Filip Ilievski: Fluorescence labelling in re-coded E. coli with non-canonical chemical entities: Single-codon labelling for single-molecule tracking

Abstract

Single-molecule tracking (SMT) enables direct observation of molecular dynamics in living cells, revealing heterogeneity hidden by in vitro ensemble measurements. However, current protein labeling strategies using self-labeling tags such as HaloTag (~33 kDa) or SNAPtag (~20 kDa) can interfere with the function of proteins that undergo large conformational changes or participate in tightly orchestrated multi-factor complexes. This thesis develops and applies FLORENCE (Fluorescence Labelling in Re-coded E. coli with Non-canonical Chemical Entities), a genetic code expansion (GCE) technology that enables site-specific protein labeling with single-codon resolution for SMT of bacterial elongation factors.

Conventional labeling with bulky tags can prevent functional ribosome binding of translation factors. To address this, in Paper I, we systematically optimized a complete GCE system in genomically re-coded E. coli (GRE) strains where all 321 UAG stop codons have been converted to UAA and release factor 1 deleted. We evaluated pyrrolysyl-tRNA synthetase variants (PylRS1–3), characterized six GRE strains for growth rate and morphology, and optimized a single-plasmid vector architecture combining the orthogonal translation system with the target gene. Using strain-promoted azide-alkyne cycloaddition (SPAAC) between BCNcontaining non-canonical amino acids and JF646-azide dye, we achieved complete labeling within 30 minutes in live cells. Validation with dual-labeled HaloTag and LacY reporters demonstrated that FLORENCE yields SMT results comparable to conventional HaloTag labeling.

In Paper II we applied FLORENCE to study elongation factor G (EF-G), an essential for ribosomal translocation. HaloTag fusions at both termini showed that bulky tags abolish EF-G function in vivo. In contrast, FLORENCE labeling at position 301 (301UAG) revealed 30–45% slow-state occupancy consistent with ribosome binding, as confirmed by tracking the catalytically inactive H92A mutant.

To improve GRE fitness for physiological studies, Paper III reports a novel GRE*, with superior growth compared to the parental GRE6. Single-cell microfluidic analysis confirmed wild-type-like phenotype, and whole genome sequencing revealed deletion of the ratA translation initiation toxin. FLORENCE-labelled EF-G and EF-Tu were tracked at 1 ms temporal resolution, with catalytically inactive mutants showing an increase in ribosome-bound states. Still, as in Paper III, optimization of the expression level of these factors remains critical.

In summary, this thesis establishes FLORENCE as a user-friendly experimental platform for SMT investigation of translation factors and other challenging targets in living bacterial cells.