Improved monomeric red, orange and yellow fluorescent proteins derived from Discosoma sp. red fluorescent protein Nathan C Shaner 1 , Robert E Campbell 1,6 , Paul A Steinbach 1 , Ben N G Giepmans 3,4 , Amy E Palmer 1 & Roger Y Tsien 1,2,5 Fluorescent proteins are genetically encoded, easily imaged reporters crucial in biology and biotechnology 1,2 . When a protein is tagged by fusion to a fluorescent protein, interactions between fluorescent proteins can undesirably disturb targeting or function 3 . Unfortunately, all wild-type yellow-to-red fluorescent proteins reported so far are obligately tetrameric and often toxic or disruptive 4,5 . The first true monomer was mRFP1, derived from the Discosoma sp. fluorescent protein ‘‘DsRed’’ by directed evolution first to increase the speed of maturation 6 , then to break each subunit interface while restoring fluorescence, which cumulatively required 33 substitutions 7 . Although mRFP1 has already proven widely useful, several properties could bear improvement and more colors would be welcome. We report the next generation of monomers. The latest red version matures more completely, is more tolerant of N-terminal fusions and is over tenfold more photostable than mRFP1. Three monomers with distinguishable hues from yellow-orange to red-orange have higher quantum efficiencies. Although mRFP1 overcame DsRed’s tetramerization and sluggish maturation and exceeded DsRed’s excitation and emission wave- lengths by about 25 nm, the extinction coefficient, fluorescence quantum yield and photostability decreased somewhat during its evolution 7 . To minimize these sacrifices, we subjected mRFP1 to many rounds of directed evolution using both manual and fluores- cence-activated cell sorting (FACS)-based screening. The properties of the resulting variants include several new colors, increased tolerance of N- and C-terminal fusions, and improvements in extinction coeffi- cients, quantum yields and photostability, although no single variant is optimal by all criteria. The red chromophore of DsRed results from the autonomous multi-step post-translational modification of residues Gln66, Tyr67 and Gly68 into an imidazolidinone heterocycle with p-hydroxybenzy- lidene and acylimine substituents 8 . Our first attempt at improving the brightness of mRFP1 involved construction of a directed library in which residues near the chromophore, including position 66, were randomized. The top clone of this library, mRFP1.1, contains the mutation Q66M, which promotes more complete maturation and provides an additional 5 nm red-shift of both the excitation and emission spectra relative to mRFP1. We then set out to reduce the sensitivity of mRFP1.1 to N-terminal fusions. Because Aequorea victoria green fluorescent protein (GFP) is relatively indifferent to N- or C-terminal fusions, we eventually generated mRFP1.3 by replacing the first seven amino acids of mRFP1.1 with the correspond- ing residues from enhanced GFP (MVSKGEE) followed by a spacer sequence NNMA (numbered 6a–d), and appending the last seven amino acids of GFP to the C terminus. mRFP1.3, unlike its pre- decessors, was found to have an equivalent high level of fluorescence regardless of fusions to its N terminus. Through additional rounds of screening random libraries based on mRFP1.3 and wavelength-shifted mRFP variants, we identified the beneficial folding mutations V7I and M182K, which were incorpo- rated into clone mRFP1.4. Randomization of position 163 in mRFP1.4 led to the identification of the substitution M163Q, which results in a nearly complete disappearance of the absorbance peak at B510 nm, present in all previous mRFP clones. The additional mutations N6aD, R17H, K194N, T195V and D196N were introduced through two further rounds of directed evolution to produce our final optimized clone, mCherry (Table 1 and Figs. 1, 2 and 3). To test whether the introduction of GFP-type termini into mRFP variants would benefit fusion proteins expressed in mammalian cells, we fused mRFP1 and mCherry to the N terminus of a-tubulin. In most HeLa cells, expression of mRFP1-a-tubulin resulted in diffuse cytoplasmic fluorescence rather than proper incorporation into microtubules (Fig. 3b). However, mCherry-a-tubulin fusions were successfully incorporated into microtubules in most cells (Fig. 3c), similar to results seen with GFP-coupled tubulin 9 . The amount of full- length mRFP1-a-tubulin expressed was similar to that of mCherry-a- tubulin as verified by in-gel fluorescence and western blot analysis (data not shown). Equivalent results were obtained with Madin-Darby canine kidney and primary human fibroblasts (data not shown). By Published online 21 November 2004; doi:10.1038/nbt1037 1 Departments of Pharmacology, 2 Chemistry and Biochemistry, and 3 Neurosciences, 4 National Center of Microscopy and Imaging Research, and 5 Howard Hughes Medical Institute, University of California at San Diego, 9500 Gilman Drive, La Jolla, California 92093, USA. 6 Present address: Department of Chemistry, University of Alberta, Edmonton, Alberta T6G 2G2, Canada. Correspondence should be addressed to (rtsien@ucsd.edu). NATURE BIOTECHNOLOGY VOLUME 22 NUMBER 12 DECEMBER 2004 1567 LETTERS © 2004 Nature Publishing Group http://www.nature.com/naturebiotechnology