Installing hydrogen bonds as a general strategy to control viscosity sensitivity of molecular rotor fluorophores

Molecular rotor‐based fluorophores (RBFs) activate fluorescence upon increase of micro‐viscosity, thus bearing a broad application promise in many fields. However, it remains a challenge to control how fluorescence of RBFs responds to viscosity changes. Herein, we demonstrate that the formation and regulation of intramolecular hydrogen bonds in the excited state of RBFs could modulate their rotational barrier, leading to a rational control of how their fluorescence can be activated by micro‐viscosity. Based on this strategy, a series of RBFs were developed based on 4‐hydroxybenzylidene‐imidazolinone (HBI) that span a wide range of viscosity sensitivity. Combined with the AggTag method that we previously reported, the varying viscosity sensitivity and emission spectra of these probes enabled a dual‐color imaging strategy that detects both protein oligomers and aggregates during the multistep aggregation process of proteins in live cells. In summary, our work indicates that installing intracellular excited state hydrogen bonds to RBFs allows for a rational control of rotational barrier, thus allow for a fine tune of their viscosity sensitivity. Beyond RBFs, we envision similar strategies can be applied to control the fluorogenic behavior of a large group of fluorophores whose emission is dependent on excited state rotational motion, including aggregation‐induced emission fluorophores.


INTRODUCTION
3][4] These molecules typically consist of a rotor group, which is an excited state single bond to allow for rotation. [5]8][9] In solvents of low viscosity, RBFs would rotate and undergo the twisted intramolecular charge transfer (TICT) process, wherein the intramolecular charge transfer rapidly occurs at excited state to effectively quench fluorescent emission. [10]hen RBFs are buried in viscous environments, the TICT process is inhibited, resulting in enhanced fluorescence (Figure 1A). [11][14][15] However, it remains a challenge to control the way how fluorescence of RBFs responds to viscosity changes, thus limiting the application and application of the large class of RBFs.Hence, it is highly desirable to develop multiple chemical principles that can rationally control the viscosity sensitivity for a broad range of scaffolds for RBFs.
The key to control viscosity sensitivity is to regulate the rotational barrier along the rotating group at excited state. [16,17]In the past few years, our group have reported two strategies that employ the π electron density or π-rich carbon-carbon linkages to influence the viscosity sensitivity.For instance, we showed that the excited state rotational barrier of 4-hydroxybenzylidene-imidazolinone (HBI) could be rationally controlled by an electron density regulator (EDR).The π electron of EDR could conjugate with the fluorophore and modulate the σ versus π component of the rotating Ibond (Figure 1B), thus altering HBI's excited state rotational barrier and viscosity sensitivity. [18]However, this method is mostly limited to the regulation of HBI but challenging for other RBF scaffolds.In another work, we reported the installation of π-rich carbon-carbon linkages as a general strategy to modulate rotational barriers of multiple scaffolds of RBFs. [19]Unfortunately, the extension of the π-rich linkers would inevitably result in red-shift of both excitation and emission spectra.Thus, it would be beneficial to introduce new chemical strategies that could both control viscosity sensitivity of multiple RBF scaffolds and minimally perturbed their spectra features.
In this work, we report the installation of hydrogen bonds (H-bonds) at the excited state to regulate the viscosity sensitivity of RBFs.Using HBI as an example, an intramolecular H-bond at the ortho position of the benzene ring (Figure 1B) is found to effectively modulate HBI's excited state rotational barriers (E a ), leading to the control of viscosity sensitivity (Figure 1B).We demonstrate that the intramolecular H-bonds primarily increase E a , whose value is positively correlated with strength of the H-bonds.In addition to HBI, we also show the effectiveness of this strategy to other RBF scaffolds, thus validating the generality of this strategy.Furthermore, the intramolecular H-bond could be jointly applied with other mechanisms, rendering a synergistic effect to modulate the viscosity sensitivity of RBFs.Finally, we applied HBI-based probes with intramolecular H-bonds to visualize the multistep protein aggregation process both in vitro and in live cells.In summary, this work provides another general strategy to modulate the viscosity sensitivity of RBFs, by both tunning the excited state rotational barriers and minimally perturbed the emission spectra features.The introduction of intramolecular H-bonds can be potentially applied to other molecular excited states towards chemical control of desired properties.

Modulating excited state rotational barriers of HBI-based RBFs with intramolecular hydrogen bonds
[22] We then hypothesize that installing H-bond with different strength to HBI would modulate its rotational motion, leading to varying viscosity sensitivity.To this end, we synthesized HBI (1a) and three derivatives (1b-1d) with different types of intramolecular hydrogen bonds as O−H⋅⋅⋅, N−H⋅⋅⋅N, and O−H⋅⋅⋅N, respectively (Figures 2A S1-S4, and Table S1).Subsequently, we measured the fluorescence intensity of compounds 1a-1d in solvents with varying viscosities, using mixtures of ethylene glycol and glycerol with different ratios. [7]The slope of the fitting plot using logarithm of fluorescence intensity as a function of logarithm of solvent viscosity was defined as viscosity sensitivity value (x), which was determined as 0.67, 0.43, 0.42, and 0.33 for 1a-1d, respectively (Figures 2B-C, S17).These results indicate that intramolecular H-bonds influence the viscosity sensitivity of HBI-based RBFs.
The decreasing x values from 1a to 1d could be attributed to the increasing excited state rotational barrier.To test this notion, we conducted density functional theory (DFT) calculations to determine the potential energy surface (PES) of the ground state (S 0 ) and the first excited singlet state (S 1 ) (Table S2-S3).In the calculation, we defined the φ angle as the dihedral angle along which the benzyl group rotates along the I bond (Figure 1B).For 1a that exhibited the largest x value, we found its S 1 excited state rotational barrier (E a ) of 0.05 eV (Figure 2C,D).Whereas, 1d with the smallest x value exhibited a much higher barrier as 0.23 eV (Figures 2C and 2E).At their TICT state (φ = 90 • ), the energy gap (E TICT ) between S 1 and S 0 states was found to be 0.3698 eV for 1a and 0.6469 eV for 1d (Figures 2D-E).For 1b and 1c with medium x values, their rotational barriers were determined to be 0.0857 eV and 0.097 eV, respectively (Figure 2B).The E a values of HBI derivatives were negatively correlated with their x values (Figure 2B), suggesting that the higher rotational barrier corresponds with the lower x value (i.e., prone to activate fluorescence at lower viscosities).Thus, the photo-excited 1d would experience a greater E a to undergo nonradiative decay via a greater E TICT , resulting in fluorescence emission even at lower viscosities.By contrast, 1a would rotate over a smaller E a and enter into TICT state to undergo non-radiative decay via a smaller E TICT .These results collectively suggest that the introduction of intramolecular H-bond effectively restricts the rotation of I bond, resulting in a higher rotational barrier at excited state and a reduced viscosity sensitivity for HBI.

Evidence of intramolecular H-bonds at the excited state of HBI-based RBFs
We next investigated the intramolecular H-bonds of 1b to 1d in both ground (S 0 ) and excited states (S 1 ).Reduced density gradient (RDG) analysis indicated that the existence of H-bonds for 1b-c in both ground state and excited state (Figures 3A, S21-S25).At the S 0 state, the hydrogen bond energies of 1b, 1c, and 1d were determined to be −7.61,−9.11, and −15.43 kcal/mol, respectively (Figure 3B).At the S 1 state, the H-bond energy of 1b, 1c, and 1d were calculated as −8.38, −9.70, and −15.74 kcal/mol, respectively (Figure 3B).At both the S 1 state of each compound, the increasing H-bond strength corresponds to increasing rotational barrier and decreasing x values (Figure 2B).In addition, strength of the excited H-bond was stronger than their ground state counterpart for all compounds, suggesting that the intramolecular H-bonds would be more effectively affecting the rotational motion along the I bond at the excited states.
H-bond has been shown to induce the process of excited state intramolecular proton transfer (ESIPT). [23,24]To rule out this possibility, we calculated the Gibbs free energy (ΔG) of the ESIPT process (Table S4-S9), which would be favoured by a negative ΔG and disfavoured by a positive ΔG.For 1b and 1c, no ESIPT products could be identified by calculation (Table S9).For 1d with the strongest H-bond, a product of ESIPT could be formed, albeit with ΔG as 18.04 kJ/mol (Figure 3C), indicating that the ESIPT process is thermo-dynamically unfavourable.These results strongly suggest that the intramolecular H-bond, which is strengthened at the excited state, plays an important role in modulating the excited state rotational barrier and viscosity sensitivity of HBI derivatives.

Installation of intramolecular H-bonds is a general strategy to modulate viscosity sensitivity of RBFs
We then asked whether the viscosity sensitivity of other RBFs could be influenced by the installation of intramolecular Hbonds.To this end, we selected two representative RBFs, 2-(oxazolo-naphthoimidazo-pyridin-2-yl)phenol (IP1, 2a) and 2-(4-aminophenyl)benzothiazoles (p-APBT, 3a).Intramolecular H-bonds were realized by introducing a hydroxyl group at the ortho position of the phenyl substituent, affording 2b and 3b (Figure 4A as 0.076, 3b as 0.032; Figures 4B and S18) that were significantly smaller than that of their counterparts without H-bonds (2a as 0.18, 3a as 0.13; Figures 4B and S18).The changes in the x values suggest that RBFs with H-bond would bear a higher excited state rotational barrier, thus resulting in a lower x value.Taken together, our results provide compelling evidence that the installation of intramolecular H-bonds could serve as a general strategy to rationally control the viscosity sensitivity of RBFs based on multiple scaffolds.

Synergistic control of viscosity sensitivity via multiple mechanisms
In the past few years, two mechanisms had been reported to control the excited state rotational barrier and viscosity sensitivity of RBFs.Particularly for HBI derivatives, one mechanism was based on the control of π electron density of the I bond to modulate the rotational barrier.Because the mechanisms of I bond π electron density and intramolecular H-bonds are orthogonal, we next asked whether they could act synergistically to modulate the viscosity sensitivity of HBI-based RBFs.
We chose compound 1d as a starting point to extend π conjugation to the the diarylethene group with an indole, resulting in compounds 4a (Figure 5A, S9).Based on our previous work, the diarylethene group is named as electron density regulator (EDR), whose π electron density should be positively correlated with the activation energy (E a ) of rotation and control the viscosity sensitivity.Accordingly, the x value of 4a was determined to be 0.26 (Figure 5B, S19), slightly smaller than that of 1d.This observation is consistent to the mechanism that conjugated π orbitals resist rotation, thus the indole group would lead to a higher rotational barrier and a lower x value.As a support to this notion, theoretical calculation determined the rotational barrier (E a ) of the S 1 excited state as 0.51 eV (Figure 5C), about two-fold higher than the barrier (0.23 eV) of 1a (Figure 2E).When the indole group was replaced by heterocycles with decreasing π electron densities (compounds 4b to 4f in Figure 5A, Figure S10-S14), we observed increasing x values as represented by 0.33 for pyrrole and 0.35 for thiazole (Figure 5B, Figure S19).These results collectively indicate that installation of the intramolecular H-bond can be jointly used with other mechanisms to modulate the viscosity sensitivity of RBFs.Lastly, we explore the fluorescence response of these probes in different solvents with varying dielectric constant, and the results demonstrating that the all compounds including probe P1 and P2 exhibit none solvatic sensitivity and exhibit excellent anti-interference capability to polarity, H-bonding, and so on (Figure S26-S30).The synthetic methods of all compounds are available in supporting information (Schemes S1-S8, Figure S32-S61).

Detecting protein aggregation kinetics via RBFs with different viscosity sensitivity
Next, we examined how RBFs with intramolecular H-bonds could be applied to detect viscosity changes in biological events, using protein aggregation as a demonstration.[31][32] In this process, microenvironment of misfolding proteins evolves when protein aggregation proceeds. [33,34]n previous reports, we have demonstrated that microviscosity of misfolded proteins continuously increase from misfolded oligomers to insoluble aggregates. [6]Furthermore, we showed that the E a value is inversely relate to viscosity sensitivity (x), akin the higher (lower) E a value corresponds with smaller (greater) x value.At excited state, the molecular rotor tends to rotate away from the planar conformation (as the bright state) to the twisted structure (as the dark state), resulting in nonradiative decay.The rate of rotation, also noted as the rate of nonradiative decay, is dependent on the height of rotational barrier (E a ).Conditions with lower viscosities stall rotation more effectively for RBFs with higher E a (smaller x) than lower E a (greater x).As a result, RBFs with higher x values are better to detect small changes in viscosity than probes with smaller x values.Thus, we envisioned that the tuneable viscosity sensitivity of RBFs could be used to detect how micro viscosity of misfolding proteins changes during their aggregation.Because RBFs with excited H-bonds exhibit lower x values, they are expected to activate fluorescence in misfolded oligomers.Whereas RBFs without H-bonds would only turn-on fluorescence when micro viscosity reaches the level of insoluble aggregates.
To this end, we employed the previously reported AggTag method (Figure 6A), which has been used to visualize aggregated protein-of-interest (POI) both in vitro and in cellular.In this method, the POI is genetically fused to a selflabeling protein tag, in this case HaloTag. [35,36]HaloTag is a modified haloalkane dehalogenase designed to covalently react with chloroalkane as a reactive warhead.Covalent bond formation between the HaloTag and the chloroalkane linker is bioorthogonal and exhibit a rapid reaction rate as 10 7 M −1 ⋅s −1 under physiological conditions.Alternating the functional module has no effect on the reactivity between chloroalkane and HaloTag. [37]When AggTag probes are covalently conjugated to the POI-Halo fusion, fluorescence is only activated when POI misfolds or aggregates.To realize this assay, the reactive warhead of HaloTag was installed on probes 1d to generate AggTag probes P1 and P2 (Figure 6B S15-S16).The viscosity sensitivity of P1 and P2 was evaluated by their x values being 0.32 and 0.28, respectively (Figure S20).
In this work, we chose the destabilized mutant of superoxide dismutase 1 SOD1-A4V as the model system. [38]isfolding and aggregation of SOD1-A4V has been associated with Amyotrophic Lateral Sclerosis (ALS).The AggTag probe P1 and P2 were conjugated with purified SOD1-A4V-Halo protein and both exhibited low fluorescence in folded state (Figure 6C,D), indicating the low background fluorescence.We next examined fluorescence intensity of P1 and P2, which were conjugated with SOD1-A4V-Halo and then subjected to heat-induced misfolding and aggregation.It has been reported that when SOD1-A4V-Halo primarily forms insoluble aggregates when incubated at 59 • C. [39,40] Using this condition, we observed a fluorescence increase for about 5.6-fold for both P1 and P2 conjugated SOD1-A4V-Halo (Figure 6C,D).By contrast, SOD1-A4V-Halo incubated at 54.5 • C was shown to undergo a two-step aggregation process, including the initial formation of misfolded oligomers and then insoluble aggregates.Under this condition, P1 exhibited a more rapid fluorescence kinetics than P2, saturating at round 14 and 22 min, respectively (Figure 6E).Notably, P1 exhibited a more rapid kinetics than that of the turbidity assay (26 min).Because the signal of turbidity assay arises from insoluble aggregates, this result indicates that P1 could detect soluble misfolded oligomers.Whereas, the fluorescence kinetic of P2 was comparable to that of turbidity assay, suggesting that fluorescence of P2 was mostly activated by insoluble aggregates.In last, it is need to clarify that although the probes can distinguish the soluble and insoluble aggregates, the limitation is that these probes are unable to distinguish the viscosity change in the same secondary structure monomer and two water-soluble structures of the oligomer.

Visualize the multistep protein aggregation process in live cells
It is desirable to visualize the multistep aggregation process of proteins in live cells. [41][44][45][46] The excitation and emission spectra of P1 and P2 minimally overlapped, lending convenience to develop a two-colour imaging strategy (Figure 7A).Using this strategy, the red fluorescence of P1 could be used to monitor both the misfolded oligomers and insoluble aggregates.Whereas, the green fluorescence of P2 could only detect insoluble aggregates.Prior to the intracellular detection assay, we confirmed that probes P1 and P2 did not stain lysosomes by lysosomal colocalization experiments (Figure S31).
To test this notion, we used HEK293T cells to express the HaloTag-fused Huntingtin exon 1 protein (Htt) with varying expansion of polyglutamine (Htt-polyQ), whose aggregation is associate with Huntington's disease. [47,48]Htt-polyQ-Halo was expressed in the presence of P1 and P2 to enable covalent conjugation.Htt-polyQ with 25 glutamine repeats (Htt-Q25-Halo) would not form misfolding oligomers insoluble aggregates, as reflected by the dark fluorescence from both P1 and P2 (Figure 7B, top panel).When Htt-Q46-Halo was expressed in HEK293T cells, we only observed fluorescence signals from P1 but not P2, suggesting formation of misfolded oligomers (Figure 7B, middle panel).By contrast, the aggregation-prone Htt-Q110-Halo formed punctate structures that emitted fluorescence from both P1 and P2 channels (Figure 7B, bottom panel).In addition to Htt-polyQ, we also examined misfolding and aggregation of SOD1-A4V-Halo in cells that were treated with the proteasome inhibitor MG132 at 2.5 μM. [49,50]Without MG132, both P1 and P2 emitted minimal fluorescence (Figure 7C, top panel).After an 8-h incubation, we observed red fluorescence from P1 in both diffusive and small granular structures, while P2 remained dark (Figure 7C, middle panel).With an elongated incubation time (24 h), fluorescence of both P1 and P2 was activated in perinuclear granules (Figure 7C, bottom panel), which corresponds to aggresomes that contain insoluble protein aggregates.Collectively, these results demonstrated that RBFs with different viscosity sensitivities could visualize and distinguish the misfolded oligomers and insoluble aggregates in live cells.

CONCLUSION
In summary, we have demonstrated that the installation intramolecular H-bonds is a general strategy to control the viscosity sensitivity of RBFs.While the H-bonds could form at the ground state (S 0 ) of RBFs, we show that their strength would increase at the singlet excited state (S 1 ).Because the structure of these H-bonds would interrupt the rotation of the photo-excited fluorophore, RBFs with intramolecular Hbonds exhibit higher excited state rotational barrier between the planar fluorescent structure and the dark twisted con-formation.Due to the higher rotational barrier, RBFs with intramolecular H-bonds emit fluorescence at lower viscosities as reflected by their smaller x values than their counterparts without H-bonds.In this work, we apply this strategy to control the viscosity sensitivity of three classes of RBFs, resulting in x values ranging from 0.04 to 0.67.In addition to this method, we show that it could be jointly used with other chemical strategies to synergistically control the viscosity sensitivity of RBFs.As an application, the resulting RBFs based on the HBI scaffold allow us to conduct a twocolour imaging experiment to visualize the multistep protein aggregation process both in vitro and in live cells.In summary, our work adds a new chemical mechanism to rationally control the viscosity sensitivity RBFs.The ability to design fluorophores with desired excited state rotational barriers provides opportunities for the community to potentiate novel applications in biological and material sciences.

F I G U R E 1
Rational regulation of the excited state rotational barrier of rotor-based fluorophores (RBFs) to control their viscosity sensitivity.(A) The correlation between height of rotational barriers (E a ) and viscosity sensitivity (x values) of RBFs; (B) modulation the excited state rotational barriers of viscosity sensitivity of HBI-based fluorophores via intramolecular hydrogen bond.

F I G U R E 2
Installing intramolecular H-bonds to modulates potential energy surface (PES) of HBI derivatives at the excited states.(A) Structure of HBI and its derivatives, 1a to 1d; (B) the correlation between viscosity sensitivity (x value) and rotational barrier (E a ) of 1a to 1d; (C) comparison of viscosity sensitivity of 1a and 1d using the slope of linear fitting plot of logarithmic intensity as a function of logarithmic intensity; (D and E) excited state PES scan of 1a (D) and 1d (E).

F I G U R E 3
Computational analysis of the H-bond energy and excited state intramolecular proton transfer (ESIPT) process.(A) Reduced density gradient (RDG) method was used to qualitatively analyse the existence of H-bonds, as indicated by the blue colour of the possible H-bond forming region; (B) H-bond energies of 1b, 1c, and 1d in ground state (S 0 ) and excited state (S 1 ); (C) Gibbs free energy (ΔG) calculation of 1d suggests ESIPT process is thermodynamically unfavourable.

F I G U R E 6
Detecting protein aggregation using the AggTag method.(A) HBI activates fluorescence in response to protein aggregation; (B) structure of the AggTag probes P1 and P2; (C and D) aggregation-induced fluorescence activation of P1 and P2 conjugated with purified SOD1(A4V)-Halo protein; (E) P1 and P2 fluorescence activation exhibits more rapid kinetics than turbidity kinetics during heat-induced aggregation.

F
I G U R E 7 P1 and P2 differentiate misfolded oligomers and insoluble aggregates in live cells.(A) The AggTag method enables detection of the multistep protein aggregation process in live cells.(B) Images of Htt-Q25-Halo (top panel), Htt-Q46-Halo (middle panel), and Htt-Q110-Halo (bottom panel) using P1 and P2.Htt-Q46-Halo forms diffusive misfolded oligomers that were detected only by P1 but not P2.Htt-Q110-Halo forms perinuclear granular aggresomes that were detected by both P1 and P2.(C) Images of SOD1(A4V)-Halo under the stress of MG132 (2 μM) using dual probe P1 (0.5 μM) and P2 (0.5 μM).Cells were treated for 8 or 24 h prior to fluorescence confocal microscopy.Dark background of P1 and P2 was observed in the absence of MG132 (top panel).At 8 h, P1 exhibited fluorescence in both diffusive and small granular misfolded oligomers while P2 remained dark (middle panel).At 24 h, P2 exhibited fluorescence only in perinuclear granular aggresomes and P1 remained to be fluorescent in both diffusive and small granular misfolded oligomers (bottom panel).Blue: Hochster 33342.Red: P1.Green: P2.Scale bar: 5 μm.