A team led by Heng Xue, Xinfeng Li, Guibo Rao, and Hang Yang, working with colleagues including Krystyna Dąbrowska of the Polish Academy of Sciences, focused on AbP20, a podovirus that infects Acinetobacter baumannii, a pathogen notorious for its resistance to last-resort antibiotics. Using a mouse infection model, the researchers asked a deceptively simple question: when the body produces antibodies against a therapeutic phage, which of those antibodies actually matter for treatment outcome? The answer, they found, is far from uniform across the viral particle.

The experimental design began with repeated intraperitoneal administration of AbP20, given once daily for seven consecutive days. This dosing regimen induced a robust phage-specific antibody response in the animals, and those antibodies measurably impaired phage therapy. Rather than treating the antibody response as a single entity, however, the team dissected it at the level of individual viral structural proteins, using a genome-guided antigen screening strategy to determine which components of the virion provoked the most therapeutically damaging immune reactions.

The screening results pointed decisively to two proteins: the nozzle and the fiber. Antibodies elicited by these two structural components, rather than those directed against the portal, the capsid, or the adaptor proteins, emerged as the dominant drivers of phage therapy failure in the mouse model. This specificity is significant because it suggests that the immunological vulnerability of a therapeutic phage is not distributed evenly across its architecture but concentrated in the machinery the virus uses to recognize and penetrate its bacterial host.

The mechanistic follow-up experiments clarified why these two proteins matter so much. Fiber-specific antibodies were found to block bacterial adsorption, the initial step in which the phage attaches to receptors on the surface of Acinetobacter baumannii. Without attachment, the infection cycle cannot begin. Nozzle-specific antibodies, by contrast, acted at a later stage, blocking genomic injection, the moment when the phage delivers its DNA into the bacterial cell. In other words, the two antibody populations sabotage the phage at two distinct and sequential checkpoints of the infection process, one preventing docking and the other preventing entry.

But neutralization inside the dish was only half of the story. The researchers discovered that both fiber-specific and nozzle-specific antibodies also promoted the formation of large phage aggregates in circulation. These antibody-mediated aggregates potentiated phagocytosis by macrophages, the scavenger cells of the immune system, and notably this enhancement occurred through an Fc receptor-independent pathway. That detail is mechanistically important: it implies that the immune clearance of antibody-coated phages does not rely on the classical Fc-mediated uptake route, pointing instead to an alternative mechanism by which macrophages recognize and engulf aggregated viral particles.

The dual action of these antibodies, blocking infection on one hand and accelerating immune clearance on the other, constitutes a synergistic assault on therapeutic efficacy. A phage that cannot attach to or inject into its target bacterium is already crippled, and the same antibodies simultaneously shorten the window of time the phage survives in the body by hastening its removal. The study’s authors describe this as a dual synergistic mechanism through which nozzle- and fiber-elicited neutralizing antibodies impair phage therapy, a framing that captures how the two effects compound rather than merely add together.

To probe whether escaping neutralization could rescue treatment, the team turned to an evolved AbP20 variant with improved neutralization escape capacity. The results were instructive but sobering. This variant exhibited antibody-accelerated phagocytosis and host immune clearance comparable to the parental phage, meaning that evolving away from antibody recognition did not spare it from the clearance arm of the immune response. Nevertheless, the variant partially rescued the therapeutic failure caused by neutralizing antibodies, indicating that the infection-blocking effect is the more decisive contributor to treatment loss, while immune clearance imposes a persistent background cost that escape mutations do not readily evade.

The broader implications for phage therapy development are considerable. Clinical phage programs typically monitor neutralizing antibody titers as a single aggregate measure, but this study suggests that the protein specificity of the antibody response is what ultimately shapes therapeutic outcomes. If antibodies against the portal, capsid, or adaptor proteins are comparatively benign while fiber- and nozzle-directed antibodies are the dominant saboteurs, then rational phage engineering, careful phage selection, or dosing strategies that minimize exposure of the receptor-binding apparatus to the immune system could meaningfully extend the viability of phage treatments in patients who mount strong antibody responses.

For a field racing to keep pace with antimicrobial resistance, the findings offer both a warning and a roadmap. The warning is that repeated phage dosing, a common feature of current treatment protocols, can provoke antibody responses that quietly dismantle therapeutic efficacy through mechanisms that standard efficacy testing may not capture. The roadmap lies in the molecular detail: by identifying the nozzle and fiber proteins as the principal immunological weak points of AbP20, the study provides a concrete framework for predicting which phages are most likely to fail under immune pressure and for designing next-generation therapeutics that either evade these antibodies or tolerate their presence well enough to complete the job.

Subject of Research: Antibody specificity in phage therapy failure against Acinetobacter baumannii

Article Title: Antibodies targeting phage fiber and nozzle proteins impair Acinetobacter baumannii phage therapy by blocking infection and promoting immune clearance

Article References: Xue, H., Li, X., Rao, G., Hu, F., Zhong, M., Miernikiewicz, P., Qiu, R., Cao, S., Dąbrowska, K., & Yang, H. (2026). Antibodies targeting phage fiber and nozzle proteins impair Acinetobacter baumannii phage therapy by blocking infection and promoting immune clearance. PLOS Biology, 24(9), e3004009. https://doi.org/10.1371/journal.pbio.3004009

Image Credits: AI Generated

DOI: 10.1371/journal.pbio.3004009

Keywords: phage therapy, Acinetobacter baumannii, antimicrobial resistance, neutralizing antibodies, fiber protein, nozzle protein, macrophage phagocytosis, podovirus, immune clearance, PLOS Biology, phage aggregates, Fc receptor-independent pathway

A team led by Dorentina Humolli and Alexander Harms at ETH Zürich has now introduced a method designed to close this gap. Writing in Nature Microbiology, the researchers present HIDEN-SEQ, short for hidden Acr-enabled transposon-insertion sequencing, a genome-wide approach that systematically links viral genes to measurable phenotypes. The technique adapts transposon-insertion sequencing, or TnSeq, a method that revolutionized bacterial genetics by quantifying the fitness consequences of thousands of random insertions across a genome through comparative deep sequencing. Until now, no broadly applicable equivalent existed for phages, largely because transposition into viral genomes was inefficient and lacked a general selection marker for identifying phage clones that had acquired an insertion.

HIDEN-SEQ solves this problem with an elegant two-part design. First, the mariner transposon Himar1, which inserts randomly at TA dinucleotide sites, was engineered to carry an anti-CRISPR gene, acrVIA1, under a strong constitutive promoter. Second, the bacterial host expresses the CRISPR effector LseCas13a together with a guide RNA targeting the transcripts of an essential phage gene, the major capsid protein. Only phage clones that have picked up the transposon, and therefore express the anti-CRISPR protein, can evade Cas13a targeting and replicate. The choice of a type VI CRISPR system is technically significant: because Cas13a recognizes and cleaves RNA rather than DNA, it bypasses the DNA-based genome protection mechanisms that tailed phages typically deploy, and its collateral cleavage of host and viral RNA aborts infection of any unprotected phage. The name HIDEN-SEQ reflects this molecular game of hide and seek between host-expressed Cas13a and the transposon-encoded anti-CRISPR protein.

To benchmark the method, the team applied it to phage T4, the classic model virus with a 168-kilobase genome encoding nearly 300 genes, around 130 of which remain uncharacterized despite decades of study. The input library achieved 97.2 percent saturation of TA sites across the genome, confirming highly efficient transposition. After outgrowth, insertions in essential genes were strongly depleted, and the analysis concluded that 190 T4 genes are dispensable for infection of the laboratory strain Escherichia coli K-12 under standard conditions, in close agreement with the essentiality map assembled over decades of prior research. Only a handful of genes appeared falsely essential, a discrepancy the authors attribute largely to polar effects, the disruption of neighboring gene expression by an inserted element. Notably, they deliberately omitted a transcriptional terminator from the transposon to minimize this artifact, and showed experimentally that adding one indeed caused strong polar effects.

The resolution of the method extends beyond binary essential or non-essential calls. Genes could be classified as showing weakly, moderately or strongly reduced fitness when disrupted, and in some cases the data resolved essentiality at subgene level. The T4 RNA ligase gene rnlA, for example, tolerated insertions only in its C-terminal domain, consistent with earlier work showing that this region directs the enzyme toward tRNA breaks in the anticodon loop. This depth of information transforms a simple viability screen into a nuanced fitness atlas of the viral genome.

The most striking application of HIDEN-SEQ lies in identifying conditionally essential genes, those required only under specific challenges. When the T4 library was propagated on hosts expressing particular bacterial defence systems, insertions in the corresponding antidefence genes were selectively depleted. The known antidefence genes rIIA and rIIB, which counteract the RexAB system, and dmd, an inhibitor of the RnlAB toxin-antitoxin system, behaved exactly as expected. The screen also recovered the anti-CBASS protein Acb1 and the anti-DarTG1 factor AdfN. Crucially, when the library was challenged with the DarTG2 defence system, a single uncharacterized gene, tk.4, showed specific depletion. The encoded protein is homologous to the DarG antitoxin of Mycobacterium tuberculosis and carries a macrodomain typical of ADP-ribosylglycohydrolases, leading the team to name it AdfM, a previously unknown anti-DarTG2 factor that probably removes ADP-ribose modifications from viral DNA.

Conditional essentiality was not limited to immunity. Comparing the library grown in rich Lysogeny Broth with growth in defined M9 minimal medium revealed that the T4 gene vs, encoding a modifier of valyl-tRNA synthetase, was the most strongly depleted under nutrient limitation, and a deletion mutant indeed failed to form plaques efficiently in minimal medium. This demonstrates that HIDEN-SEQ can expose genes needed for phage replication across different physiological states of the host, not only in the context of antiviral defence.

To test portability, the researchers applied the method to two further E. coli phages. Bas37, a close relative of T4, yielded essentiality patterns broadly similar to T4, with informative exceptions: its RNA ligase required both protein domains, and the regA orthologue, a translational repressor, was essential in Bas37 but dispensable in T4, differences validated by targeted gene disruptions. Bas54, a member of the poorly studied Vequintavirinae with 227 predicted genes, proved equally amenable, achieving 91.6 percent TA-site saturation. In Bas54, 145 genes were non-essential, several bona fide tail fibre genes tolerated disruptive insertions, and remarkably, 15 of the 44 essential genes lacked any functional annotation, underscoring how much of the essential phage genome remains obscure even beyond classical models.

The team then pushed the approach into a clinically relevant setting by infecting a panel of uropathogenic E. coli isolates with the T4, Bas37 and Bas54 libraries. For each phage-host pair, multiple viral genes showed strongly reduced insertion counts relative to the laboratory strain, marking candidate antidefence factors. Using DefenseFinder to predict defence systems in the clinical genomes, followed by targeted deletions of the most depleted phage genes and ectopic expression of individual defence systems in E. coli K-12, the researchers matched several candidates to specific branches of bacterial immunity. Among the findings, the Alt ADP-ribosyltransferase of T4 and a beta-alpha glucosyltransferase of Bas37 were implicated in evading the Mokosh type I defence system, with only the T4 version of Alt able to partially restore plaque formation when expressed in trans. Two previously uncharacterized proteins, T4 NrdC.1 and Bas37_0203, were confirmed as antidefence factors against the Septu and Druantia type III systems respectively, despite lacking any known domains or homologues. Bas37_0260, which shares no homology with T4’s type IV restriction inhibitors beyond a conserved N-terminal signal sequence, proved to be a novel inhibitor of McrBC. In Bas54, a MoxR ATPase and a gene of unknown function were both required to overcome the predicted DS-28 defence system of one clinical isolate.

Taken together, HIDEN-SEQ establishes a versatile platform that connects phage phenotypes to underlying genes and molecular mechanisms. Compared with recently developed alternatives such as knockdown-based CRISPRi approaches, which failed to detect the essentiality of roughly a quarter of known essential T4 genes, or targeted mutagenesis methods that require extensive custom construct design, HIDEN-SEQ demands no prior knowledge of the target phage’s genes, relies on a single guide RNA for selection, and produces high-density insertion libraries that can be reused across many hosts and conditions. The authors note that related transposon-sequencing approaches reported for phages of Pseudomonas and Prodigiosinella during the study’s revision confirm the broad applicability of phage transposon mutagenesis, while suggesting that individual viruses may require tailored adjustments such as phage-specific promoters or alternative transposon systems to overcome DNA modifications. With its capacity to reveal dispensable genome regions useful for phage engineering and to assign functions to the dark matter of viral genomes, HIDEN-SEQ promises to accelerate progress in phage therapy, microbial ecology and biotechnology at a moment when new weapons against antibiotic-resistant bacteria are urgently needed.

Subject of Research: A transposon-insertion sequencing method for mapping bacteriophage gene essentiality and discovering antidefence genes

Article Title: Systematic mapping of bacteriophage gene essentiality with HIDEN-SEQ

Article References: Humolli, D., Piel, D., Ransome, J., Bausch, K., Tschudin-Sutter, S., Ortelli, M., Dehio, C., Veening, J.-W., & Harms, A. (2026). Systematic mapping of bacteriophage gene essentiality with HIDEN-SEQ. Nature Microbiology. https://doi.org/10.1038/s41564-026-02455-8

Image Credits: AI Generated

DOI: 10.1038/s41564-026-02455-8

Keywords: bacteriophages, HIDEN-SEQ, transposon-insertion sequencing, gene essentiality, phage therapy, CRISPR-Cas13a, anti-CRISPR, antidefence genes, phage T4, functional genomics, bacterial defence systems, microbiology

Previous metagenomic surveys of human breast milk had revealed an abundance of viral genetic material, much of it phage DNA, hinting that these viruses might be active participants in the early-life microbiome. But metagenomics alone cannot distinguish between free-floating DNA fragments, prophage genomes quietly embedded in bacterial chromosomes, and fully formed virions capable of infection. Demonstrating the presence of actual phage particles requires a different approach: culturing a susceptible bacterial host and showing that something in the milk sample can lyse it, producing the clear plaques on a bacterial lawn that have been the calling card of phage hunters for more than a century.

That is precisely what the team, led by Steven De Soir of UCLouvain and the Queen Astrid Military Hospital in Brussels, together with colleagues at KU Leuven, the University of Minho, and Monash University, set out to do. Working with breast milk samples collected from a lactating donor under written informed consent, the researchers succeeded in isolating six distinct bacteriophages capable of infecting P. aeruginosa, along with a co-occurring strain of the bacterium itself. The identification of the Pseudomonas isolates was carried out with the support of the microbiology laboratory of the Cliniques universitaires Saint-Luc in Brussels, ensuring that the host bacterium recovered from the milk was unambiguously characterized.

The genomic characterization of the six phages revealed a striking diversity. Four of the isolates proved to be strictly lytic, meaning they replicate inside and destroy their bacterial hosts rather than integrating into the host genome. The remaining two were temperate phages, the kind that can choose between a lytic cycle and a lysogenic lifestyle in which their genome becomes a quiet passenger within the bacterium. Across the six isolates, the genomes spanned multiple genera, indicating that a single sample of human breast milk can harbor phages drawn from several branches of the viral tree of life, all targeting the same bacterial species.

Among the four lytic phages, three showed highly similar host ranges and produced comparable plaque morphologies, suggesting they may be closely related despite being recovered as separate isolates. This kind of redundancy is a familiar feature of phage ecology, where dominant viral lineages can appear repeatedly within a single ecological niche. Whether these three represent genuinely distinct entities or variants of a locally dominant phage is one of the questions the study leaves open for future work, but their coexistence in the same sample underscores how concentrated phage activity against a single host species can be within this body fluid.

Host-range profiling, the standard method for determining which bacteria a given phage can infect and kill, showed that the newly isolated phages display broad activity against P. aeruginosa. Critically, that activity included the co-occurring P. aeruginosa strain recovered from the very same milk sample. This observation is significant because it demonstrates that the phages and their bacterial host were not merely passive co-travelers; the viruses were capable of infecting the bacterium found alongside them, raising the possibility of active phage-host dynamics unfolding within the milk itself.

Perhaps the most intriguing laboratory finding concerns synergy. The researchers observed synergistic interactions between pairs of lytic phages, as well as between lytic phages and one of the temperate isolates. In practical terms, synergy means that combinations of phages can suppress bacterial growth more effectively than the sum of their individual effects, a property that underpins the design of phage cocktails in therapeutic applications. That such interactions should be detectable among phages recovered from breast milk suggests that the viral community in this fluid is not a random assortment but potentially a functionally coordinated ensemble shaped by coevolution with its bacterial targets.

The implications extend in two directions. The first concerns neonatal biology. P. aeruginosa is an opportunistic pathogen of considerable clinical importance, notorious for its antibiotic resistance and its role in hospital-acquired infections, particularly in immunocompromised patients and those with cystic fibrosis. Yet Pseudomonas species are also documented members of the early-life microbiome, and the presence of both the bacterium and its phages in breast milk raises the possibility that infants ingest a miniature predator-prey system with every feed. Such a system could influence which bacterial strains colonize the infant gut, modulate bacterial population sizes, and even shape the developing immune system’s encounter with microbial antigens. The authors note that these findings elicit questions toward phage-host dynamics in human breast milk and highlight implications for neonatal microbial colonization and immune modulation.

The second direction concerns phage therapy and biotechnology. Phages active against P. aeruginosa are of intense interest because of the pathogen’s inclusion on lists of priority antibiotic-resistant bacteria. Discovering that novel, genetically characterized lytic phages with broad host ranges can be recovered from an unexpected and easily accessible human source adds to the growing catalog of candidate therapeutic agents. The Belgian research community involved in this study, including the Laboratory for Molecular and Cellular Technology at the Queen Astrid Military Hospital, has a long-standing track record in phage therapy research, and the isolation methods demonstrated here could inform future searches for phages in other human-associated niches. The work was supported by Innoviris, the Brussels regional research funder, and by the Belgian FRS-FNRS, with fellowship support for early-career researchers.

There are, of course, important caveats. The study reports findings from breast milk of a single lactating woman, so the prevalence of active Pseudomonas phages in breast milk across the broader population remains unknown. It is also not yet clear whether the phages and the bacterium were produced locally within the mammary gland, transported there from elsewhere in the maternal body, or introduced through environmental contact. The entero-mammary pathway, by which maternal gut immune cells and microbes are thought to traffic to the breast, offers one plausible route, but confirming the origin of these phages will require further study. What the work establishes beyond doubt is that functional phage particles targeting a clinically significant pathogen exist in human breast milk, transforming what was once a metagenomic hint into a demonstrable biological reality. As researchers begin to probe the dynamics between these viruses, their bacterial hosts, and the infant recipients of both, breast milk may prove to be one of the most accessible windows into the hidden ecology of the human virome.

Subject of Research: Isolation of active Pseudomonas aeruginosa bacteriophages and their bacterial host from human breast milk

Article Title: Isolation of active Pseudomonas aeruginosa bacteriophages and of their host from human breast milk

Article References: De Soir, S., Sáez Moreno, D., Wagemans, J., Glorieux, A., Lavigne, R., Pirnay, J.-P., Merabishvilli, M., Barr, J. J., Van Bambeke, F., & De Vos, D. (2026). Isolation of active Pseudomonas aeruginosa bacteriophages and of their host from human breast milk. npj Viruses. https://doi.org/10.1038/s44298-026-00235-z

Image Credits: AI Generated

DOI: 10.1038/s44298-026-00235-z

Keywords: bacteriophages, Pseudomonas aeruginosa, human breast milk, virome, neonatal microbiome, phage therapy, lytic phages, temperate phages, host range, phage-host dynamics, microbiome development, antibiotic resistance

The research, led by Avery J. C. Noonan and colleagues at Lawrence Berkeley National Laboratory and collaborating institutions, was built on an unusually rigorous foundation. The team systematically optimized their workflow over more than 13.2 million individual training runs, evaluating over 200 parameter settings across six published phage–host interaction datasets. Together these datasets encompass 115,037 experimentally determined interactions spanning 949 bacterial strains and 518 phages, drawn from Escherichia coli, Klebsiella, Pseudomonas and Vibrionaceae hosts. The proportion of infectious interactions within each dataset ranged from as little as 2.1 percent to 36.2 percent, a severe imbalance that mirrors the reality of phage biology, where most phages fail to infect most strains they encounter.

The core technical challenge the researchers faced is one familiar to anyone working at the intersection of genomics and machine learning: dimensionality. Each bacterial or phage genome can contribute thousands of numerical features—protein families, gene fragments or sequence k-mers—while the phenotypic data describing which phage infects which strain is sparse, binary and heavily skewed toward negative outcomes. This combination invites overfitting, in which a model memorizes quirks of its training data rather than learning generalizable rules. Many earlier tools sidestepped the problem by restricting features to genes already known to mediate infection, such as phage tail fibres or host capsule-synthesis loci. That strategy works within a single, well-studied genus but cannot transfer to emerging pathogens or poorly characterized phages, and it forecloses the discovery of novel infection determinants.

The new framework, by contrast, is deliberately phylogeny-agnostic. Genomes are represented as binary presence–absence profiles of protein families, clustered with MMSeqs2 at optimized similarity thresholds. Informative features are then selected through recursive feature elimination, and predictions are generated by an ensemble of gradient-boosted decision trees using the CatBoost algorithm, with class weights computed separately for each phage to compensate for imbalance. The team compared eight machine-learning algorithms, six feature-selection methods and multiple genome-representation schemes, finding that algorithmic choices mattered more than how genomes were encoded. Notably, the simplest representation—short 3-amino-acid k-mers—performed worst, while protein families, longer k-mers and hybrid schemes captured sufficient biological signal.

Performance was assessed through nested cross-validation under three increasingly demanding scenarios: predicting infections of unseen bacterial strains by known phages, infections of known strains by unseen phages, and interactions between entirely novel phage–strain pairs. Across datasets, the area under the receiver operating characteristic curve ranged from 0.67 to 0.94, with Matthews correlation coefficients between 0.13 and 0.54. For E. coli, the model achieved an AUROC of 0.87, essentially matching the 0.86 reported by a previous species-specific method that required phage-specific models and host features based on known E. coli infection mediators. The new approach reached comparable accuracy without any of those constraints. Performance correlated strongly with dataset size, and the weakest results came from the smallest, most imbalanced dataset—a Klebsiella matrix of only 3,658 interactions with 4.9 percent positives—suggesting the limits reflect data availability rather than flaws in the method.

The framework also translated its probability scores into practical phage-cocktail design. Phages were clustered into ‘activity groups’ using HDBSCAN density-based clustering, and the top-ranked phage from each cluster was selected, balancing predicted efficacy against mechanistic diversity. In silico tests across twenty-fold cross-validation showed that model-guided selection outperformed the common heuristic of choosing the most promiscuous phages. For single-phage selection, the models identified active phages for 66.9 percent of E. coli strains, 66.7 percent of Klebsiella strains and 39.4 percent of Vibrionaceae strains—improvements of 1.1-fold, 3.1-fold and 2.7-fold over promiscuity-based choices. Five-phage cocktails achieved bacterial coverage of up to 97.5 percent for E. coli, 87.8 percent for Klebsiella and 57.5 percent for Vibrionaceae, consistently beating baseline strategies.

Crucially, the team did not stop at computational validation. Because the phages in their training data were not available for retesting, they experimentally challenged the model with 52 previously unseen phages from the BASEL collection, spotted against 25 E. coli strains from the ECOR reference collection. Of the 1,300 tested interactions, 1,240 yielded clear phenotypes, and the model’s predictions against this genuinely novel matrix achieved an AUROC of 0.84—only marginally below its cross-validation performance of 0.87. This demonstrated that the framework generalizes to phages it had never encountered and to interaction data generated in a different laboratory with different protocols, a stringent test that many published prediction tools have never undergone.

To verify that the model’s predictions rest on real biology rather than statistical artifacts, the researchers turned to genome-wide random barcode transposon-site sequencing, or RB-TnSeq. They screened a pooled library of 3,804 gene knockouts in E. coli ECOR27 against 19 phages, identifying 51 genes whose disruption altered susceptibility. When these genetically validated infection mediators were compared with the model’s top predictive features, ranked by SHAP values, 68.6 percent showed a connection: 12 mapped directly to predictive features, 10 lay within a three-gene window of one, and 13 were functionally linked through the STRING protein-association database. Direct matches were enriched 5.0-fold above random expectation. Interestingly, the model often picked out a single regulatory gene within a larger biosynthetic pathway—porin regulators such as tsx, fadL and nfrA, or individual genes in lipopolysaccharide and capsule clusters—suggesting it captures pathway-level regulation rather than merely listing every component.

Feature interpretation also surfaced both familiar and unexpected biology. Among the top 25 predictive features in the E. coli dataset were elements tied to cell wall biosynthesis, mobile genetic elements, restriction–modification systems and genes of viral origin, alongside eight features with no clear mechanistic link that may represent uncharacterized infection mediators. Network and synteny analyses pointed to candidate roles for putrescine catabolism and regulation of the known phage receptor fepA, hypotheses the authors flag for future experimental work. Defence systems generally carried negative impacts on infection probability, as expected, while anti-defence features trended positive.

The study is candid about its limitations. Cross-genus prediction remains poor, with leave-one-genus-out AUROC values hovering barely above random, indicating that interaction determinants are largely genus-specific or that protein-family features cannot capture conserved mechanisms across deep evolutionary distances. RB-TnSeq validation is restricted to non-essential genes with strong fitness effects, and the authors suggest CRISPR interference screens as a complement for essential genes. Performance also depends on large datasets of confirmed interactions, which are slow and expensive to produce. Nevertheless, the successful merging of Klebsiella datasets from different research groups points toward community-driven, standardized data generation as a viable path forward. As antimicrobial resistance escalates, the authors argue, this framework—released as the open-source GenoPHI package—offers a scalable computational platform for rational phage-therapy design and precision microbiome engineering across clinical, agricultural and industrial settings.

Subject of Research: Machine-learning prediction of strain-level bacteriophage–host interactions from bacterial and phage genome sequences

Article Title: Phylogeny-agnostic strain-level prediction of phage–host interactions from genomes using machine learning

Article References: Noonan, A. J. C., Moriniere, L., Rivera-López, E. O., Patel, K., Pena, M., Svab, M., Kazakov, A., Deutschbauer, A., Dudley, E. G., Mutalik, V. K., & Arkin, A. P. (2026). Phylogeny-agnostic strain-level prediction of phage–host interactions from genomes using machine learning. Nature Microbiology. https://doi.org/10.1038/s41564-026-02482-5

Image Credits: AI Generated

DOI: 10.1038/s41564-026-02482-5

Keywords: bacteriophages, phage therapy, machine learning, phage–host interactions, genomics, antimicrobial resistance, strain-level prediction, RB-TnSeq, phage cocktails, microbiome engineering, CatBoost, protein families

The discovery began not in a hospital ward but in an environmental survey. The team was assessing bacteria in a wetland that is home to a camp of grey-headed flying foxes, Pteropus poliocephalus, large fruit bats whose roosts make the surrounding waters a rich and underexplored microbial habitat. From those waters the researchers isolated a strain of Klebsiella pneumoniae, designated sequence type ST4919, and with it the phage that preys upon it. That phage, christened Ptero in a nod to the flying foxes, became the founding member of a new family of minimalist viruses.

Klebsiella pneumoniae is a pathogen of considerable clinical concern. It is a leading cause of hospital-acquired infections, including pneumonia, bloodstream infections, and urinary tract infections, and its growing resistance to last-line antibiotics has made it a priority target for alternative therapies. Because many drug-resistant bacteria circulate through the environment, moving between farms, gardens, waterways, and human communities, environmental surveillance serves a dual purpose: it maps the spread of dangerous strains and can uncover natural predators, the bacteriophages, that keep those populations in check.

What sets Ptero and its relatives apart is their minimized architecture. Comparative genomics revealed that the thirteen related phages characterized in the study carry substantially reduced gene sets compared with many of their contractile-tailed cousins. Rather than being a limitation, this simplicity proved to be a scientific asset. When a biological machine has fewer parts, each remaining component becomes easier to assign a function, and the Monash team exploited this by combining structural analysis of the phage virions with biochemical analysis of purified phage proteins.

Contractile phages, members of the broader class that includes the well-studied T4 phage, infect bacteria using a sophisticated injection apparatus. At the tip of the tail sits the baseplate, a multiprotein hub that senses the host cell surface and triggers a dramatic conformational change. That change drives the tail sheath to contract, driving an inner tube through the bacterial cell envelope so that the phage genome can be delivered into the host cytoplasm. In complex phages, the baseplate involves dozens of proteins whose interdependencies are difficult to untangle. The minimalist phages, by contrast, appear to retain only the essential core elements of this device.

By studying Ptero’s simplified version of this machinery, the researchers were able to define what constitutes the irreducible baseplate of a contractile phage. The analysis also brought two virion-associated enzymes into focus. Each phage particle carries two such enzymes, which the study indicates are deployed to kill specific Klebsiella hosts, most plausibly by degrading components of the bacterial cell surface to facilitate attachment and entry. Enzymes of this kind are of particular interest in phage therapy and biotechnology because they can act on the bacterial cell wall independently of the infection cycle itself.

One of the most striking implications of the study concerns host range. The thirteen related phages have evolved to infect diverse bacteria with distinct surface features. Bacterial surfaces vary enormously between species and strains, decorated with different polysaccharides, outer membrane proteins, and other appendages that phages typically use as recognition cues. A phage family that maintains a minimal infection apparatus while accommodating this diversity suggests that the core elements of the baseplate are more versatile than previously appreciated, with host specificity likely tuned by a small number of variable components.

The methodological approach underpinning the work drew on multiple platforms at Monash University. Structural analysis of the virions was performed using cryo-electron microscopy at the Monash Ramaciotti Cryo Electron Microscopy Facility, allowing the researchers to visualize the architecture of the phage particles at near-atomic resolution. Proteomic and biochemical characterization of purified phage proteins was carried out at the Monash Proteomics and Metabolomics Platform, with support from Bioplatforms Australia through the National Collaborative Research Infrastructure Strategy. The study was led by Alex Hall, Hansel Adriel, Yan Li, Pok Man Leung, and colleagues, with senior authors Fasséli Coulibaly and Trevor Lithgow coordinating the structural and microbiological arms of the project.

The broader context of the research is the global crisis of antimicrobial resistance, which is driven by the spread of drug-resistant bacteria through human communities. As conventional antibiotics lose effectiveness, phage therapy has attracted renewed attention as a means of treating infections caused by multidrug-resistant organisms. Recent years have seen a surge of studies isolating and engineering phages against Klebsiella pneumoniae, including efforts to expand host ranges through experimental evolution and to develop synthetic phages with broader coverage of resistant clinical isolates. Fundamental work such as this, which dissects the minimal machinery of infection, provides the mechanistic foundation on which such applied efforts depend.

The minimalist phages also carry lessons for evolutionary biology. Genome reduction is a recurring theme in virology, as viruses shed genes that their hosts or co-infecting viruses can supply. Yet a contractile phage cannot dispense with the machinery needed to pierce a bacterial cell envelope, one of the more demanding mechanical tasks in biology. The fact that Ptero and its relatives accomplish this with a pared-down toolkit indicates that the essential functions of host recognition, cell wall dissolution, and membrane penetration can be achieved with a remarkably small complement of proteins. For phage biologists, these thirteen viruses now serve as a reference scaffold for understanding how the more elaborate baseplates of complex phages are assembled from and built upon this core. For the growing community working to turn phages into therapeutic agents, they offer a simplified platform in which the determinants of host specificity can be mapped, engineered, and ultimately redeployed against the drug-resistant bacteria that continue to emerge from natural environments.

Subject of Research: Minimalist bacteriophages typified by phage Ptero and their infection machinery in Klebsiella pneumoniae

Article Title: A family of minimalist bacteriophages, typified by phage Ptero, have evolved to infect diverse bacteria with distinct surface features

Article References: Hall, A., Adriel, H., Li, Y., Leung, P. M., Lee, H.-C., Schittenhelm, R. B., Dunstan, R. A., Coulibaly, F., & Lithgow, T. (2026). A family of minimalist bacteriophages, typified by phage Ptero, have evolved to infect diverse bacteria with distinct surface features. npj Viruses. https://doi.org/10.1038/s44298-026-00240-2

Image Credits: AI Generated

DOI: 10.1038/s44298-026-00240-2

Keywords: bacteriophage, phage Ptero, Klebsiella pneumoniae, antimicrobial resistance, phage therapy, baseplate, cryo-electron microscopy, viral genomics, host range, cell wall enzymes, environmental microbiology, Monash University

The research team collected water samples from three sites along the Haridwar stretch of the river, all chosen for their high levels of human activity: the Daksheshwar Mahadev temple, the famous bathing ghat of Har ki Pauri, and Chandi Devi Ghat along the Eastern Ganga canal road. Samples from all three locations were gathered on the same day, transported at 4 degrees Celsius in sterile centrifuge tubes, and pooled in equal proportions before being stored at minus 20 degrees Celsius. From this pooled sample, the researchers extracted both metagenomic DNA and RNA using commercial kits, verifying quality by gel electrophoresis and spectrophotometry before constructing sequencing libraries.

The laboratory workflow followed established metagenomic protocols. DNA and RNA-derived cDNA were fragmented into roughly 600 base-pair segments using the KAPA HyperPlus kit, end-repaired, A-tailed, and ligated to sequencing adapters. After library amplification with Illumina primers and purification with Ampure beads, the libraries were quantified with a Qubit high-sensitivity assay and sequenced on an Illumina HiSeq 4000 platform. The run produced 22.4 million paired-end sequences of 151 base pairs each, with a GC content of 59 percent and low duplication rates of 4.9 percent for forward reads and 2.5 percent for reverse reads, indicating a robust and technically consistent sequencing depth suitable for downstream analysis.

Bioinformatic processing began with quality assessment using FastQC, followed by trimming and filtering with fastp and Trim Galore. Taxonomic classification was then performed with Kraken 2, a k-mer-based classifier that matches short sequence fragments against a reference database of complete bacterial, archaeal and viral genomes from NCBI RefSeq. Low-abundance taxa represented by fewer than ten reads or less than 0.01 percent of classified reads were excluded to remove likely false positives. Phylogenetic relationships among the identified phages were reconstructed using ETE 3 for visualization and VICTOR, the genome-based virus classification service of the Leibniz Institute DSMZ. The raw sequence data have been deposited in the NCBI Sequence Read Archive under BioProject accession PRJNA1207791.

The results revealed a strikingly diverse viral community. In total, the sample contained 6,543 distinct viral species, from which 549,515 sequence reads were obtained. Seventeen viral phyla were identified, dominated overwhelmingly by Uroviricota, the phylum that encompasses tailed bacteriophages, with 3,575 hits. Negarnaviricota and Pisuviricota followed with 474 and 460 reads respectively, while Artverviricota and Kitrinoviricota contributed 361 and 338. At the class level, Caudoviricetes, the tailed phages, stood out dramatically with 3,575 bacteriophages, far ahead of Revtraviricetes at 360 and Pisoniviricetes at 287. Among classified orders, Crassvirales led with 867 hits and Kirjokansivirales followed with 795, but the single largest category was the unclassified group, containing 1,488 reads, a clear sign that a substantial fraction of the river’s virome remains scientifically undescribed.

At the family level, one group dominated the dataset to a remarkable degree. Casjensviridae accounted for 516,163 generated sequences, dwarfing every other phage family; Rountreeviridae followed with 10,094 sequences, then Peduoviridae with 5,734 and Autographiviridae with 2,614. Kyanoviridae, Herelleviridae, Schitoviridae, Demerecviridae, Mesyanzhinovviridae and Crevaviridae completed the top ten. Among genera, the unclassified category again led with 300,671 sequences, followed by Chivirus with 205,535 and Copernicusvirus with 8,607. At the species level, the Salmonella phage Chivirus iEPS5 was the most abundant organism in the entire sample with 300,245 reads, followed by the Vibrio phage Valbvirus ValB1MD2 at 124,213 reads and the Staphylococcus phage Andhravirus andhra at 68,758. Abundance fell steeply beyond these top few, indicating a strongly skewed phage population structure.

Many of the dominant phages have well-characterized host ranges documented in earlier studies, and their presence in the Ganga carries ecological significance. Chivirus iEPS5 is a flagellatropic Salmonella phage that depends on the counterclockwise-rotating flagellar filament of Salmonella enterica serovar Typhimurium for adsorption and injects its DNA through the filament during infection. ValB1MD2 kills Vibrio alginolyticus, an emerging foodborne pathogen linked to contaminated seafood. Andhravirus andhra, originally classified among the podoviruses, infects Staphylococcus epidermidis, and its proteins Andhra_gp10 and Andhra_gp14 have been shown to inhibit bacterial growth and disintegrate staphylococcal cell walls. Rosenblumvirus CSA13, another Staphylococcus phage, can prevent biofilm formation across multiple environments, marking it as a candidate biocontrol agent.

The phylogenetic analysis added further texture to the picture. Five Pandoravirus species appeared among the 25 most prevalent viral species, making it the most species-rich genus in that set. Chivirus iEPS5, Zhonglingvirus SAP012, Andhravirus andhra and Rosenblumvirus CSA13 all belong to the class Caudoviricetes, with the first two sharing the family Casjensviridae and thus a closer taxonomic relationship. Other caudoviricetes including Elvirus EL, Efquatrovirus LY0322, which infects the drug-resistant bacterium Enterococcus faecalis, the Agrobacterium phage Atu_ph07 and the Pseudomonas phage Psa21 formed related clusters. Psa21 has previously been recognized as a potential biocontrol agent against bacterial canker of kiwifruit, while Atu_ph07 specifically infects Agrobacterium tumefaciens, the plant pathogen responsible for crown gall disease.

The study’s context stretches back more than a century. In 1896, the bacteriologist Ernest Hanbury Hankin reported antibacterial activity in the waters of the Ganga and Yamuna, noting that the activity was filterable and heat-labile and limited the spread of cholera, an observation made decades before the formal discovery of bacteriophages. Today, with an estimated 10 to the power of 31 viruses on Earth, phages are understood as key regulators of microbial populations and environmental processes. The authors note that the Ganga’s phages, each with particular host bacterial species, appear to reduce the potency and lethality of their pathogenic hosts, and that phage genes linked to bacterial virulence and antibiotic resistance underscore the ecological role of phages in shaping microbial dynamics and bacterial evolution in the river.

The findings arrive at a moment when phage therapy is regaining momentum as a response to multidrug-resistant bacteria, since lytic phages can selectively destroy infected bacterial cells and multiply within their hosts, advantages that become critical as antibiotic resistance genes spread through aquatic environments. The authors caution that their approach has limitations: Kraken 2 relies on exact k-mer matches and can therefore miss divergent or erroneous sequences, read-level classification offers limited resolution at the species level in complex metagenomes, and read counts as proxies for abundance are biased toward larger viral genomes that generate more mappable fragments. Even so, the work provides a genomic baseline for the phageome of the upper Ganga, a system never analyzed in this way before, and points toward phage-driven strategies for environmental monitoring, pollution management and the eventual development of bacteriophage-based therapeutic interventions rooted in the river’s own viral diversity.

Subject of Research: Bacteriophage diversity in the Ganga River ecosystem revealed by metagenomic sequencing

Article Title: Analysis of bacteriophage diversity in Ganga river ecosystem: insights from metagenome-based sequencing

Article References: Katara, A., Chand, S., Chaudhry, V., Vishwakarma, S., Dubey, R. C., Chandra, H., Shilbayeh, S. A. R., Khan, S.-U.-D., Vohra, S., Poddar, N. K., & Khan, S. (2026). Analysis of bacteriophage diversity in Ganga river ecosystem: insights from metagenome-based sequencing. International Microbiology. https://doi.org/10.1007/s10123-026-00901-5

Image Credits: AI Generated

DOI: 10.1007/s10123-026-00901-5

Keywords: bacteriophages, metagenomics, Ganga River, virome, Caudoviricetes, Casjensviridae, water microbiology, phage therapy, antibiotic resistance, viral ecology, freshwater ecosystem, Haridwar

The research team focused on the concept of a phage replication threshold, the minimum density of susceptible host cells required for a lytic phage population to sustain itself. Below this density, the rate at which new phage particles are produced through infection and lysis falls short of the rate at which virions are lost through clearance, diffusion barriers, and failed encounters with hosts. Although this threshold concept was established decades ago in simplified liquid cultures, its value in the complex, spatially structured, and immunologically active environment of the intestinal tract had never been rigorously quantified. Because the density of any invading bacterial species in the gut depends heavily on the composition of the resident microbiota, the researchers hypothesized that increasing microbial complexity should suppress phage replication and, in turn, alter the selection for phage resistance.

To test this hypothesis, the team assembled an elegant experimental system built around two clinically relevant, multidrug-resistant Escherichia coli strains. The first, designated EcST73, was isolated from a healthy traveler and belongs to sequence type ST73; the second, EcST1193, came from a case of neonatal sepsis. Each strain produces a distinct polysaccharide capsule, or K-antigen, which serves as the receptor for a specific lytic phage: φYodit, a Kayfunavirus that binds the K5 capsule of EcST73, and φMoritz, a Vectrevirus that binds the K1 capsule of EcST1193. Both phages belong to the order Autographivirales and possess short, non-contractile tails. The choice of capsule-targeting phages carried a practical advantage: phage-resistant mutants that shed their capsules form visibly translucent colonies on salt-free agar plates, allowing the researchers to track the rise and fall of resistance directly from fecal samples without any genetic engineering of the target population.

The experimental backbone consisted of three mouse lines harboring microbiotas of increasing complexity. Antibiotic-treated mice, whose colonization resistance had been destroyed by ampicillin, permitted the E. coli strains to reach enormous densities, averaging around 10¹⁰ colony-forming units per gram of feces. Gnotobiotic mice carrying a low-complexity microbiota of only a few species supported intermediate colonization levels in the range of 10⁶ to 10⁷ CFU per gram. Conventional specific-pathogen-free mice, with their full complement of resident microbes, mounted strong colonization resistance and kept the invading strains at or below roughly 10⁵ CFU per gram, with EcST1193 in fact being cleared from most animals entirely. Bacterial populations stabilized about 48 hours after oral infection, so the researchers administered a single oral dose of 10⁹ plaque-forming units of the matching phage at precisely that time point.

The results were unambiguous. In antibiotic-treated and low-complexity microbiota mice, phage titers rose and remained detectable for several days, evidence of active replication. In conventional mice, by contrast, phages became undetectable within 48 hours in most animals, indicating that the small E. coli population could not sustain viral reproduction. Strikingly, the phages had no significant effect on total E. coli loads in any of the models. Where replication did occur, resistance followed rapidly. In antibiotic-treated animals, capsule-defective mutants appeared within a day of phage administration and soon dominated the population, after which phage titers declined as susceptible hosts vanished. Sequencing of these resistant clones revealed mutations in genes governing capsule synthesis, transport, and regulation, including emrR, a transcriptional regulator of capsule production, kpsM, required for capsule polysaccharide transport, and, in EcST1193, the neuS and neuE genes involved in polysialic acid biosynthesis for the K1 capsule. Many clones also carried mutations in maltose metabolism genes, a signature of E. coli adaptation to the antibiotic-treated gut.

When the researchers plotted phage titers against bacterial densities across all animals, a clear threshold emerged. Phage replication and the consequent selection for capsule-defective mutants occurred only when susceptible bacteria exceeded approximately 10⁶ CFU per gram of feces, equivalent to roughly 1.3 × 10⁶ bacteria per milliliter of cecal content. Below this line, no resistant mutants were ever observed, because phages simply could not replicate and therefore exerted no selective pressure. Notably, this in vivo threshold is about a hundred times higher than the proliferation threshold of 10⁴ bacteria per milliliter previously determined in vitro for phage T4 and E. coli. The authors attribute this discrepancy to the harsh realities of the intestinal environment: viscous contents impede phage diffusion, structural heterogeneity creates microenvironments where susceptible cells hide, off-target adsorption diverts virions, intestinal flow washes particles away, epithelial cells uptake phages, and digestive enzymes and bile salts inactivate them. Furthermore, fecal counts may overestimate the fraction of cells that are actively growing and expressing the phage receptor, since capsule production is often heterogeneously expressed in Enterobacteriaceae.

Perhaps the most consequential finding came from a competition experiment designed to expose a hidden subpopulation of susceptible cells. The researchers co-infected antibiotic-treated mice with a one-to-one mixture of wild-type EcST73 and a phage-resistant capsule-knockout mutant, each tagged with a distinct antibiotic-resistance marker so the two could be quantified separately. Without phage, the two strains coexisted stably. With phage φYodit, the wild-type population crashed by several orders of magnitude but then stabilized at around 2.4 × 10⁶ CFU per gram, remarkably close to the replication threshold. The phage went extinct in six of ten animals, and after viral extinction the susceptible population rebounded in some mice. Crucially, mini-plaque assays confirmed that most surviving wild-type cells remained fully phage-susceptible even after ten days of coexistence with the virus. In other words, the microbiota is not the only force that can shelter susceptible bacteria; a dense population of resistant cells is sufficient on its own to dilute phage spread below the critical threshold.

This observation carries sobering implications for phage therapy. The researchers showed that flooding the gut with phages at concentrations hundreds of times higher than the susceptible population, roughly 2 × 10⁹ virions against 5 × 10⁶ bacteria per gram of feces, still failed to eliminate the targets. Passive therapy, in which an overwhelming excess of phages drives host extinction without secondary rounds of replication, is therefore unlikely to achieve complete decolonization of the intestine. Instead, phages reduce susceptible populations to a residual level just beneath the replication threshold, where they persist indefinitely and can regrow once phages disappear. The characteristics of this refuge population and the niche it occupies remain open questions, but its existence helps explain why phage cocktails in previous mouse and calf studies reduced, yet never eradicated, intestinal loads of E. coli and Klebsiella pneumoniae.

For clinical translation, the threshold offers a practical decision rule. Human intestinal loads of E. coli vary between 10² and 10⁹ CFU per gram of feces, so phages should not be expected to replicate, or to select for resistance, in patients carrying fewer than 10⁶ bacteria per gram. Conversely, patients whose microbiotas have been disturbed by antibiotics or pathogen-driven dysbiosis often harbor densities well above the threshold, making them plausible candidates for successful phage treatment. Even a transient reduction in bacterial load may suffice to improve outcomes and reduce mortality, as multiple animal studies have demonstrated. The authors acknowledge limitations: only two phage-bacterium pairs and a single phage dose were tested, and both phages were capsule-specific Autographivirales whose properties might influence the threshold value. Nevertheless, comparable effects have been reported with diverse phages using other receptors, suggesting that the microbiota-constrained replication threshold may be a general feature of phage-bacteria interactions in the gut, one that must be respected if phage therapy is to fulfill its promise as a precision tool for microbiota engineering.

Subject of Research: Phage replication threshold in the mammalian gut and its dependence on microbiota-constrained bacterial density

Article Title: Microbiota-constrained bacterial density limits phage infection in the gut and allows persistence of susceptible cells

Article References: Bertola, A., Wenner, N., Rocha, L. L., Keys, T. G., & Diard, M. (2026). Microbiota-constrained bacterial density limits phage infection in the gut and allows persistence of susceptible cells. iScience, 29(11), Article 117433. https://doi.org/10.1016/j.isci.2026.117433

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117433

Keywords: bacteriophage, phage therapy, gut microbiota, Escherichia coli, replication threshold, phage resistance, capsule, colonization resistance, antibiotic-resistant bacteria, Enterobacteriaceae, mouse model, microbiota engineering

The first line of defense is often the simplest: denying the virus entry. Phages begin their infection cycle by binding to specific surface receptors, typically proteins, lipopolysaccharides, or other cell-envelope structures that normally serve ordinary physiological functions. Bacteria can evade attachment by mutating or masking these receptors, or by producing extracellular matrices that physically shield them. While receptor blockade is conceptually straightforward, it imposes fitness costs, because receptors that phages exploit frequently carry out essential roles in nutrient uptake or cell signaling. The review emphasizes that this trade-off between viral resistance and bacterial competitiveness shapes which defense strategies dominate in natural populations and, by extension, which resistant mutants are likely to emerge during therapy.

Once phage DNA has entered the cell, bacteria deploy a second layer of protection: restriction-modification systems. These systems pair a methyltransferase, which chemically tags the bacterium’s own DNA at specific sequence motifs, with a restriction endonuclease, which cleaves any incoming DNA lacking the same protective methylation pattern. Unmodified phage genomes are therefore shredded upon injection. Recent work highlighted in the review shows that these systems are far more dynamic than once thought. Studies published in Nucleic Acids Research have revealed nonlinear regulatory dynamics within restriction-modification operons that modulate how susceptible bacteria are to horizontal gene transfer, and horizontal acquisition of Type I restriction-modification systems has been shown to enhance bacterial pathogenicity by methylating genes encoding transcription factors. In other words, these ancient immune modules do more than cut viral DNA; they influence gene regulation, genome evolution, and the movement of antimicrobial resistance genes between pathogens.

The most celebrated bacterial immune system, CRISPR-Cas, adds something the others lack: memory. When a bacterium survives a phage attack, it can capture a short fragment of the invader’s genome and store it as a spacer within the CRISPR array. Transcribed into guide RNAs, these spacers direct Cas nucleases to destroy any DNA or RNA matching the stored sequence, providing adaptive, sequence-specific immunity that can be inherited by daughter cells. The review traces the molecular logic of CRISPR adaptation, from the acquisition of new spacers to the interference stage in which effector complexes patrol the cytoplasm. Type III systems add a further twist: upon detecting foreign RNA, they synthesize cyclic oligoadenylate second messengers that activate accessory proteins, including the membrane protein Csx23, thereby linking sequence recognition to broader protective responses.

Perhaps the most striking recent discovery in the field is that bacteria possess immune signaling circuits that look uncannily like those of human innate immunity. The Thoeris system, for example, relies on a sensor that detects phage infection and produces a cyclic ADP-ribose variant, gcADPR, which activates SIR2-family effector proteins. Structural studies published in Nature showed that these effectors assemble into filaments upon activation and then relentlessly deplete the cell’s supply of NAD+, a molecule essential for energy metabolism. The result is a form of programmed cell death: the infected bacterium sacrifices itself, and by draining NAD+ it starves the replicating phage before progeny virions can be assembled. The CBASS system operates on a similar principle, using cyclic nucleotide second messengers to switch on effector enzymes such as the phospholipase CapV, which has been shown to disrupt the bacterial cell membrane. Pycsar, prokaryotic Argonaute systems, and the SPARSA defense complex all follow comparable logic of nucleotide or nucleic-acid-triggered activation.

A third broad category, abortive infection and toxin-antitoxin mechanisms, embraces the same sacrificial philosophy. Abortive infection systems, first characterized decades ago in dairy bacteria such as the AbiQ mechanism of Lactococcus lactis, halt phage replication by killing or stalling the infected cell. Toxin-antitoxin pairs consist of a stable toxin and a labile antitoxin; phage infection can degrade the antitoxin or trigger DNA damage responses that free the toxin to poison vital processes. The review highlights ShosTA, a system that disrupts purine metabolism to block phage propagation, and a toxin-antitoxin module described in Nature Communications that defends through DNA damage and repair pathways. Other work has shown that a Type II toxin-antitoxin system modulates the Hachiman defense complex, balancing the fitness cost of chronic defense activation against antiphage activity, while the Hachiman complex itself has been shown to execute DNA cleavage as its terminal defensive act.

Crucially, the review stresses that bacterial defenses do not operate in isolation. Genomic surveys reveal that defense systems cluster in so-called defense islands, hotspots of recombination where multiple immune modules accumulate, and experimental studies in Cell Host & Microbe demonstrated that different systems exhibit synergistic anti-phage activity when combined in the same cell. Even more surprisingly, anti-CRISPR proteins themselves can trigger bursts of CRISPR-Cas9 expression that paradoxically enhance phage defense, hinting at layers of cross-regulation that are only beginning to be mapped. CRISPR-Cas systems have also been shown to supervise and regulate other anti-phage defenses, suggesting that the bacterial immune arsenal functions as an integrated network rather than a collection of independent gadgets.

Phages, however, are not passive victims. The review catalogs a growing repertoire of countermeasures, including RNA-based anti-CRISPRs described in Nature that suppress CRISPR-Cas immunity without protein-coding inhibitors, and a phage-encoded anti-CRISPR protein reported in Nature Microbiology that co-opts the host metabolic enzyme enolase to prevent Type III CRISPR signaling. Viruses have also been shown to inhibit TIR gcADPR signaling to overcome Thoeris-like defenses, and helix-turn-helix proteins that bind both RNA and DNA can shut down phage anti-CRISPR control in unexpected directions. Large-scale analyses of human gut phages have uncovered diverse mechanisms for evading Type II CRISPR immunity, indicating that anti-defense strategies are widespread in the virome. On the therapeutic side, researchers are now engineering phages that carry synthetic anti-defense proteins to overcome bacterial immunity barriers, and structure-guided discovery pipelines are accelerating the identification of both anti-CRISPR and anti-defense proteins for biotechnological and clinical use.

The clinical stakes of this arms race are considerable. Phage therapy has advanced through personalized compassionate-use programs, including a multinational retrospective study of 100 consecutive cases, randomized first-in-human trials of nebulized phage cocktails for chronic Pseudomonas aeruginosa infections in cystic fibrosis, and engineered phage treatments for disseminated drug-resistant Mycobacterium abscessus. Yet bacterial resistance to therapeutic phages, often mediated by the very defense systems the review describes, remains a central obstacle, alongside the evolution of resistance through receptor modification. The authors argue that defense-informed phage selection, in which candidate therapeutic phages are matched against the known immune repertoire of the target pathogen, together with rational phage engineering and the design of phage-antibiotic combinations, offers a scientific basis for improving outcomes. Regulatory frameworks, good manufacturing practice compliance, and emerging concerns about human immune responses to phages, including phage-specific immunity that has been shown to impair efficacy in animal models, add further layers of complexity.

By integrating defense system classification, phage counter-defense evolution, and clinical therapy outcomes into a single framework, the review provides a roadmap for the next generation of anti-infective strategies. The message for the field is clear: phage therapy cannot be designed as if bacteria were defenseless. Every therapeutic cocktail is a move in an evolutionary game whose rules are written in restriction enzymes, cyclic nucleotides, SIR2 filaments, and RNA-guided nucleases. Understanding those rules, the authors contend, will determine whether phage therapy fulfills its promise against the mounting threat of antimicrobial resistance.

Subject of Research: Bacterial anti-phage defense mechanisms and their implications for phage therapy

Article Title: Anti-phage defense systems in bacteria: molecular mechanisms and their role in shaping phage therapy strategies

Article References: Khan, S., Ilyas, N., Nawaz, A., Nawaz, H., Khan, M. I., Bhatti, M. T., Hussain, U., & Khan, S. (2026). Anti-phage defense systems in bacteria: molecular mechanisms and their role in shaping phage therapy strategies. Molecular Biology Reports, 53(1), Article 1647. https://doi.org/10.1007/s11033-026-12793-9

Image Credits: AI Generated

DOI: 10.1007/s11033-026-12793-9

Keywords: bacteriophages, anti-phage defense, CRISPR-Cas, restriction-modification, Thoeris, CBASS, toxin-antitoxin, abortive infection, anti-CRISPR, phage therapy, antimicrobial resistance, phage engineering

The core of the Detectron design marries two molecular technologies that synthetic biologists have refined over the past decade. The first is the toehold switch, a de novo-designed RNA regulator first reported by Green and colleagues in 2014. A toehold switch is a carefully folded strand of RNA that hides a protein-coding or functional sequence behind a hairpin structure. When a complementary trigger RNA binds to an exposed single-stranded ‘toehold’ region, the hairpin opens and the previously sequestered sequence becomes active. Toehold switches are prized because they can be designed almost arbitrarily, letting researchers program sensors for nearly any RNA sequence of interest. Their weakness, however, is that their outputs are typically proteins or RNA molecules that are themselves transient, difficult to multiplex, and poorly suited to sequencing-based readouts.

That is where the second technology enters. Retrons are bacterial genetic elements that naturally produce large amounts of single-stranded DNA inside living cells. A retron consists of a reverse transcriptase enzyme together with a noncoding RNA scaffold, traditionally divided into the msr and msd regions, that the enzyme partially copies into DNA. The result is a covalently linked RNA-DNA hybrid molecule, a peculiar product first characterized in the 1980s and later recognized as part of bacterial anti-phage defense systems. Recent years have seen retrons repurposed as tools for genome editing and DNA recording, and the Shipman laboratory has been at the forefront of that effort. Detectrons fuse the two systems: the toehold switch controls when the retron’s reverse transcriptase activity produces its DNA product, and the sequence of that DNA product serves as a barcode unique to the RNA that triggered it.

In practice, a Detectron is a single engineered RNA molecule. In its default state, the switch region folds into a hairpin that blocks productive priming of reverse transcription. When the target RNA appears and hybridizes to the toehold, the structure rearranges, the retron reverse transcriptase gains access to its template, and a short single-stranded DNA barcode is synthesized. Cells that never encounter the target produce little or no barcode; cells that do encounter it accumulate a measurable DNA record. Because the barcode is DNA, it is stable, it survives conditions that would destroy RNA, and it can be pooled, amplified, and sequenced alongside thousands of other samples. This converts what was once a per-sample fluorescence measurement into a massively parallel counting problem, one that modern sequencing platforms are exceptionally good at solving.

Getting the architecture to work well was not trivial, and much of the new paper is devoted to systematic optimization. The authors constructed a large synthetic library of toehold-retron variants, systematically varying structural parameters such as the lengths of the two stems in the hairpin, the loop that connects them, the bulges that interrupt base pairing, and the distance between the stem and the priming guanine where reverse transcription begins. By measuring barcode production for each variant in both the presence and absence of trigger RNA, they generated a quantitative map of how each design choice affects the on/off ratio, the critical figure of merit that determines how cleanly a sensor distinguishes true signal from background.

Machine learning then extracted design principles from that map. Using XGBoost, a scalable tree-boosting algorithm, the team trained regression models on the variant library and evaluated them with cross-validated metrics including the coefficient of determination, root mean squared error, and Pearson correlation. SHAP analysis, a technique that attributes model predictions to individual features, revealed which structural parameters mattered most. The findings were sometimes counterintuitive: shortening the loop below roughly seven nucleotides sharply degraded both signal strength and on/off ratios, while specific combinations of stem lengths and hairpin folding free energy proved far more predictive of performance than any single parameter alone. Armed with these rules, the researchers built Detectron-V2, a redesigned second-generation device that achieved substantially enhanced on/off ratios and high specificity across a panel of phage transcripts.

To demonstrate the platform’s real-world utility, Han and Shipman turned to bacteriophages, the viruses that infect bacteria. Detectrons were programmed to recognize transcripts from specific phage genes, including the major capsid gene gp23 of phage T4 and sequences from phage T7. When a bacterial cell carrying the appropriate Detectron was infected by the matching phage, the viral RNA triggered barcode synthesis; unrelated phages did not. The team even showed that the response was dose-dependent, with barcode enrichment scaling with the multiplicity of infection, meaning the devices do not merely register that an infection occurred but carry quantitative information about how intense it was. Target-site selection within a viral transcript also mattered, with sites predicted to be more locally accessible by RNA folding algorithms yielding better sensor performance.

The most striking demonstration came in pooled cultures. The researchers mixed together bacterial strains, each carrying a Detectron tuned to a different phage, and infected the pooled population. After sequencing the barcode pool, they could determine which phages had infected which strains within a single pot, without any physical separation of the samples. They extended this to ratiometric susceptibility profiling, in which the relative abundance of barcodes reports quantitatively on how susceptible different host strains are to different phages. For the phage-therapy field, where matching the right phage to the right bacterial infection is the central clinical challenge and current methods require growing each phage-host pair separately, a pooled sequencing-based assay of this kind could dramatically accelerate screening.

Beyond virology, Detectrons join a growing family of molecular recording technologies that include CRISPR-based recorders and recombinase state machines, but they occupy a distinctive niche. Where CRISPR recorders typically write into the genome through DNA cleavage events, Detectrons produce their records through reverse transcription of an engineered RNA template, a process that does not require the recorder to target the cell’s own DNA. That makes the approach potentially safer and more modular, and the authors note that the framework enables alternative modes of transcript-based sensing with applications including viral infection detection more broadly. The ability to record transcriptional events in complex bacterial communities, such as microbiomes, suggests uses in environmental monitoring and diagnostics where cells must be observed in place rather than disrupted.

The work also carries the familiar caveats of early synthetic biology. The demonstrations were performed in engineered bacterial strains with inducible promoters and defined phage infections, and porting the technology to other organisms, to natural microbial communities, or to mammalian cells will require re-optimization of the switch architecture for each new context. Han and Shipman have filed a provisional patent application on the technology, signaling commercial interest, and the complete variant libraries, sequences, code, and statistical data have been released to support replication and extension by other groups. Still, the conceptual advance is clear: a programmable RNA sensor whose output is not a fading fluorescent glow but a stable, sequenceable, multiplexable DNA barcode. As the design rules mature, devices of this kind could turn living cells into their own record-keepers, quietly logging which genes spoke and when, for anyone with a sequencer to read the archive.

Subject of Research: Synthetic RNA biosensors that record transcript detection as DNA barcodes via retron reverse transcription

Article Title: Detectrons convert transient RNA sequences into stable DNA barcodes for high-throughput analysis of RNA-dependent processes

Article References: Detectrons convert transient RNA sequences into stable DNA barcodes for high-throughput analysis of RNA-dependent processes. (n.d.). https://doi.org/10.1038/s41587-026-03334-7

Image Credits: AI Generated

DOI: 10.1038/s41587-026-03334-7

Keywords: Detectrons, toehold switches, retrons, reverse transcriptase, DNA barcodes, RNA sensing, synthetic biology, bacteriophages, phage therapy, machine learning, molecular recording, Nature Biotechnology

The logic of the approach is elegantly simple. If a phage attacks only one host species, then adding that phage to a complex community is equivalent to performing a surgical deletion experiment: the targeted bacterium is depleted while the rest of the community, at least in principle, remains untouched. Any subsequent changes in the abundance or behavior of other members can then be attributed to the loss of the depleted organism, providing the kind of causal evidence that observational metagenomics alone cannot deliver. Despite this potential, phages have rarely been harnessed for this purpose, largely because isolating viruses with sufficiently narrow host ranges against ecologically relevant bacteria has been a technical bottleneck.

Hessler and colleagues overcame that bottleneck for the bacterial genus Variovorax, a group of soil and rhizosphere bacteria that has emerged as a key player in plant-associated communities. The team isolated nine phages that replicate in several ecologically important Variovorax species. Host-range testing showed that these viruses are strikingly specific: spot assays across multiple Variovorax strains revealed that only one of the nine phages, V45_66, was able to form plaques on more than a single host. That narrow specificity is precisely what makes the phages useful as experimental scalpels, since a broad-host-range virus would trigger cascading effects that would confound the interpretation of any community-level changes.

The researchers’ first test case came from a long-standing hypothesis about vitamin sharing. Variovorax had been predicted, on the basis of correlation network analyses, to engage in thiamine interdependencies with other community members—meaning that some organisms might rely on Variovorax to supply this essential B vitamin, or vice versa. To test this, the team worked with complex enrichment communities containing between 18 and 31 bacterial species, grown in the laboratory on different carbon sources including guar gum, locust bean gum, arabinogalactan, and konjac gum. These communities were allowed to stabilize over repeated passages, with pairwise Bray–Curtis dissimilarity analysis showing that their composition settled into a steady state by day eight to ten of the experiment.

Into these stabilized communities, the researchers introduced three of their Variovorax-specific phages, either individually or as a cocktail, targeting a strain designated Variovorax SCN45. Genome-resolved metagenomics—a technique that reconstructs the genomes of individual community members from mixed sequencing data and tracks their relative abundances over time—revealed a consistent reduction in Variovorax across all four enrichment cultures following phage application. Crucially, the depletion did not stop with Variovorax. One other population, a bacterium belonging to the genus Leifsonia, declined in relative abundance once its putative partner was gone, exactly as the correlation-based hypothesis of thiamine interdependence had predicted.

The decisive experiment came next. If Leifsonia was collapsing because it had lost access to Variovorax-produced thiamine, then simply adding thiamine to the culture should rescue it. That is precisely what the researchers observed: supplementation with 10 milligrams per liter of thiamine restored the Leifsonia population. Differential abundance screening, combined with statistical testing across the pre-phage, post-phage, and post-phage-plus-thiamine conditions, identified the Leifsonia population and two members of the family Microbacteriaceae as likely dependents on Variovorax-derived thiamine. Metagenome-assembled genome data for Leifsonia were consistent with the amplicon sequence variant analysis, and gene fitness profiling by RB-TnSeq showed that Variovorax SCN45 carries a complete thiamine biosynthesis operon alongside a type IV secretion system, both of which were highlighted in the community context. Together, these lines of evidence confirmed thiamine production as the mechanistic basis for the interdependence between Variovorax and Leifsonia, converting a statistical association into a verified biochemical relationship.

The team then extended the approach to a second, entirely different experimental system: a defined rhizosphere co-culture grown with Arabidopsis plants. In this system, a strain called Variovorax CL14 is known to degrade the plant hormone auxin produced by another bacterium, Arthrobacter CL28. Auxin is a central regulator of plant growth, and its degradation by Variovorax modulates how the plant responds to its microbial residents. Using phages v.cl_11 and v.cl_23, the researchers eliminated Variovorax CL14 from the co-culture. The result was a re-establishment of the stunted root phenotype, demonstrating that removing the auxin-degrading bacterium allowed the hormone to accumulate and suppress root growth. Growth curve experiments confirmed that the phages had no direct effect on Arthrobacter CL28 itself, reinforcing the specificity of the manipulation.

Beyond the specific findings about thiamine and auxin, the study carries a broader methodological significance. The enrichment communities in the experiments were monitored for stability, and haplotype frequency analysis was used to track evolutionary changes associated with phage exposure, giving the researchers a window into both ecological and evolutionary responses to viral attack. Supplementary analyses showed that the summed relative abundance of organisms encoding complete thiamine biosynthesis pathways shifted predictably across the experimental stages, and that the phage-induced perturbations were reproducible across independent carbon-source enrichments. This level of internal consistency strengthens the case that phage-mediated depletion can serve as a general-purpose tool for causal inference in microbiome science, complementing approaches such as gnotobiotic reconstruction, metabolic modeling, and isotope tracing.

The implications reach well beyond the laboratory. Microbiomes govern processes of enormous practical importance—from nutrient cycling in soils and plant health in agriculture to the functioning of bioreactors and, in medicine, the balance of human gut communities. A technique that allows researchers to remove a single species from a functioning community and observe the consequences could accelerate the identification of keystone organisms, the validation of cross-feeding networks, and ultimately the rational design of synthetic microbial consortia. The authors note that their experiments lay the foundation for research employing both wildtype and engineered phages to test interaction hypotheses and for targeted microbiome manipulation. Engineered phages, in particular, could in principle be designed to deliver payloads or to tune depletion dynamics with even finer control.

There are, of course, caveats and open questions. Phage infection itself can trigger physiological changes in the target host before lysis, and phage replication may impose selection pressure on the community, as suggested by the haplotype dynamics observed in this study. Host-range specificity, while an asset here, must be established case by case, and not every ecologically important bacterium has a culturable phage. Nevertheless, the demonstration that nine narrow-host-range phages could be isolated against Variovorax, and that three of them could cleanly test a metabolic interdependence hypothesis inside communities of up to 31 species, marks a substantial advance. The work, supported by the U.S. Department of Energy’s m-CAFEs Science Focus Area at Lawrence Berkeley National Laboratory, suggests a future in which viruses—long studied as pathogens of bacteria—become routine instruments for dissecting the hidden wiring of microbial ecosystems, one species at a time.

Subject of Research: Phage-based depletion of Variovorax to test microbial interactions in intact microbiomes

Article Title: Phage-based depletion of Variovorax reveals interactions within intact microbiomes

Article References: Hessler, T., Chiniquy, D., Adler, B. A., Hoff, J., Tucker, E., Huddy, R. J., Sachdeva, R., Lei, S., Harrison, S. T. L., Barrangou, R., Diamond, S., Deutschbauer, A. M., & Banfield, J. F. (2026). Phage-based depletion of Variovorax reveals interactions within intact microbiomes. Microbiome. https://doi.org/10.1186/s40168-026-02539-x

Image Credits: AI Generated

DOI: 10.1186/s40168-026-02539-x

Keywords: bacteriophages, microbiome, Variovorax, thiamine, cross-feeding, metagenomics, rhizosphere, Arabidopsis, auxin, Leifsonia, microbial interactions, phage therapy