Vibrational Circular Dichroism (VCD) Chiral Assignment of Atropisomers: Application to γ-Aminobutyric Acid (GABA) Modulators Designed as Potential Anxiolytic Drugs

Abstract

Atropisomers exist when axial chirality is present as a result of conformationally restricted rotation around a single bond. The interconversion rate of the individual atropisomers is critical to the assessment of chiral stability of a drug throughout scale-up, development, production, and storage as well as in vivo pharmacokinetics. We describe the application of vibrational circular dichroism spectroscopy coupled with quantum mechanics simulations to assign the absolute axial chirality and measure the racemization half-life of a series of potential anxiolytic drugs that act as γ-aminobutyric acid modulators.

Keywords

Infrared spectroscopy Atropisomer Vibrational circular dichroism Computational chemistry.

INTRODUCTION

Approximately 60% of drugs making it into the commercial market are chiral.¹ While most chiral drugs have conventional rectus (R) or sinister (S) stereocenters, some exhibit axial chirality arising from restricted rotation around single bonds such that individual conformers are stable enough to be isolated—so-called atropisomers (Fig. 1). Examples of drugs with axial chirality due to atropisomerism include the antibiotic vancomycin and the anticholinergic telenzepine (used to treat peptic ulcers).^2,3

Fig. 1.

Atropisomers arising from restricted rotation of the biphenyl group in the γ-aminobutyric acid (GABA) modulator of interest. The C₁-C₂-C₃-C₄ dihedral dictates whether the fluoro substituent is directed (i) into the plane or (ii) out of the plane.

The presence of an atropisomeric species represents more than just a stereochemistry curiosity. The degree of interconversion between atropisomers (i.e., the chiral stability) can have serious implications for production of the compound (scale-up, purity assessment, and storage) as well as its effects in vivo. Atropisomerism has been classified in terms of the energy barrier to interconversion as well as the associated half-life to racemization.^4,–6 For example, the Bristol–Meyers–Squibb compound BMS-207940, for treatment of congestive heart failure, exists as interconverting atropisomers that have concentration-dependent interconversion rates in plasma (t₁/₂ ≈ 2.5 h at 400 μg/mL versus t₁/₂ < 0.1 h at 20 μg/mL).⁷ Naturally, interconversion causes significant challenges for pharmacokinetic and safety studies of the individual atropisomers. Unlike conventional stereochemistry, there are degrees of axial stereochemistry ranging from conformations essentially locked in position to the other extreme of conformations exhibiting nearly free rotation. A general guideline has been proposed by LaPlante et al. suggesting three classes of atropisomers in a drug discovery context: Class 1 are those with an interconversion barrier <∼20 kcal/mol and are likely to have a fast rotation rate allowing for development as essentially a single freely rotating compound; Class 2 have interconversion barriers between ∼20 and 30 kcal/mol and exhibit delayed rotation with t_1/2 on the order of minutes to weeks, necessitating more challenging development as a mixture of the two formally separable atropisomers (e.g., BMS-207940 above); Class 3 have barriers > 30 kcal/mol (t_1/2 on the order of years) and likely may be developed as a single, stable atropisomer.⁶

The γ-aminobutyric acid (GABA) modulators in the present study are designed as potential anxiolytics for psychiatric indications.^8,9 The compounds contain a biphenyl substructure and may exhibit atropisomerism depending on the size and nature of the substituents. In addition to the quantification of chiral stability, the assignment of absolute axial stereochemistry is desirable to differentiate the biological activities of the individual atropisomers (i.e., if only one of the atropisomers is biologically active, which one is it and can structure-activity relationships be established?). Data for the separated atropisomers of a representative GABA compound are shown in Fig. 2. One of the atropisomers is 60 times more active than the other, and the absolute chirality and stability were unknown at the time they were tested. By combining vibrational circular dichroism (VCD) spectroscopy and quantum mechanics simulations, a comprehensive knowledge of the energetics, rotational interconversion rates, and absolute stereochemistry associated with biological activity may be built as powerful assets for medicinal chemistry teams to interpret data for existing compounds and to design new compounds with desired properties.

Fig. 2.

Biological data for the separated atropisomers. “Atropisomer A” is 60 times more active at the target relative to “Atropisomer B.” However, the absolute chirality and interconversion rate were both unknown at the outset of this study.

While traditional applications of vibrational spectroscopy are “blind” to the specific enantiomeric state or enantiomeric excess, VCD permits acquisition of infrared (IR) spectra, which are phase sensitive to the specific chiral structure present. In the VCD experiment, the differential absorbance of right-versus left-circularly polarized light is measured. Pure enantiomeric species will produce spectra that are identical in all respects, except that the spectral features of the two enantiomers will be opposite in sign, i.e., if a band is positive for one enantiomer it will be negative for the opposite enantiomer. An overview of vibrational optical activity, of which VCD is a subset, provides a detailed presentation of the technology.¹⁰

While the raw VCD spectra can be used to differentiate the two enantiomers, the spectra alone are not independently sufficient to assign absolute chirality. Absolute assignment requires comparison of the VCD spectrum to a reference spectrum for which the specific signs of bands are known to be directly linked to an established absolute chirality. While such a reference spectrum can be obtained from a compound for which the absolute chiral structure is known (i.e., either through X-ray crystal structure analysis or from a batch generated through a stereo-specific synthesis), most often no such prior compound or knowledge is available. In these cases the reference spectrum must be obtained from quantum mechanical simulations of the VCD spectrum.^11,–14

In this paper, VCD spectroscopy coupled with quantum mechanical simulations is used to assign the absolute axial chirality and measure the half-life of a series of potential anxiolytic drugs that act as γ-aminobutyric acid (GABA) modulators including the enantiomeric pair 9-amino-2-cyclopropyl-5-((P)-2-fluoro-6-methoxy-phenyl)-2,3-dihydro-pyrrolo[3,4-b]quinolin-1-one and 9-amino-2-cyclopropyl-5-((M)-2-fluoro-6-methoxy-phenyl)-2,3-dihydro-pyrrolo[3,4-b]quinolin-1-one (Fig. 1). The resultant understanding proved critical in decisions regarding suitability of these compounds as drug development candidates.

RESULTS AND DISCUSSION

Conformations and Nomenclature. Quantum mechanics optimizations were carried out on the four most energetically favorable conformers. These four conformers consisted of two pairs [labeled (P1, P2) and (M1, M2) in Fig. 3]¹⁵ that differ by the orientation of the methoxy and fluoro substituents relative to the plane of the quinoline ring. Additional conformers were located and assessed (e.g., rotation around the methoxy group and rotation around the N-cyclopropyl bond) and were included in the simulation of VCD spectra. When viewed in the orientation of Fig. 3 (with the cyclopropyl on the right), the methoxy group of the P conformers projects into the plane of the page (i.e., methoxy is “back”). Similarly, the fluoro group of the M conformers projects into the plane of the page (i.e., fluoro is “back”).

Fig. 3.

Relative energies (6-311G**/B3LYP) of optimized minima and transition states along the dihedral rotational profile. Two sets of conformers were located [labeled (M1, M2) and (P1, P2)] that differ by the orientation of the methoxy and fluoro substituents relative to the plane of the quinoline. The origin of the observed atropisomerism is the set of barriers between these pairs of conformers (at dihedral angles ∼0° and 180°). The individual M and P minima are nearly isoenergetic and have essentially free rotation within each pair. The barrier between M and P minima is more easily surmounted by passing the fluorine near the quinoline nitrogen (TS M1/P1, 27 kcal/mol) rather than passing the methoxy near the quinoline nitrogen (TS M2/P2, 50 kcal/mol).

Calculation of Rotational Barrier. The observed atropisomerism originates from the rotational energy barriers to interconversion between these pairs of conformers (at dihedral angles of approximately 0° and 180°). A rigid dihedral drive (using the 6-31G*/B3LYP level of theory) reveals a very small energy barrier for conversions between P1 ↔ P2 as well as M1 ↔ M2. Within this approximation, the compound is free to sweep a range of ∼60° of the dihedral angle (60°–120° for M1–M2 and 240°–300° for P1–P2) with a minimal change in energy (<2 kcal/mol). However, there is a significant energy barrier for conversion from P to M and vice versa.

To characterize the energy barriers more precisely, each of the minima and transition states along the torsional profile was optimized using a larger 6-311G** basis set and four different density functionals (B3LYP, CAM-B3LYP, B3PW91, and B97D). All four functionals (including the B97D, which includes an empirical accounting of dispersion) provide similar predictions of the relative energies and barriers associated with the rotational path. The 6-311G**/B3LYP data (Fig. 3) illustrate that the barrier between the P and M configurations is approximately 27 kcal/mol when taking the path in which the fluorine comes into close proximity to the nitrogen of the quinoline ring, whereas the barrier is nearly twice as high (50 kcal/mol) for conversion between P and M configurations by rotation in the opposite direction (i.e., passing the methoxy group near the nitrogen of the quinoline ring). The rate-limiting barrier to interconversion of ∼27 kcal/mol is at the upper end of Laplante's “class 2” of atropisomers,⁶ suggesting that atropisomerism likely exists, but interconversion may occur on a timescale of days to weeks such that development in clinical trials and long-term storage as a single entity might require additional investigation.

Analysis of Vibrational Circular Dichroism Spectra. The physical sample of the compound exhibited atropisomerism to the extent that it was able to be chromatographically separated into two enantiomers. The absolute chirality of each atropisomer was unknown. The theoretical estimate of interconversion barrier also required experimental confirmation by obtaining a quantitative measure of the half-life.

Figure 4 illustrates the infrared absorbance and VCD spectra collected from samples of each atropisomer. As expected, the infrared spectra of each sample appear virtually identical, while the VCD spectra of the two samples are quite obviously opposite in sign. Comparison of infrared and VCD spectra of each atropisomer also provides an additional means of checking if any contaminants or impurities are present in one of the samples, and the quality of the spectra is sufficient to attempt assignment of absolute chirality via comparison to simulations.

Fig. 4.

Experimental infrared and VCD spectra for Atropisomer A and Atropisomer B.

An overlay plot of the density functional theory (DFT) computed VCD spectra (6-311G**/B3PW91) for the individual M conformers lying within 1 kcal/mole of the global minimum energy structure is presented in Fig. 5. The individual conformation of a molecule imparts dramatic differences in both the intensity and sign of the majority of the bands in the VCD spectrum. Boltzmann weighted averaging across the conformers is required to approximate the observed spectrum. In cases where differing conformations result in inversion of a band (e.g., the bands at 1710 cm⁻¹), it is imperative that all populated conformations in the actual sample have been identified and that the relative energies used for Boltzmann weighted co-addition of the conformational spectra are accurate. When very low-lying conformers demonstrate multiple bands that are inverted (as they are across the entire simulated spectra for this compound), reliance on DFT relative energies for accurate Boltzmann weighting should be done with caution even when using large basis sets.

Fig. 5.

Calculated VCD spectra for the individual conformers of the M configuration that reside within 1 kcal/mole of the global minimum energy structure. The individual conformers generally produce bands of varying intensity and phase, and each contribution must be accurately Boltzmann averaged to produce a composite spectrum for comparison to the experimental spectrum. However, several bands (1534 cm⁻¹, 1492 cm⁻¹, 1472 cm⁻¹, and 1278 cm⁻¹) have contributions from all conformers that preserve the same phase and thus do not rely upon accurate Boltzmann weighting. These bands are more reliable in the comparison to experimental data.

However, the simulated spectra in Fig. 5 also exhibit regions where the profile and sign of a specific band is consistent across all conformations. Interpretation of these bands is less influenced by the precision of the Boltzmann weighting, and there is a higher degree of confidence that possible omissions in identified conformations and/or accuracy of calculated energy levels will not lead to erroneous assignment of absolute chirality. Bands from 1550 to 1470 cm⁻¹ and a set of bands near 1278 cm⁻¹ fall into this high-confidence category.

The direct comparison of phase for the Boltzmann weighted average of the calculated VCD spectra versus the experimental VCD spectra collected for Atropisomer A and Atropisomer B is presented in Fig. 6. Atropisomer A unambiguously corresponds to the P configuration, while Atropisomer B corresponds to the M configuration. By naming convention, Atropisomer A can be assigned as 9-amino-2-cyclopropyl-5-((P)-2-fluoro-6-methoxyphenyl)-2,3-dihydro-pyrrolo[3,4-b]quinolin-1-one and Atropisomer B as 9-amino-2-cyclopropyl-5-((M)-2-fluoro-6-methoxy-phenyl)-2,3-dihydro-pyrrolo[3,4-b]quinolin-1-one.

Fig. 6.

Direct comparison of the measured VCD spectra and the Boltzmann weighted simulated spectra. The matching of phase within this high-confidence region of the spectrum allows assignment of Atropisomer A as the P configuration and Atropisomer B as the M configuration.

Interpretation of the vibrational origin of a band also plays a role in adding confidence to the VCD assignment. These assignments were deduced from animations of the vibrational modes in the simulated spectra and verified by their consistency with standard ranges of IR mode assignments. For example, the 1646 cm⁻¹ band corresponds to deformation of the primary amine. The 1614 cm⁻¹ band corresponds to aromatic ring deformations of the fused ring system in conjugation with the pendant aromatic and involves some NH₂ deformation and will therefore be matrix sensitive. The 1534 cm⁻¹ band, however, corresponds to a highly localized in-plane fused ring vibration. As such, this band is expected to be largely insensitive to differences in the vapor phase matrix of the calculation versus the solution phase of the actual sample. Fortunately, this band also exhibited consistency of sign in the computed spectra across all conformers. The 1493 cm⁻¹ band is associated with CH₃ deformation of the aryl methoxy. Analogous to the 1534 cm⁻¹ fused ring vibration band, the CH₃ deformation should be largely insensitive to matrix.

Calculation of Enantiomeric Conversion. Calculation of enantiomeric conversion for Atropisomer A was obtained as a function of time using both the 1534 and 1471 cm⁻¹ bands in the VCD spectrum. Intensities for the 1534 and 1471 cm⁻¹ bands were divided by the corresponding peak intensity in the infrared spectrum, thereby negating any concentration or pathlength differences (due to potential evaporative losses through the cell seals, alignment, etc.). Rather than monitoring a single band, two bands were used to enhance the signal-to-noise of the experiment by essentially running two parallel experiments. The intensity ratio of the VCD-to-IR spectra at any given time (t) was then divided by the VCD-to-IR spectra ratio at time zero (t₀) to yield VCD intensities relative to the t₀ VCD intensity. The relative VCD intensities for each of the two bands at any given time were then averaged and multiplied by 100 to give a percent enantiomeric excess (ee) versus t₀ (Fig. 7).

Fig. 7.

Contains the plot of percent ee vs. time for Atropisomer A at 31 °C in d₆-DMSO. The exponential fit of the four data points has been extrapolated to 250 days but was not forced to zero.

RACEMIZATION KINETICS

First-order kinetic calculations were performed according to the rate law mathematics presented in Eq. 1, where k₁ is the rate constant and A is the relative analyte concentration, i.e., intensity. The integrated form of the first-order rate equation is presented in Eq. 2, where the subscripts denote the initial time (0) and subsequent times thereafter (t), respectively. Finally, the equation for calculation of the reaction half-life (t_1/2) is presented in Eq. 3. Results of the first-order kinetics analysis are presented in Fig. 8.

Fig. 8.

The plot of the first-order kinetic linearization for Δee vs. time. From the slope in Fig. 8 and Eq. 3, a value of t_1/2 = 31 days was obtained.

\frac{d x}{d t} = k_{1} [A]

(1)

k_{1} t = ln \frac{{[A]}_{0}}{{[A]}_{0} - {[A]}_{t}}

(2)

t_{1 / 2} = \frac{ln 2}{k_{1}}

(3)

CONCLUSION

Experimental and computational VCD studies have been applied to a representative compound within a chemical series of GABA modulators exhibiting biological activity consistent with anxiolytic benefit. The resulting absolute assignment provides drug discovery teams with data that can be used for important decisions regarding advancement into clinical trials. In this work, the atropisomer associated with stronger biological activity is assigned to the P configuration. Second, the computed barrier to interconversion (∼27 kcal/mol) suggests that development of the more active atropisomer as a single entity represents an elevated risk of technical challenges during drug development with respect to chiral stability. The measured racemization half-life via monitoring of VCD signal was 31 days in deuterated dimethyl sulfoxide (d₆-DMSO) and could conceivably be significantly shorter under biological conditions. Finally, the procedure was applied to assess related compounds within the series. The computed barriers provide a reasonable means for prospective molecular designs such that the resulting compounds either have essentially free rotation or are locked into a single atropisomer with an exceedingly slow interconversion rate.^8,9

EXPERIMENTAL

Quantum Mechanics Computations. A Monte Carlo molecular mechanics search of low-energy conformers for the structure was conducted using MacroModel within the Maestro graphical interface (Schrödinger Inc.). Using these structures as starting points, each conformer was minimized using DFT within Gaussian09. Optimized structures, harmonic vibrational frequencies/intensities, VCD rotational strengths, and free energies at standard temperature and pressure (including zero-point energies) were determined for each stationary point. Transition states between atropisomers were confirmed by the existence of a single imaginary vibrational mode (i.e., a negative force constant in the Hessian matrix), which was also animated and inspected to ensure its reasonable correspondence to the intended reaction path. The B3LYP, B3PW91, CAM-B3LYP, and B97D density functionals were used as implemented in Gaussian09. The 6-311G** basis set was used throughout. A scaling factor of 0.975 was uniformly applied to the 6-311G**/B3PW91 harmonic frequencies to approximate the fundamental frequencies. A more detailed review of the theory and implementation of vibrational circular dichroism simulations is also provided.¹⁴

Simulations of infrared and VCD spectra for each conformer were generated using an in-house program to extract spectral data from the Gaussian09 output file and fit Lorentzian line shapes (5 cm⁻¹ line width) to the computed spectra. The software also provided for interactive selection of conformers and calculation of Boltzmann averages. Finally, the in-house software facilitated automated upload into commercial Fourier transform infrared spectroscopy software packages (e.g., Thermo Electron Omnic®) accepting the .csv format. In this manner, direct comparisons between simulated and experimental spectra can be made using the flexibility and tools of the commercial software.

Spectral Acquisition. Approximately 15 mg of atropisomers A and B, respectively, were dissolved in 0.28 ml d₆-DMSO. Solutions were individually loaded into a 0.1 mm BaF₂ infrared cell for analysis at 4 cm⁻¹ resolution using a 5 h, dual source, VCD scan protocol. All analyses were conducted using the BioTools ChiralIR® instrument. The instrument incorporated a single photo-elastic modulator set for polarization modulation at 37.024 kHz with λ/4 retardation (optimized for acquisition of the spectral region centered around 1400 cm⁻¹). Lock-in amplification with a 30 μs time constant, and a 20 kHz high-pass and a 4 kHz low-pass filter was used.

Footnotes

ACKNOWLEDGMENTS

The authors thank Keith Herzog for synthesizing the racemate and Dr. John McCauley for performing the chiral separation into its atropisomers. We also thank Kathy Knappenberger for testing each atropisomer in the GABAA2 binding assay. Additional thanks to Dr. Marc Chapdelaine for leading the chemistry efforts in this project, and for being a tireless champion (along with Dr. Jim Empfield) of the utilization of VCD at AstraZeneca. Finally, special thanks to Pat Capobianco and Lars Jonsson for building and maintaining a computing system capable of handling hundreds of QM simulations per week.

References

Lin

Zhang

Cheng

. “Overview of Chirality and Chiral Drugs”. In: Lin

Zhang

Cheng

, editors. Chiral Drugs: Chemistry and Biological Action. Hoboken, NJ: John Wiley and Sons, 2011. Pp. 3–28.

Williamson

M.P.

Williams

D.H.

. “Structure Revision of Antibiotic Vancomycin. Use of Nuclear Overhauser Effect Difference Spectroscopy”. J. Am. Chem. Soc. 1981. 103(22): 6580–6585.

Eveleigh

Hulme

E.C.

Schudt

Birdsall

N.J.M.

. “The Existence of Stable Enantiomers of Telenzepine and Their Stereoselective Interaction with Muscarinic Receptor Subtypes”. Mol. Pharmacol. 1989. 35(4): 477–483.

Clayden

Moran

W.J.

Edwards

P.J.

Laplante

S.R.

. “The Challenge of Atropisomerism in Drug Discovery”. Angew. Chem. Int. Ed. 2009. 48(35): 6398–6402.

Laplante

S.R.

Edwards

P.J.

Fader

L.D.

Jakalian

Hucke

. “Revealing Atropisomer Axial Chirality in Drug Discovery”. ChemMed-Chem. 2011. 6(3): 505–513.

Cross

Hawkes

G.E.

Kroemer

R.T.

Liedl

C.R.

Loerting

Nasser

Pritchard

R.G.

Steele

Watkinson

Whiting

. “ The Structure, Modeling and Dynamics of Hindered 5,6-Diarylacenaphthenes”. J. Chem. Soc. Perkin Trans. B. 2001. 4: 459–467.

Zhou

Y.S.

Tay

L.K.

Hughes

Donahue

. “Simulation of the Impact of Atropisomer Interconversion on Plasma Exposure of Atropisomers of an Endothelin Receptor Antagonist”. J. Clin. Pharmacol. 2004. 44(7): 680–688.

Liu

X.F.

Chang

H.-F.

Schmiesing

R.J.

Wesolowski

S.S.

Knappenberger

K.S.

Arriza

J.L.

Chapdelaine

M.J.

. “Developing Dual Functional Allosteric Modulators of GABAA Receptors”. Bioorg. Med. Chem. 2010. 18(23): 8374–8382.

Alhambra

Becker

Blake

Chang

Damewood

J.R.

Daniels

Dembofsky

B.T.

Gurley

D.A.

Hall

J.E.

Herzog

K.J.

Horchler

C.L.

Ohnmacht

C.J.

Schmiesing

R.J.

Dudley

Ribadeneira

M.D.

Knappenberger

K.S.

Maciag

Stein

M.M.

Chopra

Liu

X.F.

Christian

E.P.

Arriza

J.L.

Chapdelaine

M.J.

. “Development and SAR of Functionally Selective Allosteric Modulators of GABAA Receptors”. Bioorg. Med. Chem. 2011. 19(9): 2927–2938.

10.

Nafie

L.A.

Dukor

R.K.

. “Applications of Vibrational Optical Activity in the Pharmaceutical Industry”. In: Pivonka

D.E.

Chalmers

J.M.

Griffiths

P.R.

, editors. Application of Vibrational Spectroscopy in Pharmaceutical Research and Development. West Sussex, Chichester, UK: John Wiley and Sons, 2007. Pp. 129–154.

11.

Stephens

P.J.

. “Theory of Vibrational Circular Dichroism”. J. Phys. Chem. 1995. 89(5): 748–752.

12.

Rauk

. “Vibrational Circular Dichroism Intensities: Ab Initio Calculations. Understanding Chemical Reactivity”. In: Mezey

P.G.

, editor. New Developments in Molecular Chirality. Dordrecht, the Netherlands: Kluwer Academic Publishers, 1991. Pp. 57–92.

13.

Nafie

L.A.

Freedman

T.B.

. “Vibronic Coupling Theory of Infrared Vibrational Transitions”. J. Chem. Phys. 1983. 78(12): 7108–7116.

14.

Crawford

T.D.

. “Ab Initio Calculation of Molecular Chiroptical Properties”. Theor. Chem. Acc. 2000. 115(4): 227–245.