The Oldest War on Earth: Bacteriophages and the Return of Phage Therapy

The Hook: A Man in a Coma and a Phone Call to the Navy

In November 2015, Tom Patterson, 68, a professor of psychiatry at the University of California San Diego, fell ill while on holiday in Egypt. What began like food poisoning turned out to be necrotising pancreatitis. A pseudocyst the size of a grapefruit formed in his pancreas, and growing inside it was a bacterium called Acinetobacter baumannii — a pathogen the World Health Organization now ranks at the very top of its priority list.

Patterson was flown to Frankfurt, then to San Diego. Over four months he received essentially everything modern infectious-disease medicine has to offer. The isolate was resistant to every antibiotic tested, including the drugs of last resort. Drains were placed, the cyst leaked into his abdominal cavity, he went septic, and he slipped into a coma. His doctors began discussing withdrawal of care with his wife, the epidemiologist Steffanie Strathdee.

Strathdee did something that sounds absurd in twenty-first-century medicine: she started searching the literature for a treatment the West had abandoned seventy years earlier. She wrote to laboratories in Texas and Maryland, including the U.S. Navy's Naval Medical Research Center. She asked for viruses. Specifically, viruses that eat this exact bacterium.

Within weeks, three laboratories were screening sewage samples, treatment-plant sludge and environmental water for viruses active against Patterson's particular isolate. They found nine. On 15 March 2016 the FDA granted emergency authorisation, and Patterson received a cocktail of bacteriophages — first into the abscess cavities, then intravenously. Three days later he woke from the coma and recognised his daughter. The case was published in 2017 in Antimicrobial Agents and Chemotherapy and is widely regarded as the moment phage therapy returned to Western medicine.

This article is about the viruses in that cocktail. It is about why they are the most numerous biological entities on this planet, why they have been running an arms race with their hosts for billions of years that makes any human security architecture look like a toy — and why, despite spectacular individual cases, there is to this day not a single approved phage drug in Europe or the United States. It is a story about biology, but just as much about evidence, regulation, and the question of what to do with a medicine that is alive and changes.

Part 1: The Most Abundant Life Form Nobody Sees

Ten to the Thirty-First

Bacteriophages — literally "bacteria eaters", usually just phages — are viruses that exclusively infect bacteria. They cannot infect a human cell; they simply lack the machinery. That makes them interesting as a therapeutic and overwhelming as a natural phenomenon.

The common estimate for the total number of phages on Earth is around 10³¹. For comparison: the estimated number of stars in the observable universe is about 10²². There are therefore roughly ten billion times more phages on Earth than stars in the visible cosmos. A millilitre of ocean surface water typically contains on the order of 10⁷ virus particles. Taken together, marine viruses cause something in the range of 10²⁸ to 10²⁹ new infections per day — which works out to roughly 10²³ infection events every second.

These numbers are not mere trivia. Phages kill a substantial fraction of marine bacterial biomass daily and return its contents to the dissolved nutrient pool. They are a principal driver of the global carbon and nutrient cycles. And they are the strongest selective pressure bacteria have faced for something like three billion years — more consequential than heat, cold, desiccation or any antibiotic nature ever invented.

Anatomy of an Injection Machine

The best-known body plan is the tailed phage (order Caudovirales in the older taxonomy). It consists of an icosahedral head — a protein capsule in which the genome is packed under considerable pressure — plus a tail with a baseplate and tail fibres. The pressure is not a metaphor: the internal pressure inside a T4 phage head is on the order of tens of atmospheres, comparable to a champagne bottle, and part of the DNA injection is driven by it.

The course of an infection has an almost mechanical clarity:

1. Adsorption. The tail fibres probe the bacterial surface and bind a specific receptor — a surface protein, a sugar residue of the lipopolysaccharide, a pilus, a flagellum.
2. Injection. The baseplate changes conformation, the tail sheath contracts, an inner tube punches through the cell envelope, and the genome is discharged into the cytoplasm. The capsid stays outside — precisely this observation was the core of the 1952 Hershey–Chase experiment, which showed that heredity resides in DNA and not in protein.
3. Takeover. Within minutes the phage shuts down host transcription and redirects ribosomes, nucleotides and energy to its own replication. It brings no ribosome of its own; it hijacks the host's.
4. Assembly and lysis. New heads and tails are built separately and snapped together. Then two proteins open the cell: a holin perforates the membrane, and a hydrolase called an endolysin cuts the cell wall from inside. The cell bursts, typically releasing 50 to 200 new particles. In E. coli under laboratory conditions the whole cycle takes about 20 to 30 minutes.

Read with a software engineer's eyes, this is a remarkably compact attack pattern: target identification through a specific interface, injection of a payload, redirection of the existing runtime to foreign purposes, self-replication, exfiltration by destroying the host. The analogy is old and not accidental — early virology and early computing borrowed vocabulary from each other.

Lytic, Lysogenic, and Why the Difference Decides Everything Therapeutically

Not every phage kills immediately. Two life strategies are distinguished:

Lytic (virulent) phages always run the cycle described above: infect, replicate, kill. They are the only candidates for therapy.

Lysogenic (temperate) phages have a second option. Instead of replicating at once, they integrate their genome into the bacterial chromosome or maintain it as a plasmid. The integrated phage is called a prophage, and the cell is lysogenic. It keeps dividing and passes the prophage to all descendants, often for thousands of generations. Only when the cell comes under stress — DNA damage, the SOS response — does the switch flip, the prophage excises itself and kills its host.

André Lwoff shared the 1965 Nobel Prize with François Jacob and Jacques Monod in part for elucidating lysogeny. The switch between the two states, the regulation of phage lambda, became the textbook model of gene regulation itself.

For therapy, lysogeny is a disqualifier, for two very concrete reasons. First, a temperate phage may not kill the pathogen at all, merely making it immune to further infection by related phages. Second — and this matters more — prophages frequently carry genes that benefit the host. Cholera toxin of Vibrio cholerae, diphtheria toxin, the Shiga toxin of enterohaemorrhagic E. coli, the scarlet-fever toxin: all are encoded on prophages. A temperate phage can turn a harmless bacterium into a lethal pathogen. Phages can also accidentally package host DNA and deliver it into another cell at the next infection — transduction, one of the main routes by which antibiotic-resistance genes travel between bacteria. An impure phage preparation can therefore aggravate precisely the problem it is meant to solve. Every serious therapeutic phage bank consequently sequences its candidates completely and excludes anything carrying integrase, toxin or resistance genes.

Specificity: Curse and Blessing

Antibiotics are broad-spectrum weapons. A carbapenem hits a large number of Gram-negative species and, incidentally, a substantial part of the gut flora. A phage is the opposite: it often does not even recognise all strains of a species, only a few percent of them. The receptor it binds varies between isolates.

Therapeutically this is both things at once. A blessing, because the microbiome remains essentially untouched — no Clostridioides difficile colitis, no fungal overgrowth, no collateral selective pressure on uninvolved species. A curse, because you must know before treating which phage works against this particular isolate. That requires an antibiogram equivalent, the phagogram, and a sufficiently large bank of characterised phages. It was precisely this step that cost weeks in Patterson's case.

Part 2: Discovered, Celebrated, Forgotten, Survived

Two Discoverers, One Dispute

In 1915 the British bacteriologist Frederick Twort published in The Lancet the observation that micrococcus cultures sometimes turned "glassy", and that the phenomenon could be transmitted through bacteria-free filtrates. Twort formulated several hypotheses, settled on none, and did not pursue the matter — the First World War intervened.

In 1917 the French-Canadian autodidact Félix d'Hérelle described the same phenomenon in dysentery bacilli at the Institut Pasteur, but interpreted it differently: as "un microbe invisible antagoniste du bacille dysentérique", an invisible living antagonist. He coined the name bacteriophage. And, crucially, he drew the therapeutic conclusion immediately: in 1919 he treated four children with bacterial dysentery at the Hôpital des Enfants-Malades in Paris; all recovered. Twort and d'Hérelle disputed priority for years, and the argument was heated enough to earn its own name in the history of science: the d'Hérelle controversy.

The 1920s and 1930s were the first golden age. Phage preparations against dysentery, cholera, staphylococci and streptococci were marketed by companies including Eli Lilly and L'Oréal (then via its subsidiary Laboratoire du Bactériophage). Sinclair Lewis's novel Arrowsmith (1925), awarded the Pulitzer Prize a year later, made the phage-researching physician a literary figure.

The Collapse in the West

Then the field fell apart, for three interlocking reasons.

First: missing fundamentals. Until the 1940s nobody knew exactly what a phage was. Electron micrographs only became available from 1940 onward. Nobody knew phages were host-specific, so preparations were produced against "staphylococci" in general, and people were puzzled by erratic efficacy.

Second: methodological sloppiness. Many preparations had been "preserved" with heat or oxidising agents — which inactivated the phages. An influential report by the American Medical Association's Council on Pharmacy and Chemistry judged the evidence base inadequate in the 1930s, and measured against the studies actually submitted, that verdict was not unfair.

Third, and decisively: penicillin. From 1943 onward there was an agent that was broad-spectrum, reliable, chemically defined, shelf-stable and industrially producible by the ton. Against that profile, a fragile, host-specific, self-replicating viral preparation stood no chance.

Survival in Tbilisi

In the Soviet Union, history took a different course. The Georgian bacteriologist Giorgi Eliava had worked with d'Hérelle in Paris and founded a bacteriological institute in Tbilisi in 1923 devoted to phage research — today's Eliava Institute. D'Hérelle himself spent time there. Eliava was executed in 1937 during the Great Terror, but the institute survived.

Because antibiotics were scarce and expensive in the Eastern bloc, phage therapy was never abandoned there. The Red Army used phage preparations against wound infections and dysentery during the Second World War. To this day, phage preparations are routinely available in Georgia, Russia and Poland, and both the Eliava Institute and the Hirszfeld Institute in Wrocław treat patients from all over the world.

This legacy is ambivalent. It preserved practical know-how, phage banks and manufacturing routine for seventy years. At the same time, most of the evidence collected there is unusable by today's standards: case series without control groups, published in Russian, unblinded, without standardised endpoints. I am of the opinion that this circumstance has contributed more to the persistent scepticism of Western regulators than any biological objection.

Part 3: The Arms Race — Three Billion Years of Security Architecture

Here the topic becomes interesting for anyone who works professionally on attack and defence. Because bacteria are anything but defenceless against phages. On the contrary: a modern bacterial genome typically contains several defence systems, often clustered in so-called defence islands — genomic regions where defence genes accumulate. This statistical observation itself served as a search heuristic from around 2018 onward and led to the discovery of more than a hundred previously unknown immune systems.

The lines of defence can be ordered by the point at which they intervene:

The last row deserves a comment, because it is evolutionarily spectacular. The CBASS system (cyclic-oligonucleotide-based anti-phage signalling system) uses an enzyme that, upon phage infection, produces a cyclic dinucleotide, which then activates an effector that destroys the cell's own membrane or genome. That enzyme is the evolutionary ancestor of human cGAS, the central sensor of innate immunity against cytosolic DNA. Likewise, the TIR domain of the bacterial Thoeris system is the ancestor of the TIR domains of our Toll-like receptors. Our innate immune system is, in essential parts, a descendant of bacterial anti-phage defence. The origin of our immune signalling lies in this war.

And the Other Side Keeps Up

Phages, for their part, are not passive. They chemically modify their own DNA to evade restriction enzymes (T4, for instance, hydroxymethylates and glucosylates its cytosines). They encode anti-CRISPR proteins that block Cas complexes directly — over sixty families have now been described. Some build a proteinaceous compartment in the cytoplasm, a "phage nucleus", that shields the replicating DNA from nucleases.

In 2023 a paper from Rotem Sorek's laboratory at the Weizmann Institute, published in Nature, showed that phages also carry countermeasures against the newly discovered systems: four families of anti-defence proteins (Gad1, Gad2, Tad2, Had1) that specifically disable the Gabija, Thoeris and Hachiman systems. Homologues of these proteins are found in hundreds of phages across many bacterial species. A follow-up study from the same milieu described seven further inhibitor families comprising thousands of genes in 2025.

The picture that emerges: every defence system has a countermeasure, every countermeasure has a counter-countermeasure. There is no end state, no solved security, only an ongoing exchange unconcluded for billions of years. Anyone doing software security knows this dynamic — see The Backdoor at the Heart of Linux: The XZ Attack and the Anatomy of a Supply-Chain Compromise — but biology runs it with an incomparably longer breath and an incomparably larger sample.

And incidentally: CRISPR-Cas, today the most important tool in genetic engineering, is nothing other than one of these defence lines. The "spacers" between the repeats are keepsakes of phage attacks survived. The story of that realisation is the subject of The Programmable Scissors: CRISPR and the Rewriting of Life.

Part 4: What Phages Can Do That Antibiotics Cannot

Four properties distinguish phages categorically from chemical antibiotics.

First: self-amplification. An antibiotic is administered and then degraded; its concentration falls monotonically. A phage replicates where its host bacteria are, and only there. The dose rises at the site of infection and disappears once the target is eliminated. In pharmacokinetic terms this is an entirely different regime — the term is auto-dosing — and it breaks the usual models used to analyse dose-finding studies.

Second: co-evolution. An antibiotic is a fixed molecule; resistance to it is a one-way street. A phage can mutate along. In practice this is exploited through "training": phages are repeatedly cultured on the patient's newly resistant isolate, and variants that bind again are selected.

Third: biofilms. Many phages encode depolymerases that degrade the extracellular polysaccharide matrix of biofilms — precisely the barrier on which antibiotics fail on implants, catheters and joint prostheses. That is why orthopaedic prosthetic joint infections are among the most-studied indications.

Fourth, and this is the field's most elegant idea: phage steering. You select a phage not for how well it kills, but for what price resistance to it exacts.

The canonical example is the phage OMKO1, which binds the outer membrane protein OprM in Pseudomonas aeruginosa. OprM is a component of the multidrug efflux systems MexAB and MexXY — the molecular pumps the bacterium uses to expel antibiotics. The bacterium thus faces a genuine trade-off: mutate OprM to escape the phage, and it simultaneously damages its efflux pumps and becomes susceptible to antibiotics again. Leave OprM intact, and the phage eats it. Chan and colleagues demonstrated this in Scientific Reports in 2016; Burke, Turner and colleagues systematically tested the approach in 2020 against tetracycline, erythromycin and ciprofloxacin.

Phages, then, are not merely a replacement antibiotic. They are a tool for influencing the direction of a pathogen's evolution. I am of the opinion that this is the conceptually most important idea in modern phage research — more important than any individual case success — because it shifts the frame from "killing" to "steering selective pressure".

This perspective becomes more interesting still when one remembers that bacteria coordinate their collective behaviour through chemical signals anyway; see When Bacteria Take a Vote: Quorum Sensing and the Secret Language of Microbes.

The Downsides

For completeness, the drawbacks belong on the same list.

The immune system sees phages. A phage is a large particle of foreign protein. With repeated intravenous administration, neutralising antibodies develop and can reduce efficacy. Manageable for short courses; an open problem for chronic infections.

Endotoxin release. Lysis of Gram-negative bacteria releases lipopolysaccharide. At high bacterial loads this can trigger an inflammatory response — though the same risk exists with bactericidal antibiotics.

Manufacturing. Phages are propagated on their host bacteria. The final product must therefore be purified of bacterial debris and endotoxin, and the preparations are sensitive to temperature, pH and solvents. This, as we shall see, is exactly where the best-known negative trial went wrong.

Part 5: The Evidence, Soberly Assessed

Public perception of phage therapy is shaped by spectacular individual cases. Science needs more. Here is the state of play, ordered by strength of inference.

The Individual Cases

Tom Patterson (2016/2017). Schooley et al. described in Antimicrobial Agents and Chemotherapy the production and use of personalised phage cocktails in a 68-year-old diabetic patient with necrotising pancreatitis complicated by disseminated multidrug-resistant A. baumannii infection. Two laboratories identified nine lytic phages active against the patient's isolate; intravenous and percutaneous administration was associated with reversal of the clinical trajectory and clearance of the infection. Importantly: the patient received antibiotics concurrently, and this is a single uncontrolled case. It is no proof of causation; it is, however, a solid proof of feasibility for personalised manufacture within weeks.

The mycobacteriophages (2019). Dedrick and colleagues reported in Nature Medicine on a 15-year-old cystic fibrosis patient with a disseminated, drug-resistant Mycobacterium abscessus infection following bilateral lung transplantation. Candidates active against her isolate were identified from a collection of more than 10,000 phages; two of them were temperate and were genetically engineered into purely lytic variants by deleting the repressor gene. The three-phage cocktail was given intravenously every 12 hours for 32 weeks. Treatment was well tolerated and associated with objective clinical improvement: closure of the sternal wound, improved liver function, substantial resolution of the infected skin nodules. It was the first administration of genetically engineered phages to a human.

The Large Case Series

Pirnay et al., Nature Microbiology 2024. A Belgian-led consortium centred on the Queen Astrid Military Hospital in Brussels, KU Leuven and Sciensano analysed the first 100 consecutive cases of personalised phage therapy it had facilitated between January 2008 and April 2022 — across 35 hospitals in 29 cities and 12 countries. The results:

• clinical improvement in 77.2% of infections,
• eradication of the targeted pathogen in 61.3%,
• most common indications: lower respiratory tract, skin and soft tissue, bone,
• 26 individual phages and 6 defined cocktails were used, in some cases pre-adapted to the patient's isolate,
• and the decisive finding: without concomitant antibiotics, eradication was about 70% less probable.

That last point deserves emphasis, because it contradicts the popular narrative. On this evidence, phages are not the successor to antibiotics but their partner. The combination beats either monotherapy — plausible mechanistically too, since resistance pathways against the two are often in opposition (see OMKO1).

At the same time, the study is retrospective, uncontrolled, and assembled over 14 years and many centres. These were patients in whom everything else had failed — a selection that can bias in both directions.

The Randomised Trials

This is where it gets uncomfortable.

PhagoBurn (Jault et al., The Lancet Infectious Diseases 2019). The first substantial European randomised, controlled, double-blind phase 1/2 trial: a phage cocktail against P. aeruginosa in infected burn wounds, nine centres in France and Belgium. Result: the phage cocktail was inferior to standard care (silver sulfadiazine). The primary endpoint was reached at a median of 144 hours in the phage group versus a median of 47 hours in the control group.

That is a negative result and must be named as such. The obvious explanation, however, is not a biological but a pharmaceutical failure: the preparation lost titre massively during storage, so that patients received a dose several orders of magnitude too low. The trial was also substantially under-recruited. It illustrates that manufacturing and stabilisation — not biology — are the field's bottleneck.

ELIMINATE, Part 1 (Kim et al., The Lancet Infectious Diseases 2024). The randomised, open-label first part of a two-part phase 2 trial of LBP-EC01, a cocktail of six phages enhanced with CRISPR-Cas3, against E. coli in uncomplicated urinary tract infections. A regimen of two days of intraurethral plus three days of concurrent intravenous dosing (1 × 10¹⁰ PFU) together with oral trimethoprim-sulfamethoxazole was well tolerated, showed consistent pharmacokinetic profiles in urine and blood, and produced a rapid and durable reduction of E. coli with corresponding resolution of symptoms. The blinded, controlled second part is underway.

BX004 (BiomX). An inhaled phage cocktail against chronic P. aeruginosa colonisation in cystic fibrosis. The phase 2b trial in roughly 60 patients (2:1 randomised, double-blind, placebo-controlled, eight weeks) began in July 2025. Following a safety review by the independent Data Monitoring Committee, the study continued with an adjusted dosing regimen; topline results were most recently expected in the second quarter of 2026. As of this article, they are not yet available.

Summary of the Evidence

The honest balance: there is strong plausibility, convincing individual cases, a large uncontrolled series with good numbers — and to date not a single completed, adequately powered, placebo-controlled trial demonstrating efficacy. Whoever calls phage therapy "proven" overstates the case; whoever dismisses it as quackery ignores the data just as much.

Part 6: The Real Obstacle Is Not Biological

Why, after more than a hundred years, is there no approved phage medicine in Europe or the United States? The answer lies in a structural conflict between the nature of the agent and the architecture of drug regulation.

Pharmaceutical approval is designed around a defined, unchanging product tested in a large, homogeneous population against a defined indication. Personalised phage therapy violates all three:

• The agent is not unchanging. It replicates and mutates; "training" on patient isolates is part of the procedure. A batch definition in the classical sense is difficult.
• The population is not homogeneous. Each patient potentially receives a different cocktail, tailored to their isolate.
• The indication is not the disease but the individual strain.

A phage cocktail for a single patient is, in this logic, something like an n-of-1 trial — methodologically legitimate, but regulatorily homeless.

Belgium's answer: the magistral formula. From 2018 onward, Belgium found a route that requires no change to European pharmaceutical law. Phages are treated there as active pharmaceutical ingredients (APIs) whose quality is defined by a national monograph. From these ingredients a hospital pharmacist prepares, on a physician's prescription, a magistral preparation — the classical compounded formulation that exists in every European pharmacy law. This removes the need for a central marketing authorisation for each preparation, while quality and patient safety are secured through the monograph and manufacturing controls. It was precisely this framework that enabled the 100 cases of the Pirnay study.

The model is elegant but no panacea. It does not scale to millions of patients, it generates no registration-grade evidence, and it shifts responsibility onto individual pharmacists and physicians. Broad application still requires defined, fixed cocktails going the classical route — and those are exactly the ones being tested in the randomised trials.

There is also an economic problem that phage therapy shares with all new anti-infectives. A medicine that cures an infection in five days and should, for resistance-stewardship reasons, be used as rarely as possible is a poor business model. Several antibiotic start-ups have gone insolvent in recent years despite having approved products. Phage therapy additionally suffers from the fact that natural phages are hard to patent — one reason why the commercially active companies almost without exception work with genetically engineered phages.

Part 7: What Remains Open

So that this article does not end as a promotional piece, here are the unresolved points, named as precisely as possible:

1. Dose finding. For a self-replicating agent it is unclear what "dose" even means. There are indications of a lower threshold (too few phages never reach the focus of infection) and possibly of an upper one as well (massive lysis, endotoxin, immune reaction). Robust PK/PD models are largely missing; ELIMINATE Part 1 is one of the first clean contributions.
2. Immunogenicity in long-term use. How strongly neutralising antibodies reduce efficacy in chronic infections has not been quantified.
3. Resistance development within the patient. Bacteria become phage-resistant over the course of treatment — that is observed. Whether and how fast cocktails, training and phage steering can counter this is an open question of clinical practice.
4. Standardisation of phagograms. To this day there is no internationally harmonised procedure for determining whether a phage is "active" against an isolate — nothing equivalent to the EUCAST standard for antibiotics.
5. The role of the microbiome. The human gut hosts its own phage virome. What a therapeutic phage dose does there is barely studied.

The Central Takeaway

The core insight here is neither "phages will save us from the antibiotic crisis" nor "phage therapy is unproven nostalgia". It is structural: an agent that lives, replicates and mutates does not fit a system built for unchanging molecules — and the obstacle is not the biology but the architecture around it.

Phage therapy did not fail in the West because it did not work, but because in 1940 there was no theory to explain its variability, and because a simpler competitor existed. It did not fail in PhagoBurn in 2019 because of biology, but because of a preparation's stability in a refrigerator. It does not fail today because of efficacy, but because drug regulation has no concept for a personalised, changing medicine. Three defeats, not one of them for the reason you would assume.

From this follows a broader lesson that reaches far beyond microbiology. When a technology fails repeatedly, it is worth asking whether it really fails at its core function, or at the infrastructure, the measurement methods, the business model or the rulebook. In software this is the same question as: is the approach wrong, or is the toolchain merely missing? Belgium answered it for phages not by changing the biology but by reinterpreting the legal frame — a medicine became an ingredient plus a compounded formulation, and suddenly a hundred patients could be helped.

The second, more concrete prompt concerns security thinking. Bacteria have been fighting a defensive war for three billion years that they never win, and their strategy is not the perfect wall but defence in depth plus diversity: several independent systems per genome, different systems in neighbouring cells of the same population, and as a last resort the willingness to sacrifice the individual cell to save the population. No monoculture, no single point of trust, no belief in finality.

The practical prompt: take one defensive measure in your own system — an authentication, a validation, a backup concept — and ask not whether it is strong, but whether it is the only one of its kind. If an attacker bypasses it, what is left? And is there an instance prepared to sacrifice a component deliberately before the damage goes lateral? Bacteria have been answering both questions in the affirmative for aeons.

A Question to Sit With

Phage therapy was not refuted eighty years ago; it was displaced — by a solution that was easier to produce, store and market, not necessarily by one that was better. Which approach in your own field disappeared the same way: not falsified, merely less convenient than the alternative that happened to be available at the time? And what would have to change in the infrastructure — not in the idea itself — for it to work today?

Cross-References in the Vault

• The Programmable Scissors: CRISPR and the Rewriting of Life – CRISPR-Cas is originally one of the bacterial defence systems against exactly the phages discussed here; the spacers are memories of attacks survived.
• When Bacteria Take a Vote: Quorum Sensing and the Secret Language of Microbes – how bacteria coordinate collective behaviour; some anti-phage decisions are likewise density-dependent.
• Billions from a Few Genes: V(D)J Recombination and the Invention of Antibody Diversity – vertebrate adaptive immunity compared with the adaptive memory of CRISPR; two independent solutions to the same problem.
• The Enemy That Became a Power Plant: Endosymbiosis and the Bacterial Origin of Complex Life – another case in which an intruder became a permanent component; the prophage is the viral variant of the same motif.
• The Machine Made of RNA: The Ribosome, the Ribozyme, and the Translation of the Genetic Code – the ribosome is precisely the machine a phage hijacks, because it brings none of its own.
• The Backdoor at the Heart of Linux: The XZ Attack and the Anatomy of a Supply-Chain Compromise – the same dynamic of attack, camouflage and countermeasure, only in software and on a considerably shorter timescale.

Sources

• R. T. Schooley et al., "Development and Use of Personalized Bacteriophage-Based Therapeutic Cocktails To Treat a Patient with a Disseminated Resistant Acinetobacter baumannii Infection", Antimicrobial Agents and Chemotherapy 61(10), e00954-17 (2017): https://journals.asm.org/doi/10.1128/aac.00954-17
• R. M. Dedrick et al., "Engineered bacteriophages for treatment of a patient with a disseminated drug-resistant Mycobacterium abscessus", Nature Medicine 25, 730–733 (2019): https://www.nature.com/articles/s41591-019-0437-z
• J.-P. Pirnay et al., "Personalized bacteriophage therapy outcomes for 100 consecutive cases: a multicentre, multinational, retrospective observational study", Nature Microbiology (2024): https://pmc.ncbi.nlm.nih.gov/articles/PMC11153159/
• P. Jault et al., "Efficacy and tolerability of a cocktail of bacteriophages to treat burn wounds infected by Pseudomonas aeruginosa (PhagoBurn): a randomised, controlled, double-blind phase 1/2 trial", The Lancet Infectious Diseases 19(1), 35–45 (2019): https://www.thelancet.com/journals/laninf/article/PIIS1473-3099(18)30482-1/abstract
• P. Kim et al., "Safety, pharmacokinetics, and pharmacodynamics of LBP-EC01, a CRISPR-Cas3-enhanced bacteriophage cocktail, in uncomplicated urinary tract infections due to Escherichia coli (ELIMINATE): the randomised, open-label, first part of a two-part phase 2 trial", The Lancet Infectious Diseases (2024): https://www.thelancet.com/article/S1473-3099(24)00424-9/abstract
• B. K. Chan et al., "Phage selection restores antibiotic sensitivity in MDR Pseudomonas aeruginosa", Scientific Reports 6, 26717 (2016): https://www.nature.com/articles/srep26717
• K. E. Burke, P. E. Turner et al., "Phage steering of antibiotic-resistance evolution in the bacterial pathogen, Pseudomonas aeruginosa", Evolution, Medicine, and Public Health 2020(1), 148–157 (2020): https://academic.oup.com/emph/article/2020/1/148/5870287
• E. Yirmiya, A. Leavitt, R. Sorek et al., "Phages overcome bacterial immunity via diverse anti-defence proteins", Nature (2023): https://www.nature.com/articles/s41586-023-06869-w
• Review of nucleotide immune signalling (CBASS, Pycsar, Thoeris, CRISPR): Annual Review of Microbiology (2024): https://www.annualreviews.org/content/journals/10.1146/annurev-micro-041222-024843
• Current review: "Phage therapy: from basic biology to clinical application", Nature Reviews Microbiology (2026): https://www.nature.com/articles/s41579-026-01352-5

← All articles