Improving the Brightness of Pyronin Fluorophore Systems through Quantum-Mechanical Predictions

Abstract

The pyronin class of fluorophores serves a critical role in numerous imaging applications, particularly involving preferential staining of RNA through base pair intercalation. Despite this important role in molecular staining applications, the same set of century-old pyronins (i.e. pyronin Y (PY) and pyronin B (PB)), which possess relatively low fluorophore brightness, are still predominantly being used due to the lack of methodology for generating enhanced variants. Here, we use TD-DFT calculations of interconversion energies between structures on the S₁ surface as a preliminary means to evaluate fluorophore brightness for a proposed set of pyronins containing variable substitution patterns at the 2, 3, 6, and 7 positions. Using a nucleophilic aromatic substitution/hydride addition approach, we synthesized the same set of pyronins, and demonstrate that quantum-mechanical computations are useful for predicting fluorophore performance. We produced the brightest series of pyronin fluorophores described to date, which possess considerable gains over PY and PB.

Xanthene-based structures with 3,6-diamino substitution, more commonly referred to as pyronins (Figure 1),¹ have origins in the late 19^th century^{2, 3} and became popular as dyes originally in textiles then later in the scientific community for imaging applications.^{4, 5} Pyronins, like structurally similar rhodamines, typically exhibit fluorescent properties and are commonly employed in microscopy,^{6, 7} cell cycle studies,⁸ electrophoresis,^{9, 10} flow cytometry,^{11, 12} paraffin section analysis,^{13, 14} mitochondrial studies,^15–17 detecting protein-protein interactions,¹⁸ chemo/ion sensors,^{16, 19} among other uses.^20–24 Most applications utilize commercially available pyronin Y (PY) or pyronin B (PB) and to a lesser extent acridine red (AR) which preferentially intercalate RNA allowing for imaging/detection at an excitation wavelength (λ_ex) of ~530 nm (AR)/~550 nm (PY/PB) and an emission wavelength (λ_em) of ~560 nm (AR)/~570 nm (PY/PB), however, numerous synthetically tailored pyronins have been generated for specialized uses.^{15, 18, 21} While PB, AR, and particularly PY, remain popular dyes, the quantum yields (Φ_f) for these fluorophores in aqueous solvents are relatively low in comparison to other xanthene-based fluorophores such as rhodamines and fluoresceins limiting the sensitivity of these fluorophore systems. As a result, pyronins are often applied in high concentrations (~100 μM) to stain cells which can lead to cytotoxicity.

Figure 1.

Structures and quantum yields¹ of alkyl substituted pyronins.

In general, fluorophores containing arylamino groups, like that of PB, PY, and AR, are prone to undergo twisted intramolecular charge transfer (TICT) mechanisms in the excited singlet state which can reduce the quantum efficiency of the system (Figure 2).²⁵ In this process, alkyl substituted arylamino groups rotate out of conjugation with the π-system resulting in an electron transfer that produces a diradical form of the molecule. Recently, efforts have been made to mitigate TICT mechanisms as a means to produce fluorophores that possess particularly high quantum yields and molar absorptivities.²⁶ This concept has been applied by Lavis et al,^{27, 28} in which N,N-dialkylamino groups were replaced with azetidinyl substitution to dramatically increase the quantum efficiency and photon yield for a range of classical fluorophores; N-arylazetidine systems possess larger energetic barriers for electron transfer processes needed to form the TICT state in comparison to N,N-aryldialkylamino systems.²⁸

Figure 2.

Jabłoński diagram illustrating an excited singlet state twisted intramolecular charge transfer (TICT).

We became interested in optimizing the photophysical properties of pyronins, to generate more photogenic variants that could be used in considerably lower cellular staining concentrations and/or could provide higher resolution and more information rich data. Our envisioned approach, was the replacement of traditional N,N-substituted amino groups at the 3 and 6 position with groups that produce relatively large energetic barriers for TICT processes in the excited state, in order to maximize molar absorptivities and quantum yield efficiencies. We were particularly interested in the replacement of dimethylamino (PY)/diethylamino (PB) groups with an azetidinyl group, due to gains observed in structurally similar rhodamine-based systems,²⁸ as well as monoalkylamino substitution patterns due to the >2-fold increase in quantum yield between PY/PB and AR.¹ Additionally, we were interested in introducing 2,7-difluoro-substitution into the pyronin ring structure, for some variants, to attenuate hydrophobicity and investigate the potential of NH-F intramolecular hydrogen bonding as a means deter TICT processes; analogous fluorination substitution has been successfully incorporated into fluorescein (i.e. Oregon Green fluorophore)²⁹ and rhodamine derivatives³⁰ to improve photophysical, chemical, and biological properties while preserving fluorophore brightness. Furthermore, we aimed to retain the relative polarity properties of PY, PB, and AR, as these pyronins possess generally good aqueous solubility and cellular permeability which is critical for effective cellular staining. To aid our design, we performed Log P and quantum-mechanical time-dependent density functional theory (TD-DFT) calculations as well as calculated vibronically-resolved fluorescence spectra to evaluate the lipophilicity and propensity to undergo excited singlet state TICT for a range of 3,6-diamino substituted pyronins.

We envisioned computationally evaluating a set of pyronins (1-13, Figure 3) containing various amino-substitution patterns at the 3,6-positions, in addition to the incorporation of hydrogen (1-5) or fluorine (6-12) substituents at the 2,7-positions; PY, PB, and AR were used for comparison. The proposed series contains azetidine-substituted pyronins 1 and 8, monoalkylamino members variants of AR (2-5), and 2,7-difluorinated analogs including fluorinated versions of PY (7), PB (8), and AR (9) along with 10–13 which are capable of forming H-F intramolecular hydrogen bonds. To evaluate the hydrophobicity, calculated Log P (cLog P) values were generated³¹ for 1-13, PY, PB, and AR (Table 1). To ensure retention of the relative lipophilicity of PY, PB, and AR a cutoff of <1 cLog P was employed for 1-13 removing butylamino substituted pyronins 5 and 13 from further analysis.

Figure 3.

Structures of pyronins 1-13 and PY, PB, and AR with calculated Log P (cLog P) values³¹ (in blue) with H-F hydrogen bonding illustrated for 9-13 (red dashes).

Table 1.

Energy differences and interconversion barriers (kcal/mol) between S₁-min and TICT states of pyronins 1-13, with PY, PB, and AR from TD-DFT calculations. Dihedral angles (DA) between the aromatic plane and alkylamino groups are given in degrees (°) for TICT structures in the first excited state.

	ΔG	ΔG‡	DA		ΔG	ΔG‡	DA
PY	−3.2	5.1	89.6	6	−8.2	1.0	90.4
PB	−4.2	3.1	88.5	7	−6.6	1.4	89.6
AR	+0.1	8.8	0.2	8	+0.3	7.1	89.7
1	+2.4	9.1	89.9	9	+2.8	8.5	1.4
2	−0.0	7.2	0.1	10	+3.0	7.7	2.3
3	−0.1	8.4	0.1	11	+3.0	8.5	7.1
4	−0.2	7.4	2.2	12	+3.3	7.3	2.3

For the remaining pyronins, geometry optimizations were performed with the B3LYP-D3BJ DFT^32–35 functional and def2-TZVP³⁶ basis sets while solvation effects in water were described with CPCM^37–39 using ORCA 5.0.2 package.^40–42 The locally-excited structures, labelled as S₁-min, as well as the TICT structures were optimized with Tamm-Dancoff approximation⁴³ TD-DFT on the first excited state surface. The S₁-min structures retain the structural symmetry of the S₀ lowest-energy structures, in other words, the side groups have significant overlap in both states. Vibrational frequency analyses were performed for all structures to determine their local minima or saddle point characters. Gibbs free energy differences, ΔG, between S₁-min and TICT structures as well as interconversion barriers, ΔG^‡, were determined (Table 1). A positive ΔG value shows the S₁-min structure is lower in energy than the TICT structure. By contrast, a negative ΔG value shows the TICT structure is lower in energy. We emphasize that the ΔG^‡ barriers arise from rotation of the alkylamino side chains (Table 1). Transition states between the S₁-min and TICT structures were obtained with climbing-image nudged-elastic band, CI-NEB, calculations on the first excited-state surface.

On the first excited state surface, the arylamino alkyl groups are fully rotated between the xanthene core for 2-4 and 9-12 (monoalkylamino species), restoring near-planarity (Table 1 and Figure S38). By contrast, for the dialkylamino species, the TICT structures represent partial (approximately 90°) rotation of the arylamino alkyl groups (Table 1). S₁-min and TICT structures of the monoalkylamino species (AR, 2, 3 and 4) are isoenergetic. The rigid azetidinyl group (1) favors S₁-min over the TICT structure by 2.4 kcal/mol, deterring S₁-min/TICT interconversion. Likewise, 2,7-difluoro groups (9, 10, 11 and 12) favor S₁-min geometries by 2.8–3.3 kcal/mol. Interestingly, when the azetidinyl and 2,7-difluoro strategies are combined (8), the TICT structure is isoenergetic with S₁-min. ΔG for 8 is only +0.3 kcal/mol, compared with +2.4 kcal/mol for 1.

Comparing PY and PB respectively to 6 and 7, shows impact of 2,7-difluoro substitutions for dialkylamino species. For 6 and 7, TICT structures are lower in energy than S₁-min geometries by 6.6–8.2 kcal/mol. Without fluorine substitution, the S₁-min structures of PY and PB are stabilized (−3.2 to −4.2 kcal/mol, Table 1), although TICT structures remain lower in energy. Overall, Table 1 shows that the S₁-min → TICT interconversion free energies are increased by 2,7-difluoro substitution in monoalkylamino species (AR, 2, 3 and 4 versus 9, 10, 11 and 12). The opposite is true for the dialkylamino species (1, 6 and 7 versus 8, PY and PB). Thus, fluorination is only useful for deterring S₁-min → TICT interconversion in the monoalkylamino species. Based on ΔG values in Table 1, contributions of TICT geometries to the population on the first-excited state surface follow the trend: 6 and 7>> PY and PB >> AR, 2-4 and 8 > 1 and 9–12. Thus, TICT geometries are least accessible for 1 and 9-12.

Comparison of the vertical excitation energies of 10 and 7 were carried out (see Figure S38 for a visualization of this comparison). These are used as examples as they belong to groups on opposite extremes of the data in Table 1. For 10, the first absorption, S₀ → S₁, is a π-π* excitation dominated by the configuration of HOMO/LUMO transition, with both orbitals centered at the xanthene core. This is the case for the S₀ minimum structure as well as a structure with a twisted side-group. For 7, rotation of a diethylamino side group by about 90° leads to a twisted structure similar to the TICT geometry on the first excited state surface. Crucially, the HOMO is now centered at the amino group although the LUMO remains centered at the xanthene core. Natural transition orbitals are provided in Figure S38. Thus, the HOMO-LUMO transition in this twisted structure of 7 leads to a dark state (Figure 4). The low oscillator strength for the TICT-like geometry of 7 reveals that it can only relax via non-radiative decay. This is also the case for 6 and is fully consistent with earlier reports on rhodamine dyes by Rega et al.⁴⁴ who reported that structural deformation of the xanthene and carboxyphenyl groups can switch the energy orderings of two excited states with dark and bright characters. We conclude by emphasizing that Table 1 indicates that non-radiative decay will be highest for PY, PB, 6 and 7 for which the TICT structures are lower in energy on the first excited state surface. Conversely, we expect improved quantum yields for AR, 1-4, and 8-12.

Figure 4.

Optimized ground state geometry of 7 with a twisted diethylamino side group. Frontier orbitals for this structure are also shown.

Encouraged by our computational results, we pursued the synthesis of 1-4 and 6-12 through a two-staged synthetic route (Figure 5), initiating from hexafluorobenzophenone 14 or difluoroxanthone 15. The first stage utilizes a nucleophilic aromatic substitution (S_NAr) approach to substitute fluorine atoms, situated para to the carbonyl group of 14⁴⁵ or 15,⁴⁶ with amino groups; additional hydroxyl substitution of 14 at an ortho substituted fluorine provides the xanthone ring structure via intramolecular cyclization. The second stage uses hydride addition to the carbonyl, which subsequently undergoes elimination to provide the pyronin product.

Figure 5.

S_NAr/hydride addition approach to generating pyronins.

First stage reaction conditions for the substitution of 14 gave generally high reaction yields for the two-step sequence involving the treatment with amine followed by heating in hydroxide to provide the xanthone (Table 2). While most reaction yields for this conversion were >87%, the conversion of 14 to 19 underwent substantial N-demethylation during treatment with hydroxide. As a consequence of this low-yielding step, we decided to remove pyronin 9 from our synthetic library. Amino substitution conditions were applied to xanthone 15, with heating, which provided 3,6-diamine substituted products 23-26 in good to excellent yields (Table 3).

Table 2.

Synthesis of xanthones 16-22 using S_NAr conditions.

Entry	HNR2a	Yield (%)d	Product
1	HNMe2b	94	16
2	HNEt2	87	17
3	azetidinec	84	18
4	H2NMe	<5e	19
5	H2NEt	92	20
6	H2NPr	91	21
7	H2NiPr	94	22

^aAll amines were used neat and were reacted at 26 °C with the exception of HNMe₂ and azetidine.

^bDimethylamine was generated via decomposition of DMF in KOH at 60 °C.

^cAzetidine was reacted as a solution in acetonitrile at 60 °C.

^dIsolated yield.

^eProduct of the second step decomposes in presence of hot KOH.

Table 3.

Synthesis of xanthones 23-26 using S_NAr conditions.

Entry	HNR2a	Yield (%)c	Product
1	azetidineb	76	23
2	H2NEt	97	24
3	H2NPr	94	25
4	H2NiPr	89	26

^aAll amines were used as aqueous mixtures at 200 °C in a sealed-vessel with the exception of azetidine.

^bAzetidine was reacted in a sealed-vessel as a solution in acetonitrile at 150 °C.

^cIsolated yield.

Second stage reaction conditions, involved the reduction of xanthones 16-18 and 20-26 using borane-dimethylsulfide complex (BH₃-DMS). Similarly to previously reported reductions of this type,³² two equivalents of hydride are added to the 9-position to provide the over-reduced xanthene product, which slowly decomposes upon standing, in the presence of atmospheric oxygen, to the pyronin product. This oxidation process is expediated in the presence of 2,3-dichloro-5,6-dicyano-1,4-benzoquinone (DDQ), to provide pyronin product as a trifluoroacetic acid salt, upon acidic workup. Table 4 summarizes the yield for this two-step reduction/oxidation process, which produces pyronin products 1-4, 6-8, and 10-12 in good to excellent yields.

Table 4.

Synthesis of pyronins 1-4, 6-8, 10-12 and comparison of photophysical properties^a with PY, PB, and AR.

	Yield (%)b	λmax (nm)	εmax (cm−1M−1)	λem (nm)	Φf	εmax X Φf
PY	-	545	42,600	563	0.18	7,700
PB	-	551	65,600	568	0.20	13,100
AR	-	519	44,900	535	0.62	26,000
1	82 (23)	545	70,600	563	0.72	50,800
2	78 (24)	519	66,000	537	0.81	53,500
3	84 (25)	522	66,800	539	0.82	54,800
4	86 (26)	522	43,500	539	0.87	37,800
6	90 (16)	554	34,400	573	0.01	3,400
7	91 (17)	561	61,400	578	0.02	12,200
8	86 (18)	549	71,700	567	0.63	45,200
10	75 (20)	516	57,800	534	0.76	43,900
11	77 (21)	519	56,200	536	0.93	52,300
12	75 (22)	520	49,200	536	0.75	36,900

^aPhotophysical properties were measured in pH 7.0 PBS buffer.

^bIsolated yield from reactant (in parenthesis).

The photophysical properties of pyronins 1-4, 6-8, and 10-12 were acquired and compared to PY, PB, and AR as summarized in Table 4 (see Figure S1 for experimental spectra). The maximum absorbance values (λ_max) for dialkylamino-substituted pyronins 1, and 6-8 are analogous to PY and PB and structurally similar rhodamines (i.e. rhodamine B) and can be excited with lasers in the 550 nm range.

Vibronically-resolved absorption and fluorescence spectra of 1-4, 8, 10-12 and AR were computed with the adiabatic Hessian (AH) approach, with the ESD module of ORCA version 5.0.2. Details are provided in the Supporting Information, SI.⁴⁷.^48–51 To test dependence of our results on the DFT functional, we used ωB97X-D3,^{52, 53} PBE-D3BJ^{33, 34, 54} and BLYP-D3BJ.^{33–35, 55} We also considered dependence of the calculated spectra on the percentage of Hartree-Fock (HF) exchange in B3LYP, specifically we tested 0.05 (B3LYP-5%), 0.10 (B3LYP-10%) or 0.15 (B3LYP-15%). Vibronic spectra are compared to experimental data and vertical excitations obtained with the STEOM-DLPNO-CCSD approach in the SI.^{48, 56} We did not compute the vibronic spectra for PY, PB, 6 and 7. This is because our analysis that TICT structures are lower in energy than S₁-min geometries on the first excited state surface (Table 1). Additionally, the nature of the S₀ → S₁ transition is significantly different for TICT structures of 6 and 7 (Figure 4). Treatment of these systems will likely require multiconfigurational treatment of the excited state wavefunctions as well as methodologies to account for the structural relaxations on the excited state surface, far beyond the scope of the current work.

Comparing absorption band maxima, λ_max, obtained from TD-DFT spectra to experimental data (Table 4) shows that B3LYP-D3BJ consistently underestimates the experimental wavelength by 27–33 nm (Table 5). ωB97X-D3 shifts to even shorter wavelengths, by as much as 72–96 nm, while PBE-D3BJ overestimates the experiment to longer wavelengths, by 16–40 nm. Better agreement is obtained by reducing the amount of Hartree-Fock exchange in the B3LYP functional, to either 5% or 10%.,

Table 5.

Deviation of absorption and fluorescence λ_max (nm) from TD-DFT from experimental data.

	Absorption		Fluorescence
	MSD	MAD	MAD	MSD
PBE-D3BJ	+23	23	+11	11
BLYP-D3BJ	+25	25	+12	12
B3LYP-5%	+5	5	−10	10
B3LYP-10%	−10	10	−24	24
B3LYP-15%	−22	22	−37	37
B3LYP-D3BJ	−31	31	−46	46
ωB97X-D3	−83	83	−94	94
ωB97X	−85	85	−99	99
CAM-B3LYP	−71	71	−85	85

Experimentally, monoalkyl-substituted pyronins 2-4 and 10-12 possessed λ_max values of approximately 525 nm, and demonstrate similar bathochromic shifts to AR in comparison to PY/PB. The TD-DFT absorption data indeed shows that the band maxima for 2-4 and 10-12 are indeed within 5 nm of AR’s band maximum, with all DFT functionals (see Figure S39). Experiments also indicate that the band maxima of the azetidinyl compounds, 1 and 8, are bathochromically shifted by 26–30 nm versus AR. All hybrid functionals replicate these bathochromic shifts, 31–44 nm (Figure S39).

For fluorescence band maxima, B3LYP-D3BJ and ωB97X-D3 consistently underestimate experimental wavelengths (Table 5 and Figure S40). The bathochromic shifts of 1 and 8, versus AR, 2-4 and 9-12 are however captured by both functionals, although the lowest MADs are obtained for B3LYP-5% and BLYP-D3BJ.⁵⁷

Pyronins 1-4, 6-8, and 10-12 all possess Stoke shift values in the range of 16–19 nm, again baring strong resemblance to the photophysical trends observed for PY, PB, and AR. However, in most cases, the molar absorptivity at λ_max (ε_max) and the quantum yields (Φ_f) are elevated in comparison to PY, PB, and AR, indicating that these pyronin variants, as was predicted from our computational studies, are brighter fluorophores than PY, PB, and AR. Notice that pyronins 6 and 7, have very low Φ_f values, due to a more stabilized TICT structure in the excited state, as described earlier in our computational studies, Table 1. In general, experimental Φ_f values are in moderately decent alignment with our predictions (based on the integrated vibronic fluorescence spectra, see SI), with 11 > 4 > 2 ~ 3 > 10 ~ 12 > 1 > 8.

In considering ε_max x Φ_f, a quantitative measure of fluorophore intensity, other trends emerge. Most notably, n-propylamino substitution results in highly fluorescent pyronins 3 and 11, though unexpectantly pyronin 3 demonstrated slightly brighter fluorescence than 11. Monoalkyl-substituted pyronins 2-4, in general, result in strong fluorophore performance, and all out-perform fluorinated counterparts (10-12). Interestingly, isopropylamino substitution lowers molar absorptivity which adversely affects fluorophore brightness, despite relatively high quantum yields. Azetidinyl substitution, as noted in previous studies,^{24, 58} produced highly fluorescent pyronins in the 550 nm excitation range; nevertheless, both 1 and 8 are less fluorescent than monoalkylamino species 2, 3, and 11, which appear to better mitigate TICT transitions states due to larger energetic barriers of rotation about the amino group in the S₁-min structures. Overall, fluorophore intensities follow the trend 3 > 2 > 11 > 1 > 8 > 10 > 4 > 12 >> PB > PY.

In summary, using molecular modeled interconversion energies between S₁ and TICT states and calculated vibronic adsorption and fluorescence spectra, we were able to predict large differences in fluorophore performance for the pyronin family of fluorophores. In particular, the S₁ ↔ TICT interconversion energies are very useful for separating species that either do or do not possess lossy internal conversion to the dark TICT states. The integrated areas for calculated vibronic fluorescence spectra and absorption oscillator strengths are also useful for estimating quantum yield trends, which often correlates to overall fluorophore brightness. Methodologically, we found that while TD-DFT generally reproduces the trend in absorption and fluorescence band maxima, the best agreement, for these set of compounds, is found with DFT functionals with small amounts of exact Hartree-Fock exchange, 0–10%. Through the computational methods adopted in this work, we were able to identify the 3,6-dipropylamino substitution patterns as a means for generating highly fluorescent pyronins 3 and 11 in the 525 nm excitation range. Additionally, 3,6-diazetinyl substitution of 1 and 8 offer bright pyronin fluorophores in the 550 nm excitation range. Together, pyronins 1, 3, 8, and 11 offer considerable gains (~2–6.5 fold increases in ε_max x Φ_f) over traditional pyronins (i.e. AR, PY, and PB), and are the most intensely fluorescent set of pyronin fluorophores reported to date. It is likely that information obtained from molecular modeling of energetic barriers between S₁ and TICT states could be used as a means to guide the discovery of other high performance xanthene-based fluorophores such as within the acridine, rhodamine, and rhodols families.

ABBREVIATIONS

PY

pyronin Y

PB

pyronin B

AR

acridine red

TICT

twisted intramolecular charge transfer

TD-DFT

quantum-mechanical time-dependent density functional theory

cLog P

calculated Log partition coefficient

ΔE

energy differences between S1 and TICT structures

ΔE^‡

interconversion energy barriers between S1 and TICT structures

ΔH

enthalpy differences between S1 and TICT structures

ΔH^‡

interconversion enthalpy barriers between S1 and TICT structures

HOMO

highest occupied molecular orbital

LUMO

lowest unoccupied molecular orbital

S_NAr

nucleophilic aromatic substitution

DDQ

2,3-dichloro-5,6-dicyano-1,4-benzoquinone

λ_max

maximum absorbance wavelength

ε_max

molar absorptivity at the maximum absorbance value

Φ_f

quantum yield

λ_em

maximum emission wavelength