What If We Designed Antibiotics for the Resistance They Will Face?

What If We Designed Antibiotics for the Resistance They Will Face?
Author: David Alvaro, PhD
Published date: 8 September 2026
Category:
Human Health Innovation
Share

Antibiotic development is usually optimized around whether a candidate can kill a pathogen safely and effectively. But every new antibiotic also creates a new evolutionary problem: once bacteria encounter the drug, what routes are available to escape it?

Researchers have begun exploring whether antimicrobial resistance (AMR) evolution can be forecast prospectively rather than understood mainly after resistance has appeared. The goal is not to predict every mutation that will emerge in patients, but to better understand the evolutionary possibilities created by a treatment and use that knowledge to inform how antimicrobials are developed and deployed.1

A study of programmable antisense antibiotics published in Nature Communications in August 2026 offers an unusually clear example of why that could matter.2 Researchers experimentally evolved reduced susceptibility to antisense peptide nucleic acids (PNAs) in four major Gram-negative pathogens. They found that the resistance outcome depended strongly on the cell-penetrating peptide used to deliver the PNA into bacterial cells. One delivery architecture repeatedly opened an uptake-based escape route. Another generated much more modest resistance associated with changes in the sequence targeted by the antibiotic.

That result raises a larger possibility. Resistance susceptibility may not simply be an unavoidable property discovered after an antimicrobial is developed. In some cases, the evolutionary routes available to bacteria could become another variable that developers deliberately evaluate and, potentially, design around.

A Programmable Antibiotic Gives Evolution More Than One Place to Push

Antibacterial antisense oligomers (ASOs) use sequence-specific binding to bacterial RNA to interfere with expression of selected genes. PNAs and phosphorodiamidate morpholino oligomers are among the major synthetic chemistries investigated for this purpose. Because the recognition sequence can be changed, the molecular target can be programmed.3

Recognition, however, is only half of the problem. PNAs do not efficiently penetrate bacterial cells unaided, so antibacterial PNA constructs have frequently been linked to cell-penetrating peptides (CPPs) that facilitate uptake. This creates two separable design variables: the sequence that determines what the therapeutic recognizes and the carrier that helps it reach that target. 

The 2026 resistance study used that modularity to examine how both targeting and delivery influence bacterial evolution. Laboratory evolution experiments exposed strains of Escherichia coli, Klebsiella pneumoniae, Salmonella enterica, and Pseudomonas aeruginosa to PNA constructs incorporating different CPPs, followed by genomic analysis of the resulting resistant populations. The study did not test every carrier-target combination in every organism, but the comparisons revealed a strong relationship between the delivery peptide and the resistance mechanisms that emerged.2

For PNAs conjugated to a CPP called KFF (a.k.a. (KFF)₃K, comprising lysine (K) and phenylalanine (F) residues), reduced uptake became a recurring route to resistance. Mutations affecting sbmA, which encodes a protein involved in uptake of these peptide-PNA conjugates, appeared consistently in the species treated with KFF-linked PNAs, alongside additional changes involving peptide transport, translation, and the cell envelope. 

A construct delivered with a different CPP, RXR (a.k.a. (RXR)₄XB, comprising arginine (R), 6-aminohexanoic acid (X) and β-alanine (B)), produced a different resistance profile in E. coli. Eight of 10 evolved populations showed no increase in the minimum inhibitory concentration, while the two with measurable increases showed only modest shifts associated with mutations in the PNA-binding site. Because the antibiotic’s recognition sequence is programmable, those mutations represent a type of resistance that could potentially be addressed by redesigning the oligomer.2

The target alone did not determine the resistance outcome. The way the drug entered the bacterium helped determine where evolution found an opening.

When Delivery Becomes an Evolutionary Vulnerability

The sbmA result matters because it changes how delivery should be understood. For antibacterial antisense agents, delivery is not only a pharmacological problem. It can also determine which evolutionary escape routes are available.

If a drug depends strongly on a bacterial uptake pathway, changing that pathway may offer bacteria a comparatively direct route to reduced susceptibility. Dependence on SbmA for uptake of some peptide-antisense conjugates was already known, but the new study demonstrates systematically how that dependency can influence resistance evolution across multiple bacterial backgrounds.

The contrast with target-sequence resistance is particularly informative. Mutating the RNA sequence recognized by an antisense agent can reduce susceptibility. Still, programmability creates a potential response that does not exist for every drug class: change the therapeutic sequence to match the altered target. In a 2023 study published in Antimicrobial Agents and Chemotherapy, researchers experimentally introduced mutations into the target sequence of peptide-conjugated phosphorodiamidate morpholino oligomers directed against P. aeruginosa and found that the mutations reduced efficacy.4 Redesigning the oligomers to complement the mutant sequences restored activity, including in an animal infection model.

That does not mean target-sequence resistance is trivial or that a new antisense antibiotic could simply be redesigned and immediately substituted whenever resistance appears clinically. Development, manufacturing, safety evaluation, regulation, diagnostics, and deployment all stand between an updated sequence and routine treatment. The 2026 study also does not establish that antisense antibiotics can be updated faster than bacterial resistance can spread.

However, it establishes a narrower point with potentially broader significance: different resistance mechanisms can present very different engineering problems. A mutation that undermines an essential feature of a delivery strategy is not equivalent to a mutation that alters the sequence recognized by a programmable therapeutic. Understanding which type of escape a candidate tends to select could therefore matter during drug design.

Evolutionary Robustness Is a Design Property, Not a Label

This idea is not unique to programmable antibiotics. Researchers have been asking for years whether some therapeutic designs leave bacteria fewer viable evolutionary exits than others.

Experiments examining two antivirulence treatments against P. aeruginosa found substantially different resistance outcomes, even though both interventions targeted the same general bacterial system. Resistance to one treatment spread rapidly, while resistance to the other arose much less readily. The researchers argued that evolutionary robustness should therefore be understood as a relative property that varies among specific treatments, rather than something that can be assigned categorically to an entire therapeutic strategy.5

More recent antibiotic research has begun asking what molecular properties might produce that relative robustness. A 2025 study in Nature Communications examined experimental antibiotics designed to affect membrane integrity while also disrupting another cellular function.6 Resistance evolution remained limited in tested strains of E. coli, K. pneumoniae, Acinetobacter baumannii, and P. aeruginosa. At the same time, comparator antibiotics that also acted on two targets but lacked the same membrane-disrupting architecture remained susceptible to resistance development. The results suggested that having more than one target was not sufficient on its own; the drug’s specific architecture affected the resistance routes available to bacteria.

The new antisense work extends that principle into a particularly flexible therapeutic platform. The target sequence matters, but so does the carrier. Changing one component can alter the mechanisms through which resistance develops.

Programmability therefore does not make asobiotics immune to evolution. It provides additional design variables that could potentially be used to influence the resistance landscape.

From Resistance Testing to Resistance Forecasting

This way of thinking points toward a different role for experimental evolution in antimicrobial development.

Resistance studies often establish whether resistance can emerge and characterize mechanisms once resistant organisms have been generated. Resistance forecasting asks whether those experiments can do more: identify likely evolutionary liabilities early enough to influence which drugs move forward, how they are designed, or how they are ultimately used. Although forecasting remains challenging and has not become a standard component of AMR control, it has been proposed as a complement to new drug development, diagnostics, and antimicrobial stewardship.1

The antisense study provides a concrete development-stage example. Researchers did not simply observe that bacteria could become less susceptible to PNAs. By evolving populations under different antisense constructs and sequencing the resulting mutants, they could see that one carrier repeatedly directed evolution toward an uptake pathway, while another produced a different and more limited resistance profile.2

That suggests a different development philosophy: characterize a candidate’s evolutionary vulnerabilities while selecting the drug, rather than waiting for resistance to reveal them after deployment.

The questions could become more specific. Can bacteria escape simply by disabling an uptake mechanism? How many accessible mutations meaningfully reduce susceptibility? Does changing the carrier close one escape route only to create another? If resistance appears, can the therapeutic architecture be modified to restore activity?

Those questions do not require the assumption that laboratory evolution will perfectly predict clinical evolution. Bacteria encounter far more complicated environments during infection, and evolutionary outcomes will depend on drug exposure, microbial populations, host conditions, transmission, existing resistance determinants, and other factors that laboratory systems cannot fully reproduce. Forecasting resistance is therefore better understood as mapping plausible evolutionary vulnerabilities than as predicting the future with certainty. 

Even an incomplete map could be useful if it distinguishes candidates that offer bacteria an obvious, repeatable escape mechanism from those that present a more constrained evolutionary landscape. Resistance susceptibility could then become something developers characterize during candidate selection, alongside antibacterial activity and other performance characteristics.

Designing for Evolution Without Pretending We Can Defeat It

The strongest lesson from this emerging research is not that scientists have found a way to prevent bacterial evolution. The antisense experiments themselves generated resistance, and earlier work on evolutionary robustness shows why claims that a particular therapeutic concept is intrinsically resistant to resistance should be treated cautiously.

The more realistic opportunity is to make the evolutionary consequences of antimicrobial design visible earlier.

For asobiotics, the separation between targeting and delivery makes that possibility especially clear. The recognition sequence can be changed. The carrier can be changed. Experimental evolution can reveal whether particular combinations repeatedly select uptake defects, target mutations, or other adaptations. Related work has already demonstrated that some sequence-based resistance can be overcome experimentally through corresponding sequence redesign.2,4

Potency, spectrum, pharmacology, safety, and manufacturability will remain fundamental requirements for any successful antimicrobial. Evolutionary robustness could become another design criterion: not an assurance that resistance will never emerge but an assessment of how readily it can emerge, by what routes, and whether those routes leave room for an effective response.

The goal may never be an antibiotic to which resistance cannot evolve. A more realistic goal is to develop antibiotics with a better understanding of the evolutionary routes bacteria are likely to take, and whether those routes can be constrained or engineered around.

References

  1. 1. Rolff, Jens, et al. “Forecasting antimicrobial resistance evolution.” Trends in Microbiology. 32: 736–745 (2024). https://doi.org/10.1016/j.tim.2023.12.009
  2. 2. Mulkern, Adam J., et al. “A systematic identification of resistance determinants to antisense antibiotics suggests adaptation strategies dependent on the delivery peptide.” Nature Communications. 12 Aug. 2026. https://doi.org/10.1038/s41467-026-76357-y
  3. 3. Moustafa, Dina A., et al. “Developing antisense oligomer biotics ‘asobiotics’ as precision antibacterials: designs, strategies, and considerations for future success.” FEMS Microbiology Reviews. 2 Jan. 2026. https://doi.org/10.1093/femsre/fuaf059
  4. 4. Nanayakkara, Amila K., et al. “Sequence specificity defines the effectiveness of PPMOs targeting Pseudomonas aeruginosa.” Antimicrobial Agents and Chemotherapy. 23 Aug. 2023. https://doi.org/10.1128/aac.00245-23
  5. 5. Rezzoagli, Chiara, et al. “Probing the evolutionary robustness of two repurposed drugs targeting iron uptake in Pseudomonas aeruginosa.” Evolution, Medicine, and Public Health. 2018: 246–259 (2018). https://doi.org/10.1093/emph/eoy026
  6. 6. Maharramov, Elvin, et al. “Exploring the principles behind antibiotics with limited resistance.” Nature Communications. 21 Feb. 2025. https://doi.org/10.1038/s41467-025-56934-3

Subscribe to
disAMR Matters

Receive newly released thought leadership articles and news in your inbox twice a month.

Name(Required)