Scientists Build a Cephalosporin Triggered Prodrug That Delivers an Antibiotic Nucleoside in Resistant Bacteria

Chemists in France and Italy have designed and synthesized a family of prodrugs that hijack the very enzymes resistant bacteria use to defend themselves, using a cephalosporin scaffold to trigger the release of an antibiotic nucleoside inside Gram-negative pathogens. The study describes cephalosporin-based pronucleotides of the nucleoside analog zidovudine (AZT) that are cleaved by β-lactamases, the enzymes that normally destroy penicillins and cephalosporins. In laboratory tests, the compounds were rapidly processed by the widespread extended-spectrum β-lactamase CTX-M-15, and one derivative, a pyridine-containing phosphoramidate diester designated 16b or JMV8596, showed modest but measurable antibacterial activity against β-lactamase-producing strains of E. coli. The authors describe this as an initial proof of concept for an enzyme-triggered antibacterial pronucleotide, while stressing that many optimization steps remain before the approach could approach a drug candidate.

The motivation comes from a well-documented therapeutic gap. Antibiotic resistance is already associated with 1.27 million deaths annually worldwide, and the World Health Organization lists carbapenem-resistant Enterobacterales as a critical priority for antibiotic discovery. The most common mechanism of resistance to β-lactams, the most widely used class of antibacterial drugs, is the production of β-lactamases, enzymes that irreversibly hydrolyze the β-lactam ring.

The choice of cargo was equally deliberate. AZT, best known as an antiviral drug, has long shown potent activity against enterobacterial species such as E. coli. Its antibacterial mode of action depends on successive phosphorylation steps by bacterial kinases, culminating in the triphosphate form, which acts as a chain terminator during nucleic acid biosynthesis. The first phosphorylation step is generally rate-limiting and, in the case of AZT, is mediated by a thymidine kinase expressed only by Enterobacterales, which gives the drug a narrow spectrum. Previous clinical interest in AZT as a standalone antibiotic was undermined by a high mutation rate leading to decreased expression of that kinase. A pronucleotide strategy, of the kind that produced marketed antivirals such as sofosbuvir and remdesivir, masks the negative charges of the monophosphate so the molecule can diffuse into cells, after which masking groups are removed to release the 5′-monophosphate.

Earlier attempts to apply known pronucleotide chemistries, including HepDirect, CycloSal and SATE approaches, to antibacterial nucleoside analogs had failed, showing an overall lack of activity, apparently because the enzymatic machinery needed to unmask them was missing in bacteria. The new design therefore relies on enzymes that resistant bacteria are guaranteed to possess: their own β-lactamases. The team attached AZT monophosphate, through a SATE-like self-immolative linker, to the 3-position of a cephalosporin core. When a β-lactamase opens the β-lactam ring, a spontaneous chemical rearrangement releases the 3-substituent, which then fragments further to liberate a phosphoramidate monoester of AZT. From there, the designers explored two routes to the final active monophosphate: enzymatic cleavage of the phosphorus-nitrogen bond by bacterial phosphoramidases, or purely chemical hydrolysis favored by low-basicity amines such as morpholine and proline.

Synthesis centered on a pivotal H-phosphonate diester intermediate built from the commercially available cephalosporin 7-ACMA, acylated with a penicillin G-like side chain and substituted with 2-mercaptoethanol. A panel of amines was introduced at phosphorus using the Atherton-Todd reaction, whose conditions had to be carefully optimized because the nitrogen must be nucleophilic enough to react, yet not so reactive that it isomerizes the sensitive double bond of the cephalosporin. The series included benzyl, isopropyl and indole derivatives aimed at phosphoramidase recognition, glycine and pyridine analogs, a permanently charged choline-related cation intended to improve solubility and membrane penetration, and the chemically labile morpholine and proline compounds. Amines with very low pKa values, such as aniline and trifluoroethylamine, failed to react at all, and the acidic deprotection step needed to unmask the final compounds repeatedly eroded yields because the β-lactam ring itself is acid-sensitive.

Stability testing in HEPES buffer at pH 7.2 and 37 °C produced a nuanced picture. Most derivatives were far more stable than expected: half-lives ranged from 8 to more than 600 hours, with the proline compound 22b and the phosphodiester and phosphorothioate analogs essentially stable over the observation window. Surprisingly, the main degradation product under purely chemical conditions was the intact cephalosporin ester, meaning the β-lactam ring stayed intact in water and the observed hydrolysis occurred at a phosphorus-oxygen bond instead. Neither free AZT nor the AZT monophosphate was detected chemically, so enzymatic activation is required for activity. To explain the unexpected phosphorus-oxygen lability, the team ran extensive computational analyses, including conformer generation with RDKit, DFT optimization with the r2SCAN-3c method, and NBO and QTAIM analyses. These supported an intramolecular hydrogen bond between a phosphate oxygen and the cephalosporin carboxylic acid, forming a pseudo-12-membered ring that increases the electrophilicity of the phosphorus atom, although the authors emphasize that both proposed mechanisms remain hypothetical and unproven experimentally.

When the purified CTX-M-15 β-lactamase was added, the picture changed dramatically. Almost all pronucleotides were substrates of the enzyme, with complete disappearance of starting material within 15 minutes, the first time point measured. The enzyme opened the β-lactam ring, and a subsequent elimination reaction released the phosphoramidate monoesters with half-lives of roughly 2 to 6 hours, remarkably similar regardless of the amine substituent on phosphorus. The team confirmed the degradation profile independently by 31P NMR, monitoring compound 16b over 24 hours with an internal standard and verifying by LC-MS that no free AZT appeared. Testing against the metallo-β-lactamase NDM-1, the most clinically concerning enzyme of that class, showed rapid hydrolysis of the phosphoramidate 16b in under 15 minutes, but the phosphodiester 11b proved more stable, taking an hour for full hydrolysis. That stability turned out to be inhibition: the unmasked phosphoric acid function of 11b binds the catalytic zinc of NDM-1 with a measured Ki of 2.40 ± 0.22 μM in a competitive mechanism, while showing no specific inhibition of CTX-M-15. The authors suggest this finding could inform the design of metallo-β-lactamase inhibitors or β-lactam agents resistant to them.

Microbiological evaluation used isogenic E. coli laboratory strains each producing a single β-lactamase, including CTX-M-15, KPC-3, NDM-1, IMP-1 and AmpC, with AZT as a positive control. Among all the pronucleotides, only compound 16b showed activity, with MIC values of 32 μg/mL against the KPC-3 producer and 64 μg/mL against the IMP-1 producer, and more than 128 μg/mL elsewhere. Adding a sub-inhibitory concentration of colistin, a membrane-permeabilizing polymyxin, did not change the MIC values, indicating that the outer membrane is not the main barrier to activity, although efflux and other permeability mechanisms could not be excluded. Against three contemporary clinical or environmental isolates producing multiple β-lactamases, 16b showed no significant activity except against an NDM-5-producing E. coli strain, B28, where an MIC of 128 μg/mL was measured. Because NDM-5 has become the most prevalent NDM variant in numerous countries, the authors consider this an encouraging, if modest, first demonstration of antibacterial activity for their platform.

The work also clarified where the activation cascade slows down. The β-lactamase recognition step and the self-immolative release both worked well in vitro, but the self-immolative intermediate, often described in the literature as decomposing instantaneously, actually persisted for several hours, and the nature of the amine substituent had little influence on that rearrangement rate. The released phosphoramidate monoester was chemically stable under physiological conditions, meaning that cleavage of its phosphorus-nitrogen bond must be enzymatic, most plausibly by bacterial phosphoramidases in the cytoplasm. Those enzymes remain poorly characterized in bacteria, with limited knowledge of their substrate recognition and expression, which currently limits rational design. The closest natural substrate mimic in the series, the amino derivative 25b, showed no antibacterial activity, suggesting poor phosphoramidase recognition or insufficient intracellular accumulation. The pyridine derivative 16b may benefit from partial zwitterionic character that aids inner membrane crossing while preserving phosphoramidase susceptibility, but this remains an interpretation rather than a demonstrated mechanism.

To support the conformational analysis, the team developed and freely released a Python program, IIP.py, built entirely on freeware and designed to identify intramolecular hydrogen bonds in any molecule, validated against literature phosphate compounds. The authors caution that the downstream steps of their delivery cascade, membrane translocation of the monoester and phosphoramidase-mediated cleavage in the cytoplasm, remain to be established, and they call for systematic exploration of bacterial phosphoramidase substrate recognition as a research priority. Even so, the study delivers a working chemical demonstration that β-lactamases can be repurposed as intracellular triggers for pronucleotide activation, and a lead compound, JMV8596, that converts that chemistry into a modest but measurable antibacterial effect against β-lactamase-producing Gram-negative bacteria; the authors state that this derivative potentially represents the first enzyme-triggered antibacterial pronucleotide.

Subject of Research: Cephalosporin-based β-lactamase-triggered pronucleotide prodrugs of zidovudine as a delivery strategy against antibiotic-resistant Gram-negative bacteria

Article Title: Cephalosporin-based Pronucleotides as β-lactamase-triggered antibiotic delivery Systems in Gram-Negative Bacteria

Article References: Faure, F., Cancade, S. M. I., Sansarricq-Meterie, M., Martin-Ghoteimi, C., Carretero, E., Hantova, S., Vo-Hoang, Y., Licznar-Fajardo, P., Lionne, C., Docquier, J.-D., Peyrottes, S., & Gavara, L. (2026). Cephalosporin-based Pronucleotides as β-lactamase-triggered antibiotic delivery Systems in Gram-Negative Bacteria. Results in Chemistry, 32, Article 103989. https://doi.org/10.1016/j.rechem.2026.103989

Image Credits: AI Generated

DOI: 10.1016/j.rechem.2026.103989

Keywords: antibiotic resistance, Gram-negative bacteria, β-lactamases, pronucleotides, zidovudine, cephalosporin, prodrugs, CTX-M-15, NDM-1, phosphoramidates, drug delivery, antimicrobial resistance