The search for alternatives to conventional antibiotics often comes with an appealing possibility: perhaps a sufficiently different form of attack could leave bacteria with no viable route to resistance. Bacteriophages can replicate within bacterial hosts, predatory bacteria can invade and consume other bacteria, and clustered regularly interspaced short palindromic repeats (CRISPR)-associated (Cas) systems can be programmed to eliminate cells carrying selected genetic sequences.1,2
These approaches create different selection pressures from those imposed by conventional antibiotics, but selection remains. A recent study of bacterial predation published in ISME Journal provides a clear example.3 Researchers repeatedly exposed Escherichia coli to Bdellovibrio bacteriovorus, a predatory bacterium being investigated as a potential biological antimicrobial. The prey populations consistently evolved resistance, and selection became stronger as predation pressure increased. However, escape was not free: the resistant bacteria lost fitness when the predator was absent.
For development purposes, it is useful to separate two dimensions of resistance risk. Resistance frequency describes how readily bacteria escape. Resistance quality describes what those survivors have become. A durable antimicrobial should ideally limit both the number of escape routes and the advantages available to bacteria that take them.
What Makes a Bacterial Predator Different?
B. bacteriovorus is a small predatory bacterium that attacks a broad range of Gram-negative bacteria. During its predatory cycle, it locates a suitable prey cell, attaches to its surface, enters the periplasmic space between the inner and outer membranes, and establishes a protected compartment in which it grows. It consumes resources from the prey, produces progeny, and eventually lyses the cell to release them.
The predator therefore depends on a sequence of successful biological interactions with its prey. Its close engagement with several parts of the Gram-negative cell envelope has helped generate interest in its potential development as a therapeutic agent, including against bacteria that no longer respond to existing antibiotics.
The complexity of this attack might suggest that resistance would be difficult to achieve. However, the prey may escape by altering an interaction required early in the predatory cycle rather than countering every subsequent stage.
Evidence of this capacity already exists among naturally occurring E. coli strains. In one survey of the 72-strain E. coli Reference Collection, 27 strains were highly resistant to B. bacteriovorus. Production of curli, extracellular amyloid fibers associated with the bacterial surface, accounted for much of that protection. Curli can provide a physical barrier against predation, even when bacteria are growing in suspension rather than within a biofilm.4
One resistant strain, ECOR29, retained protection after the major curli component was deleted. A genetic screen implicated enzymes involved in lipopolysaccharide (LPS) modification, outer-membrane regulation, and other aspects of the cell envelope. The investigators did not identify a single, complete mechanism explaining this curli-independent defense, but the results showed that prey susceptibility depends heavily on properties of the bacterial surface.4
These examples concern defenses already present in diverse bacterial strains. The more pressing therapeutic question is what happens when a susceptible population encounters a predator repeatedly and has an opportunity to adapt.
Watching Resistance Evolve Under Predation
In the recent ISME Journal study, researchers used experimental evolution to examine whether E. coli K-12 could acquire genetically determined resistance to B. bacteriovorus. It did so consistently. Selection for resistance increased with predation pressure, and resistance was widespread after two cycles of exposure.3
The genetic results again directed attention to the outer membrane. A common resistance route involved mutations that led to reduced expression of OmpF, an outer-membrane porin. A less common mutation affected WaaF, an enzyme involved in modifying cell-envelope LPS. That mutation also produced several effects, including downregulation of OmpF.
The findings indicate that changes to the cell envelope can alter susceptibility to predation, demonstrating that prey bacteria can answer a complex biological threat through focused genetic and physiological changes.
Predation pressure also shaped the strength of selection. In this experimental system, greater pressure produced stronger selection for resistance. Whether the same relationship holds across other predators, pathogenic strains, and treatment environments will require direct testing.
The scope of the experiment remains important. E. coli K-12 is a laboratory model, and these findings do not establish how frequently comparable resistance would emerge in pathogenic bacteria, infected patients, or complex microbial communities. They show that genetically determined escape is possible and can appear after limited repeated exposure. Resistance characterization should therefore begin during development, before diminished activity is observed in more advanced settings.
Escape Came at a Price
The resistant E. coli did not gain an unconditional advantage. In the absence of B. bacteriovorus, predation resistance reduced fitness. The mutations that helped the bacteria survive one environment made them less competitive in another.
This trade-off shifts attention from resistance frequency to resistance quality. If escape allows a pathogen to preserve its growth, virulence, transmissibility, and resistance to other treatments, it poses a direct threat to durability. If escape compromises an important bacterial function, the resistant population may become easier to control through another intervention or less able to persist after treatment ends.
The focal study does not establish that resistance to B. bacteriovorus reduces virulence or restores antibiotic susceptibility. Its demonstrated trade-off concerns fitness in the absence of the predator. Determining the clinical value of that cost will require studies across relevant pathogens, environments, and resistance mechanisms.
Even so, the result exposes a limitation in how resistance risk is often discussed. Counting resistant colonies or measuring a change in susceptibility captures only part of the evolutionary outcome. Two antimicrobial candidates could produce similar rates of resistance while creating different risks if escape from one preserves the pathogen’s capabilities and escape from the other imposes a substantial handicap.
Resistance profiling for biological antimicrobials should therefore continue beyond the initial demonstration of escape. Developers need to determine whether resistance remains stable after treatment is removed, whether it changes growth or competitive fitness, and whether it creates susceptibility to another form of attack. The objective is to understand the organism that treatment leaves behind.
The same distinction between resistance frequency and resistance quality becomes visible in other nontraditional platforms, although the mechanisms and consequences of escape differ.
Different Technologies Create Different Escape Routes
Predatory bacteria are not unique in confronting this problem. Phages and CRISPR-Cas antimicrobials can also generate or encounter resistance, although the routes reflect their distinct mechanisms.
Bacteriophages depend on interactions with bacterial receptors to initiate infection. Altering or losing a receptor can therefore protect the bacterium, but the value of that defense depends on the receptor’s normal function. A well-known example involves the lytic phage OMKO1 and multidrug-resistant (MDR) Pseudomonas aeruginosa. OMKO1 uses OprM, a component of the MexAB and MexXY multidrug efflux systems, as its receptor. These systems help the bacterium remove antibiotics from the cell.5
When P. aeruginosa evolved resistance to OMKO1, changes to the efflux mechanism increased susceptibility to drugs from several antibiotic classes. Resistance to the phage interfered with a system contributing to antibiotic resistance, making conventional drugs more effective against the resulting bacteria.
That outcome is not a universal property of phage resistance. It depends on selecting a phage whose receptor performs a clinically important function and on the changes produced under selection. It nevertheless illustrates how antimicrobial design can influence the quality of resistance. A receptor that the pathogen can modify without a meaningful loss offers a less demanding route of escape than one tied to a valuable bacterial function.
CRISPR-Cas antimicrobials move the evolutionary contest inside the bacterial cell. These systems can be programmed to recognize particular genetic sequences, enabling selective killing or removal of antimicrobial resistance genes. Their specificity may limit effects on unrelated bacteria, but the targeting system also contains components that can fail or change.
In an experimental study of an episomally encoded CRISPR-Cas9 antimicrobial targeting E. coli, the most frequent resistance mutations affected the plasmid-encoded Cas9 protein.6 Bacterial genome rearrangements involving mobile genetic elements inactivated Cas9 and allowed cells to escape killing. Restoring an intact Cas9 copy overcame that resistance mechanism.
Other possible routes include changes to the target sequence, guide RNA, or CRISPR-associated machinery, as well as DNA repair, anti-CRISPR proteins, and resistance to the delivery vehicle. Multiplexing several targets may reduce some routes to escape, while using different CRISPR-Cas variants may complicate resistance mediated by anti-CRISPR proteins. Delivery remains part of the evolutionary problem because bacteria may evade the vehicle even when the intracellular antimicrobial remains active.
Predators, phages, and CRISPR-Cas systems therefore present bacteria with different evolutionary problems. Predatory resistance can involve the cell envelope, phage resistance can involve receptors, and CRISPR-Cas escape can affect the targeting system, target sequence, or delivery mechanism. Each platform therefore requires resistance studies tailored to its mechanism; novelty alone does not establish a high barrier to escape.
From Evolution-Proof to Evolution-Aware Development
“Evolution-proof” sets an absolute standard that experiments can rarely establish. Failure to observe resistance under one set of conditions does not show that escape is impossible across other strains, environments, population sizes, treatment intensities, or durations. A more informative development program would define the conditions under which resistance appears and characterize the resulting bacteria.
Experimental evolution can contribute to this work before a candidate reaches clinical testing. Repeated exposure can reveal whether resistance emerges consistently or occasionally, whether independent populations converge on the same mechanism, and whether treatment pressure changes the strength of selection. Genetic and phenotypic analysis can then identify what the organism altered and what those alterations cost, as demonstrated in studies of predatory bacteria and CRISPR-Cas antimicrobials.3,6
These studies should extend across the intended target population. Naturally occurring resistance within the E. coli Reference Collection shows that the behavior of one susceptible strain cannot define an entire species.4 Testing genetically diverse clinical isolates could expose pre-existing defenses, alternative routes of escape, and subpopulations that a treatment may fail to reach.
Developers can then investigate whether the available routes can be constrained. A combination might attack structures or processes that cannot be changed simultaneously without a severe fitness penalty. Sequential treatment could exploit a susceptibility created by the first intervention. A biological agent could be selected or engineered to use a receptor whose modification compromises another important bacterial function. The OMKO1 example provides proof of the underlying principle, although each product and pathogen pairing will require direct experimental support.5
Resistance frequency and resistance quality should both inform development decisions. A low rate of escape is valuable, as is an escape phenotype that grows poorly, loses an important defense, or becomes susceptible to another treatment. By contrast, efficient initial killing may provide limited durability if the surviving population retains its fitness and other resistance mechanisms.
This approach treats evolution as an input to product design. Resistance studies can guide target selection, combination strategy, dosing, and surveillance planning. They can also establish the boundaries of defensible claims. A therapy that has not produced detectable resistance under defined laboratory conditions may have a high barrier to resistance, but those conditions cannot demonstrate that it is resistance-proof.
Making Escape Costly
The available evidence does not prove that an evolution-proof antimicrobial can never exist. However, it does show why the description should be treated cautiously. Bacteria have demonstrated routes of defense or escape against phages, predatory bacteria, and programmable CRISPR-Cas antimicrobials.
The B. bacteriovorus study shows both sides of the problem: E. coli found routes to escape, but those routes reduced fitness when the predator was absent. That cost may ultimately matter more than a claim of evolutionary invulnerability.
An antimicrobial that leaves bacteria no route of escape would be ideal. A treatment that makes each accessible route damaging could also provide durability. The practical goal is to narrow the choices available to the pathogen, measure the consequences of each choice, and design treatment around the vulnerabilities that remain.
