Experimental identification and theoretical analysis of a thermally stabilized green fluorescent protein variant

In this study, we aim to relate experimentally measured macroscopic properties to dynamic and structural changes as calculated by molecular dynamics (MD) simulations. We performed the analysis on four GFP (green fluorescent protein) variants, which have amino acid replacements or insertion in a flexible region on the protein surface and which resulted from a previous protein splicing reaction optimization experiment. The variants are a reference GFP (CEGFP), GFP-N144C, GFP-N144C/Y145F, and a GFP with five residues inserted between Y145 and N146 (GFP-5ins). As a result, we identified a single Y145F mutation that increased the thermal stability of GFP-N144C/Y145F by 3-4 °C. Because circular dichroism measurements indicated that the overall GFP β-barrel fold was maintained in all variants, we presumed that the fluorescence activity and thermal stability related to local changes that could be detected by standard MD simulations. The 60 ns MD simulations indicated that the Y145's hydroxyl group, which is straight and buried in the crystal structure, was bent avoiding the hydrophobic core during the simulation in both CEGFP and GFP-N144C. This local strain was relieved in GFP-N144C/Y145F, where the tyrosine's hydroxyl group was replaced with the F145 hydrophobic aliphatic carbon. F145 remained indeed buried during the simulation maintaining local compactness, which presumably reflected the improved thermal stability of GFP-N144C/Y145F. Furthermore, the analysis of internal water molecules localized within the GFP's β-barrel suggested that a change in the local hydrogen bonding pattern around the chromophore correlated with a strong fluorescence activity decrease in GFP-5ins. Although relating experimental observation with calculated molecular features proved to be delicate, this study suggested that some microscopic features could be useful reporters for redesigning GFPs and other proteins. The newly identified GFP-N144C/Y145F was among the most stable GFP variant and demonstrates the potential of such computer-aided design.