Mechanistic Insights into the Racemization of Fused Cyclopropyl Isoxazolines

Article abstract of DOI:10.1021/acs.orglett.9b04236

Experimental and computational studies of the unexpected racemization of enantiopure fused cyclopropyl isoxazolines are reported. These studies offer insights into the mechanism of racemization, quantify the position of the transition state on the dipolar

Full text of DOI:10.1021/acs.orglett.9b04236

Letter
pubs.acs.org/OrgLett
Cite This: Org. Lett. XXXX, XXX, XXX−XXX
Mechanistic Insights into the Racemization of Fused Cyclopropyl
Isoxazolines
Kyle W. Quasdorf,*,† Adam B. Birkholz,† Michael D. Bartberger,*,†† John Colyer,† Stephen Osgood,†
Kevin Crossley,† and Seb Caille†
†Departments of Chemical Process Research and Development and Therapeutic Discovery, Amgen, Inc., One Amgen Center Drive,
Thousand Oaks, California 91320, United States
††1200 Pharma LLC, 844 East Green Street, Suite 204, Pasadena, California 91101, United States
S
* Supporting Information
ABSTRACT: Experimental and computational studies of the
unexpected racemization of enantiopure fused cyclopropyl
isoxazolines are reported. These studies offer insights into the
mechanism of racemization, quantify the position of the
transition state on the dipolar−diradical continuum, and
establish a relationship between the structure and stability of
this class of compounds. Experimental and computed energy
barriers for racemization are also presented.
ince the discovery that aspartyl protease BACE was shown
was obtained. After conducting a series of control experiments,
it was discovered that batches of enantiopure cyclopropane 1
had, in fact, completely racemized upon being stored as
solutions in toluene under ambient conditions at room
temperature.
There have been several reports describing the synthesis of
fused cyclopropyl isoxazolines,4 including enantioenriched
variants,5 and, to the best of our knowledge, the susceptibility
of these compounds to racemization has not been reported.
We considered the fragmentation of the central carbon−
carbon bond of the cyclopropane as the most likely source for
this mode of racemization, as shown in Figure 2.
S
to lead to the increased formation of β-secretase cleavage
products,1 the pursuit of a BACE inhibitor compound has been
an important area of research over the past decades as a
potential treatment for Alzheimer’s disease.2 AMG 978 (4)
emerged as a lead compound from our BACE program, with
the key step in the synthetic sequence being the
diastereoselective addition of aryl lithium compound 2 to
cyclopropyl fused isoxazoline 1, as shown in Figure 1.3 This
We envisioned that the racemization could potentially occur
through a dipolar, zwitterionic pathway (Figure 2, top),
involving heterolytic cleavage of the cyclopropane carbon−
Figure 1. Reaction and racemization of cyclopropane 1.
reaction typically proceeds in good yield producing 3 in high
diastereomeric ratio (dr) and enantiomeric excess (ee). Having
access to enantiopure cyclopropane 1 is essential for being able
to ultimately deliver 4 with the required chemical and optical
purity. However, during our process development efforts for
this reaction, we observed an erosion of ee in various lots of
AMG 978 over several batches, and, eventually, only racemic 4
Figure 2. Possible zwitterionic or diradical pathways for racemization
of 1 to rac-1.
Received: November 25, 2019
© XXXX American Chemical Society
A
DOI: 10.1021/acs.orglett.9b04236
Org. Lett. XXXX, XXX, XXX−XXX

Organic Letters
Letter
carbon bond, proceeding through transition structure zwit-
TS-5 with or without the presence of an intermediate along the
potential surface. An alternative pathway (Figure 2, bottom)
would involve homolytic opening of the cyclopropane through
a diradicaloid TS to form diradical intermediate dirad-int-6
and subsequent ring closure. Note that rearrangement of the
all-carbon bicyclo[3.1.0]hex-2-ene substrate has been studied
experimentally6 by the Baldwin and Doering groups, and, more
recently, computationally by Houk,7 establishing the diradical
pathway as the dominant mode of isomerization.
to reproduce pericyclic reaction barrier heights in recent
benchmarks11−13) to determine which functional best models
the experimental data. As discussed below, the ωB97x-D/6-
311+G(d,p) model chemistry was ultimately utilized through-
out the study. For 1, the calculated barrier for racemization is
found to diminish with increasing solvent dielectric (see Figure
3), in accord with experimental results (vide infra). Increasing
Given the overall structural similarity, yet potential
electronic differences between our fused cyclopropyl isoxazo-
lines and the parent hydrocarbon, a series of control
experiments and computational investigations were conducted.
Key experimental findings were as follows: (i) the rate of
racemization was increased at higher temperatures; (ii) the rate
of racemization was increased as the solvent dielectric constant
increased; (iii) the presence of oxygen or light had no effect on
the rate of racemization; and (iv) the inclusion of radical
inhibitors such as BHT had no effect on the rate of
racemization. While dipolar intermediates such as zwit-TS-5
are known to react with dipolarophiles,8 the addition of species
such as styrene or methyl acrylate did not allow for the
trapping of any adducts arising from dipolar cycloadditions. In
fact, throughout all of our control experiments, no side
products were observed whatsoever.
Figure 3. Computed geometric parameters (gas phase) and relative
free energies of activation at 45 °C in various solvents for the
racemization of 1 at the IEFPCM-ωB97x-D/6-311+G(d,p) level of
theory. Semitransparent atoms at the transition structure illustrate
motion along the imaginary frequency vibrational mode (see the
Supporting Information for an animated graphic).
Our initial computational imperative was to characterize the
structural and electronic nature (zwitterionic versus diradical,
presence of intermediates, etc.) of the racemization process
depicted in Figure 2. All calculations were performed using the
Gaussian 16 program.9 To begin, CASSCF(4,4)/6-31+G*
calculations with an active space analogous to the previously
studied hydrocarbon (CN π/π* and cyclopropyl-like σ/σ*)
were utilized to locate a TS for the racemization of 1. This TS
is structurally characterized by a flattening of the out-of-plane
methylene with concomitant lengthening of the breaking C−C
σ-bond, along with the equalization of the C=N and C−C ring
bonds constituting the “aza-allyl” moiety at the TS, versus the
corresponding distances in the reactant. IRC calculations
originating from this transition structure progress smoothly to
product or reactant without the presence of intermediates
along the reaction path.
solvent dielectric was observed to stabilize the formation of a
dipole at the transition state (as gauged by dipole moment
magnitude relative to reactant, see the Supporting Information
for details), as would be expected for a zwitterionic reaction
mechanism.14 The electrostatic potential-derived charges
exhibit clear differences between substrate 1 and its transition
structure for racemization (Figure 4). In particular, the buildup
In contrast to the bicyclo[3.1.0]hex-2-ene archetype, the
(4,4) CAS occupancies of the transition state for the
racemization of 1 were found to be 1.92, 1.68, 0.32, and
0.08,10 descriptive of a process possessing a small amount of
diradical character but predominantly closed-shell in nature.
Furthermore, geometry optimization and stability calculations
utilizing single-reference density functional theory (DFT)
(RωB97x-D/6-311+G(d,p); vide infra) resulted in a TS
geometry with a stable closed-shell DFT solution. Finally,
while a stable intermediate on the triplet diradical surface
corresponding to dirad-int-6 was located by DFT calculations,
this species was found to be ∼10 kcal/mol higher in energy
than the analogous, closed-shell singlet TS at the same level of
theory. Taken together, these data suggest that the
racemization of 1 is best described as proceeding through a
predominantly zwitterionic mechanism devoid of intermedi-
ates.
Figure 4. Electrostatic potential (a.u.) mapped onto the total electron
density (0.001 a.u. contour level) with selected heavy atom Merz−
Kollman atomic charges (electrons, fit to dipole moment; Debye) of
substrate and TS of 1 (ε = 1) illustrating the shift in charge
distribution. ωB97x-D/6-311+G(d,p). Image rendered with
PyMOL15
of a partial positive charge at the oxocarbenium-like oxygen
atom and increased negative charge at the terminal C and N
atoms of the aza-allyl-like moiety is readily evident.
We reasoned that the energetics of the reaction could be
modulated by varying the functional group at the position of
the difluoromethane unit present in cyclopropane 1. As
expected, electron-withdrawing groups that stabilize the
nascent accumulation of overall negative partial charge at the
aza-allyl moiety of the TS were predicted to lower the barrier
for racemization, compared to those bearing substituents that
Having established the suitability of this system for study by
single-reference methods, we then investigated the effect of
implicit solvent on the description of reaction energetics, while
also evaluating DFT functionals (selected based on the ability
B
DOI: 10.1021/acs.orglett.9b04236
Org. Lett. XXXX, XXX, XXX−XXX

Organic Letters
Letter
are more electron-donating. Utilizing the compounds shown in
Figure 5, we set out to experimentally determine ΔG⧧ values
and perform further studies on the racemization of these
compounds.16,17
calculated barrier heights are presented in Table 2, with the
correlation highlighted in Figure 6.
Table 2. Experimental and Computational ΔG⧧ Values for
a
Cyclopropane Substrates in Acetonitrile at 30 or 70 °C
Figure 5. Cyclopropane substrates for racemization studies.
We began by studying the racemization of cyclopropane 1 in
MTBE, acetone, and acetonitrile. The racemization rate is
slowest in MTBE and occurs most rapidly in acetonitrile. This
is consistent with our earlier observations and computational
predictions that the rate of racemization increases as the
solvent polarity increases. A comparison of the experimentally
determined18 and computationally predicted ΔG⧧ values for
the racemization of cyclopropane 1 in those solvents is shown
in Table 1. There is excellent agreement in the solvent trends
between the experimental and computational data.
Table 1. Experimental and Computational ΔG⧧ Values for
Racemization of Cyclopropane 1 in Various Solvents at 30
a
°C
a
Note the following footnotes cited in this table: 1ΔG⧧ values
reported in units of kcal/mol. 2IEFPCM-ωB97x-D/6-311+G(d,p);
3values shifted by −2.7 kcal/mol for observed systematic error;
a
1
Note the following footnotes cited in this table: calculated using
4measured and calculated at 30 °C.; and measured and calculated at
5
IEFPCM-ωB97x-D/6-311+G(d,p); and 2includes a correction of
−2.7 kcal/mol, to account for observed systematic mean signed error
in the DFT calculations.
70 °C.
Having explored the effects of solvent on cyclopropane 1, we
then set out to determine experimental ΔG⧧ values for each
substrate shown in Figure 5, in order to compare to the
computationally predicted value. For this study, with
acetonitrile as the solvent of choice, temperatures of either
30 °C or 70 °C were selected, depending on the stability of the
substrate. Experiments for substrates with lower barriers of
racemization were conducted at 30 °C, and those with higher
barriers were performed at 70 °C. All of the tested DFT
functionals showed good correlation with experiment, giving
free-energy R2 values of >0.95 (see the Supporting
Information) but systematically underestimated or over-
estimated the barrier height by 2−5 kcal/mol, compared to
experiment. CCSD(T)/aug-cc-pVTZ single-point calculations
on cyclopropane 1 were performed using the ωB97x-D/6-
311+G(d,p) geometries. The difference between the DFT
barrier and the CCSD(T) barrier was approximately equal to
the systematic error observed in the DFT calculations, relative
to the experiment (see the Supporting Information for further
discussion). Among the functionals tested, IEFPCM-ωB97x-
D/6-311+G(d,p) gave the best combination of relative and
absolute agreement with experiment and are used throughout
this work. (R2 = 0.96, slope = 1.10, 2.7 kcal/mol MUE, where
MUE denotes mean unsigned error). The experimental and
Figure 6. Comparison between experimental and calculated ΔG⧧
values for the eight cyclopropane substrates in acetonitrile.
The difluoromethane, benzothiazole, and pivalate substrates
(entries 1, 2, and 3, respectively, in Table 2) possess
experimentally measured ΔG⧧ values of 25.6, 26.4, and 27.8
kcal/mol at 30 °C. The (comparatively less electron-with-
drawing) methyl- and aryl-substituted compounds proved to
be significantly more stable. Thus, the racemization of these
compounds was performed at 70 °C in order to achieve
reasonable rates. The p-CN phenyl substrate exhibited the
most facile racemization of this group, with an experimentally
C
DOI: 10.1021/acs.orglett.9b04236
Org. Lett. XXXX, XXX, XXX−XXX

Organic Letters
Letter
determined ΔG⧧ value of 28.5 kcal/mol (Table 2, entry 4)
followed by the p-Cl substrate (Table 2, entry 5) at 29.0 kcal/
mol. The methyl and phenyl substrates had almost identical
stabilities, with both compounds having a measured ΔG⧧ value
of ∼29.5 kcal/mol (Table 2, entries 6 and 7). Finally, the p-
OMe phenyl substrate proved to be the most stable substrate
investigated, with an experimental racemization energy barrier
of 29.8 kcal/mol (Table 2, entry 8). All experimentally
determined ΔG⧧ values demonstrate an excellent correlation
with computationally determined ΔG⧧ values, as highlighted in
Table 2 and Figure 6.
To further support the proposed dipolar mechanism, a
Hammett plot was constructed from data collected for the p-
OMe phenyl, phenyl, p-Cl phenyl, and p-CN phenyl
substituted cyclopropane compounds, as shown in Figure
7.19 The slope of the Hammett plot is consistent with buildup
Seb Caille: 0000-0001-5434-5483
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
Stimulating discussions with Drs. Jason Tedrow and Narbe
Mardirossian (Amgen, Inc.) are gratefully acknowledged.
REFERENCES
■
(1) Vassar, R.; Bennett, B.; Babu-Khan, S.; Khan, S.; Mendiaz, E.;
Denis, P.; Teplow, D.; Ross, S.; Amarante, P.; Loeloff, R.; Luo, Y.;
Fisher, S.; Fuller, J.; Edenson, S.; Lile, J.; Jarosinski, M.; Biere, A.;
Curran, E.; Burgess, T.; Louis, J.-C.; Collins, F.; Treanor, J.; Rogers,
G.; Citron, M. Science 1999, 286, 735−741.
(2) Low, J. D.; Bartberger, M. D.; Cheng, Y.; Whittington, D.; Xue,
Q.; Wood, S.; Allen, J. R.; Minatti, A. E. Bioorg. Med. Chem. Lett.
2018, 28, 1111−1115.
(3) Uno, H.; Terakawa, T.; Suzuki, H. B. Bull. Chem. Soc. Jpn. 1993,
66, 2730−2737.
(4) (a) Burrowes, T. G.; Jackson, W. R.; Faulks, S.; Sharp, I. Aust. J.
Chem. 1977, 30, 1855−1858. (b) Sheshenev, A. E.; Baird, M. S.;
Bolesov, I. G.; Shashkov, A. S. Tetrahedron 2009, 65, 10552−10564.
(c) Nesi, R.; Giomi, D.; Papaleo, S.; Dapporto, P.; Paoli, P. J. Chem.
Soc., Chem. Commun. 1990, 23, 1675−1676. (d) Baird, M. S.; Li, X.;
Al Dulayymi, J. R.; Kurdjukov, A. I.; Pavlov, V. A. J. Chem. Soc., Perkin
Trans. 1 1993, 21, 2507−2508.
(5) Baird, M. S.; Huber, F. A. M.; Clegg, W. Tetrahedron 2001, 57,
9849−9858.
(6) (a) Baldwin, J. E.; Keliher, E. J. J. Am. Chem. Soc. 2002, 124,
380−381. (b) von E. Doering, W.; Zhang, T-h.; Schmidt, E. K. G. J.
Org. Chem. 2006, 71, 5688−5693.
Figure 7. Hammett plot for aryl-substituted cyclopropanes.
(7) (a) Suhrada, C. P.; Houk, K. N. J. Am. Chem. Soc. 2002, 124,
8796−8797. (b) Doubleday, C.; Suhrada, C. P.; Houk, K. N. J. Am.
Chem. Soc. 2006, 128, 90−94.
of a negative charge (i.e., the existence of a partial aza-allyl-like
anion moiety) in the transition state for racemization. The
highly linear nature of the plot is supportive of a consistent
mode of racemization over a range of electronically diverse
substrates.
In conclusion, we have performed both experimental and
computational studies on the unexpected racemization of
enantiopure cyclopropyl fused isoxazoline compounds. Both
the experimental and computational data are consistent with
the racemization proceeding primarily through a transition
state possessing predominantly dipolar character. The stability
of these compounds can be significantly modulated by
temperature, choice of solvent, and substituents.
(8) Partridge, K. M.; Guzei, I. A.; Yoon, T. P. Angew. Chem., Int. Ed.
2010, 49, 930−934.
(9) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.;
Robb, M. A.; Cheeseman, J. R.; Scalmani, G.; Barone, V.; Petersson,
G. A.; Nakatsuji, H.; Li, X.; Caricato, M.; Marenich, A. V.; Bloino, J.;
Janesko, B. G.; Gomperts, R.; Mennucci, B.; Hratchian, H. P.; Ortiz, J.
V.; Izmaylov, A. F.; Sonnenberg, J. L.; Williams-Young, D.; Ding, F.;
Lipparini, F.; Egidi, F.; Goings, J.; Peng, B.; Petrone, A.; Henderson,
T.; Ranasinghe, D.; Zakrzewski, V. G.; Gao, J.; Rega, N.; Zheng, G.;
Liang, W.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.;
Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Vreven, T.;
Throssell, K.; Montgomery, J. A., Jr., Peralta, J. E.; Ogliaro, F.;
Bearpark, M. J.; Heyd, J. J.; Brothers, E. N.; Kudin, K. N.; Staroverov,
V. N.; Keith, T. A.; Kobayashi, R.; Normand, J.; Raghavachari, K.;
Rendell, A. P.; Burant, J. C.; Iyengar, S. S.; Tomasi, J.; Cossi, M.;
Millam, J. M.; Klene, M.; Adamo, C.; Cammi, R.; Ochterski, J. W.;
Martin, R. L.; Morokuma, K.; Farkas, O.; Foresman, J. B.; Fox, D. J.
Gaussian 16, Revision C.01; Gaussian, Inc.: Wallingford, CT, 2016.
(10) (a) CAS occupancies at CASSCF (4,4)/6-31+G* for 2-
difluoromethyl-bicyclo[3.1.0]hex-2-ene were found to be 1.91, 1.01,
0.99, and 0.09. (b) Inclusion of oxygen lone pairs in the active space
(i.e., CASSCF (6,5) for the transition structure for racemization of 1)
does not significantly change the form of the CAS orbitals nor the
observed CAS occupancies. See the Supporting Information for an
elaborated discussion.
ASSOCIATED CONTENT
* Supporting Information
■
S
The Supporting Information is available free of charge at
https://pubs.acs.org/doi/10.1021/acs.orglett.9b04236.
Experimental procedures, characterization data, and
computational data (PDF)
AUTHOR INFORMATION
Corresponding Authors
■
(11) Goerigk, L.; Hansen, A.; Bauer, C.; Ehrlich, S.; Najibi, A.;
Grimme, S. Phys. Chem. Chem. Phys. 2017, 19, 32184−32215.
(12) Mardirossian, M.; Head-Gordon, M. Mol. Phys. 2017, 115,
2315−2372.
(13) See the Supporting Information for further details on the
computational results.
*E-mail: quasdorf@amgen.com (K. W. Quasdorf).
*E-mail: michael.bartberger@1200pharma.com (M. D. Bart-
berger).
ORCID
Michael D. Bartberger: 0000-0002-5167-3139
D
DOI: 10.1021/acs.orglett.9b04236
Org. Lett. XXXX, XXX, XXX−XXX

Organic Letters
Letter
(14) Reichardt, C.; Welton, T. Solvents and Solvent Effects in Organic
Chemistry; Wiley−VCH Verlag & Co. KGaA, Weinheim, Germany,
2011.
(15) The PyMOL Molecular Graphics System, Version 2.1.1;
̈
Schrodinger, LLC.
(16) For compounds 1, 7, and 8, absolute stereochemistry was
established from an advanced intermediate in the AMG 978 synthesis.
For compounds 9, 10, 11, 12, and 13 absolute stereochemistry was
determined by comparison to known literature compounds. See the
Supporting Information for details.
(17) Sakai, T.; Mitsutomi, H.; Korenaga, T.; Ema, T. Enantio-
enriched aryl substrates where obtained via an enzymatic resolution.
Tetrahedron: Asymmetry 2005, 16, 1535−1539.
(18) See the Supporting Information for details on the experimental
determination of ΔG⧧ values.
(19) Hansch, C.; Leo, A.; Taft, R. W. Chem. Rev. 1991, 91, 165−195.
E
DOI: 10.1021/acs.orglett.9b04236
Org. Lett. XXXX, XXX, XXX−XXX