Why Phages Fail to Clear Bacterial Populations: Inside Cambridge’s €2.37 Million PHYSIOPHAGE Project
A bacteriophage can be genetically capable of infecting a bacterial strain and still perform very differently depending on the physiological state of the individual cell it encounters. A €2.37 million European Research Council project at the University of Cambridge is investigating this neglected dimension of phage biology, asking how metabolism, stress, cellular heterogeneity and local environmental conditions determine whether infection progresses efficiently or leaves behind bacterial subpopulations capable of surviving phage exposure.
PHYSIOPHAGE, formally titled “Systems Analysis of The Physiological Determinants of Phage Infection Dynamics,” is led by Professor Somenath Bakshi at the University of Cambridge Department of Engineering. The ERC Consolidator Grant has a total EU contribution of €2,374,178 and officially began on April 1, 2026, following signature of the grant agreement on March 16. The project is scheduled to continue until March 31, 2031.
The project addresses a fundamental limitation in the way phage activity is often measured. Conventional experiments usually evaluate large bacterial populations and report averaged quantities such as adsorption rate, burst size, lysis time or overall bacterial killing. Those measurements are extremely useful, but they can obscure what is happening to individual bacterial cells. Two cultures producing the same average result could contain very different underlying infection dynamics, including small subpopulations that respond unusually slowly, resist progression of infection or survive long enough to repopulate the culture.
This problem is especially important because bacteria that are genetically identical are not physiologically identical. Individual cells can differ in growth rate, metabolic activity, energy availability, stress responses, gene expression and access to nutrients. Those differences become even greater in natural and clinical environments, where bacteria encounter oxygen gradients, nutrient limitation, immune pressure, antibiotics and physical structures such as biofilms rather than the homogeneous nutrient-rich conditions generally used in laboratory cultures.
Bacteriophages are completely dependent on their bacterial hosts for replication. Unlike an antibiotic molecule that can retain an intrinsic chemical activity independently of the metabolic machinery of its target, a phage must hijack the bacterial cell to reproduce. Genome injection, transcription, viral protein synthesis, genome replication, assembly and eventual lysis all rely to different degrees on resources provided by the infected bacterium.
A bacterium growing rapidly with abundant energy and biosynthetic resources may therefore provide a very different environment for viral replication from a slowly growing or stressed cell. The same phage–bacterium combination could consequently produce different infection outcomes under different physiological conditions even when neither the bacterial genome nor the phage genome has changed.
PHYSIOPHAGE is designed to determine exactly how those differences arise. According to the European Commission’s project description, the researchers will investigate three connected questions: how metabolism and stress responses determine the progression of infection within individual cells, how physiological variation between bacterial cells influences phage efficacy across entire populations, and how the physical and biochemical environment surrounding bacteria alters the final outcome of infection.
The programme builds on single-cell technologies already developed by Bakshi and colleagues at Cambridge. Rather than observing only the final collapse or survival of a bacterial culture, the researchers can track individual infected cells through the phage life cycle and record the timing of distinct stages of infection.
Their high-throughput imaging platform combines microfluidics with time-lapse microscopy and fluorescent reporters that reveal both the progression of viral infection and properties of the bacterial host. Thousands of individual bacterial cells can therefore be followed while phages inject their genomes, express viral genes and eventually cause host-cell lysis.
This level of resolution is important because the average behaviour of an infected population may not correspond closely to the behaviour of any particular cell. A population described as having a lysis time of a certain number of minutes could actually contain cells lysing across a broad temporal distribution. Similarly, an average burst size can conceal substantial variation in the number of viral particles produced by different infected bacteria.
Recent work from Bakshi's group and collaborators has begun demonstrating how extensive that variability can be. A single-cell study of the bacteriophage T7 life cycle developed a time-resolved imaging assay capable of following infection from genome entry through growth arrest, viral gene expression and cell lysis in individual bacteria. The experiments revealed substantial heterogeneity between infection events even when genetically similar phages infected genetically similar bacterial hosts.
These observations raise a fundamental question for therapeutic phage development. If phage efficacy depends strongly on bacterial physiology, measuring whether a phage forms plaques or reduces turbidity under one optimized laboratory condition may provide only part of the information needed to predict its activity inside an infection.
Clinical bacterial populations are rarely uniform. Chronic lung infections, urinary tract infections and persistent wounds, which Bakshi has specifically identified as examples of dynamically changing environments relevant to the project, can contain bacteria occupying very different physiological states. Some may be actively dividing while others experience nutrient starvation, oxidative stress, altered oxygen concentrations or exposure to antibiotics.
Biofilms amplify this problem because bacteria separated by only a small physical distance can encounter very different microenvironments. Cells near the surface of a biofilm may receive abundant nutrients and oxygen, while bacteria deeper inside the structure can be metabolically slow and exposed to accumulated waste products. The biological environment encountered by an infecting bacteriophage can therefore change over micrometre-scale distances.
PHYSIOPHAGE will investigate whether these differences create what the project describes as phage-tolerant subpopulations. Phage tolerance in this context should be distinguished from classical genetic phage resistance. Resistance generally involves heritable changes such as receptor mutations or acquisition of antiviral defence systems that prevent successful infection. Physiological tolerance can instead arise when the state of a bacterial cell temporarily reduces the effectiveness of infection without necessarily changing its genome.
That distinction could have important therapeutic consequences. If a bacterial cell survives because of a reversible physiological state, descendants produced after conditions improve may once again become susceptible to the same phage. However, those surviving cells could provide a reservoir from which the bacterial population recovers after an apparently successful initial treatment.
Similar phenomena are well established in antibiotic research. Bacterial persister cells can survive high antibiotic concentrations without carrying conventional resistance mutations because they occupy transient physiological states that make the antibiotic less effective. Bakshi's laboratory has extensively investigated antibiotic persistence and transient tolerance, and PHYSIOPHAGE extends related systems-level questions into bacteriophage infection.
The parallel is not exact because phages and antibiotics operate through fundamentally different mechanisms. A phage must replicate within its host, meaning that host physiology can influence not only whether bacterial killing occurs but also viral amplification itself. A poorly productive infection may consequently affect subsequent rounds of treatment by reducing the number of progeny phages available to infect neighbouring bacteria.
The timing of lysis provides a good example of why population averages can become misleading. A phage that lyses bacteria rapidly may appear advantageous because it releases progeny sooner. However, lysis that occurs too quickly can reduce the time available to manufacture new viral particles, potentially decreasing burst size. Conversely, longer infection cycles may produce more phages but delay their release and increase the probability that environmental conditions change before replication is complete.
A 2026 study involving Bakshi and colleagues explored this problem computationally by examining variability in lysis time and burst size. Published in PLOS Computational Biology, the work showed that a phage's average growth rate in isolation did not necessarily predict its success in simulated competition with another phage.
The models found that variability itself could sometimes provide an advantage. Phages with more heterogeneous lysis times could remain competitive under resource-limited conditions even when their average growth characteristics appeared less favourable. The study demonstrated that mean values alone may not be sufficient to predict phage fitness and reinforced the importance of measuring distributions of infection outcomes rather than only population averages.
PHYSIOPHAGE aims to connect these theoretical insights with experimental measurements of bacterial physiology. The programme will generate detailed single-cell datasets and use them to construct data-driven models capable of predicting how infection progresses according to the physiological state of the host.
A central component of this strategy is microCOSM, a platform developed to connect behaviour observed in single bacterial cells with the dynamics of larger populations. Single-cell microscopy can reveal mechanistic relationships with exceptional precision, but therapeutic questions ultimately concern whether a bacterial population is controlled or eradicated. PHYSIOPHAGE therefore needs a framework capable of linking events occurring within individual cells to outcomes emerging across thousands or millions of bacteria.
The project will use microCOSM to test whether models built from individual infection events can predict population-level behaviour in real time. If a model identifies a particular bacterial physiological state as strongly associated with delayed infection or survival, researchers can examine whether the abundance of that state predicts failure of phage-mediated population clearance.
This approach also creates an opportunity to manipulate bacterial physiology rather than treating it simply as an uncontrollable source of variability. One of the questions defined by PHYSIOPHAGE is whether physiological perturbations can be used to steer infection toward a desired outcome.
A phage that performs poorly against slow-growing bacteria might potentially become more effective if environmental conditions are modified to shift those cells into a more permissive physiological state. Alternatively, particular stresses could expose vulnerabilities that enhance viral replication or reduce bacterial tolerance.
Such physiological steering could eventually become relevant to phage–antibiotic combination therapy. Antibiotics do more than kill susceptible bacteria; they can change bacterial morphology, metabolism, stress responses and resource allocation. Depending on the phage and antibiotic involved, these physiological effects could enhance viral replication, interfere with it or create heterogeneous responses across the bacterial population.
Understanding those interactions mechanistically could make combination therapy more predictable. Instead of testing large numbers of phage–antibiotic pairs empirically and identifying combinations that appear synergistic, researchers could eventually select treatments according to the physiological state they create within the target bacterial population.
The project is also relevant to phage engineering. Much contemporary phage engineering focuses on changing characteristics intrinsic to the virus, such as receptor-binding proteins, host range, anti-defence systems or antimicrobial payloads. PHYSIOPHAGE introduces a complementary design principle: a therapeutic phage should also be optimized for the physiological environments it is likely to encounter.
A candidate that performs exceptionally against rapidly growing bacteria in nutrient-rich culture could be less valuable than another phage that maintains reliable replication across a broader range of host metabolic states. Robustness to physiological variability could therefore become a desirable therapeutic property in its own right.
This does not mean that a universal “stress-proof” phage can necessarily be engineered. Different stages of viral replication require host resources, and severe metabolic limitation may fundamentally restrict what any phage can accomplish. The value of the project lies partly in defining those boundaries and identifying which physiological constraints can realistically be overcome.
The research may also help explain discrepancies between laboratory susceptibility tests and treatment outcomes. A phagogram showing that a bacterial isolate is susceptible to a particular phage is an important piece of information, but the test is generally conducted under a defined set of laboratory conditions. If bacterial physiology strongly alters infection dynamics, susceptibility measured in one environment may not perfectly predict activity in another.
That does not invalidate phage susceptibility testing. Instead, PHYSIOPHAGE could eventually identify which physiological conditions need to be represented in assays designed to predict therapeutic activity more accurately. Tests might incorporate nutrient limitation, stress conditions or other parameters relevant to particular infection sites rather than evaluating phage activity only during rapid bacterial growth.
The work also intersects with another increasingly important theme in microbiology: phenotypic heterogeneity as a driver of treatment failure. Microbial populations can use variability as a form of risk distribution. While most cells may adopt a physiological state optimized for current conditions, small subpopulations can occupy alternative states that become advantageous when the environment suddenly changes.
From the perspective of phage therapy, such diversity means that eliminating the majority of susceptible cells may still leave behind a minority capable of surviving infection for reasons unrelated to genetic resistance. Those survivors can become disproportionately important once the sensitive majority has been removed.
Understanding when these phage-tolerant populations appear is therefore one of PHYSIOPHAGE's explicit objectives. Single-cell imaging is particularly well suited to this problem because rare cells can disappear within averages generated from millions of bacteria but become obvious when individual lineages are tracked directly.
Bakshi's previous work on bacterial persistence provides the methodological background for this approach. A 2021 Nature Microbiology study developed methods for following bacterial lineages across changing environmental conditions and demonstrated how single-cell tracking can uncover relationships between growth history, dormancy and persistence that bulk measurements would otherwise obscure.
PHYSIOPHAGE applies the same general philosophy to viral infection: treatment outcomes emerge from individual cells whose physiological histories and responses are not identical. Understanding those individual trajectories may provide information that cannot be recovered simply by increasing the precision of population-level measurements.
The project also has applications beyond human medicine. Phages are being developed for targeted microbiome modification, food safety, agriculture and industrial biocontrol. In all of these settings, environmental conditions can differ substantially from standard laboratory media, making physiological robustness equally relevant.
A phage deployed against a foodborne pathogen on an agricultural surface, for example, may encounter nutrient-starved bacteria rather than rapidly growing cells. A phage used to reshape the gut microbiome encounters complex nutrient gradients and competition between microbial species. Predicting activity across such environments requires an understanding of how bacterial physiology changes the viral life cycle.
The €2.37 million ERC Consolidator Grant therefore supports a project that is less concerned with discovering another collection of therapeutic phages than with developing a predictive framework for understanding why the same phage can succeed or fail under different conditions.
This distinction makes PHYSIOPHAGE particularly relevant as the phage field moves toward increasingly rational therapeutic development. Genomic screening can identify bacterial defence systems, machine-learning models can help predict host range, and protein engineering can modify phage specificity. Physiological state represents another layer of information that may need to be incorporated into those predictive systems.
A future phage-selection process could therefore involve much more than asking whether a virus recognizes a bacterial species. Researchers may need to determine whether it can infect the specific strain, overcome its intracellular antiviral systems and replicate reliably across the physiological states expected at the infection site.
PHYSIOPHAGE began in April 2026, so it should not be interpreted as a newly announced project or as evidence that Cambridge researchers have already solved the problem of phage tolerance. The ERC award was announced in 2025, while the formal European grant started the following spring and will continue for five years.
Its significance lies instead in the scientific question being addressed. Many failures of bacterial clearance may not arise because the chosen phage is fundamentally incapable of infecting its host, but because infection occurs within a heterogeneous bacterial population in which individual cells provide very different environments for viral replication.
By combining real-time single-cell imaging, microfluidics, microCOSM, quantitative modelling and controlled manipulation of bacterial physiology, Bakshi's team aims to establish how those microscopic differences scale into population-level treatment outcomes. If successful, the project could provide a framework for designing phages and treatment strategies that remain effective not only under ideal laboratory conditions, but across the metabolically diverse and constantly changing environments encountered during real bacterial infections.
Sources :
European Commission CORDIS — PHYSIOPHAGE: Systems Analysis of The Physiological Determinants of Phage Infection Dynamics. Grant Agreement 101229139. CORDIS confirms the €2,374,178 EU contribution, March 16, 2026 signature date, April 1, 2026 start, March 31, 2031 end date, Cambridge host institution and the project's three principal research questions. CORDIS
https://cordis.europa.eu/project/id/101229139
University of Cambridge Department of Engineering — “Cambridge researcher awarded ERC funding to explore virus-bacteria interactions to combat antibiotic resistance,” December 10, 2025. The article describes Bakshi's high-throughput microfluidic infection assay, SyMBac-based image analysis, microCOSM and the intended application to urinary, pulmonary and persistent wound infections. Department of Engineering
https://www.eng.cam.ac.uk/news/cambridge-researcher-awarded-erc-funding-explore-virus-bacteria-interactions-combat-antibiotic
University of Cambridge Department of Engineering — Professor Somenath Bakshi research profile. His research programme includes phage infection kinetics, antimicrobial persistence, high-throughput dynamic phenotyping and therapeutic bacteriophage engineering. Department of Engineering
https://www.eng.cam.ac.uk/profiles/sb2330
Wedd C., Li R., Yunusov T. et al. — “Single-cell imaging of the lytic phage life cycle in bacteria.” The study introduces high-throughput time-resolved imaging of individual T7 infection events and reports substantial heterogeneity between infected cells. The current bioRxiv version is available under CC BY 4.0 and is listed by the Fusco laboratory as accepted in Cell Systems. BioRxiv
https://doi.org/10.1101/2024.04.11.588870
Smith A., Hunter M., Bakshi S. & Fusco D. — “Live and let die: Lysis time variability and resource limitation shape lytic bacteriophage fitness.” PLOS Computational Biology 22, e1013340 (2026). The study shows computationally that variation in lysis time can influence phage fitness and that population averages may not reliably predict competitive performance under resource limitation. PLOS
https://doi.org/10.1371/journal.pcbi.1013340
Bakshi S., Leoncini E., Baker C. et al. — “Tracking bacterial lineages in complex and dynamic environments with applications for growth control and persistence.” Nature Microbiology 6, 783–791 (2021). This earlier work provides methodological background for tracking physiological heterogeneity and persistence at single-cell resolution. Nature
https://doi.org/10.1038/s41564-021-00900-4
.png){kind=logo}
Comments
Post a Comment