Improved Chemical-Genetic Fluorescent Markers for Live Cell Microscopy
Inducible chemical-genetic fluorescent markers are promising tools for live cell imaging requiring high spatiotemporal resolution and low background fluorescence. The Fluorescence-Activating and absorption Shifting Tag (FAST) was recently developed to form fluorescent molecular complexes with a family of small, synthetic fluorogenic chromophores (so-called fluorogens). Here, we use rational design to modify the binding pocket of the protein and screen for improved fluorescence performances with
... e performances with four different fluorogens. The introduction of a single mutation results in improvements in both quantum yield and dissociation constant with nearly all fluorogens tested. Our improved FAST (iFAST) allowed the generation of a tandem td-iFAST that forms green and red fluorescent reporters 1.6-fold and 2-fold brighter than EGFP and mCherry, respectively, while having comparable size. The ubiquitous implementation of fluorescence microscopy to address biological questions has led to the development of a series of markers based on small molecules, proteins, and hybrid techniques. The discovery of the green fluorescent protein (GFP) and the subsequent expansion of the color palette of fluorescent proteins have revolutionized biological imaging due to being genetically encoded and their absolute labeling specificity. 1 In parallel, a suite of small molecule probes with excellent brightness and photostability complements the toolbox of fluorescent markers. 2 However, both approaches suffer from drawbacks -fluorescent proteins are large and have slow maturation times, while small molecule probes suffer from off-target labeling. Thus, hybrid techniques have sought to combine the advantages of the two approaches for live cell imaging, while circumventing their respective drawbacks. 3 A particularly promising hybrid approach is the use of inducible chemical-genetic fluorescent markers, which are comprised of a small protein tag that interacts specifically with a small molecule ligand, termed fluorogen, whose fluorescence is increased upon interaction with its protein tag. 4-6 A recently developed system named the Fluorescence-Activating and absorption Shifting Tag (FAST) 7 is comprised of a small, monomeric, 14 kDa protein that was engineered from the photoactive yellow protein (PYP) to specifically and reversibly bind a series of hydroxybenzylidene rhodanine (HBR) derivatives (Figure 1A,B). Upon interaction with FAST, the fluorogen is locked into an emissive conformation. The resulting fluorescent complex provides many of the advantages of traditional autofluorescent proteins for live cell imaging while circumventing slow maturation times and oxygen dependence. The bipartite nature of this type of system allows one to optimize the fluorogen and the protein tag independently to improve or fine-tune the properties of the system. Previously, optimization of the fluorogen generated a suite of brighter and redder fluorescent markers, while still retaining the same protein tag. 8 To further improve the brightness, we optimized the protein tag through rational design of the binding pocket. The resulting variants were found to have increased quantum yields, particularly with the recently developed fluorogens HBR-3OM and HBR-3,5DOM. 8 We compared our rationally designed variants to new FAST variants that were recently identified through a new image-based, in-cell library screen to be brighter than FAST, but were not characterized as purified proteins. 9 Finally, we present tandem constructs of FAST variants as an effective strategy to increase the overall brightness of the system. The parent protein of FAST, PYP, is a bacterial photoreceptor with a covalently attached 4-hydroxycinnamoyl (HC) chromophore, which participates in hydrogen bonding with Y42 and E46 ( Figure 1C ). 10-12 FAST was engineered through directed evolution of PYP by saturation mutagenesis of the loops near the chromophore binding pocket, resulting in replacement of the loop 94-101 in conjunction with the mutation of C69, which tethers HC, to glycine. However, the rest of the chromophore pocket of PYP remains intact, which can inform our understanding of how HBR derivatives bind to FAST. The observed 2 Figure 1. FAST principle and improvement. (A) FAST variants bind and activate the fluorescence of HBR analogs. (B) The structures of the HBR analogs used in this study. (C) Detailed view of the chromophore binding pocket in PYP including the relative positions of V107 and V122 (PDB: 1nwz). (D) Proposed fluorogen binding pocket in iFAST. characteristics of FAST, namely the absorption red-shift and fluorescence activation upon fluorogen binding are due to the binding mode of the fluorogen to the protein. The absorption red-shift is in agreement with the fluorogen being bound in its deprotonated form and stabilized by the same amino acids that interact with the chromophore of PYP, Y42 and E46. Furthermore, binding restricts the conformational freedom of the chromophore, leading to fluorescence activation. In seeking to improve the fluorescence quantum yield of FAST, we sought to further increase the conformational locking of the chromophore in the binding pocket. We identified the residues V107 and V122 as being potential targets for rational design because they are about 8 and 6 Å from the benzyl ring of 4-hydroxycinnamic acid in the PYP crystal structure (PDB: 1NWZ). 13 Since the most successful fluorogens yet developed for FAST include modifications to the benzyl moiety to increase bulk, 8 we hypothesized that further introduction of bulky residues at these positions in the binding pocket of the protein would further improve the physicochemical properties of the complex ( Figure 1D ). The proteins FAST V107I , FAST V122I , and double mutant FAST V107I,V122I were generated by site-directed mutagenesis and expressed in E. coli. The mutants reported by Emanuel et al, 9 mutant1, mutant2, and mutant3, were also expressed and purified. The thermodynamic dissociation constant KD, fluorescence quantum yield (Φ), and molar absorption coefficient were measured for the 24 pairs made of the six proteins with the four fluorogens: HMBR, HBR-3,5DM, HBR-3OM, and HBR-3,5DOM (Table 1, Figures 2 and S1 ). The mutation V107I was observed to slightly increase the affinity of the protein for the fluorogens. Furthermore, for certain fluorogens, particularly HBR-3OM and HBR-3,5DOM, which have bulky methoxy substituents on the benzyl ring, the quantum yield was dramatically increased, while the quantum yield for HBR-3,5DM exhibited a more modest increase in quantum yield. On the other hand, the mutation V122I decreased the binding affinity for the fluorogens by a factor of 3 to 8, depending on the fluorogen. This loss of affinity might result from V122I being closer to the fluorogen binding site relative to V107I. Nonetheless, this mutation also increased the quantum yield of several of the 3 Table 1 . Physico-chemical properties of mutant:fluorogen complexes in PBS pH 7.4. Abbreviations are as follows: labs, wavelength of maximal absorption; lem, wavelength of maximal emission; e, molar absorption coefficient at labs (standard error is typically 10%); Φ, fluorescence quantum yield; KD thermodynamic dissociation constant. Mutant Fluorogen µM) FAST HMBR 481 540 45 0.23 ± 0.03 0.13 HBR-3,5DM 499 562 48 0.49 ± 0.05 0.08 HBR-3OM 494 561 40 0.36 ± 0.04 0.31 HBR-3,5-DOM 518 600 39 0.31 ± 0.03 0.97 iFAST HMBR 480 541 41 0.22 ± 0.03 0.07 HBR-3,5DM 499 558 46 0.57 ± 0.08 0.06 HBR-3OM 495 560 39 0.49 ± 0.07 0.2 HBR-3,5-DOM 516 600 38 0.40 ± 0.04 0.41 FAST V122I HMBR 480 542 41 0.35 ± 0.05 0.35 HBR-3,5DM 499 559 43 0.52 ± 0.07 0.57 HBR-3OM 495 560 38 0.47 ± 0.07 1.4 HBR-3,5-DOM 516 596 ~30 0.32 ± 0.03 8.3 FAST V107I,V122I HMBR 480 541 35 0.24 ± 0.03 0.3 HBR-3,5DM 499 558 50 0.48 ± 0.07 0.29 HBR-3OM 495 560 33 0.49 ± 0.07 2.0 HBR-3,5-DOM 516 598 ~30 0.33 ± 0.04 6.4 mutant1 HMBR 486 545 56 0.23 ± 0.03 0.024 HBR-3,5DM 507 565 48 0.25 ± 0.03 0.58 HBR-3OM 501 555 51 0.26 ± 0.03 0.17 HBR-3,5-DOM 515 595 36 0.26 ± 0.03 2.0 mutant2 HMBR 486 544 59 0.23 ± 0.03 0.024 HBR-3,5DM 504 565 47 0.34 ± 0.04 0.49 HBR-3OM 501 555 48 0.26 ± 0.03 0.2 HBR-3,5-DOM 516 595 41 0.27 ± 0.03 2.5 mutant3 HMBR 488 547 39 0.27 ± 0.04 0.012 HBR-3,5DM 504 565 47 0.34 ± 0.04 0.21 HBR-3OM 502 555 50 0.24 ± 0.03 0.2 HBR-3,5-DOM 515 596 38 0.25 ± 0.03 3.0 complexes. The effect of the two mutations was not additive, as seen from the characterization of the double mutant FAST V107I,V122I , which displayed lower binding affinities as might be expected from the introduction of V122I and similar increases in quantum yield as previously observed in both single mutants. The overall brightest complex is FAST V107I :HBR-3,5DM, while the largest improvement in quantum yield was achieved with FAST V107I :HBR-3OM, thus we termed this protein iFAST for improved-FAST. Interestingly, the mutant1-3 that were reported previously as brighter than FAST:HMBR by in-cell measurement in E. coli showed no significant improvement in fluorescence quantum yield relative to FAST, although the binding affinity for HMBR was increased by an order of magnitude. When tested with the three other fluorogens, mutants1-3 showed poorer performance, exhibiting lower quantum yields overall. The reported improvement in mutant1-3 brightness with HMBR in E. coli cells is probably due to the increased binding affinity favoring reporter formation in bacteria. Next, we hypothesized that dimerizing FAST could generate significantly brighter markers for imaging through increase of the apparent absorption coefficient, as 4 Figure 2. Spectral characteristics of FAST and iFAST. Absorption spectra of FAST (A) and iFAST (C) and fluorescence spectra of FAST (B) and iFAST (D) with HMBR, HBR-3,5DM, HBR-3OM, and HBR-3,5DOM. Fluorogen concentration was 3 µM and FAST mutant concentration was 40 µM. Spectra were recorded in pH 7.4 PBS at 25°C. has previously been done with some autofluorescent proteins such as tdTomato. 14 A tandem of FAST (td-FAST) results in a probe of about 28 kDa, which is about the same size as fluorescent proteins, with a brightness 1.4fold higher than EGFP 15 when used with HBR-3,5DM. Furthermore, to capitalize on the mutations that improve the quantum yield of FAST, we generated a tandem of iFAST (td-iFAST) as well, which provides an additional improvement in overall brightness depending on the fluorogen chosen. The brightness of td-FAST and td-iFAST along with FAST and iFAST was tested in live cell confocal microscopy with HBR-3OM, which exhibits the largest change in quantum yield between FAST and iFAST, to evaluate their relative brightness in the context of cell imaging (Figure 3 ). We observed that tandems of FAST are expressed homogenously in the cell, confirming that neither the size nor the construction of the tandem impedes diffusion into the nucleus ( Figure 3A ). Relative to FAST, iFAST was found to be 1.7-fold brighter, which is slightly higher than predicted based on the brightness measured in vitro ( Figure 3B ,C and Figures S2-S3) . Furthermore, td-FAST and td-iFAST were found to be 2.8fold and 3.8-fold brighter than FAST, respectively ( Figure 3B ,C and Figures S2-S3 ). Next we examined the photobleaching of FAST, iFAST, FAST V122I , and mutant1 with HMBR in living cells. It has previously been reported that FAST exhibits a biphasic photobleaching behavior composed of a rapid phase and a slow phase. 9,16 Long acquisitions by confocal microscopy allowed us to compare the photobleaching behavior of the four proteins with 10 µM HMBR ( Figure S4A ). In the conditions used, iFAST, FAST V122I , and mutant1 show a reduction of the amplitude of the rapid bleaching phase compared to FAST. After this rapid initial fluorescence decrease, FAST, iFAST, FAST V122I showed almost steady fluorescence, while mutant1 seemed to be slightly more prone to photodamage ( Figure S4A ). Mutant1 contains a cysteine residue, which is prone to photooxidation, in its sequence that is absent in the three other variants, which might explain the observed difference of behavior. The photobleaching of FAST and iFAST were also examined in the presence of 10 µM HBR-3,5DM, HBR-3OM, and HBR-3,5DOM ( Figure S4B -D). FAST and iFAST showed similar photobleaching curves for all three fluorogens. The reporters with HBR-3,5DM and HBR-3OM, while being brighter, appear to be slightly more sensitive to photobleaching than with HMBR, particularly in the case of HBR-3OM. Imaging with 10 µM HBR-3,5DOM revealed that both FAST and iFAST are as photostable as mCherry. One major advantage of fluorogenic and hybrid systems comprised of a small molecule and a protein tag is that further development of the system can be done by 5 Figure 3 . Relative in-cell brightness of FAST, iFAST, td-FAST, and td-iFAST. (A) Representative confocal micrographs of HEK 293T cells expressing FAST, iFAST, td-FAST, and td-iFAST. The images were all acquired at 16-bit with the same settings, each pixel defined by 65,536 grey levels. Given the difference of brightness between FAST, iFAST, td-FAST and td-iFAST, the intensity scale used for the images of tdFAST and td-iFAST did not allow the proper visualization of FAST and iFAST. Therefore we provided one panel (top) that is optimized for visualizing td-FAST and a second (bottom) that is optimized for visualizing FAST. Intensity (grey level) scales are given on the right of each panel. Cells were labeled with 20 µM HBR-3OM. Scale bars are 20 µm. (B) Relative brightness in cells (mean ± sem of 4 experiments, each with an average of 80-90 cells quantified). Fluorescence levels were normalized by the protein expression level determined by quantitative western blotting as shown on Figure S2 . Figure S3 presents the individual measurements. (C) Comparison of the brightness of FAST, iFAST, td-FAST and td-iFAST labeled with HBR-3OM in mammalian cells (in vivo brightness) and solution (in vitro brightness). optimizing either the protein or the small molecule. Here, we have presented work to improve the fluorescence performance of the FAST system by manipulation of the protein tag by rational design to introduce a single mutation that increases steric hindrance around the chromophore. The use of rational design was complementary to the directed evolution approach used originally for developing FAST, as it has facilitated the finetuning of properties difficult to further improve by high-throughput approaches. Not only has rational design provided us with improved tags but also supports our hypothesis that HBR derivatives bind to FAST in a similar fashion as 4-hy-droxycinnamic acid to PYP. We also showed that dimerizing FAST is an effective way to increase the overall brightness, particularly when the size of the tag is less of an issue. Optimized td-iFAST is 1.6-fold brighter than EGFP, while having comparable size. Since the color of FAST-based systems can be changed at-will, td-iFAST presents a particularly attractive alternative to red fluorescent proteins. In particular, td-iFAST with HBR-3,5DOM is 2-fold brighter than mCherry. These new variants of FAST exhibit quantum yields on par with fluorescent proteins while retaining the advantages of the FAST system -namely, no maturation time, on-demand imaging, small tag size, and the ability to change the color at-will.