Antibiotic resistance is usually treated as the loss of an option. A bacterium acquires a mutation or other resistance mechanism, an antibiotic loses activity, and treatment shifts to another drug or combination. But becoming resistant can also change the organism in ways that extend beyond protection from a particular antibiotic. Those changes may create new physiological costs, dependencies, or liabilities.
A new study in Microbiology Spectrum offers a striking example.1 Researchers testing drug-resistant clinical isolates of Mycobacterium tuberculosis found that adding vancomycin could restore substantial activity to isoniazid, a first-line antibiotic used to treat tuberculosis (TB), despite resistance to the latter drug in the bacterial population. In a mouse model infected with a multidrug-resistant (MDR) clinical strain, the combination also reduced bacterial burden where either drug alone did not produce a significant reduction. Mechanistic analyses pointed toward changes involving cell-wall biology and stress responses that may help explain the effect.
The finding remains preclinical, and it does not establish vancomycin plus isoniazid as a treatment for drug-resistant TB. Its larger significance is conceptual. It adds to evidence that resistance can create a new biological state with vulnerabilities of its own.
When Isoniazid Resistance Creates Something New to Target
The researchers began with 19 clinical M. tuberculosis strains isolated from sputum samples of patients with drug-resistant TB. Three did not grow under the relevant isoniazid condition and were excluded from the principal comparison, leaving 16 isolates spanning different resistance profiles. These included strains with genetic changes associated with high- and lower-level isoniazid resistance, as well as MDR strains carrying resistance-associated mutations affecting both isoniazid and rifampicin.
Across those 16 isolates, the combination of isoniazid and vancomycin significantly reduced bacterial growth compared with either drug alone. Fourteen showed greater than 90% growth reduction relative to at least one of the single-drug conditions. That breadth matters because a liability identified in one resistant strain may not persist across different genetic backgrounds. Here, the effect extended across multiple resistance backgrounds.
The researchers then tested the combination in an acute mouse infection model using an MDR clinical strain. After 10 days of treatment, neither isoniazid nor vancomycin alone significantly reduced bacterial burden. The combination produced an approximately tenfold reduction in bacterial burden in the lungs and a similar reduction in dissemination to the spleen, along with improved lung histopathology relative to the untreated control. The authors interpreted those findings as evidence that vancomycin restored sensitivity to isoniazid in vivo.
That interpretation should remain bounded by the experiment. No human efficacy data were reported, and the study does not show that the combination will prove safe, practical, or effective as part of clinical TB therapy. What it does show is that resistance does not necessarily make an older antibiotic biologically irrelevant. Under the right conditions, features associated with the resistant phenotype may be exploitable.
What Changed in the Resistant Bacterium?
The next question is why.
Transcriptomic analysis and systems-level modeling identified changes involving cell-wall lipid pathways during drug exposure. In particular, the researchers observed downregulation of genes involved in synthesis of phthiocerol dimycocerosates, lipid components of the mycobacterial cell envelope. They proposed that changes in cell-wall lipid composition could increase vancomycin access to its target and contribute to the activity of the combination.
The distinction between cause and response is important here. The study examines bacteria that already carry resistance-associated changes, but some of the molecular patterns observed emerged during treatment. The data therefore do not establish that every relevant cell-wall alteration was created directly by the original resistance mechanism. The authors present the pathway as a possible mechanistic basis for the observed combination activity rather than a complete causal explanation.
The broader point is that resistance mutations do not operate in isolation. Bacteria may accommodate them through changes in metabolism, transcription, envelope structure, redox balance, or other cellular systems. Those adaptations can change what the resistant organism depends on for survival. A useful question therefore becomes not only which target has become resistant, but what new constraints have appeared as a consequence.
Resistance Can Create New Dependencies
The new study extends a line of investigation that has been developing for years. In 2016, researchers studying isoniazid-resistant mycobacteria used genomic data and network analysis to search for what they called “emergent vulnerabilities.” Redox response emerged prominently from that work. Resistant bacteria displayed increased sensitivity to several compounds that interfered with antioxidant responses, and combinations of isoniazid with compounds including vancomycin showed activity against M. tuberculosis and a small number of drug-resistant clinical strains. The resistant strains also showed diminished capacity to counter oxidative stress.2
A 2024 study approached the same general problem using genome-wide CRISPR interference (CRISPRi), transcriptomics, and metabolomics. Researchers mapped genetic liabilities in an isoniazid-resistant M. tuberculosis strain carrying an altered katG gene and found that compensatory metabolic and transcriptional remodeling created dependencies involving respiration, ribosome biogenesis, and nucleotide and amino-acid metabolism. Some of those liabilities also translated to clinical isolates.3
That progression is important. Resistance can trigger secondary adaptations, and those adaptations can generate dependencies that were less consequential in the susceptible organism. Methods that map genetic and metabolic vulnerabilities can therefore be used not simply to explain how resistance works but to identify what the resistant bacterium has become less able to do without.
Collateral Sensitivity Is Related but Not Identical
This idea overlaps with collateral sensitivity, but the terms should not be treated as interchangeable.
Collateral sensitivity describes an evolutionary trade-off in which resistance to one antibiotic causes increased susceptibility to another. In its clearest form, the resistance mechanism itself changes the bacterium in a way that increases the effectiveness of a second drug.4
That differs from antibiotic synergy. Two drugs can become more effective when administered together because of physiological responses generated by simultaneous exposure, even if resistance to the first drug did not independently increase susceptibility to the second. Related phenomena can therefore produce similar treatment effects through different biological routes.
That distinction is useful for interpreting the vancomycin–isoniazid result. The study is built around the premise that drug resistance creates exploitable liabilities, but the reported activity also involves a simultaneous combination and additional physiological changes during treatment. The broader framework of resistance-associated vulnerability is therefore more precise than assuming that every element of the effect represents classical collateral sensitivity.
That framing also accommodates other types of resistance-associated liabilities. A resistant organism might acquire a metabolic dependency, a weakened stress response, an altered cell envelope, or another weakness that can be targeted without fitting the strict definition of increased susceptibility to a second antibiotic.
Can Treatment Be Designed Around Evolutionary Trade-Offs?
A more ambitious extension of the same logic asks whether treatment can influence which resistant state appears next.
In a 2013 study, researchers evolved Escherichia coli resistance to 23 clinically used antibiotics and mapped patterns of collateral resistance and sensitivity. Those relationships formed networks in which resistance to one drug could sometimes increase susceptibility to another. The investigators proposed using compatible antibiotics sequentially so that adaptation to one treatment would make the next more effective, and they demonstrated proof of principle using cyclic deployment of gentamicin and cefuroxime.5
That approach differs from exploiting a liability that already exists. One strategy identifies a weakness in an established resistant phenotype; the other attempts to steer evolution toward a phenotype that preserves a future treatment option. Both depend on the same underlying premise: resistance evolution can impose biological constraints, and some of those constraints may be predictable enough to use therapeutically.
The Problem: Bacterial Evolution Does Not Always Cooperate
Predictability is the difficult part.
Collateral-sensitivity patterns can vary among strains that have become resistant to the same antibiotic. A relationship that appears robust in one genetic background may weaken, disappear, or reverse in another. Recent reviews have therefore identified strain-to-strain variation as a major obstacle to translating collateral sensitivity and genetic-vulnerability mapping into treatment strategies.6
Large-scale clinical surveillance data reinforce that caution. A 2026 analysis examined more than 5 million minimum inhibitory concentration (MIC) measurements across three antimicrobial resistance surveillance datasets, encompassing 86 antibiotics and 30 pathogen species. Among 12,024 species–antibiotic relationships evaluated, the investigators identified collateral sensitivity in 364 (3.0%). Collateral resistance — in which resistance to one antibiotic was associated with reduced susceptibility to another — appeared in 5,044 relationships (42.0%).7
In those surveillance data sets, resistance was therefore far more often associated with reduced susceptibility to additional antibiotics than with collateral sensitivity. The existence of collateral sensitivity cannot justify assuming that resistance will routinely generate a useful trade-off.
The same study nevertheless identified six collateral-sensitivity relationships conserved across four bacterial species. The translational question is therefore not simply whether collateral sensitivity exists. It is which relationships are sufficiently strong, reproducible, and recognizable to support a treatment decision.
The same threshold applies to resistance-associated liabilities more broadly. Finding a vulnerability in one laboratory strain establishes possibility. Showing that it persists across genetically diverse resistant bacteria and remains detectable in vivo begins to establish usefulness.
When Does a Vulnerability Become Clinically Useful?
For a resistance-associated liability to become more than an experimental observation, several problems must be solved.
First, it must be reproducibly linked to a defined resistance state or mechanism. A broad label such as isoniazid resistance can encompass different genetic routes, and those routes do not necessarily impose the same physiological consequences. A treatment strategy that depends on a liability created by one mechanism may fail when resistance arises through another.
The effect must also survive the diversity present in patient-derived bacteria. The progression from engineered mutants to clinical isolates in the 2024 CRISPRi study, and from earlier mycobacterial models to the larger clinical-isolate panel tested in the 2026 vancomycin study, illustrates the importance of that step.1,3
Researchers must then determine whether the dependency is pharmacologically accessible and whether its mechanism is understood well enough to predict when the strategy should work. A vulnerability is considerably more useful if a laboratory can identify the relevant resistance mechanism and infer the associated liability than if every bacterial isolate requires extensive experimental screening.
Durability presents another problem. Applying a second antibiotic creates a new selective pressure. Bacteria can acquire compensatory changes that reduce the cost of the original resistance mechanism, eliminate a collateral weakness, or create resistance to the new treatment. Genetic-vulnerability mapping may therefore be valuable not only for identifying what resistant organisms depend on now, but also for anticipating routes through which those dependencies might disappear.
Clinical development would also need to distinguish between simultaneous combinations and sequential evolutionary strategies. A combination such as isoniazid plus vancomycin attempts to exploit a resistant state that already exists. Collateral-sensitivity cycling seeks to influence which state emerges next. Their biological logic overlaps, but the evidence needed to establish clinical utility will not necessarily be the same.
Can Resistance Become a Therapeutic Map?
Drug-resistant TB alone remains a substantial problem. The World Health Organization estimated that approximately 390,000 people developed MDR or rifampicin-resistant TB in 2024.8 New antibiotics remain essential, but resistance biology may also reveal ways to recover or extend the utility of existing agents.
The emerging opportunity is to extract more information from resistance itself. A mutation that protects a bacterium from one drug may force compensatory changes elsewhere. Those adaptations can produce liabilities that are measurable, targetable, and, in some cases, shared across resistant strains.
The vancomycin–isoniazid study does not establish a general rule, and it does not show that an older TB drug can simply be restored by pairing it with another antibiotic. It does provide a concrete example of a broader strategy: examine what the bacterium had to change in order to become resistant, then determine whether those changes created something new to attack.
If those liabilities can be linked reliably to recognizable resistance mechanisms and shown to persist across diverse infections and under continued treatment pressure, resistance could become more than a sign that one therapeutic option has been lost. It could also provide a map to the next vulnerability.
