Bridging the spectral gap in fluorescent proteins through directed evolution

Summary

Fluorescent proteins are used as non-invasive tags for protein trafficking, structure, and action. In this issue of Chemistry and Biology, Hoi et al. (2013) present a new optimized zFP538 yellow fluorescent protein called mPapaya1 that is bright, monomeric, and an excellent fusion partner for cellular proteins.

Imaging proteins fused to fluorescent proteins (FP) is a cornerstone of modern cellular biology and biophysics(Crivat & Taraska, 2012).. This method allows proteins to be observed and studied in living cells, tissues, and organisms. FPs that are bright, structurally stable, and monomeric are generally the best choice to non-invasively tag a protein partner for live cell imaging. A tremendous amount of work over the last 15 years through protein engineering, directed evolution, and the discovery new FPs from different organisms, has led to the development of many excellent FPs that span the visible spectrum(Crivat & Taraska, 2012).

Some notable holes, however, in the FP toolset have been difficult to fill. For example, no stable bright monomeric FPs whose emission sits in the yellow region between ~530 nm and 560 nm have been found(Davidson & Campbell, 2009).. This spectral gap rests between the most red-shifted AequoreavictoriaGreen Fluorescent Protein (GFP) mutants and the most blue-shifted DiscosomaRed Fluorescent Protein (dsRed) variants. Red-shifted GFP variants such as mVenus and mCitrine have the closest emission peaks at 528 and 529 nm respectively. These proteins have a tyrosine at position 203 that interacts via pi-pi stacking with the Tyr66-derived phenol ring of the chromophore. The pi-electron orbital of Tyr203 is thought to affect the electronic transition of the chromophore to red-shifted the emission wavelength(Davidson & Campbell, 2009). Along with other folding and stabilization mutations, these proteins have become useful bright yellow fluorescent proteins (YFPs). Their utility as fluorescence resonance energy transfer (FRET) acceptors, however, has been controversial. Some reports have shown that photo-bleaching of YFP-like protein scan generate a cyan fluorescent protein-like species (Kirber et al., 2007; Valentin et al., 2005). This spurious color switching could seriously skew quantitative FRET measurements. Others, however, have failed to show this behavior (Thaler et al., 2006; Verrier & Soling, 2006). Cleary, the development of new FPs further into the yellow part of the visible spectrum would benefit both multi-color and FRET imaging studies.

In this issue of Chemistry and Biology, Hoi et al.(2013) present the development of a new yellow FP derived from the Zoanthus button polyp protein zFP538 (Figure 1a–b) (Zagranichny et al., 2004). Native zFP538 contains a unique three-ring chromophore not found in other FPs (Figure 1c). Specifically, the polypeptide backbone around the central chromophore breaks at the 65–66 amino acid position and generates a six membered ring (Remington et al., 2005). The addition of a double bond in this system likely extends the conjugation of the chromophore and lowers the peak emission wavelength into the yellow part of the visible spectrum (Figure 1d). The native protein also known as ZsYellowis, however, relatively dim and strongly tetrameric. These characteristics have limited its use as a fusion partner for imaging studies.

Figure 1.

Yellow fluorescent protein from Zoanthus sp. A) The Zoanthus button polyp. B) The crystal structure of the yellow fluorescent protein zFP538 and its chromophore (C). D) The emission spectrum of the mPapaya1 protein (white line). The emission peak is at 541 nm. Photo of Zoanthus sp. courtesy of Coral Morphologic.

Here, through a series of rational and directed evolution procedures, along with a set of stringent screening assays, the authors develop a new zFP538 mutant called mPapaya1. Through the addition of 18 mutations, mPapaya1 has an emission maximum at 541 nm, is brighter than EGFP, is stable, monomeric, does not visibly perturb the behavior of several fusion partners, and should provide another excellent choice as an in vivo probe for microscopy and FRET. To arrive at the final protein the authors performed eight rounds of mutagenesis targeting unique aspects of the zFP538’s structure or behavior. The first two rounds used rational mutations informed by the crystal structure of zFP538 to disrupt the tetramerization interface of the four subunits. Similar mutations have been used to monomerize other FPs. After these rounds, however, several features of the protein including brightness were compromised. This required further mutagenesis and screening to return desirable characteristics to the protein. Specifically, through directed evolution, libraries of error-prone PCR-generated mutants were made, expressed in bacteria, and the resultant colonies were screened for color, brightness, and bleaching. Similar directed evolution/screening methods have been used in the past to create the dsRed-based mFruit fluorescent proteins (Shaner et al., 2004). The success of this method illustrates the power of combining the best aspects of structure-guided mutations with saturated directed evolution. Similar success has been seen in computational protein design where binding sites or catalytic sites in a protein are screened in silico with every possible combination of amino acids to find supportive interactions or geometries (Fleishman & Baker, 2012). To develop mPapaya1, instead of the computer doing the work, the bacteria, combined with massive fluorescent screening, accomplished the goal. It is clear from the recent success of both these approaches(at the bench or on the computer) that the ability to evaluate all possible mutations—some not necessarily obvious to rational design—facilitates successful protein engineering (Romero & Arnold, 2009).

While many studies have used mutagenesis to develop new FPs colors, the ability of a protein to perform as a non-invasive partner is one of the most important characteristics of a fusion tag. In this study, after the protein was fully evolved for brightness, bleaching, and color, the authors go on to test the protein’s performance in vivo. First, to evaluate the monomeric nature of mPapaya1 the protein was evaluated using the recent protocol of Constantini and coworkers (Costantini et al., 2012). In this assay a test protein (mPapaya1 for example)is fused to the cytoplasmic end of the endoplasmic reticulum (ER) membrane protein CytERM. If the fusion pair homo-oligomerizes, the structure of the ER is changed from a tubular network into a visible network of organized smooth ER whorls. Thus, counting the percentage of cells with whorls indicates the level of oligomerization of the test protein. In this study, 83% of cells with the mPapaya1 fusion had a tubular ER morphology, compared to 77% of cells expressing the monomeric control protein mEGFP, suggesting that mPapaya1 is essentially monomeric. As further support for its monomeric nature, mPapaya1 was fused to multiple partners (histone H2B, connexins, tubulin, and others), observed with microscopy, and the fusion was seen to not interfere with the normal trafficking or localization of these proteins.

As a final test the authors evaluated mPapaya1 as a partner for FRET. This was important because one of the motivations to create mPapaya1 was to find an alternative yellow FRET acceptor. To test the response of mPapaya1 in a FRET assay the authors created a proteolytically-sensitive mPapaya/mTFP1 fusion. This donor/acceptor pair should have a Forster radius of 5.1 nm. This fusion showed a strong FRET signal and a large decrease in FRET upon activation of the protease both in vitro and in vivo. Thus, mPapaya1 should act as an excellent yellow FRET acceptor for structural studies or the creation of new biosensors.

Completely rational protein design has seen limited success. Perhaps this is the result of the complex and interconnected nature of a protein’s structure and function (Fleishman & Baker, 2012). Often, non-intuitive mutations are needed to improve folding, activity, or the stability of a protein. A major boon for fluorescent protein design is the ability to rapidly screen for function— namely color and intensity. Similar screening methods to look for other features in proteins such as enzyme activity, ligand binding, or optical qualities, are also being combined with massive directed evolution methods to produce new engineered proteins with optimized behaviors (Fleishman & Baker, 2012; Romero & Arnold, 2009). In addition to providing an exciting new fluorescent protein, the work presented here demonstrates the ability of directed evolution to fill gaps in the collection of nature’s protein library.