A fluorophore's electron-deficiency does matter in designing high-performance near-infrared fluorescent probes

The applications of most fluorescent probes available for Glutathione S-Transferases (GSTs), including NI3 which we developed recently based on 1,8-naphthalimide (NI), are limited by their short emission wavelengths due to insufficient penetration. To realize imaging at a deeper depth, near-infrared (NIR) fluorescent probes are required. Here we report for the first time the designing of NIR fluorescent probes for GSTs by employing the NIR fluorophore HCy which possesses a higher brightness, hydrophilicity and electron-deficiency relative to NI. Intriguingly, with the same receptor unit, the HCy-based probe is always more reactive towards glutathione than the NI-based one, regardless of the specific chemical structure of the receptor unit. This was proved to result from the higher electron-deficiency of HCy instead of its higher hydrophilicity based on a comprehensive analysis. Further, with caging of the autofluorescence being crucial and more difficult to achieve via photoinduced electron transfer (PET) for a NIR probe, the quenching mechanism of HCy-based probes was proved to be PET for the first time with femtosecond transient absorption and theoretical calculations. Thus, HCy2 and HCy9, which employ receptor units less reactive than the one adopted in NI3, turned out to be the most appropriate NIR probes with high-sensitivity and little nonenzymatic background noise. They were then successfully applied to detecting GST in cells, tissues and tumor xenografts in vivo. Additionally, unlike HCy2 with a broad isoenzyme selectivity, HCy9 is specific for GSTA1-1, which is attributed to its lower reactivity and the higher effectiveness of GSTA1-1 in stabilizing the active intermediate via H-bonds based on docking simulations.


Fluorescence imaging in Cells and Tissues
Fluorescence microscopic images in cells and tissue sections were captured by a FV1000 confocal laser scanning microscope (Olympus, Japan). Direct imaging was implemented without washing. For imaging related to EA and NEM, operation conditions were as follows: (1) To eliminate any interferences induced by long-time incubation in HEPES buffer, HepG2 cells were pretreated with HEPES buffer (20 mM, 5% glucose, pH 7.4) for 40 min and then incubated with 10 μM HCy-based probes in HEPES buffer for 30 min before subjected to imaging; (2) HepG2 cells were pretreated with 100 μM EA in HEPES buffer for 30 min and then incubated with 10 μM HCy-based probes in HEPES buffer in the presence of EA for another 30 min before subjected to imaging; (3) HepG2 cells were pretreated with 50 μM NEM in HEPES buffer for 40 min and then incubated with 10 μM HCy-based probes in HEPES buffer in the absence of NEM for another 30 min before subjected to imaging. For comparison of fluorescence intensities of various cell lines incubated with HCy2 or HCy9, the incubation time was 30 min or 1.5 h for HCy2 or HCy9, respectively. For fluorescence imaging in tissues, mice liver or lung tissue sections were incubated with 20 μM HCy2 in HEPES buffer (20 mM, 0.5% DMSO, 5% glucose, pH 7.4) for 1 h before imaging.

Establishment of tumor model in mice
All animal experiment procedures were performed according to the protocols approved by the Animal Ethics and Use Committee of Dalian Medical University. Female BALB/c-nu mice with 5 weeks were obtained from the animal center of Dalian Medical University and were housed under standard conditions with 12 h light/dark cycle. The tumors were grafted by subcutaneous injection of 5×10 6 HepG2 cells in 200 μL of PBS-matrigel (1:1) mixture into the front flanks of each mouse. Subsequently, when tumor size reached about 300 mm 3 , in vivo fluorescence imaging was carried out.
Femetosecond TA spectroscopy TA experiment was conducted with a home-made femtosecond pump-probe setup. 2 Briefly, laser pulses (800 nm, 50 fs pulse length, 1 kHz repetition rate) were generated by a Ti: sapphire femtosecond laser source (Spitfire, Spectra-Physics). An optical parametric amplifier was used to change the laser wavelength. The white light-continuum generation from a thin CaF 2 plate was used for the probe. By placing a Berek compensator in the pump beam, the mutual polarization between pump and probe beams was set to the magic angle (54.7°). The wavelength and the power of the pump beam were 630 nm and 100 μW, respectively, and the spot size was 0.3 mm in diameter. A solution of ca. 1 mM HCy9 in DMSO (Spectrum Pure, >99.9% GC) was subjected to the measurement. The kinetic of the repeated scans remained the same, indicating no sign of degradation.
Calculations of ω, spin density distribution, P + k and ω k The global elctrophilicity ω values were obtained based on the calculations of vertical ionization potential (I) and vertical electron affinity (A) with ORCA 4.0.1 program at the PWPB95/def2-QZVPP level 3 according to the formulas for the valance state parabola model in the literature 4 , utilizing RIJCOSX acceleration with auxiliary basis sets def2/J 5 and def2-QZVPP/C 6 and involving the atompairwise dispersion correction with the Becke-Johnson damping scheme (D3BJ) 7 . The integral grid was set to 4, and the convergence limit was set as 'tightSCF'. In light of the analysis in the literature 4 and the probes' applications in biological systems, all these calculations were implemented in gas phase. The global elctrophilicity ω in the valence state parabola model was calculated according to the Equations (2) - (4) 4 , in which μ means the electronic chemical potential and η represents the chemical hardness: (2) (3) (4) The calculation method for P + k . Specifically, geometry optimization and vibrational frequency analysis were carried out at the B3LYP/6-31G(d) level 8 of DFT theory. The P + k values were obtained based on natural population analysis (NPA) charge by singlepoint energy calculations over the optimized neutral geometries (since the positive charge in π-conjugated organic dye like cyanine or HCy always brings about problems in accuracy, 9 and considering the indolium moiety bearing the positive charge not being a reactive site, a simplified model for HCy-based probes with the deletion of methyl on the nitrogen of indolium ring was used) using the unrestricted UB3LYP formalism for the radical anion of compounds at the same level. And the results of these calculations were analysed by Multiwfn software 10 to give spin density distributions. Above calculations were conducted with Gaussian 09 suite of programs 11 . The ω k values were calculated according to the Equations (5).

Docking Simulations
Homology modeling of GSTM1-1 protein structure. The starting structure was homologously modeled on the basis of single template using Modeller 9.20 software. The template (PDB code: 6GSV) was obtained from the RCSB PDB. The alignment result was then used as the input for Modeller, and a total of ten models were generated and evaluated by the default DOPE potential in Modeller. The best GSTM1-1 model was chosen to be applied for docking simulations. The crystallographic structures of GSTA1-1 (PDB code: 1K3L) and GSTP1-1 (PDB code: 18GS) were retrieved from the RCSB PDB. Afer adding the explicit hydrogen atoms, deleting all water molecules and removing the ligands, we processed the protein structure using AutoDock Tools with Grid box size of 40/54/45 around center 80. 5/26/18.5, 45/42/45 around center 20.5/17.5/31 and 54/45/50 around center 8/8/26 for GSTA1-1, GSTM1-1 and GSTP1-1, respectively. Geometrically optimized HCy9-GS and HCy2-GS σ complexes were prepared for docking by using AutoDock Tools to assign AD4 atom types, calculate Gasteiger charges and set all rotatable bonds as active torsions except the one in rigid fluorophore moiety. The complexes were docked into the protein using AutoDock 4. The exhaustiveness parameter was set as default. The maximum number of evals was set as long (25,000,000).

TD-DFT calculations
The calculation method was same with the one in the case of NI-based probes. 2 Briefly, in order to mimic the real process of photoexcitation in femetosecond TA spectroscopy, the solvent effect in DMSO solution was employed in the SCRF calculations by using the SMD solvation model 12 . Geometrical optimizations of HCy9, HCy and HCy10 in the ground state and vibrational frequency analysis were carried out at the B3LYP/6-31G(d) level 8 of DFT method. The electron transitions of HCy9, HCy and HCy10 were studied at the B3LYP/aug-cc-pVDZ level [13][14][15] based on TD-DFT method conducted with Gaussian 16 suite of programs 16 .

Deduction I
Since light intensity is essentially a kind of energy flux density (as an approximation, herein emission peak broadening is disregarded), ∫ Any way, Similarly, for a nonenzymatic reaction between GSH and a substrate (i.e. probe): : fluorescence intensity of enzymatic reaction mixture : the overall fluorescence collection efficiency of the apparatus : Planck constant : frequency of the emitted light : frequency of the excitation light : energy of per photon : the number of fluorescent photons through a unit area emitted by the dye per unit time : fluorescence quantum yield : the number of photons absorbed by the dye per unit time and per unit area : the initial intensity of the incident light, namely the initial intensity of the excitation light : the remaining intensity of the light which has underwent absorption by the dye molecules : absorbance : molar extinction coefficient : optical path length : the molar concentration of the luminescent substance (the dye molecules) : the brightness of the dye : the initial reaction rate : reaction time : catalytic constant : Michaelis constant : the initial molar concentration of the enzyme : the initial molar concentration of the substrate (herein the probe) : apparent second-order reaction rate constant : the initial molar concentration of GSH : fluorescence intensity of nonenzymatic reaction mixture Notes: [a] In the derivation of " ", the definition of absorbance " " is used.
[b] and are the equivalent infinitesimal when , so . And the working concentration of the probe is generally several to tens of μM, which makes rather small (close to zero), and hence meet the requirement for aforementioned approximation.
[c] In the derivation of " ", the Lambert-Beer law " " is used.
[d] In the derivation of "∫ ", the hypothesis that the initial rate remains unchanged in a short period since the initiation of reactions and Michaelis-Menten equation are applied. For the former, no matter in the study of kinetic measurements or in the incubation of cells with probes, administration in a relatively short time assures this deduction reasonable.
[e] The fluorescence intensity is proportional to signals arising from both photoluminescence mechanism (PM) and recognition or reaction mechanism (RM). For the former mechanism, a specific dye with high molar extinction coefficient and/or high fluorescence quantum yield, namely high brightness, is favorable to probe design as the fluorophore. For the latter mechanism, a probe with large catalytic constant and small Michaelis constant is beneficial. In practice, is more important than . And certainly, increasing the intensity of excitation light , improving the overall fluorescence collection efficiency of the apparatus or prolonging the reaction or incubation time can also afford a brighter fluorescence signal, and this is the very reason why one must keep these parameters consistent strictly in a set of experiments addressing the comparison of the fluorescence intensities from different samples.
[f] One of the main conclusions in this work is that although herein the property of a probe can be artificially divided into two parts, namely the PM and RM, the probe should be treated as an entirety. In other words, the RM is related not only to the receptor unit but also to the fluorophore, and so is the PM.

Deduction II
If stands for the portion of a certain probe dissolved in the water, and stands for the portion dissolved in the octanol, then and . Thus, we get , namely Then, what if the fluorophore's hydrophilicity affects remarkably the encounter between probe and GSH molecules and thus the apparent nonenzymatic reaction rate constant k nonc ? Since GSH is a hydrophilic and polar substance, the effective initial concentration of the probe that can be "felt" readily by GSH molecules ( ) should multiply the apparent one ( ) by a factor that is proportional to , and herein for convenience we can just directly take . Hence, the nonenzymatic reaction rate and , in which is the true nonenzymatic reaction rate constant that is determined only by the nature, namely the chemical structure, of the receptor unit of one certain probe and is the apparent one that is measured by the kinetic study, and is the initial concentration of glutathione. So, . It should be noted that, this deduction is based on the hypothesis that it is the hydrophilicity instead of the electron-deficiency of the fluorophore that bring about the differences in reactivity between corresponding HCy-based and NI-based probes. Therefore, if this hypothesis were true, when regarding different receptor units, the amplifications of induced by the replacement of NI with the more hydrophilic fluorophore HCy should be positively correlated with the amplifications of . In other words, R 1 = k nonc (HCy-)/k nonc (NI-) should be positively correlated with R 3 = x w (HCy-)/x w (NI-) = (1 + P (NI-))/(1 + P (HCy-)), which is not the actual situation as shown in relevant tables. Hence, it is not the fluorophore's hydrophilicity that causes the increased reactivities of HCy-based probes.

Deduction III
As an approximation, suppose regardless of the specific structure of the receptor unit, the increment of the local electrophilicity ω k (i.e. Δω k ,) induced by the replacement of NI with the more electron-deficient fluorophore HCy is a constant, and this assumption is demonstrated to be reasonable as follows: If Δω k is a constant, when increasing the electriphilicity of the receptor unit, namely enlarging ω k , R 4 should decrese monotonically, which is none other than the actual case reflected by the trend in Table 3 in the main text. Then, without loss of generality, we assume the variation of k nonc induced by the replacement of either a fluorophore or a receptor unit can be described by formulas of the same pattern. For example, one may choose the formula pattern describing HCy-based probes (refer to Figure S9), where a and b are constants, and a > 0.
Hence, when increasing the electriphilicity of the receptor unit, namely enlarging ω k , R 1 should decrese monotonically, which is exactly consistent with the actual case reflected by the trend in Table 1 in the main text. Alternatively, one can also choose the formula pattern describing NI-based probes, 2 where α and β are positive constants.
And the same conclusion can be drawn. Taken together, it's rational to infer that the larger reactivities of HCy-based probes relative to NI-based ones are probably caused by the larger electron-deficiency of HCy. Figure S1. Chemical structures of naphthalimide (NI)-based probes NI1-10. Figure S2. Spin density distributions of the anion radicals of probe models intimately related to this work. Respective positive P+ k value (amplified by a factor of 1,000) of the carbon atoms in the nitrobenzene ring is marked, with positive and negative spin density colored by green and purple, respectively. Negative spin density or P + k values are herein regarded as meaningless. 18,19 Isodensity value = 0.002. Figure S3. Respective time course of fluorescence intensity of probes HCy1−10 (20 μM) in HEPES buffer (50 mM, 0.5% DMSO, pH 7.4) with or without GSTs (12.5 μg/mL) from equine liver in the presence of GSH (10 mM). λ ex/em = 650/700 nm. GSTs were premixed with GSH for 10 min prior to the addition of probes.

Supplementary Figures, Schemes and Tables
Scheme S1. GST detection mechanism of the HCy-based probes. Figure S4. UPLC-MS analysis of the GST-catalyzed reaction between HCy2 and GSH. (a) UPLC traces of (1) HCy2, (2) reaction mixture at 2 min after addition of GSTs (6.875 μg/mL) from equine liver, (3,4) reaction mixture at 10 h after addition of GSTs (12.5 μg/mL) from equine liver and (5) Figure S8. Spin density distributions of the anion radicals of (a) DNs-AcRh, (b) DNs-Coum and (c) DNs-CV appeared in the literature 1 . Respective positive P + k value (amplified by a factor of 1,000) of the carbon atoms in the nitrobenzene ring is marked, with positive and negative spin density colored by green and purple, respectively. Negative spin density or P+ k values are herein regarded as meaningless. 18,19 Isodensity value = 0.002.                  Table S1. Electron affinity A, Electronic chemical potential μ, chemical hardness η, global electrophilicity ω, the electrophilic Parr function P + k of the α-carbon and the local electrophilicity ω k of the α-carbon for probes [a] A (NI-) [b] A (HCy-) μ (HCy-) η (HCy-) ω (HCy-) [a] The list was sorted by ω k from the lowest to the highest; apart from P + k , the unit of all the other parameters is eV.
[b] R 3 = x w (HCy-)/x w (NI-) = (1 + P (NI-))/(1 + P (HCy-)), and refer to Deduction II in this Supporting Information for more about the parameter x w . The list was sorted by the values of R 2 or R 3 from the highest to the lowest.  [a] The unit for k nonc is s -1 . These data were drawn from that literature 1 . [a] N.D. = Not determined, which means the data are unavailable due to too large nonenzymatic background noise and/or too high reaction rate hindering the data acquisition ahead of the reaction completeness. UD = undetectable, which means the data are unavailable due to the low sensitivity of corresponding probes to this isoenzyme.   [a] N.D. = Not determined, which means the data are unavailable due to too large nonenzymatic background noise and/or too high reaction rate hindering the data acquisition ahead of the reaction completeness. UD = undetectable, which means the data are unavailable due to the low sensitivity of corresponding probes to this isoenzyme.

HCy2
0.392 ± 0.041 2.0 ± 0.6 0.199 ± 0.045 [a] N.D. = Not determined, which means the data are unavailable due to too large nonenzymatic background noise and/or too high reaction rate hindering the data acquisition ahead of the reaction completeness. UD = undetectable, which means the data are unavailable due to the low sensitivity of corresponding probes to this isoenzyme.  7.79449700 -3.48094500 -0.16435900 C 5.39364600 -0.88332600 -0.02746000 H 4.19275700 -5.02799600 -0.01829400 H 6.19577600 -6.49582900 -0.12414300 H 8.46800000 -5.