Christodoulos Mitrogiannis defends thesis on finding antibiotic candidates in billion-molecule libraries

Christodoulos Mitrogiannis successfully defended his PhD thesis, "From Library Design to Hit Optimization: Leveraging DNA-Encoded Chemical Libraries to Target LpxH," at Uppsala University on 11 September 2026. Christodoulos is the ninth member of the second cohort of UAC's research school to defend their thesis.

New antibiotics need new targets and new ways to search

The bacteria at the top of the WHO's critical priority list, including Acinetobacter baumannii, Pseudomonas aeruginosa and the Enterobacterales, are Gram-negative. Their double membrane keeps most drugs out, and few new antibiotic classes against them have reached patients in decades. Getting truly new antibiotics is difficult in two ways: finding what to hit, and finding a molecule that hits it.

Christodoulos's thesis tackles both. The target is LpxH, an enzyme that bacteria need to build lipid A, an essential part of their outer membrane. Blocking LpxH in theory does double damage. First, the membrane cannot be built properly, and second, toxic intermediates pile up inside the cell. LpxH is used by around 70 percent of Gram-negative bacteria, including the critical priority pathogens, and has no close equivalent in human cells. Additionally, no approved antibiotic targets it, so in principle there is no existing resistance.

The search method that Christodoulos used was DNA-encoded chemical library (DECL) technology. Each molecule in the library carries a short DNA barcode recording how it was built. Millions or billions of molecules can be mixed with the target protein in a single tube. Those that stick to the target are kept, while the rest are washed away, and sequencing the barcodes reveals which molecules bound.

From library design to drug candidates

A first library proved the method but could not engage LpxH. Christodoulos designed and synthesised a library of 403,200 peptide-like molecules, using computational analysis to choose building blocks that spread the library across as much chemical diversity as possible. It recovered exactly the chemical groups known to bind a model enzyme, confirming that both the synthesis and the barcoding were working as expected, yet when tested against LpxH from two pathogens nothing bound. The result pointed to a design problem: the flexible, chain-like molecules in this library were not rigid enough to fit the enzyme's pocket.

A more rigid library found a new family of binders. Using that lesson, he built a second library of 1.3 million molecules designed to be more rigid and more three-dimensional. Screened against P. aeruginosa’s LpxH, it produced a clear, target-specific hit series of molecules. Computer modelling of how these molecules sit in the enzyme's active site gave a plausible binding pose to guide the next round of validation and optimisation.

A 7.2-billion-molecule library delivered compounds that kill bacteria through LpxH. In the final project, Christodoulos screened a commercial collection of 7.2 billion molecules against LpxH from E. coli and P. aeruginosa. Two families emerged. The first, sulfonyl piperazines, resembled known LpxH inhibitors, which confirmed the screen was finding the right site. The second, built on a pyrimidine core, had never been described as an LpxH inhibitor. He then resynthesised the hits without their DNA tags and redesigned them step by step, using computer models of the binding pocket and rules for what lets a molecule cross the Gram-negative membrane. The early hits were poorly soluble and could not get into bacterial cells. Adding bulkier, three-dimensional side groups improved both. The best compounds stopped the growth of E. coli, one of them in a wild-type strain, and several were active against Klebsiella pneumoniae strains with weakened defences. To prove the compounds worked by blocking LpxH rather than some other mechanism, he tested them in engineered E. coli in which LpxH levels could be reduced. With less LpxH, the bacteria became up to 32 times more sensitive, the signature of a compound acting on that enzyme.

Why this work matters for AMR

Antibiotic discovery against Gram-negative bacteria has stalled because two things are hard at the same time: finding an unexploited target and finding molecules that can reach it inside the cell. Christodoulos's work establishes LpxH as a target that DECL screening can hit, and delivers a previously unknown chemical family that inhibits it and can kills bacteria in vitro.

The thesis also leaves a practical lesson for anyone building these libraries. Comparing all three projects shows that library design, not only size, is essential for the outcome. A small, flexible library failed against LpxH, while a more rigid library succeeded. Then, a very large collection spanning many different chemical families gave the highest hit rate and the most novel starting points.

We warmly congratulate Christodoulos on a thesis that turns an underexplored target into real chemical starting points for new antibiotics, and we look forward to following his next steps.