A bacterium can become more resistant to one antibiotic while simultaneously becoming more vulnerable to another. Although antibiotic resistance is generally associated with the acquisition of defenses that make infections harder to treat, those defenses can carry biological costs. A mutation that protects bacteria against one antimicrobial threat may change their metabolism, alter their outer surface, or interfere with physiological processes that contribute to survival under other conditions. Such trade-offs raise an intriguing possibility: Could the adaptations that enable bacteria to withstand antibiotics also provide opportunities to defeat them?
Researchers have begun investigating how to exploit these vulnerabilities through antibiotic combinations, interventions that modify bacterial metabolism, and strategies that influence the evolution of resistance. Central to this effort is collateral sensitivity, a phenomenon in which resistance to one antibiotic is associated with increased susceptibility to another. Understanding why these relationships emerge, and whether they can be predicted reliably, could help identify new uses for existing antimicrobial agents.
A recent study of Klebsiella pneumoniae offers a revealing example. Researchers found that losing a component of the bacterial outer surface helped the pathogen survive exposure to the carbapenem antibiotic meropenem, but also made it substantially more susceptible to aminoglycosides, another class of antibiotics. The investigators traced these opposing effects to changes in bacterial metabolism, oxidative stress responses, and cell-envelope properties, and demonstrated that the resulting vulnerabilities could be exploited experimentally. The findings provide a starting point for considering how the biological consequences of resistance might inform future treatment strategies.
How Surface Remodeling Changes Antibiotic Resistance
Carbapenems are among the antibiotics used to treat serious infections caused by multidrug-resistant Gram-negative bacteria. Their primary mechanism involves disrupting the synthesis of peptidoglycan, the structural material that helps maintain bacterial cell integrity. K. pneumoniae, an important cause of difficult-to-treat healthcare-associated infections, can resist these drugs through several established mechanisms, including production of carbapenemase enzymes that destroy the antibiotics, reduced drug penetration across the outer membrane, and increased antibiotic efflux. However, changes in bacterial physiology can also influence how organisms respond to antimicrobial exposure.
Researchers led by Jiangqing Huang, Da-Wei Wei, and Jie Feng at the Institute of Microbiology of the Chinese Academy of Sciences, working with collaborators including scientists at the University of Queensland, investigated a different route to enhanced carbapenem resistance. In their study, published in Nature Communications in October 2026, the team examined how K. pneumoniae populations responded to repeated exposure to subinhibitory concentrations of meropenem.1 They initially monitored changes in susceptibility to bacteriophages, viruses that infect bacteria, to investigate whether antibiotic exposure was altering the bacterial surface.
The researchers found that meropenem exposure selected variants lacking O antigen, a carbohydrate component of the lipopolysaccharide (LPS) molecules that form part of the outer membrane of Gram-negative bacteria. O antigen contributes to the properties of the bacterial surface, influencing interactions with external molecules, bacteriophages, and host defenses. Its biosynthesis also consumes cellular resources. In the evolved populations, O-antigen loss frequently resulted from insertion sequences, mobile genetic elements that disrupted genes involved in its production.
The initial experimental strain already possessed the NDM-1 carbapenemase and was resistant to meropenem. Nevertheless, O-antigen disruption further increased resistance. Deleting the wbbM gene, which participates in O-antigen biosynthesis, increased the meropenem minimum inhibitory concentration (MIC) and produced approximately 100-fold greater survival on meropenem-containing agar. The mutant also gained a pronounced competitive advantage over the parental strain during antibiotic exposure, although it was less competitive in the absence of meropenem. Additional experiments demonstrated increased meropenem survival following O-antigen disruption in other K. pneumoniae strains, including initially carbapenem-susceptible backgrounds.
Unexpectedly, this effect did not result from increased carbapenemase activity or reduced drug penetration. The researchers observed a modest increase in outer-membrane permeability, and their experiments did not identify changes in carbapenemase hydrolytic activity or relevant antibiotic efflux that explained the enhanced resistance. Instead, losing O antigen triggered extensive changes in bacterial metabolism and stress responses.
Transcriptomic and metabolomic analyses revealed alterations in central carbon metabolism, respiratory electron transport, and pathways involved in maintaining the cellular redox balance. Under meropenem exposure, the O-antigen-deficient bacteria exhibited reduced expression of several pathways associated with the tricarboxylic acid cycle, lower intracellular adenosine triphosphate (ATP) levels, and changes in alternative metabolic pathways. These changes coincided with reduced accumulation of reactive oxygen species (ROS), chemically reactive molecules that can damage cellular components. The researchers found that the altered oxidative response contributed to bacterial survival during meropenem exposure, although the antibiotic’s primary bactericidal action remains disruption of cell-wall synthesis.
The consequences of O-antigen loss extended beyond meropenem resistance, however, creating a strikingly different response to another class of antibiotics.
Collateral Sensitivity Creates Opportunities for Antibiotic Strategies
When the investigators tested their O-antigen-deficient K. pneumoniae variants against other antibiotics, they found that the effects of surface remodeling varied considerably among drugs. The mutants exhibited greater resistance to additional carbapenems, and their susceptibility to several unrelated antibiotics remained unchanged. However, they became substantially more sensitive to the aminoglycosides kanamycin, tobramycin, and amikacin. In the reported exposure experiments, survival of the parental strain was more than 1,000-fold higher than that of the O-antigen-deficient mutant.
The researchers attributed this collateral sensitivity to additional changes in the bacterial envelope. The O-antigen-deficient mutants displayed a more negatively charged outer surface and altered expression of outer-membrane proteins. Because aminoglycosides are positively charged, increased surface electronegativity may promote their association with the bacterial envelope and facilitate uptake. Changes in expression of the outer-membrane protein OmpA also contributed to the proposed explanation for enhanced susceptibility.
These observations prompted experiments exploring whether the vulnerability could be exploited through antibiotic combinations. In checkerboard assays, which evaluate bacterial growth across different concentrations of two drugs, meropenem and kanamycin demonstrated synergy against the O-antigen-deficient mutant. The fractional inhibitory concentration index (FICI), a measure of antibiotic interaction, was 0.375, below the study’s threshold of 0.5 for synergy. In a separate laboratory evolution experiment, adding a low concentration of kanamycin during meropenem exposure prevented the detectable emergence of O-antigen-associated, phage-resistant variants over three passages.
The experiments illustrate two related but distinct strategies for exploiting resistance-associated vulnerabilities. One involves identifying an adaptation that has already occurred and selecting an antibiotic that takes advantage of the resulting susceptibility. The other involves using drug combinations to reduce the evolutionary advantage of particular resistance pathways before they become established. Both depend on understanding how bacterial adaptations influence susceptibility to multiple agents.
Researchers have also investigated whether treatment sequences could exploit predictable relationships between resistance to one antibiotic and susceptibility to another. In a review published in Nature Ecology & Evolution, researchers at KU Leuven examined the evolutionary processes that generate collateral sensitivity and the factors that make these responses variable.2 They emphasized that genetic background, environmental conditions, and evolutionary history can influence whether resistance to one antibiotic produces increased susceptibility or cross-resistance to another. Consequently, an antibiotic pairing that generates an advantageous trade-off in one bacterial population may behave differently in another.
A third possibility involves deliberately inducing a transient physiological state that increases susceptibility, rather than relying on stable resistance mutations. In a 2023 study, researchers at Spain’s National Centre for Biotechnology demonstrated that the antiseptic dequalinium chloride could induce a temporary physiological response associated with ciprofloxacin resistance in Pseudomonas aeruginosa.3 This response simultaneously increased susceptibility to tobramycin, including in some isolates already resistant to that antibiotic. When the researchers combined dequalinium chloride with tobramycin during short-term laboratory evolution experiments involving 10 clinical isolates, 39 of 40 bacterial populations showed extinction under the study’s experimental conditions over three days, whereas all populations exposed to either compound individually continued to grow. The results provided experimental evidence that deliberately induced collateral sensitivity could suppress resistant bacterial populations under controlled laboratory conditions.
For any of these strategies to become clinically useful, researchers must identify susceptibility relationships that persist across relevant bacterial backgrounds and treatment conditions. The challenge is substantial. In a study published in The Lancet Microbe in 2026, researchers at Leiden University and the Netherlands’ National Institute for Public Health and the Environment systematically analyzed multicenter antimicrobial surveillance data.4 Among 12,024 evaluated species–antibiotic pairs, 364 (~3%) exhibited collateral sensitivity, while 5,044 (~42%) exhibited collateral resistance. Although these findings do not directly measure how susceptibility evolves during treatment, they suggest that opportunities to exploit collateral sensitivity may be substantially less common than the cross-resistance relationships that complicate antibiotic selection.
Studies of bacterial populations during chronic infection provide another important qualification. Researchers at the Georgia Institute of Technology and Emory University examined 300 P. aeruginosa isolates collected from four individuals with cystic fibrosis, where bacterial populations can experience prolonged exposure to multiple antibiotics.5 Despite considerable variation in antibiotic resistance phenotypes, the investigators found little evidence of consistent collateral sensitivity among the aminoglycosides, β-lactams, and fluoroquinolones they examined. Their results underscore the difficulty of extending susceptibility relationships identified under controlled experimental conditions to heterogeneous bacterial populations in patients.
The therapeutic opportunity therefore lies in identifying specific, reproducible vulnerabilities rather than assuming that increased resistance generally creates a compensating weakness. A rational development strategy would need to connect particular resistance mechanisms or physiological states with consistent susceptibility patterns, establish whether those relationships remain stable during treatment, and determine whether the resulting antibiotic combinations are clinically feasible. For aminoglycosides, considerations such as toxicity, achievable drug exposure, and infection site would also influence whether a promising laboratory interaction could be translated into an effective regimen.
Reprogramming Bacterial Metabolism to Restore Susceptibility
Collateral sensitivity is not the only way to exploit the physiological consequences of resistance. Researchers are also investigating whether they can directly manipulate bacterial metabolism to increase the effectiveness of antibiotics. This approach builds on evidence that antimicrobial susceptibility depends partly on the metabolic state of the target organism, including processes that influence drug uptake, energy production, and responses to cellular stress.
The Huang study provides an example of how a protective metabolic state might be disrupted. To test whether the physiological changes associated with O-antigen loss contributed directly to enhanced meropenem survival, the investigators used targeted gene repression to interfere with selected metabolic pathways. Reducing expression of rutF, which participates in respiratory electron transfer, or pflB, which contributes to an alternative route for pyruvate metabolism, increased ROS accumulation and reduced survival under meropenem exposure. These experiments provided functional evidence that metabolic reprogramming contributed to the resistance-associated phenotype.
The researchers also found that supplementing cultures with glycine increased ROS accumulation and restored the O-antigen-deficient mutants’ meropenem susceptibility toward parental-strain levels. The intervention appeared to interfere with the protective redox state associated with O-antigen loss. Although the experiments used a glycine concentration of 100 mM and did not establish clinical feasibility, they demonstrated that the resistance-associated physiological state could be modified experimentally.
The use of metabolites to enhance antibiotic activity has been investigated through several approaches. In a 2015 study, investigators at Sun Yat-sen University demonstrated that supplying alanine or glucose could increase susceptibility to the aminoglycoside kanamycin in resistant bacteria.6 The investigators found that the metabolites stimulated central metabolic activity and increased the proton motive force, an electrochemical gradient across the bacterial membrane that contributes to aminoglycoside uptake. Their experiments included a mouse urinary tract infection model, providing evidence that metabolic potentiation could extend beyond bacterial cultures.
More recent research has examined this approach in clinically relevant multidrug- and carbapenem-resistant organisms. In a 2025 study in Nature Microbiology, researchers from the same group at Sun Yat-sen University investigated metabolic differences among antibiotic-sensitive, multidrug-resistant, and carbapenem-resistant Escherichia coli isolates.7 They identified a role for pyruvate formate-lyase in determining susceptibility to the aminoglycoside micronomicin and found that manipulating formate metabolism influenced antibiotic uptake and killing. Supplementing treatment with formate enhanced micronomicin activity against resistant bacteria. In a mouse infection model, the combination reduced bacterial growth and dissemination and improved survival relative to either intervention alone.
Although the studies involve different metabolic processes, they establish a common therapeutic principle: Bacterial physiology can influence antibiotic effectiveness in ways that may be amenable to intervention. Huang and colleagues disrupted a protective state associated with reduced oxidative stress during meropenem exposure, whereas the Sun Yat-sen University group manipulated metabolic conditions that influence aminoglycoside activity.
The research points toward two possible therapeutic approaches. One involves using metabolites directly as adjuvants to enhance existing antibiotics. The other involves identifying pharmacological agents that reproduce beneficial metabolic effects without requiring the administration of large quantities of metabolites. Targeted pharmacological intervention might eventually offer a more practical alternative to direct metabolite supplementation, although identifying suitable targets and demonstrating selective activity remain substantial development challenges. The central challenge is translating experimentally observed changes in bacterial physiology into interventions that achieve sufficient antibacterial activity without unacceptable effects on the host.
Phage Resistance and the Broader Costs of Bacterial Adaptation
Changes that influence antibiotic susceptibility can also affect bacterial interactions with phages and host defenses. Because the bacterial surface participates in interactions with both antimicrobial agents and the surrounding environment, adaptations that alter its structure may have consequences beyond the treatment that initially selected them. These effects are particularly relevant to efforts to use evolutionary pressures deliberately, favoring bacterial adaptations associated with increased susceptibility to other interventions.
Phages infect bacteria by recognizing and interacting with particular surface structures. Bacteria can escape infection by altering or eliminating these structures, although the resulting changes may affect other physiological functions. This relationship has prompted interest in phage steering, an approach that seeks to use phages to favor bacterial adaptations that create useful vulnerabilities to antibiotics.
The Huang study illustrates the interconnected nature of these responses. The researchers initially identified O-antigen-deficient variants through changes in phage susceptibility. Loss of O antigen prevented successful adsorption by several phages that depended on this surface structure, while other phages retained their activity. Because the same adaptation also changed susceptibility to meropenem and aminoglycosides, the experiments demonstrated how an alteration in a phage receptor could influence the bacterium’s response to multiple antimicrobial pressures. However, these variants emerged under antibiotic selection, not through the deliberate use of phages to steer bacterial evolution. The study therefore establishes a mechanistic relationship relevant to phage steering without demonstrating that such a strategy would be therapeutically effective.
Research from Monash University in Australia provides a closely related example. In a 2026 study, researchers investigated resistance to a newly isolated phage in a multidrug-resistant clinical K. pneumoniae strain.8 They found that mutations affecting galE, a gene involved in LPS biosynthesis, interfered with phage adsorption and altered the bacterial outer membrane. The resulting variants became more susceptible to aminoglycosides, chloramphenicol, and polymyxins, but less susceptible to meropenem and tetracycline.
The contrast with Huang’s findings is instructive. Both studies implicated changes in bacterial surface structures and demonstrated increased susceptibility to aminoglycosides. However, the effects on meropenem susceptibility differed, indicating that the consequences of surface remodeling depend on the particular genetic and physiological alterations involved.
A 2024 study published by members of the same research group reinforced this point.9 The team investigated several genetic pathways associated with phage resistance in K. pneumoniae. They found that different mutations could produce opposing effects on antibiotic susceptibility, with some increasing sensitivity to particular drugs while others increased resistance. The findings demonstrate why selecting a phage solely for its ability to kill a bacterial strain may not provide sufficient information about the evolutionary consequences of treatment.
For phage steering to become a reliable therapeutic strategy, researchers would need to identify phages that constrain bacterial evolution toward favorable adaptations. This requires understanding which receptors a phage uses, how bacteria can alter those receptors to escape infection, and what consequences those mutations have for susceptibility to accompanying antibiotics. The objective would be to make escape from one antimicrobial threat increase vulnerability to another, rather than inadvertently selecting variants with broader resistance.
Surface remodeling may also alter bacterial interactions with host immune defenses. Huang and colleagues observed that O-antigen-deficient K. pneumoniae exhibited substantially reduced survival in healthy human serum, suggesting increased vulnerability to serum-associated defenses. In a mouse infection model, the mutants also showed attenuated virulence and reduced bacterial burdens in several organs compared with the parental strain. The same surface alteration that improved survival during meropenem exposure therefore compromised functions relevant to bacterial survival in these experimental infection settings.
These additional consequences widen the range of characteristics that researchers might consider when evaluating resistance-associated adaptations. A mutation’s effects on antibiotic susceptibility may be only part of its biological significance; changes in phage susceptibility, interaction with immune defenses, and fitness during infection could also influence whether that adaptation presents an exploitable vulnerability. Understanding these relationships may help inform the selection of phages and accompanying antibiotics, as well as the assessment of bacterial escape pathways likely to emerge during treatment.
Beyond Resistance Genes: Identifying Vulnerabilities That Genomic Tests May Not Predict
Exploiting resistance-associated vulnerabilities requires more than identifying which antibiotics a bacterial isolate currently resists. Researchers must also determine why those susceptibility patterns exist, whether they reflect particular physiological adaptations, and how those adaptations might influence responses to other interventions.
Molecular approaches to antimicrobial resistance (AMR) detection have substantially improved the ability to identify known genetic resistance determinants. Targeted assays can detect important resistance genes, while whole-genome sequencing (WGS) provides a broader view of the genetic variation within an isolate. However, detecting a resistance determinant, predicting a susceptibility phenotype, and identifying the biological basis of a potentially exploitable vulnerability are distinct objectives.
The Huang findings illustrate these differences. The original K. pneumoniae strain already carried a carbapenemase gene, but disruption of O-antigen biosynthesis further enhanced meropenem resistance through metabolic and redox changes. A genetic assay focused on common carbapenemase genes would not necessarily identify the additional resistance-associated adaptation. Conversely, phenotypic antimicrobial susceptibility testing could detect differences in susceptibility without establishing that they arose from a particular surface alteration or metabolic state.
In a 2025 review, Norelle Sherry and colleagues, researchers at the University of Melbourne’s Peter Doherty Institute for Infection and Immunity examined advances and remaining challenges in using bacterial genomics for AMR surveillance and clinical applications.10 They emphasized that predicting resistance phenotypes from genomic data remains difficult for many pathogen–antibiotic combinations. Reliable prediction depends on sufficiently comprehensive knowledge of resistance mechanisms and access to high-quality datasets linking genomic variation with measured susceptibility.
The clinical-isolate research by the teams from the Georgia Institute of Technology and Emory University provides another illustration of the limitations of straightforward genotype-to-phenotype relationships.5 The investigators found that bacterial populations with relatively limited genetic diversity could exhibit substantial variation in antibiotic resistance phenotypes. Such observations complicate efforts to infer susceptibility patterns solely from the amount or nature of detectable genetic variation, particularly in populations evolving during chronic infections.
Identifying collateral sensitivity may present an additional challenge because a useful susceptibility pattern can reflect the combined effects of several physiological processes. Genetic information might indicate that a particular metabolic or surface-associated pathway has changed, but its functional consequences may depend on gene expression, environmental conditions, or interactions with other resistance determinants. These factors could help explain why the same general resistance mechanism does not always produce a consistent susceptibility trade-off.
Future characterization approaches might therefore benefit from integrating genomic data with phenotypic susceptibility testing and, where appropriate, measurements of bacterial physiology. In research settings, such integration could help identify biomarkers associated with particular vulnerabilities, improve predictions of collateral sensitivity, or guide the selection of antibiotic combinations for further evaluation. Establishing clinical value would require demonstrating that these additional measurements provide actionable information beyond existing diagnostic methods.
This information could be particularly valuable when designing treatments. A conventional susceptibility test may establish that an isolate is susceptible to a particular aminoglycoside, but it does not necessarily reveal whether that susceptibility results from an evolutionary adaptation that also changes the bacterium’s response to another antibiotic. Understanding the basis of such relationships could help researchers move from selecting drugs according to their individual activity toward investigating combinations and sequences that exploit predictable biological interactions.
From Biological Trade-Offs to Therapeutic Strategies
Researchers have demonstrated several resistance-associated vulnerabilities, but exploiting them clinically will require different strategies. Collateral sensitivity-based antibiotic strategies depend on identifying susceptibility relationships that are reproducible across relevant bacterial backgrounds. Metabolic interventions require ways to manipulate bacterial physiology at achievable concentrations without unacceptable effects on the host. Phage steering depends on anticipating how bacteria will escape infection and whether those escape pathways create vulnerabilities to accompanying treatments.
Each approach also faces a common challenge: Bacterial populations are genetically and physiologically diverse, and the adaptations that emerge under one set of conditions may not produce the same consequences elsewhere. Studies of clinical isolates and antimicrobial surveillance data demonstrate why results obtained in laboratory evolution experiments cannot be assumed to predict treatment responses in patients. Establishing clinically useful strategies will require testing susceptibility trade-offs across diverse isolates, evaluating their stability in relevant infection models, and determining whether proposed interventions improve outcomes without accelerating the emergence of alternative resistance pathways
The Huang study illustrates why these efforts may be worthwhile. A single change in bacterial surface architecture influenced carbapenem resistance, aminoglycoside susceptibility, metabolism, phage interactions, and vulnerability to host defenses. The findings reveal several experimentally accessible consequences of an adaptation that initially appeared advantageous under antibiotic pressure. Other research demonstrates that some of these consequences can be manipulated deliberately, including through metabolic supplementation and the induction of transient collateral sensitivity.
Turning these discoveries into effective antimicrobial strategies will require closer integration of mechanistic microbiology, evolutionary biology, pharmacology, and susceptibility prediction. Rather than studying resistance mechanisms only to understand why antibiotics fail, researchers may increasingly be able to use that knowledge to identify combinations, adjuvants, and evolutionary interventions that exploit the liabilities accompanying bacterial adaptation. The central opportunity is to make those vulnerabilities predictable enough to inform drug development and, ultimately, treatment decisions.