At the Interface: Unlocking Enzyme Partnerships in Bacterial Cell-Wall Remodeling

Gram-negative bacterial pathogens rely on a dynamic peptidoglycan (PG) cell wall to maintain viability, withstand antibiotic stress, and interface with their hosts. Although numerous PG-active enzymes and PG-sensing proteins have been identified, the principles that govern how their activities are structurally coordinated, acutely perturbed, and encoded in primary sequence remain incompletely defined. This dissertation addresses these questions by integrating high-resolution structural biology, nanobody-based perturbation, and machine learning to elucidate mechanisms of PG remodeling and recognition across multiple scales, from a single enzyme complex to proteome-wide families of PG-binding domains.

Chapter 1 delineates the structural and biochemical basis of conformational gating in the LtgA–Ape1 PG-remodeling complex from Neisseria meningitidis. Single-particle cryo–electron microscopy, structure-guided mutagenesis, and kinetic analyses reveal how Ape1 binding reorganizes the LtgA active site and interface to control substrate access and product release. Targeted mutations validate key gating elements and decouple catalysis from complex assembly, illustrating how a non-essential complex can fine-tune cell-wall processing in response to local context.

Chapter 2 establishes a high-throughput nanobody discovery and characterization pipeline targeting the soluble lytic transglycosylase Slt from Pseudomonas aeruginosa. Optimized Slt production, yeast-display selections, and functional profiling of Slt-binding nanobodies identify binders that differentially modulate Slt activity and binding. This workflow yields a modular toolkit for acute perturbation of cell-wall enzymes and provides tractable starting points for the development of inhibitors and mechanistic probes.

Chapter 3 presents a global, sequence-based perspective on PG recognition. Curated positive and negative sets of experimentally supported PG-binding domains are used to train a recurrent neural network on k-mer representations of protein sequences, coupled to a high-throughput motif-discovery framework. This approach uncovers over-represented sequence patterns associated with PG binding across diverse folds and taxonomic groups, predicts previously unannotated PG-binding candidates, and highlights convergent solutions to recognizing a chemically conserved yet structurally heterogeneous polymer.

Collectively, these studies define concrete molecular mechanisms, generate versatile tools for perturbing cell-wall enzymes, and propose testable sequence features that can guide the discovery of new PG-interacting proteins. By integrating structural biology, biochemistry, and machine learning, this work offers a multi-scale perspective on bacterial envelope biology and suggests novel strategies for sensitizing problematic Gram-negative pathogens to existing antibiotics.