Engineering Phages For Precision Infection Control

[THEO] Okay, picture this. You've got a bacterial infection that laughs at every antibiotic you throw at it. And instead of reaching for another drug, you reach for a virus. A specific virus, one that hunts that exact bacterium and nothing else. That's the promise sitting under today's story, Mara.

[DR. MARA] It's an old promise, actually. Phage therapy predates penicillin. People were using bacteriophages — viruses that infect bacteria — to treat dysentery in the 1920s. Then antibiotics arrived, and the West mostly dropped it. What's changed is antimicrobial resistance is now a real crisis, and we finally have the molecular tools to treat phages as something we can engineer, not just something we fish out of sewage and hope.

[THEO] And that's the arc we want to trace today — turning viruses into tools. Not one paper, but a field figuring out how to go from "found a phage" to "designed a phage." But let's set the table. Mara, when a phage attacks a bacterium, what actually happens at the surface?

[DR. MARA] The phage carries a receptor-binding protein — the RBP. Think of it as the key. The bacterial surface has a specific molecule — a receptor — that's the lock. Could be a sugar on the outer membrane, could be a protein. If the key fits the lock, the phage latches on and injects its genome. If it doesn't fit, nothing happens. That lock-and-key specificity is why a phage might kill one strain of E. coli and completely ignore the strain sitting right next to it.

[THEO] So the specificity is the feature and the bug. Great if you want a precision weapon, a nightmare if you're trying to match a phage to a patient's particular infection in time to help.

[DR. MARA] Right. And there's a second problem. Even if the key fits, bacteria carry immune systems. Restriction-modification systems that chop up incoming DNA, abortive infection systems that make the cell self-destruct rather than let the phage win. So a phage can bind, inject, and still fail. To engineer phages sensibly, you need to understand both layers — the receptor and the immunity.

[THEO] Which is exactly where our story starts. 2021, the BASEL collection out of Harms's group. This one's a foundation-layer paper.

[DR. MARA] It's infrastructure, and I mean that as high praise. They isolated over 120 E. coli phages, but the clever move was the host they used — K-12 MG1655 with the restriction-modification systems deleted. A restriction-free host. That removes the bacterium's DNA-chopping defenses so you don't bias your collection toward only the phages that happen to evade them.

[THEO] So you catch a fairer sample of what's out there.

[DR. MARA] Exactly. They curated 68 phages into a distributable set, plus classical reference phages. Then the real labor: for every single phage, they identified the essential host receptor by testing against more than fifty single-gene mutant bacteria. And they quantified how well each phage plates against eleven different immunity systems — six restriction-modification, five abortive infection.

[THEO] Fifty-plus mutants, times sixty-eight phages, three replicates minimum. That's a wall of Petri dishes.

[DR. MARA] And out of that wall, real discovery. They found LptD — a protein you'd normally think about for outer membrane assembly — acting as the terminal receptor for seven small siphoviruses. Nobody had it pegged as a phage receptor. And they mapped swappable RBP loci to seven different receptors. That word "swappable" is the seed of everything that comes after.

[THEO] Because if the key is a modular part you can swap, you can imagine changing which lock it opens.

[DR. MARA] That's the through-line. BASEL gave the field a characterized parts list. Now — how does the key actually work at the atomic level?

[THEO] 2023, the T5 story. Degroux and colleagues get a cryo-EM structure of the T5 receptor-binding protein, pb5, locked onto its receptor, FhuA. And this one genuinely surprised me. The business end of the RBP — the part that grabs the receptor — is intrinsically disordered before binding. It's floppy.

[DR. MARA] Which is counterintuitive. You'd expect a rigid key. Instead the distal half only folds into shape when it contacts FhuA. Disordered-to-ordered binding.

[THEO] It's like a key that's soft and noodly until you push it into the lock, and only then snaps rigid. And once it does, they see it imposes a 45-degree kink — and they propose that kink is the trigger that starts DNA ejection down into the cell.

[DR. MARA] I'd flag that the kink-cascade-to-ejection part is a proposed mechanism from the structure, not a fully demonstrated causal chain. But the folding-on-binding is right there in the density. And mechanistically it matters: if binding is coupled to a conformational change that fires the genome, then retargeting an RBP isn't just about grip. You might perturb the trigger.

[THEO] So now we've got parts and mechanism. Next the field goes computational.

[DR. MARA] 2024, PhageHostLearn, on Klebsiella — a serious clinical pathogen, lots of carbapenem-resistant strains. They took receptor-binding proteins and the bacterial K-locus proteins — the K-locus encodes the capsule, which is often the receptor here — ran them through ESM-2 protein language model embeddings, and trained an XGBoost classifier to predict, at the strain level, which phage infects which host.

[THEO] Strain level. Not "attacks Klebsiella" but "attacks this patient's Klebsiella."

[DR. MARA] About 82% ROC AUC in cross-validation, and 79% on 28 real carbapenem-resistant clinical isolates. And the number that matters for a clinician — a matching phage in the top-five candidates almost 94% of the time. That turns a fishing expedition into a shortlist.

[THEO] Same year, PhageMaP goes the other direction — inward, into the phage genome itself.

[DR. MARA] A Cas9-RecA system to build barcoded knockout libraries inside phages T7 and Bas63. Genome-scale. You knock out each gene and read the barcodes to see which genes are essential under which conditions across 44 hosts. And they found modular defense inhibitors — phage genes that counter bacterial immunity — that you can transfer between phage genomes.

[THEO] Transferable counters. So it's not just the RBP that's modular now. The anti-immunity gear is a swappable part too.

[DR. MARA] Both layers I mentioned at the top — receptor and immunity — are becoming engineerable. That's the payoff of BASEL characterizing both.

[THEO] Which brings us to 2025, two papers, both about actually rebuilding host range. First, the T3/T7 nanobody retargeting.

[DR. MARA] They engineered T3 and T7 to recognize a nanobody as an artificial receptor. And they learned two things. Whether the phage can evolve to escape depends on how much receptor the host displays — receptor-expression-dependent. And the capsid itself contributes to host range, independent of the RBP.

[THEO] That last bit's a twist. We've been saying the RBP is the key. Turns out the capsid — the container — has its own say in which hosts work.

[DR. MARA] It complicates the tidy lock-and-key picture. Retargeting the RBP alone may not be sufficient. And the second 2025 paper leans right into that engineering challenge — Meta-SIFT.

[THEO] This one I love. They start with deep mutational scanning of the T7 RBP, then use those results to weight a search through metagenomic sequence — real phage diversity out in the environment — for useful motifs. Built 17,000 variants.

[DR. MARA] And 24.5% were active — a strikingly high hit rate for engineered RBPs. Most strategies give you mostly dead variants. Crucially, they got T7 infecting STEC O121 — a Shiga-toxin E. coli — at high salt, an activity you couldn't reach from natural sequence diversity alone.

[THEO] So the arc lands here: from cataloguing wild phages, to seeing the key fold onto its lock, to predicting matches, to editing phage genomes, to designing receptor-binding proteins that reach targets nature never offered.

[DR. MARA] And the honest edge of the field — the capsid contribution, the proposed-but-unproven ejection trigger — tells you where the next work goes. We can design keys now. We're still learning what else the lock is connected to.

[THEO] A good place to leave it. More viruses-as-tools ahead. Stick with us on The Dish.