop DNA into a specific spot, in specific species, inside the community, without isolating anything. Editing bacteria in their native context.

[DANIEL] The immune machinery becomes the delivery vehicle. That's the inversion the whole arc was heading toward.

[SOFIA] Then the 2023 Liu review maps endogenous CRISPR systems across probiotic genera and points at therapies — antimicrobial resistance, IBD, oncology, using engineered live bacteria.

[DANIEL] And 2024, Vajente, is almost a love letter to restriction-modification as an engineering problem. Cupriavidus necator, a nonmodel bacterium, chews up incoming DNA. So they natively methylate the plasmid to carry the right passport, and they build a strain with the restriction systems deleted — the ΔRM strain.

[SOFIA] Seventy-fold better DNA delivery. You give the DNA the right stamp, the immune system waves it through.

[DANIEL] Which closes the loop with that 2009 footprinting idea — methylation as the signal that decides what's self. Fifteen years later it's a knob we turn on purpose.

[SOFIA] From watching the arms race to joining it. And where it's heading is community-scale editing, non-model chassis, live therapeutics — all of it riding on reading these ancient defenses well enough to slip past them.

[DANIEL] Understand the immune system, then borrow its weapons. That's the story.

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