HaloTag Engineering for Enhanced Fluorogenicity and Kinetics with a Styrylpyridium Dye

Abstract HaloTag is a small self‐labeling protein that is frequently used for creating fluorescent reporters in living cells. The small‐molecule dyes used with HaloTag are almost exclusively based on rhodamine scaffolds, which are often expensive and challenging to synthesize. Herein, we report the engineering of HaloTag for use with a chemically accessible, inexpensive fluorophore based on the dimethylamino‐styrylpyridium dye. Through directed evolution, the maximum fluorogenicity and the apparent second‐order bioconjugation rate constants could be improved up to 4‐fold and 42‐fold, respectively. One of the top variants, HT‐SP5, enabled reliable imaging in mammalian cells, with a 113‐fold fluorescence enhancement over the parent protein. Additionally, crystallographic characterization of selected mutants suggests the chemical origin of the fluorescent enhancement. The improved dye system offers a valuable tool for imaging and illustrates the viability of engineering self‐labeling proteins for alternative fluorophores.

Fluorescent reporters are a critical tool for examining cellular structure and processes. HaloTag is a self-labeling, nonfluorescent protein that can be covalently and selectively tagged with a fluorescent dye ( Figure 1). [1,2] This reactivity was engineered by extensive mutation of a dehalogenase enzyme to yield HaloTag version 7 (hereafter HaloTag7), which reacts particularly well with rhodamine dyes. [2] Since then, tremendous effort has been placed on chemical modification of the fluorescent, small molecules to improve the photophysical properties of the dye-HaloTag7 constructs. [3,4] Within this context, the repertoire of small molecules used with HaloTag is still mostly limited to various rhodamine dyes. [3][4][5][6][7][8][9] Although these dyes are bright and bind rapidly to HaloTag7, they are also expensive, and their synthesis requires a sophisticated command of synthetic chemistry. Moreover, focusing on a single class of dyes limits the diversity of chemistry available to users. Different dyes that are also less synthetically demanding exist. [10][11][12] However, the HaloTag-fluorophore constructs that are formed with these simpler dyes often bind slowly to HaloTag, exhibit weak fluorescent enhancement upon HaloTag binding, or are not suitable for imaging in mammalian cells. An example is F1 (Figure 1), which contains a dimethylamino-styrylpyridium core functionalized with a chloroalkane chain. Previous reports identified F1 from a library of dimethylamino-styrylpyridium compounds functionalized with chloroalkane chains of varying lengths. [12] The authors report that F1 exhibits a fluorescent enhancement of 27-fold when mixed with purified HaloTag7, indicating that F1-HaloTag7 is fluorogenic. [12] However, the accumulation of F1 in the mitochondria prevented its use in mammalian cells.
Point mutations of HaloTag7 have shown improvements in the properties of the dye-HaloTag7 constructs. [10,13,14] Thus, we posited that the properties of F1-HaloTag7 might be improved by enzyme engineering. Such a construct would be an alternative to the typical rhodamine dyes and elucidate the potential fluorogenicity of HaloTag variants with other small molecules. To this end, we engineered several variants of HaloTag7 that rapidly bind and enhance the fluorogenicity with F1. The work herein demonstrates that mutations of HaloTag7 significantly improved the performance of the dye-HaloTag construct and allowed for reliable mammalian cell imaging.
We first characterized the interaction between HaloTag7 and F1 (Figures 2 and S1). We found this interaction to be very similar to the previously reported characteristics. [12] To improve the properties of F1-HaloTag7, six positions within HaloTag7 were identified for mutagenesis based on crystallographic data ( Figure 2A): E133, E143, F144, M175, V245, and L271. [10,11,15] Sitesaturation mutagenesis at these six positions was conducted to yield an initial library of single mutants. These mutants were expressed and purified for functional screening. The mutants were not screened directly in E. coli cells to eliminate bias resulting from protein expression profiles in E. coli. The resulting protein libraries were screened for two factors, rate of fluorescence enhancement and maximum fluorogenicity.
Within this single mutant library, four mutations were particularly beneficial (Figures 2 and S2): 133C, 175Y, 245A, and 271D. The apparent second-order kinetics and the maximum fluorogenicity of these variants were determined ( Figure 2B). The maximum fluorogenicity was improved for all four of the top single-mutant variants. The fluorescence response of mutant 245A was red-shifted from that of HaloTag7 and the other mutants ( Figure S3). Mutation 271D demonstrated the most rapid and largest fluorogenic response upon binding to F1. Based on previous crystallographic data, we theorized that residue 271D might engage in an electrostatic interaction with the pyridinium of F1.
None of the variants from the second round were improved for maximum fluorescence enhancement over 271D, but several mutants displayed improved kinetics. Most of the improved variants contained mutation 271D ( Figure S4), which seemed to increase the apparent second-order rate constant most dramatically. Interestingly, recombination of the most improved mutants, R133C/M175Y/V245A/L271D, resulted in complete loss of fluorogenicity ( Figure S4). Additionally, two mutations that were not beneficial in the first rounds were found to be beneficial when combined with 271D: 143M and 144H (HT-SP3). HT-SP5 was the fastest binding variant with an apparent second-order rate constant of 2.4 · 10 3 M À 1 s À 1 . These results correspond to a 42-fold improvement in the rate constant. Nonetheless, the overall second-order rate constant for bioconjugation of F1 with HT-SP5 is lower than that for rhodamine dyes with HaloTag or HaloTag7 (2.7 · 10 6 M À 1 s À 1 and 1.9 · 10 7 M À 1 s À 1 , respectively). [1,2] Based on the positive results with purified protein, we examined F1 with the HaloTag7 variants expressed in E. coli. Each variant was expressed in BL21(DE3) cells under a T7 promoter. The cells were incubated with F1, and the fluorescence in each culture was monitored as a function of time (Figures 4 and S5). The maximum fluorescence   To each E. coli culture, F1 was added to a final concentration of 3.0 μM. The fluorescence was monitored as a function of time (λ ex = 535 nm and λ em = 610 nm). The pseudo-first-order rate constants were determined by a single exponential fit of the data. The studies were repeated in at least six biological replicates. The error bars represent the standard deviation from the mean. enhancement for HaloTag7 was 1.77 � 0.13. All of the mutants tested were improved over HaloTag7. The top variants were 271D, HT-SP3, and HT-SP4, which exhibited fluorescent enhancements (and standard deviation) of 5.46 � 0.32, 6.70 � 0.53, and 6.28 � 0.54, respectively.
The kinetic traces were modeled to pseudo-first-order kinetics. Comparison between the apparent second-order rate constants in purified protein and the pseudo-first-order rate constants in E. coli displayed a similar trend. The notable difference was HT-SP5, which is one of the best mutants in the study with purified protein. HT-SP5 expresses in E. coli much worse than HaloTag7 and other mutants (approximately 8-fold lower). The low expression of HT-SP5 results in reduced concentration of the variant, which may prevent this experiment from being in the pseudo-first order range. This factor would explain why the apparent second-order rate constant appears lower in this E. coli-based assay.
Based on the dramatic improvements in both reaction rate and fluorescence enhancement, we tested HaloTag7 and two mutants-HT-SP4 (the highest fluorogenicity) and HT-SP5 (the fastest mutant for bioconjugation)-in HeLa cells ( Figure 5). The results were compared with the commercial rhodamine-based dye Janelia Fluor® 549 HaloTag® (hereafter JF549, Promega, Scheme S1). Previous work and chemical intuition would suggest that the cationic F1 might accumulate in the mitochondria. [12] However, we were hopeful that the improved reaction rate and fluorogenicity with the HaloTag7 variants might make this system suitable for imaging in other cellular locations. Both HaloTag7 and the variants were tagged with a localization target for actin (Lifeact). [16] The cells transfected with HaloTag7, HT-SP4, and HT-SP5 were compared upon incubation with JF549 (Figures 5 and S7) or F1 (Figures 5 and  S8). For all F1 experiments, a washing step was applied prior to imaging.
The fluorescence was high for JF549 with HaloTag7 and the two variants. No significant difference in the fluorescence was observed for the different protein conjugates with JF549 ( Figure 5B, left). These results indicate that all of these constructs are capable of stabilizing the zwitterionic form of the rhodamine dye. The minimal of change in the fluorescence response is not surprising, given that the linker length of JF549 is much longer than F1 (Scheme S1). [17] In contrast, dramatically different fluorogenicities are observed for the HaloTag7 variants conjugated to F1 (Figures 5 and S8).
As expected with HaloTag7, F1 accumulates significantly in mitochondria and displays a weak fluorescent signal, excluding this construct as a viable fluorescent reporter (Figures 5 and S8). HT-SP4, the most fluorogenic variant in purified protein, exhibited a 31-fold fluorescence enhancement over HaloTag7 ( Figure 5B). Although actin labeling can be observed with HT-SP4, the signal is weak (Figures 5A and S8). HT-SP5, the most rapidly binding variant, was 113-fold more fluorogenic with F1 than the parent HaloTag7 ( Figure 5B). This strong fluorogenicity is particularly remarkable given that, unlike rhodamine-based dyes, F1 does not undergo a spirocyclization equilibrium that can be exploited to induce large fluorogenicities. For HT-SP5, cell images could be obtained that showed clear actin fibers ( Figure 5A). These results suggest that the rate of bioconjugation may be the primary factor for improvement of cell imaging. Notably, it was still necessary to wash the cells with HT-SP5 and F1 ( Figure S9).
To elucidate the chemical origin of the improved fluorescence response, we attempted to obtain crystallographic data of the variants bound to F1. Crystallographic data with well-resolved F1 were obtained for HaloTag7 and HT-SP2 ( Figure 6). In the crystal structure of F1-HaloTag7, the dimethylamino-styrylpyridium core is loosely packed by the F144, A145, T148, T172, and M175. In contrast, mutations in HT-SP2 reposition the dimethylamino-styrylpyridium core within the wide opening of HaloTag. In HT-SP2, the primary packing interactions with F1 are formed by W141, P142, 144H, A145, P243, 245A, and L271. Residue 144H of F1-HT-SP2 adopts two conformations. Notably, in both structures, no hydrogen bonding or electrostatic interactions were observed between the protein and F1. These structural data highlight that HaloTag7 can be mutated to favor large changes in the orientation of F1, underscoring the plasticity of the scaffold.
Although a crystal structure could not be obtained for F1-HT-SP5, chemical intuition suggests that an electrostatic interaction could be formed between 271D and the pyridinium of F1. This interaction may be key in improving two factors: i) rate of bioconjugation and ii) the rigidity of the dimethylaminostyrylpyridium, which improves intramolecular charge transfer necessary for fluorescence with this small molecule.
The work herein presents an evolved protein for fluorescent labeling in mammalian cells with a chemically accessible dye. Additionally, we demonstrate that an easily synthesized probe can be made useful through the careful engineering of HaloTag. Although much work in the field has focused on fluorescence optimization through synthetic chemistry, [3,4] this work suggests that an equally suitable route is the genetic optimization of the scaffold proteins (such as HaloTag7) for use with alternative fluorophores beyond the rhodamine family. We expect that this strategy could be extended to additional fluorophore families enabling the use of a wider range of tools for live-cell imaging.