FACS-optimized mutants of the green fluorescent protein (GFP)

A codon-based (NNK) mutagenesis scheme that randomizes amino acids 55–74 of GFP — the 20 residues flanking the Ser65-Tyr66-Gly67 chromophore — covers all possible single-amino-acid substitutions across this region in roughly 2400 clones, avoiding the bias toward single-nucleotide codon changes inherent in standard doped-oligonucleotide mutagenesis. Expressing this mutant gfp library (about 6×10⁶ independent clones) from a tightly IPTG-regulated tac promoter in E. coli produces negligible fluorescence prior to induction; fluorescence-activated cell sorting (FACS) with a standard FITC filter set (488 nm excitation), after 2.5 hours of induction, isolates a small subpopulation of brightly fluorescent mutants for sequencing.

Sequencing the 12 most fluorescent isolates recovered from this screen identifies three distinct classes of GFP mutant: GFPmut1 carries the double substitution F64L/S65T, GFPmut2 carries the triple substitution S65A/V68L/S72A, and GFPmut3 — represented by 10 of the 12 isolates — carries the double substitution S65G/S72A. Ser65 is mutated in all three mutant classes, always in the context of at least one other substitution; no single Ser65 substitution alone appears among the brightest recovered variants.

The GFPmut3 double substitution S65G/S72A occurs in the starting mutagenesis library at a theoretical frequency of only 1 in 3.6×10⁵ clones, yet it is recovered independently at least twice among the sequenced isolates, indicating that S72A specifically potentiates the fluorescence effect of S65G rather than the two substitutions co-occurring by chance.

GFPmut1, GFPmut2, and GFPmut3 each carry excitation maxima shifted roughly 100 nm from wild-type GFP's native 395 nm peak, to 488 nm, 481 nm, and 501 nm respectively, while their emission maxima remain essentially unchanged from wild type (507–511 nm versus 508 nm for wild type).

Equalized for the amount of soluble GFP protein, GFPmut1 fluoresces 35-fold, GFPmut3 fluoresces 21-fold, and GFPmut2 fluoresces 19-fold more intensely than wild-type GFP when excited at 488 nm; at 395 nm excitation (wild-type GFP's own peak), all three mutants are dimmer than wild-type GFP, consistent with their excitation maxima having moved away from that wavelength.

The S65T substitution alone increases GFP fluorescence roughly 6-fold over wild type; combining S65T with F64L, as in GFPmut1, increases fluorescence roughly 30-fold over wild type, showing that F64L contributes additional brightness beyond the spectral shift produced by S65T alone.

Wild-type GFP expressed in E. coli accumulates substantially in insoluble inclusion bodies, whereas GFPmut1 is approximately 90% soluble and GFPmut2 and GFPmut3 are virtually completely soluble under identical expression conditions, even though all four proteins are produced at equivalent levels.

Although GFPmut2 and GFPmut3 each fluoresce, per unit of soluble protein, at only about half the intensity of GFPmut1, whole bacteria expressing GFPmut2 or GFPmut3 fluoresce more than twice as intensely as bacteria expressing GFPmut1 — a discrepancy attributable to the greater solubility and more efficient folding of GFPmut2 and GFPmut3 relative to GFPmut1.

Bacteria expressing GFPmut1, GFPmut2, or GFPmut3 show an approximately 3-fold increase in fluorescence within 8 minutes of IPTG induction, reach half-maximal fluorescence by 25 minutes, and become roughly 100-fold more fluorescent than bacteria expressing wild-type GFP within 1 hour of induction. Under the same log-phase, 37°C growth conditions, wild-type GFP is not easily detectable until 1–2 hours after induction.