ARTICLE pubs.acs.org/JPCA
Substituted Benzoxadiazoles as Fluorogenic Probes: A Computational Study of Absorption and Fluorescence Alex Brown,*,† Tsz Yan Ngai,† Marie A. Barnes,† Jessie A. Key,‡ and Christopher W. Cairo*,‡ †
Department of Chemistry and ‡Alberta Ingenuity Centre for Carbohydrate Science, Department of Chemistry, University of Alberta, Edmonton, Alberta, Canada, T6G 2G2 ABSTRACT: General chemical strategies which provide controlled changes in the emission or absorption properties of biologically compatible fluorophores remain elusive. One strategy employed is the conversion of a fluorophoreattached alkyne (or azide) to a triazole through a copper-catalyzed azide alkyne coupling (CuAAC) reaction. In this study, we have computationally examined a series of structurally related 2,1,3-benzoxadiazole (benzofurazan) fluorophores and evaluated changes in their photophysical properties upon conversion from alkyne (or azide) to triazole forms. We have also determined the photophysical properties for a known set of benzoxadiazole compounds. The absorption and emission energies have been determined computationally using time-dependent density functional theory (TD-DFT) with the Perdew, Burke, and Ernzerhof exchange-correlation density functional (PBE0) and the 6-31+G(d) basis set. The TD-DFT results consistently agreed with the experimentally determined absorption and emission wavelengths except for certain compounds where charge-transfer excited states occurred. In addition to determining the absorption and emission wavelengths, simple methods for predicting relative quantum yields previously derived from semiempirical calculations were reevaluated on the basis of the new TD-DFT results and shown to be deficient. These results provide a necessary framework for the design of new substituted benzoxadiazole fluorophores.
’ INTRODUCTION Small molecule fluorophores are essential probes for cell biology and chemical biology.1 An ideal probe will possess properties including the following: (i) spectral resolution from other fluorophores, (ii) a large extinction coefficient and quantum yield (brightness), and (iii) resistance to photobleaching. A wide variety of organic frameworks are currently employed as fluorophores for biochemical studies,2 and the toolkit for introducing these labels into biomolecules continues to expand.3 Yet, chromophores with tailored properties could enable multicolor, FRET, and single-molecule experiments.4,5 Fluorogenic probes—chromophores that become fluorescent under specific conditions—are exceptionally useful but are difficult to design a priori.69 An improved theoretical understanding of fluorescent and fluorogenic probes would allow for the design of optimized chromophores. Our groups have been interested in developing a theoretical understanding of commonly employed fluorogenic probes using electronic structure calculations. A number of groups have examined the utility of Cu-catalyzed azidealkyne coupling (CuAAC), and the resulting triazole functionality, as a method for perturbing fluorophore properties.10 Zhou and Farhni first reported the application of the CuAAC as a fluorogenic strategy using a coumarin dye.9 The triazole product of the CuAAC reaction,1114 in some systems, provides a means to alter the spectral properties of the initial azide or alkyne.9,1517 Subsequent formation of the triazole can increase emission intensity or change emission wavelength—properties desirable in fluorogenic probes. In some examples, fluorogenic probes were identified using combinatorial strategies,15 highlighting the need for improved predictive strategies. We previously generated a series of coumarin-derived r 2011 American Chemical Society
triazoles and quantified their photophysical properties.18 We found that these dyes were very sensitive to groups at the 7-position as exhibited by hydroxy- to ethynyl-substitution and subsequent triazole formation—sensitivity at the 7-position has been reported previously.15 We found that the absorbance properties of this series were well-described by time-dependent density functional theory (TD-DFT)—perhaps not surprising considering the benchmark TD-DFT study of substituted coumarins.19 In subsequent work, we have also generated a series of benzoxadiazole-derived fluorophores with pendant azide, alkyne, alkyl, and triazole groups.20 This set of fluorophores contains a unique range of photophysical properties. In this study, we identified dyes which were fluorogenic upon CuAAC, and, remarkably, we also identified dyes which became quenched after CuAAC. Thus, these compounds present an ideal data set for the theoretical description of both excited-state electronic structure and corresponding geometry. The electronic structure of benzoxadiazole (often referred to as benzofurazan in the literature) chromophores have been previously studied by a variety of computational methods.2127 Uchiyama et al. have experimentally measured the photophysical properties of a large number of 4-substituted, 5-substituted, and 4,7-disubstituted benzoxadiazole compounds in a wide variety of solvents, and they have supported their experimental work with semiempirical electronic structure computations.2126 Experimental and computational studies of eleven 4-monosubstituted Received: August 17, 2011 Revised: November 14, 2011 Published: December 02, 2011 46
dx.doi.org/10.1021/jp2079296 | J. Phys. Chem. A 2012, 116, 46–54
The Journal of Physical Chemistry A
ARTICLE
benzoxadiazole compounds in various solvents suggest that the differences in quantum yield in nonpolar solvents are due to the probability of nonradiative coupling between S1 and T2, which corresponds to the S1 to T2 energy gap.21 The energy gaps were determined at the parametrized model no. 3, complete-active space configuration interaction (PM3-CAS/CI) level of semiempirical theory. For nine 5-substituted benzoxadiazoles the main nonradiative process was proposed to be photoreaction from an excited singlet state; in this case, the quantum yield increased with increasing S1 to S2 energy gaps as determined at the Austin model 1-CAS/CI (AM1-CAS/CI) level of theory.22 For one set of ten 4,7-disubstituted benzoxadiazoles, fluorescence quantum yields were found to be correlated to the S1 to T2 energy gap computed at the PM3-CAS/CI level of theory.23,24 Other experimental and theoretical work on a larger number of 4,7-disubstituted benzoxadiazole compounds (70 in total) suggested a relationship between the relative fluorescence intensity and the Hammett constants of the substituent groups at the 4- and 7-positions.25 A follow-up study on the same set of 4,7-disubstituted benzoxadiazole compounds suggested that the fluorescence characteristics are determined by the total of electron densities on the skeleton and the dipole moment directed from the 4- to the 7-position.26 More recently, Tsuji et al. used TD-DFT, configuration interaction singles (CIS), and symmetry-adapted cluster configuration interaction (SAC-CI) methods to examine theoretically the gas-phase absorption and emission of nine benzoxadiazole compounds. These values were compared with experimental measurements in cyclohexane.27 Tsuji et al. suggest using CIS excitedstate structures with either TD-DFT or SAC-CI energies as the most suitable method to determine fluorescence wavelengths. While each of these approaches has provided a working model for subsets of the benzoxadiazoles, we sought to develop a uniform approach to modeling the electronic structure of these compounds on the basis of TD-DFT. On the basis of the utility and prevalence of CuAAC, we chose to emphasize azide-, alkyne-, and triazolesubstituted compounds relevant for biomolecular conjugations. Theoretical determination of excitation and emission spectra by TD-DFT requires a judicious choice of functional; see, for example, the benchmark studies of Jacquemin and co-workers.19,2834 Selection must balance the nature of the excitation (valence, Rydberg, and/or charge transfer) and the properties in which one is interested in estimating, e.g., absorption only or absorption and emission. Here, we chose to utilize the hybrid Perdew, Burke, and Ernzerhof exchange-correlation density functional (PBE0)3537 as it has been shown to perform well for a variety of excited-state properties and can equally well describe ground-state properties. TD-DFT results appear to be less basis set dependent than postHartreeFock methods,19,29 and, thus, the relatively modest 6-31+G(d) basis38 has been chosen in order to provide a compromise between accuracy and computational speed. This paper reports a computational study of the photophysical properties of a number of mono- and disubstituted benzoxadiazole compounds (see Figure 1 for the backbone structure). In the present work, we intended to assess the use of TD-DFT for predicting the properties of known compounds to validate its use in guiding the design of new substituted benzoxadiazole compounds for labeling strategies.
Figure 1. Chemical structure of a 2,1,3-benzoxadiazole (benzofurazan) skeleton.
PBE0 functional3537 with the standard Pople-style double-ζ basis sets,38 6-31+G(d), for all atoms. The vertical excitation energies at the ground-state equilibrium geometries were calculated with TD-DFT4042 using the PBE0 functional and the 6-31+G(d) basis set. To determine the fluorescence wavelengths, the geometries of S1 excited states were optimized at the TD-DFT PBE0/6-31+G(d) level of theory. The S1fS0 energy gap at the optimized excited-state geometry was taken as the fluorescence wavelength. All calculations were carried out in the gas-phase and utilized C1 symmetry. The energies of the next highest lying singlet state, S2, and the second lowest lying triplet state, T2, were also determined using TD-DFT PBE0/6-31+G(d) in order to assess possible deactivation mechanisms leading to high (or low) quantum yields as suggested by Uchiyama and co-workers.2123,25,26 While one could determine the lowest triplet-state energies using standard DFT, we wanted to determine a straightforward analysis method analogous to that using PM3-CAS/CI or AM1-CAS/ CI calculations as suggested earlier for benzoxadiazoles,2123,25,26 and this analysis focused on the T2 energies rather than those of T1. All ground- and excited-state calculations were carried out with Turbomole43 using default convergence parameters except for finer numerical grids (designated as m4 in Turbomole) in the DFT and TD-DFT calculations. To verify the comparison of the gas-phase theoretical results with the experimental data, computations were carried out for a subset of all compounds considered using the polarizable continuum model (IEF-PCM)44,45 with parameters for n-hexane as implemented in Gaussian09.46 The ground-state geometries were optimized at the PCM-PBE0/6-31+G(d) level of theory. The absorption wavelengths were determined using TD-DFT with the same functional and basis set. For the determination of the fluorescence wavelengths, the geometry of the S1 excited state has been optimized47 using the state-specific PCM model.48,49
’ RESULTS AND DISCUSSION Computationally determined photophysical properties for compounds 1a6b were compared with experimental absorption and fluorescence values as measured in n-hexane (dielectric constant, ε = 1.9). Additional experimental values were derived from a series of 4-substituted, 5-substituted, and 4,7-disubstituted benzoxadiazoles analyzed computationally using the PM3-CAS/CI and AM1-CAS/CI methods by Uchiyama and co-workers.2123,25,26 We chose to focus on the 4-substituted and 5-substituted benzoxadiazoles which exhibit fluorescence in n-hexane.21,22 Absorption and emission properties for many of the monosubstituted and a single disubstituted (NBD-NHMe) species had been determined previously using TD-DFT and SAC-CI methods.27 In the present work, we have examined a set of new azide- (-N3) and ethynyl-substituted (-CtCH) benzoxadiazoles and their triazole forms produced via a CuAAC reaction.12,13 The structures of the substituents for these new compounds are given in Figure 2 and Figure 3 for the 5-substituted and the 4,7-disubstituted compounds, respectively. The synthesis and full characterization of
’ COMPUTATIONAL METHODS Ground-state geometries of the substituted benzoxadiazoles were optimized with density functional theory39 using the 47
dx.doi.org/10.1021/jp2079296 |J. Phys. Chem. A 2012, 116, 46–54
The Journal of Physical Chemistry A
ARTICLE
Figure 3. Substituents for 4,7-disubstituted benzoxadiazole compounds. R5 = R6 = H.
Table 1. Experimental21,22,24 (in n-Hexane) and Theoretical (Gas-Phase TD-DFT PBE0/6-31+G(d) and SAC-CI/6-31 +G(d)27) Absorption Wavelengths for a Set of Previously Reported 4-Substituted, 5-Substituted, and 4,7-Substituted Benzoxadiazoles (NR = Not Reported)
Figure 2. Substituents for 5-monosubstituted benzoxadiazole compounds. R4 = R6 = R7 = H.
these compounds has been reported elsewhere.20 To distinguish between the two sets of compounds, we will refer to the compounds of Uchiyama and co-workers as set A, and the azide-, ethynyl-, and triazole-substituted compounds of Key and Cairo will be called set B. While solvent plays an important role in influencing the photophysical properties of the benzoxadiazoles,2023,25,26 we sought to determine whether computational methods provide an accurate assessment and prediction of these properties for the simplest comparative case of gas-phase calculations versus experimental measurements in n-hexane, a nonpolar solvent. We found that the absorption wavelengths were well-described by TD-DFT; however, the emission wavelengths were more challenging to interpret. We discuss previously explored models2123,25,26 to predict the relative quantum yields in the context of the new TDDFT PBE0/6-31+G(d) results.
substituent
λexpt/nm
λTD‑DFT/nm
λSAC‑CI a/nm
4-NHCOMe
348
349
376
4-NMe2 4-SMe
403 362
387 384
408 400
4-NH2
376
367
408
4-SPh
362
363
NR
4-OMe
331
351
NR
5-NH2
349
336
374
5-NMe2
373
348
398
5-OMe
316
304
327
5-SMe 5-NHAc
345 329
336 320
333 NR
NBD-NHMe
422
383
451
—
14.7
22.8
MAE a
SAC-CI excitation energies calculated at DFT PBE0/6-31+G(d) ground-state geometries.
were compared with previously measured experimental values (set A; see Table 1). Table 1 also contains recently reported SACCI results for a smaller subset of benzoxadiazole compounds included in set A.27 Table 2 contains the theoretically determined, and experimentally measured, absorption wavelengths for the newly synthesized azide-, ethynyl-, and triazole benzoxadiazoles (set B). We include in our analysis the experimentally measured molar absorption coefficients (ε) for comparison to the
’ ABSORPTION WAVELENGTHS Computationally determined absorption wavelengths of 4-substituted, 5-substituted, and NBD-NHMe benzoxadiazoles 48
dx.doi.org/10.1021/jp2079296 |J. Phys. Chem. A 2012, 116, 46–54
The Journal of Physical Chemistry A
ARTICLE
Table 2. Experimental (in n-Hexane) and Theoretical (Gas-Phase TD-DFT PBE0/6-31+G(d)) Absorption Wavelengths for the 5-Substituted and 4,7-Substituted Benzoxadiazoles Synthesized in the Present Studya substituent
λexpt/nm
ε
λTD‑DFT/nm
f
5-Substituted Benzoxadiazoles 1a
289
24 300
292
0.0749
1b
289
11 200
290 b
0.0525
1c
278
500
301 c
0.0681
1d
277
2 400
296 c
0.0746
2a
325
9 000
325
0.1037
2b
318
18 900
314
0.1132
2c 3a
299 298
400 4 800
291 c 313
0.0884 0.0885
3b
324
9 000
321
0.1144
4a
375
30 500
367
0.3515
4b
380
21 900
398
0.2880
5a
410
17 100
383
0.3138
5b
504
28 300
374
0.3330
6a 6b
420 not soluble
12 200 NM
385 386
0.3218 0.3434
4,7-Disubstituted Benzoxadiazoles
MAE
Figure 4. Correlation between the experimentally observed absorption wavelengths (in n-hexane) and the gas-phase TD-DFT(PBE0)/ 6-31+G(d) values. Previously studied (circles) and new (squares) compounds are illustrated along with a linear fit to the data (R2 = 0.928).
We found that there is an excellent linear correlation between the calculated and experimental absorption λmax values; using the formula
12.6 d (21.0)
a
Also included are the experimental molar absorption coefficients and the theoretical oscillator strengths (f). See Figure 2 and Figure 3 for the identity of the compounds. b Absorption to S2. c Absorption to S3. d Excluding compound 5b.
λexpt ¼ 48:105 þ 1:1502λTD
DFT
ð1Þ
provides an excellent fit to the data (R2 = 0.928; see Figure 4). The linear correlation can be used to correct the computed values to experimental ones. Indeed, when the corrected values are used, the MAE is 12.8 nm over all 25 compounds (with 5b excluded). Clearly, this does not represent a significant improvement over the “raw” TD-DFT results. By including the newly synthesized compounds found in set B, we increase the span of experimental absorption wavelengths from 316 to 422 nm (set A), to 277 to 504 nm (sets A and B). Although the TD-DFT results fail to reflect the longest wavelength absorption correctly, vide supra. Since one goal of the present study is to determine a useful computational method for screening potential fluorophores for labeling strategies, the use of gas-phase absorption values should be validated by comparison with values determined using PCM for solvation—although the high correlation (R2 = 0.928) with the experimental data suggests the use of gas-phase values is warranted. In Table 3 the absorption wavelengths as determined in n-hexane are presented for eight of the set B compounds. The average shift (relative to the theoretical gas-phase values) is 9 nm (0.08 eV) and the discrepancy is dominated by compounds 3b and 5b, which are both triazole-substituted compounds containing benzene rings. The main reason for the large shifts for these two compounds is due to changes in the molecular geometry in going from the gas to solution phase. However, the MAE for the n-hexane results is 11.4 nm and is slightly greater than the MAE of the gas-phase results,which is 9.4 nm for these eight compounds. There is also no change in the trends upon inclusion of solvent effects—for the gas-phase versus the PCM theoretical data R2 = 0.925. Therefore, for initial screening of compounds by absorption wavelength, gas-phase data are suitable. However, for detailed study of particular compounds, the solvent effects should be taken into account.
theoretically determined oscillator strengths. The reported theoretical absorption wavelengths correspond to the transitions with the longest wavelength of significant oscillator strength (f > 0.025). In most cases, these transitions are to the lowest energy bright excited states (S0fS1). There are four notable exceptions where the strongest absorption is not to the lowest lying excited state: 1b with excitation to S2, 1c to S3, 1d to S3, and 2c to S3. Our TD-DFT results were consistent with previously reported27 TD-DFT results which included a subset of the same compounds (Table 1). The absorption wavelengths determined using TD-DFT PBE0/6-31+G(d) were in good agreement with the experimental data with a mean absolute error (MAE) of 14.7 nm for the previously reported compounds (set A; see Table 1) and 12.6 nm for the new compounds (set B; see Table 2). Extending the number of 4- and 5-substituted compounds considered in the present study results in a decreased MAE relative to a previous report.27 In evaluating the MAE for the newly synthesized compounds, we have omitted compound 5b. The maximum absorption coefficient of 5b is for an absorption at 504 nm, while TD-DFT predicts an absorption at 374 nm. Interestingly, there is an experimental absorption at 347 nm. Note that computation using an alternate range-separated hybrid functional (CAM-B3LYP) predicts an absorption at 371 nm; see the discussion regarding fluorescence for the choice of this functional. Therefore, we interpret the strong absorption at 504 nm to correspond to a state that is not well captured by DFT, e.g., involving double excitation. It has been suggested that a statistical analysis (linear regression) of TD-DFT calculated values can improve the agreement between computed and experimentally measured absorption wavelengths for particular classes of compounds.29 49
dx.doi.org/10.1021/jp2079296 |J. Phys. Chem. A 2012, 116, 46–54
The Journal of Physical Chemistry A
ARTICLE
Table 5. Experimental (in n-Hexane) Quantum Yields (Φf) and Theoretical (Gas-Phase TD-DFT PBE0/6-31+G(d)) Fluorescence Wavelengths for 5-Substituted and 4,7-Disubstituted Benzoxadiazoles Analyzed in the Present Studya
Table 3. Theoretical (n-Hexane PCM-TD-DFT PBE0/6-31 +G(d)) Absorption Wavelengths for a Subset of 5-Substituted and 4,7-Disubstituted Benzoxadiazoles Analyzed in the Present Studya λTD‑DFT/nm
substituent
shift/nm (eV)
substituent
5-Substituted Benzoxadiazoles
λexpt/nm
Φf
λTD‑DFT/nm
5-Substituted Benzoxadiazoles
2a
326
1 (0.02)
1a
NM
0
365
2b b
317
3 (0.03)
1b
NM
0
470
3a
319
6 (0.08)
1c
NM
0
572
3b
335
14 (0.16)
1d
NM
0
296
2a
424
0.01
376
2b
370
0.002
424
2c 3a
NM 406
0 0.008
489 365
3b
375
0.008
381
4,7-Disubstituted Benzoxadiazoles 372
4a 4b 5a
401 388
5b
407
MAE
11.4
5 (0.04) 3 (0.03) 5 (0.04) 33 (0.27) c
4,7-Disubstituted Benzoxadiazoles
a
Also, presented are the shifts relative to the gas-phase TD-DFT PBE0/ 6-31+G(d) results. See Figure 2 and Figure 3 for the identity of the compounds. b -C6H15 chain replaced with -CH3 c Excluding compound 5b.
Table 4. Experimental (in n-Hexane) and Theoretical (Gas-Phase TD-DFT PBE0/6-31+G(d) and SAC-CI/6-31 +G(d)27) Fluorescence Wavelengths for a Set of Previously Reported 4-Substituted, 5-Substituted, and 4,7-Substituted Benzoxadiazoles (NR = Not Reported) 21,22,24
substituent
λexpt/nm
λTD‑DFT/nm
4-NHCOMe
439
429
479
4-NMe2 4-SMe
513 433
486 484
544 389
4-NH2
500
470
479
4-SPh
457
597 b
NR
412
437
NR
5-NH2
422
406
475
5-NMe2
452
609 b,c
498
5-OMe
365
360
372
5-SMe 5-NHAc
392 383
443 373
397 NR
NBD-NHMe
500
443
532
MAE
28.9 d
522
0.03
4b
501
