Improving the Fluorescent Probe Acridonylalanine Through a Combination of Theory and Experiment

Abstract

Acridonylalanine (Acd) is a useful fluorophore for studying proteins by fluorescence spectroscopy, but it can potentially be improved by being made longer wavelength or brighter. Here, we report the synthesis of Acd core derivatives and their photophysical characterization. We also performed ab initio calculations of the absorption and emission spectra of Acd derivatives, which agree well with experimental measurements. The amino acid aminoacridonylalanine (Aad) was synthesized in forms appropriate for genetic incorporation and peptide synthesis. We show that Aad is a superior FRET acceptor to Acd in a peptide cleavage assay, and that Aad can be activated by an aminoacyl tRNA synthetase for genetic incorporation. Together, these results show that we can use computation to design enhanced Acd derivatives which can be used in peptides and proteins.

Introduction

Fluorescence spectroscopy can be a valuable tool for studying the structural dynamics of proteins and protein/protein interactions.^[1] There are several common types of protein experiments that employ fluorescence spectroscopy: folding/conformational change experiments, binding experiments, and proteolysis experiments (Fig. 1).^[2] Changes in fluorescence can be used to track protein structural change on the ns timescale using distance dependent chromophore interactions: either Förster resonance energy transfer (FRET) or quenching by photo-induced electron transfer (eT).^[3] Such studies require fluorescent probes that enable accurate measurement on a variety of distance ranges between the two chromophores. FRET ranges are characterized by the Förster radius, R₀, the distance of half-maximal energy transfer for any chromophore pair.^[4] For example, the common FRET pair fluorescein (Fam)/ tetramethylrhodamine (Tmr) has an R₀ of 47 Å and is useful to measure distances in the 30 to 90 Å range.^[5] Since many inter-residue distances in proteins are shorter than this, one needs to complement the Fam/Tmr FRET pair with other probe pairs that are better suited to shorter interactions. With this in mind, the Petersson laboratory has developed a methoxycoumarinylalanine (Mcm, 1)/ acridonylalanine (Acd, 2) FRET pair for monitoring distances in the 15 to 40 Å range, and thioamide/Mcm or thioamide/Acd eT quenching pairs for short distance (<15 Å) measurements.^[6]

Figure 1.

FRET experiments and fluorescent amino acids. Top: Protein conformational changes, protein–protein interactions, and proteolytic cleavage can be monitored by changes in the intra- or intermolecular distance between two FRET (Förster resonance energy transfer) probes. Bottom: Fluorescent amino acids based on 7-methoxycoumarin and acridone cores.

In addition to being better suited to short distance ranges than Fam or Tmr, Acd is small enough to be directly genetically incorporated, rather than post-translationally attached, and is less likely to disrupt protein folding.^[7] This allows one to place a chromophore on the interior of a protein and to label proteins that cannot be reversibly unfolded and refolded (e.g., our published labeling of LexA).^[8] Having the chromophore attached by a short sidechain rather than a Cys maleimide or “click” chemistry triazole linker also reduces the positional uncertainty of the FRET probe with respect to the protein backbone. Thus, distance measurements from FRET should more reliably report on changes in protein conformation. We and others have previously shown that Acd can be a valuable probe for protein study because of its small size (222 ų), high quantum yield in water (ϕ = 0.95), unusually long fluorescence lifetime (τ ~15 ns), and high photostability.^[9] We have developed an engineered aminoacyl tRNA synthetase (RS)/orthogonal tRNA pair for selective Acd incorporation by unnatural mutagenesis.^[9b] This has allowed us to label proteins and peptides with methoxycoumarin/Acd FRET pairs either through Mcm incorporation by solid phase peptide synthesis (SPPS) or by attachment of methoxycoumarin-maleimide to a Cys residue in a protein.^[6b] While many aspects of the Mcm/Acd FRET pair are optimal, such as significant spectral overlap and a high Mcm extinction coefficient at 325 nm where Acd has a minimum, one disadvantage is the small Acd Stokes shift which leads to significant overlap of their emission spectra. This overlap necessitates a challenging deconvolution of the Mcm/Acd spectra in order to determine FRET efficiencies and distance measurements. Thus, an Acd derivative with a larger Stokes shift would be desirable for Mcm FRET.

In addition to improving Mcm FRET, we also wish to alter other fluorescent properties of Acd, such as red-shifting excitation and emission, increasing the extinction coefficient, and altering solvatochromic effects to make brighter derivatives that are better suited to microscopy or single-molecule fluorescence applications. Sisido and coworkers have previously shown that some of these effects can be achieved by simply homoligating Acd with a benzene ring (benzoacridonylalanine or Bad, 3) to extend the π-system.^[10] However, we wish to make derivatives that can still be incorporated by the ribosome, and Sisido’s laboratory also showed that Bad was not incorporable during in vitro translation with chemically-charged tRNA. Therefore, we will use crystal structures and computational models of our evolved AcdRS and “rules” for ribosomal permissivity established in previous in vitro translation studies^{[7, 11]} to restrict Acd substitutions to positions that will allow in vivo tRNA charging and incorporation into proteins.

Identifying Acd derivatives by making amino acid analogs is synthetically laborious and unnecessary given that Acd and Bad spectroscopic properties are identical to the properties of the respective chromophore cores.^{[9a, 9c, 10]} Thus, to improve Acd fluorescence, we set out to make a series of acridone (5) core derivatives in order to identify derivatives with sufficiently improved properties to warrant synthesis of the amino acid form for incorporation into peptides. Previous studies of acridone derivatives have shown that many of the spectroscopic properties can easily be modulated through substituent effects, providing strong precedent for our work.^{[10, 12]} Moreover, focusing on the acridone core makes computational modeling more tractable, with the potential to further narrow the scope of synthetic work by predicting absorption and emission spectra.

Here, we prepare a series of acridone derivatives in order to validate the accuracy of our electronic structure calculations and identify a derivative, aminoacridonylalanine (Aad, 4) with substantially red-shifted emission. In addition, our calculations help to explain the origin of fine structure in acridone spectra, an explanation which is in conflict with the conclusions of previous computational and experimental Acd spectroscopy studies, but is consistent with the larger body of acridone literature. Finally, we synthesize Aad and perform initial trials toward its genetic incorporation.

Results and Discussion: Acridone derivative syntheses

To synthesize acridone cores that were not commercially available, we used two general strategies, functionalization of acridone and cross-coupling of pre-functionalized aryl units. Nitration of acridin-9(10H)-one 5 was performed with HNO₃ in CH₃COOH at 50 °C, giving a mixture 2-nitroacridin-9(10H)-one (6) and 4-nitroacridin-9(10H)-one (7) (variations in reaction conditions produced a 74% yield of 6 or 38% yield of 7, see ESI for details).^[13] Reduction of 6 and 7 using Na₂S and NaOH in refluxing EtOH/water gave 2-aminoacridin-9(10H)-one (8) and 2-aminoacridin-9(10H)-one (9), respectively, in quantitative yields (Na₂S used for small scale, Pd/C also used to reduce 7, see ESI for details).^[13] (Scheme 1)

Scheme 1.

Synthesis of nitroacridones (6 and 7) and aminoacridones (8 and 9).

We also used Buchwald-Hartwig cross-coupling amination between methyl 2-aminobenzoate (10) and aryl bromides 11-12, followed by Friedel-Crafts cyclization, to generate several acridone derivatives.^[14] Cross-coupling was accomplished using Pd(OAc)₂, racemic 2,2′-bis(diphenyl-phosphino)-1,1′-binaphthyl (rac-BINAP), and Cs₂CO₃ to generate methyl 2-(phenylamino)benzoates 13 and 14 in 86 and 78% yield, respectively. Then the acridone cores were achieved by saponification with LiOH, followed by Friedel−Crafts cyclization, to give 4-fluoroacridin-9(10H)-one (15, 93%) and 4-methoxyacridin-9(10H)-one (16, 96%). (Scheme 2)

Scheme 2.

Synthesis of acridones 15 and 16 and benzoacridones 18–20.

The same strategy was used to form benzoacridone derivatives. Buchwald-Hartwig cross-coupling with methyl 3-amino-2-naphthoate (17) and aryl bromides (11–12 and bromobenzene) gave the corresponding methyl 3-(phenylamino)-2-naphthoates in 97–98% yields (see ESI). Then the benzo[b]acridin-12(5H)-one derivatives (18–20) were formed by hydrolysis and Friedel−Crafts ring-closing reactions in 85–89% yields. (Scheme 2)

These methods provided access to many acridone derivatives for this study and are sufficiently general to provide access to most desired derivatives for future studies. In particular, the cross-coupling route allows us to access derivatives, such as 3-aminoacridone, that may be difficult to synthesize using nucleophilic aromatic substitution, as the directing effects of the ortho carbonyl and amino groups on each benzene ring in the acridone core would strongly favor reactions at the 2- or 4- positions.

Results and Discussion: Spectroscopic characterization

Following production of the various Acd derivatives, we determined the absorbance and emission profiles, as well as the extinction coefficient, of each core. These data are summarized in Figure 2 and Table 1, and the spectra of each compound are reported separately in the ESI. The absorption and emission spectra are normalized in Figure 2 for clarity, raw spectra are shown in the ESI and extinction coefficients and emission intensities are given in Table 1. Since we ultimately endeavor to utilize these derivatives as fluorescent unnatural amino acids, we attempted to perform the spectroscopic characterization in phosphate-buffered saline (PBS), pH 7.4. However, due to the relatively low solubility of the benzoacridone compounds in water, all measurements were performed in 1:1 acetonitrile/PBS. Extinction coefficients were obtained through serial dilutions of stocks of 300 μM, which is approximately the solubility limit of benzoacridone in acetonitrile. Fluorescence emission spectra were acquired under the same solvent conditions.

Figure 2.

Absorption and emission spectra of acridone derivatives. Spectra determined in 1:1 acetonitrile/PBS, pH 7.4. Spectra are shown normalized to enable comparison of changes in absorption and emission maxima. Raw spectra are shown in ESI.

Table 1.

Calculated and Observed Photophysical Parameters of Acridone Derivatives.^a

	Calculated		Observed
Compound	λex/εb	λem/Int.c	λex/εb	λem/Int.c
Acd 5	383/2.43	00411/1.00	398/0.70	412/1.00
2-NO2 6	408/1.87	-	402/0.37	-
4-NO2 7	464/2.06	-	434/0.60	-
2-NH2 8	419/0.47	508/0.21	420/0.52	527/0.16
4-NH2 9	400/0.61	492/0.59	400/0.59	540/0.01
4-F 15	365/1.19	397/0.48	378/0.51	412/0.65
2-OMe S1	400/0.95	444/0.47	396/0.82	447/0.63
4-OMe 16	375/0.62	421/0.35	384/0.69	431/0.37
Bz 18	446/0.81	495/0.31	438/0.58	507/0.42
4-F-Bz 19	446/0.61	496/0.24	448/0.54	508/0.21
4-OMe-Bz 20	456/0.67	512/0.25	454/0.61	512/0.14

^aLowest energy absorption and highest intensity emission data reported here. A full list of absorption and emission peaks are found in Table S2 in ESI.

^bExtinction coefficients (ε) reported as 10⁴ M⁻¹cm⁻¹.

^cEmission intensity (Int.) normalized the intensity of the highest emission peak of acridone for both calculated and observed spectra.

The parent acridone (5) absorbance spectrum features two major peaks in the near UV region with maxima at 382 and 398 nm as well as additional features below 300 nm. Derivatization of the acridone core resulted in changes of the absorbance profile that were both functional group and position dependent. For example, introducing a methoxy group in the 2 position in S1 resulted in a ~15 nm red-shift in the absorbance maximum, with minimal modulation of the line shape, while introduction at the 4 position in 16 resulted in a very minimal shift, with the multi-peak profile becoming less defined. Amino modification at either the 2 or 4 position produces a singular broad feature that is significantly red-shifted. Moreover, the absorbance profile of 2-aminoacridone (8) is pH sensitive (see ESI, Fig. S5). The spectrum takes on a single broad absorbance feature very similar to that of 4-aminoacridone (9) at high pH, which is dramatically reduced at low pH. It is important to note that both the unmodified and the 4-amino Acd chromophores are pH insensitive (see ESI, Figs. S4 and S6). Lastly, we observe that extension of the conjugated system in the case of benzoacridone (18) results in an expected shift of the absorbance to higher wavelengths, but displays an unexpected reduction in the extinction coefficient. When modified with either fluoro (19) or O-methyl (20) substituents, the absorbance profile is minimally shifted and the peaks become less well-resolved, similar to what was observed when modifying the parent acridone scaffold (15 and 16).

Acridone substitution also elicited changes in the emission profile and Stokes shift. The unmodified acridone emission features two major peaks at 412 and 435 nm with a minor peak around 460 nm. As in the absorbance profiles, the maxima of these peaks move to lower energy upon modification with the O-methyl group (S1 or 16) with the multi-peak profile becoming less defined. Similarly, introduction of the amine functionality resulted in a significant red shift as well as a reduction of the multi-peak line shape to a single broad emission. Both compounds 8 and 9 exhibited a dramatic increase in Stokes shift of ~100 nm compared to the acridone core. Both also displayed a sensitivity to pH, manifested as a decrease in emission intensity with decreasing pH, compared to the unmodified core whose emission is insensitive to pH changes in the 2–10 range. Finally, the benzo modified acridone analogs all feature a nearly identical emission profile ~100 nm red shifted from the emission of the parent acridone compound. The combination of benzo and fluoro or methoxy substitution further red shifts the absorption spectrum, but does not appreciably change the emission spectrum.

Results and Discussion: Electronic structure calculations

In order to understand the nature of the substituent effects on acridone fluorescence, we performed ab initio electronic structure calculations of the absorption and emission spectra for all of the core derivatives. The calculations employed the APF-D density functional as implemented in the Gaussian16™ suite of programs with the 6–311+G(2d,p) basis set which has been recommended for calculations of fluorescence spectra.^[15] The close agreement between the calculated (Calc.) and experimental (Obs.) spectra for the parent species, acridone, clearly justifies this choice (Figure 3). For microscopy and spectroscopy applications, the useful transition is not the expected intense C=O n → π* excitation at 253/255 nm (Calc./Obs.). Rather, it is the weaker π → π* excitation with peaks at 366/376 nm (Calc./Obs.) and 384/402 nm (Calc./Obs.). The emission spectrum shows a ~10 nm Stokes shift with π* → π emission peaks at 396/412 nm (Calc./Obs.) and 418/436 nm (Calc./Obs.). We note that although the agreement of the calculated and observed extinction coefficients is excellent for the n → π* transition, the calculations overestimate the intensity of the π → π* absorption. However, what is most important is that throughout the series of acridone derivatives, the wavelengths and intensities of the π → π* absorption and emission track very well with the experimental data, showing that the calculations can clearly predict acridone derivative properties to guide synthetic efforts (see Table 1 and ESI, Table S2 and Figs. S4-S9).

Figure 3.

Calculated Acridone Spectra. Top Left: Experimental and calculated, Franck-Condon corrected spectra, with vertical transitions used to determine the calculated spectrum shown. Bottom Left: Primary molecular orbitals (MOs) involved in the n → π* (~250 nm) and π → π* (~380 nm) transitions of acridone. Right: Jablonski diagram showing strategy for using frozen ground or excited state solvation in geometry optimization and frequency calculations to determine electronic spectra.

For acridone, we performed more sophisticated calculations, combining Franck-Condon integrals from vibrational calculations to generate spectra representing acridone absorption and emission spectra in aqueous solution. Bonding in excited states is generally not as strong as in ground states, which reduces the zero-point-energy (ZPE) from molecular vibrations. For example, the ZPE for the lowest-lying π→ π* excited state of acridone (115.76 kcal/mol) is 2.21 kcal/mol less than the ZPE for the ground state (117.97 kcal/mol). This reduces the vertical excitation energy to 78.20 kcal/mol (λ_ex = 366 nm). The electronically excited state then relaxes through internal conversion to its vibrational ground state, lowering the energy by 3.85 kcal/mol. This is followed by solvent relaxation which lowers the energy of the excited state by an additional 1.19 kcal/mol, and raises the ground state energy by 0.94 kcal/mol (Fig. 4). The vertical emission (68.61 kcal/mol or λ_em = 418 nm) is followed by relaxation through internal conversion to the vibrational ground state, lowering the energy by 3.61 kcal/mol. Solvent relaxation then lowers the energy of the ground state by an additional 0.94 kcal/mol. We found that these Franck-Condon corrected spectra have excellent agreement with the observed acridone spectra, particularly for the features in the π→ π* absorption and emission peaks. We note that previous computational studies of acridone by Oshima, accompanied by gas phase electronic spectra measurements, concluded that the various vibronic modes that they observed for the π→ π* absorption and emission were the result of different hydration clusters.^[16] While this explanation may be valid for gas phase measurements, solution phase measurements show a similar spectral shape for acridone absorption and emission in polar, hydrogen-bonding solvents like water, and in non-polar solvents like THF.^[9c] Thus, the explanation arising from our Franck-Condon corrected calculations (determined with continuum solvation), that the multiple peaks at ~380 nm arise from internal acridone vibrational modes, seems to fit better with solution phase data.

Figure 4.

Aad is a Superior FRET Acceptor for Mcm. Top: Absorbance (solid lines) and fluorescence emission (dashed lines) of Mcm (purple), Acd (blue), and Aad (orange) with spectral overlap shaded. Both Acd and Aad have low absorbance at the excitation maximum of Mcm (325 nm) and comparable spectral overlap (shaded areas), but Aad emission can more easily be distinguished from Mcm emission. Bottom Left: Normalized changes in FRET ratio (380/480 nm for Acd, 380/530 nm for Aad) of 5 µM Xxx-Leu-Leu-Lys-Ala-Ala-Ala-Mcm (Xxx = Acd or Aad) peptides upon rapid mixing with 1 µM trypsin protease. Data points from three trials shown to demonstrate reproducibility, with average shown as a smooth curve. Bottom Right: Fluorescence emission spectra, with excitation at 325 nm, of Acd or Aad peptides before and after proteolysis. Spectra normalized based on area under the curve.

Results and Discussion: Synthesis of aminoacridonylalanine for protein studies

We chose to generate the amino acid form of 2-aminoacridone (8) since its emission was as red-shifted as any of the benzo derivatives, but its smaller size and greater polarity make it more favorable for eventual genetic incorporation. While the absorption and emission spectra of 2- and 4-aminoacridone are relatively similar, the 2-amino isomer was chosen because it is brighter and our modeling studies (below) suggest that it is likely to be genetically incorporable. Since the 2-aminoacridone core shows a larger Stokes shift than the acridone core with comparable spectral overlap and selective excitability (i.e., the maximum in Mcm absorbance coincides with a minimum in Acd or Aad absorbance) it can function as a superior FRET partner for Mcm. Our spectral overlap calculations indicate that acridonylalanine (Acd, 2) and aminoacridonylalanine (Aad, 4) have similar R₀ values (Acd: 25.3 Å, Aad: 26.5 Å), but the emission of Aad at 530 nm will be easier to monitor in ratiometric FRET experiments and no deconvolutions with Mcm emission will be necessary (Fig. 4). Although 2-aminoacridone is commonly used in glycan experiments, and acyl 2-aminoacidone peptides have been used previously in protease sensors, to our knowledge, amino acid forms such as Aad have not been made previously or used in FRET applications.^{[12b, 17]}

Acd (2) was synthesized as we have previously reported. Nitration of Acd was carried out similarly to nitration of acridone, affording a single nitro isomer in 39% isolated yield (>90% conversion based on integration of the chromatogram). The nitration reaction mixture was basified and sodium sulfide added to perform the reduction to give Aad (4) in 86% yield. (see ESI) The position of the amino group was confirmed by NMR and by comparison of the absorption and emission spectra to the spectra of 2- and 4-aminoacridone (8 and 9). Protected forms of Acd and Aad were generated for solid phase peptide synthesis, and the peptides Xxx-Leu-Leu-Lys-Ala-Ala-Ala-Mcm (Xxx = Acd or Aad) were synthesized. The peptide sequence was chosen as a protease substrate based on previous studies in our laboratory so that cleavage should result in a decrease in FRET.^[18] Cleavage by the proteases trypsin and cathepsin B was tested by exciting at 325 nm and monitoring emission at 380 nm (Mcm) and either 480 nm (Acd) or 530 nm (Aad). After confirming that trypsin proteolysis proceeded at a reasonable rate in a multi-well plate assay (see ESI, Figs. S10 – S11), we measured cleavage rates in a stopped-flow fluorometer. As one can see in Figure 4, while both peptides give usable data, the change in FRET ratio is two-fold greater for the Aad peptide than the Acd peptide.

Site-specific genetic incorporation of Aad requires a tRNA synthetase, or RS, that efficiently and selectively adenylates Aad and then charges an orthogonal tRNA with this activated Aad ester.^[19] As a first step toward achieving an AadRS, we tested Acd and Aad adenylation by AcdRS1 (also referred to as clone G2), a permissive Methanocaldococcus jannaschii TyrRS derivative that incorporates Acd, and also incorporates N-phenyl p-aminophenylanine (Npf) and 4-(2′-bromoisobutyramido) phenylalanine (Brb).^{[8, 9b, 20]} Using a purified form of AcdRS1 in a Malachite green assay that we have previously used to characterize the relative activity of the enzyme toward Acd and Npf, we find that Aad is activated by AcdRS1, although less efficiently than Acd.^[8] (Fig. 5) These results can be understood with a simple model in which Aad is aligned with Acd in our published AcdRS1 docking model, based on the crystal structure of AcdRS1 with Brb.^{[8, 20b]} As one can see in Figure 5, the Aad exocyclic amine makes contact with the sidechains of Val164 and Ala167. Thus, mutation of these residues to smaller or more polar residues may improve Aad charging. Generating a library of randomized mutants of the 164–167 loop, followed by selection of the most active clones, is also likely to result in an efficient AadRS for in vivo incorporation of Aad.

Figure 5.

Aad Activation by AcdRS1. Left: Acd and Aad adenylation measured based on inorganic pyrophosphate (PP_i) formation in an enzyme-coupled Malachite green assay. Control experiments lacking ATP show that the low levels of Aad activity are significant. Right: Aad docked in the AcdRS1 active site by alignment with Acd in an energy-minimized model generated from AcdRS1 PDB coordinates 4PBR.^[20b] See ESI for details.

Conclusions

We can draw several conclusions from our results. Firstly, as anticipated from previous literature reports, we are able to modulate the fluorescence of the acridone core through relatively simple substitutions. These substitutions, introduced through direct modification of acridone or through cross-coupling and cyclization, are compatible with eventual usage in generating Acd amino acid derivatives. Secondly, the close correlation (% differences λ_ex: 2.35, λ_em: 2.40) between our calculated and observed absorption and emission spectra give us confidence that we can predict the spectra for acridone derivatives, providing guidance for future synthetic efforts. This will be particularly valuable for targeting multiply-substituted derivatives, where the number of possibilities is geometrically larger and the synthesis will be more challenging. Finally, Aad appears to be superior to Acd for several fluorescence applications. As demonstrated here, the red-shifted emission of Aad provides superior dynamic range in ratiometric FRET measurements with an Mcm donor. Furthermore, its absorption and emission wavelengths make Aad amenable to use in microscopy applications using typical filter sets for cyan fluorescent protein excitation and yellow fluorescent protein emission. Aad can also serve as a FRET donor to red wavelength dyes such as indodicarbocyanine or sulforhodamine G.^[21] Using the strategy outlined here, we can further improve the spectroscopic properties of Aad, such as its relatively low quantum yield. While Aad can be incorporated into peptides, and native chemical ligation can be used for incorporation into proteins if desired, our ultimate goal is genetic incorporation of Acd derivatives. Our initial in vitro activity results for Aad are very promising, and computational models based on the crystal structure of AcdRS1 indicate that we may be able to obtain an efficient AadRS with just one or two mutations. These rationally-designed efforts, as well as screening of random mutant libraries, are currently under way. If we can successfully generate an AadRS, this would enable the furthest red-shifted fluorescent amino acid to be genetically incorporated in living cells.