Miniaturizing GFP

Abstract

The green fluorescent protein (GFP) has enabled researchers to visualize a wide range of cellular processes, from protein expression to metastasis. However, its size (27 kDa) can disrupt localization and association of GFP-tagged proteins. Here, we aim to address this limitation by designing a miniature GFP (<20 kDa) that conserves its chromophore-forming pocket within a shortened beta-barrel fold. Using machine learning-assisted protein design, we have produced nineteen miniature GFPs, averaging 19 kDa each, that display varying levels of expression. These small GFPs display similar excitation and emission wavelengths to wild-type GFP, albeit with fluorescence reduced by four orders of magnitude due to low quantum yield and inefficient chromophore maturation. To improve brightness, we utilized random mutagenesis but were unable to isolate improved variants due to the detection limit of our selection method (FACS) being too high to distinguish miniaturized GFP fluorescence from background cellular fluorescence at the desired wavelengths. Our results show that while machine learning can be used to miniaturize GFP, this process leads to impaired function.