Absolute stereochemistry of a 4 a-hydroxyriboflavin analogue of the key intermediate of the FAD-monooxygenase cycle

Article abstract of DOI:10.1002/chem.201304393

The biological action of flavoenzymes, such as flavin adenine dinucleotide (FAD)-containing monooxygenase, involves the formation of oxygenated flavin derivatives, such as 4 a-hydroperoxyflavin and 4 a-hydroxyflavin, in which a new center of chirality is created at the 4 a position during the enzymatic reactions. So far, the absolute configuration of this center of chirality in natural 4 a-oxygenated flavins has remained unknown in spite of its key importance for the diverse functions of flavoenzymes. Herein, we report the 4 a-hydroxy adduct 3 of 3-benzyl-5-ethyl-10-(tetraacetyl-D-ribityl)flavinium (1), one of the key intermediates involved in the enantioselective organocatalytic oxidation of sulfides to sulfoxides. The 4 a-hydroxyflavin diastereomers (+)-3 and (-)-3, separated by HPLC, were characterized by electronic circular dichroism (CD) spectroscopy. Their absolute configurations at the 4 a position were, for the first time, determined by comparing experimental CD spectra with those calculated by means of time-dependent density functional theory (TDDFT) on DFT-optimized structures obtained after an extensive conformation analysis. A conformational challenge: The absolute stereochemistry of 4 a-epimeric compounds (+)-1 and (-)-1, analogues of 4 a-hydroxyflavin generated during the flavin adenine dinucleotide (FAD)-monooxygenase enzymatic cycle, was established by spectroscopic techniques, extensive conformation analysis, and theoretical calculations of circular dichroism spectra (see figure).

Full text of DOI:10.1002/chem.201304393

DOI: 10.1002/chem.201304393
Full Paper
&
Stereochemistry
Absolute Stereochemistry of a 4a-Hydroxyriboflavin Analogue of
the Key Intermediate of the FAD-Monooxygenase Cycle**
Soichiro Iwahana,[a] Hiroki Iida,[a] Eiji Yashima,*[a] Gennaro Pescitelli,[b] Lorenzo Di Bari,*[b]
Ana G. Petrovic,[c, d] and Nina Berova*[c]
Dedicated to Professor Koji Nakanishi on the occasion of his 88th birthday
Abstract: The biological action of flavoenzymes, such as
flavin adenine dinucleotide (FAD)-containing monooxyge-
nase, involves the formation of oxygenated flavin deriva-
tives, such as 4a-hydroperoxyflavin and 4a-hydroxyflavin, in
which a new center of chirality is created at the 4a position
during the enzymatic reactions. So far, the absolute configu-
ration of this center of chirality in natural 4a-oxygenated fla-
vins has remained unknown in spite of its key importance
for the diverse functions of flavoenzymes. Herein, we report
the 4a-hydroxy adduct 3 of 3-benzyl-5-ethyl-10-(tetraacetyl-
d-ribityl)flavinium (1), one of the key intermediates involved
in the enantioselective organocatalytic oxidation of sulfides
to sulfoxides. The 4a-hydroxyflavin diastereomers (+)-3 and
(À)-3, separated by HPLC, were characterized by electronic
circular dichroism (CD) spectroscopy. Their absolute configu-
rations at the 4a position were, for the first time, deter-
mined by comparing experimental CD spectra with those
calculated by means of time-dependent density functional
theory (TDDFT) on DFT-optimized structures obtained after
an extensive conformation analysis.
Introduction
This process is accompanied by the formation of 4a-hydroxy-
flavin (FlOH), which simultaneously undergoes dehydration to
regenerate Flox (Scheme 1a).[2] This flavin-catalyzed oxidation
mechanism has been widely accepted based on kinetic and
spectroscopic studies on flavoenzymes,[2] but the key inter-
mediate FlOOH has not yet been isolated as a structurally de-
fined active species despite a number of flavoenzyme struc-
tures determined by X-ray analyses because it is extremely un-
stable outside of enzymes.[3]
Riboflavin (vitamin B2) and its biologically active forms, flavin
mononucleotide (FMN) and flavin adenine dinucleotide (FAD),
play a central role in the oxidative degradation of xenobiotic
substances as cofactors of flavoenzymes, such as the hepatic
microsomal FAD-containing monooxygenase (FADMO), a class
of ubiquitous redox-active enzyme.[1] The catalytic activity of
FADMO relies on the formation of the key intermediate 4a-hy-
droperoxyflavin (FlOOH), generated from the oxidized flavin
(Flox) through conversion into the reduced flavin and the sub-
sequent reaction with molecular oxygen, which promotes the
monooxygenation of substrates, such as sulfides to sulfoxides.
Inspired by such intriguing and diverse functions of flavoen-
zymes, in particular FADMO, the design and synthesis of artifi-
cial flavin-based organocatalysts has become an attractive
challenge not only to mimic and elucidate the flavin-mediated
monooxygenation processes, but also to develop environ-
ment-friendly organocatalysts.[4] Since the pioneering work by
Kemal and Bruice[1a,5] and Murahashi et al.,[6] who developed
a versatile method to generate the oxidatively active 5-ethyl-
4a-hydroperoxyflavin in situ from the corresponding flavinium
cation in the presence of hydrogen peroxide (H2O2), which pro-
motes the catalytic oxidation of sulfides and amines in a way
similar to FADMO, a variety of chiral and achiral flavin-based
organocatalysts have been synthesized for diverse transforma-
tion reactions, including the catalytic asymmetric oxidation of
sulfides and the Baeyer–Villiger oxidation of ketones.[7] Howev-
er, the key intermediates of the FlOOH type were not isolated
and merely characterized by spectroscopic techniques,[5,6] and
their molecular structures, including the stereochemistry at the
4a position, have never been determined.
[a] S. Iwahana, Dr. H. Iida, Prof. E. Yashima
Department of Molecular Design and Engineering
Graduate School of Engineering
Nagoya University, Nagoya 464-8603 (Japan)
E-mail: yashima@apchem.nagoya-u.ac.jp
[b] Dr. G. Pescitelli, Prof. L. Di Bari
Department of Chemistry
University of Pisa, Via Risorgimento 35, 56126 Pisa (Italy)
E-mail: lorenzo.dibari@unipi.it
[c] Dr. A. G. Petrovic, Prof. N. Berova
Department of Chemistry, Columbia University
3000 Broadway 3114, New York, NY 10027 (USA)
E-mail: ndb1@columbia.edu
[d] Dr. A. G. Petrovic
Department of Life Sciences, New York Institute of Technology
1855 Broadway, New York, NY 10023 (USA)
[**] FAD =flavin adenine dinucleotide.
We have recently reported an optically active riboflavin-
based organocatalyst 1 readily derived from commercially
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/chem.201304393.
Chem. Eur. J. 2014, 20, 4386 – 4395
4386
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
In the present study, we synthesized and isolated the opti-
cally active 4a-hydroxyflavin diastereomers (+)- and (À)-3 ac-
cording to the previously reported method,[8] and investigated
the effects of the stereogenic center at the 4a position (OH
group) and the tetraacetyl-d-ribityl residue on the chiroptical
properties of (+)- and (À)-3. Electronic circular dichroism (CD)
spectra were measured and compared with CD calculations
based on the time-dependent density functional theory
(TDDFT) to determine the absolute configurations at the 4a
positions of the (+)-3 and (À)-3 diastereomers.
Despite the key importance of 4a-oxygen adducts as active
intermediates in a large family of ubiquitous flavoenzymes for
diverse functions, the stereostructures of the 4a-oxygen ad-
ducts, in particular, the absolute configurations at the 4a posi-
tion of flavoenzymes and those of the synthetic analogues,
have never been unambiguously determined.[9] The present
molecular-level stereochemical information will allow us to ra-
tionalize the previously reported enantioselectivity in the
asymmetric oxidation of sulfides catalyzed by 1,[8] to provide
an important clue toward the elucidation of the mechanisms
of flavin-catalyzed reactions, and to design novel chiral flavin-
based organocatalysts with higher enantioselectivity.
Scheme 1. a) Biological catalytic cycle for the FADMO-mediated oxidation of
sulfides with molecular oxygen through the formation of FlOOH and FlOH.
b) Biomimetic catalytic cycle for the 5-alkylated flavinium perchlorate 1-
mediated oxidation of sulfides with H2O2 through the formation of 4a-hy-
droperoxy 2 and 4a-hydroxy 3 adducts. The nucleophilic addition of H2O to
1 also produces 4a-hydroxy 3. c) Diastereoselective hydration of 1 with
Me4N+OHÀ in CD3CN at À408C. The prefixes + and À of 3 denote the signs
of the optical rotation at the sodium d-line.
Results
Chromatographic separation and characterization of (+)-3
and (À)-3
The 4a-hydroxy adduct 3 was successfully separated into dia-
stereomers (+)-3 and (À)-3 by means of a chiral HPLC column
(Chiralpak IB, Daicel Co., Ltd.) with n-hexane/CHCl3 (1:1, v/v) as
the eluent or an HPLC column (Cosmosil 5C18-MS-II, Nacalai
Tesque) with acetonitrile (ACN)/H2O (11:9, v/v) as the eluent
(see Figure S1A in the Supporting Information).
available riboflavin, which also promotes the asymmetric oxi-
dation of methyl para-tolyl sulfide catalyzed by 4a-hydroperox-
yflavin 2 generated in situ from the 5-ethylflavinium cation
1 by the nucleophilic addition of H2O2, thus giving the corre-
sponding optically active (S)-methyl para-tolyl sulfoxide with
up to 30% enantiomeric excess (ee) along with 4a-hydroxyfla-
vin 3, which further undergoes dehydration to regenerate
cation 1 in the presence of HClO4 to complete the biomimetic
catalytic oxidation cycle (Scheme 1b).[8]
1
In the H NMR spectra, diastereomers (+)-3 and (À)-3 may
be easily identified in their mixtures by means of the aromatic
protons H6 and H9, which appear as singlets at d=7.03 and
7.23 ppm for (+)-3 and d=7.05 and 7.18 ppm for (À)-3, re-
spectively (see Figure S1B in the Supporting Information). In
this way, the de values of (+)-3 and (À)-3 separated as de-
scribed above were estimated to be 94% (see Figure S1C and
S1D in the Supporting Information). In view of the subsequent
theoretical conformation analysis, variable-temperature (VT)
1H NMR spectra and nuclear Overhauser effect spectroscopy
(NOESY) spectra were recorded in CD3CN to obtain information
about the preferential conformations assumed by (+)-3 and
(À)-3 in solution. In particular, we were interested in any clue
about the possible formation of intramolecular hydrogen
bonds between the 4a-OH group and one or more acetyl
groups on the tetraacetyl-d-ribityl moiety, as evidenced by pre-
liminary modeling results (see below). As expected, the 4a-OH
protons of (+)-3 and (À)-3 in CD3CN (2 mm) were monotonical-
ly shifted downfield upon heating; the chemical shifts changed
from d=4.58 ppm at 303 K to d=4.41 ppm at 343 K for both
(+)-3 and (À)-3. A linear d (in ppm) versus T (in K) relationship
was observed in both cases with a temperature coefficient
Dd/DT (linear-fit slope) of À4.3 and À4.4 ppbKÀ1 for (+)-3 and
The enantioselectivity observed in the asymmetric oxidation
was anticipated to closely link to the stereochemistry (diaste-
reoselectivity) of 4a-hydroperoxyflavin 2 generated in situ. Vari-
ous attempts to isolate 2 to determine the structure and esti-
mate its diastereomeric excess (de) were unsuccessful because
of its labile characteristics in solution. The 4a-hydroxyflavin
compound 3 was also produced from 1 through a similar nu-
cleophilic addition of H2O (Scheme 1B) and was quite stable
while maintaining its absolute configuration at the 4a position,
which was identical to that of 2. On the basis of this assump-
tion, we performed the hydroxylation of 1 at the 4a position
in the presence of Me4N+OHÀ in CD3CN at À408C and found
that this reaction proceeded in a diastereoselective fashion,
thus yielding a diastereomeric mixture of 3 (64.5:35.5; 29% de)
1
as estimated from its H NMR spectrum (Scheme 1c).[8]
Chem. Eur. J. 2014, 20, 4386 – 4395
www.chemeurj.org
4387
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
(À)-3, respectively (see Figure S2A and S2B in the Supporting
Information). These data exclude the possibility that the 4a
-OH protons of (+)-3 and (À)-3 are involved in any intramolec-
ular hydrogen bonds.[10] This assumption was supported by the
facts that tert-butanol (40 mm in CD3CN) gave a similar coeffi-
cient of À4.8 ppbKÀ1 (see Figure S2C in the Supporting Infor-
mation) and no NOE correlations could be detected between
the protons that lie on the 4a-hydroxyflavin moiety and any of
the acetyl protons of 3.[8] As we shall discuss more in detail
below, such intramolecular hydrogen bonds would involve the
formation of a medium-size ring, which is disfavored in itself
and becomes even less likely on account of the polarity and
the fairly good hydrogen-bond-acceptor character of the sol-
vent ACN (b=0.40).[11]
of (À)-3 to highlight possible aggregation phenomena. Howev-
er, the magnitudes of the CD and absorption signals of (À)-3
(in mÀ1 cmÀ1) were hardly changed over the concentration
range from 0.0125 to 0.5 mm in ACN, thus indicating that the
formation of aggregates could be excluded (see Figure S3 in
the Supporting Information). With hope of detecting a chiropti-
cal property more sensitive to the diasteromeric nature of the
compounds, we attempted to measure vibrational circular di-
chroism (VCD) spectra of (+)-3 and (À)-3 in ACN. However, no
diagnostic signals could be detected in the IR regions of the
4a-hydroxyflavin residue.
Molecular modeling study of (S)-3
The absorption and electronic CD spectra of (+)-3 and (À)-3
were measured in ACN to investigate the effects of the chirali-
ty at the 4a position on their chiroptical properties (Figure 1).
Following a common protocol for the simulation of CD spectra
of flexible organic molecules,[12] we first carried out an exhaus-
tive molecular-modeling investigation to establish the most
probable solution conformations of 3, which were to be used
as input structures in the CD calculations. This protocol first
encompasses a comprehensive conformational search run with
molecular-mechanics calculations, followed by DFT geometry
optimizations of the various identified conformers.
Structures with an arbitrarily chosen configuration (4aS)
were considered in the modeling. The starting structure of
(S)-3 was built based on the combination of two reported X-
ray crystallographic structures, one with coordinates provided
for the 5-ethyl-4a-hydroxyflavin segment of 5-ethyl-4a-hy-
droxy-3-methyl-lumiflavin and the second with coordinates
taken from the tetraacetyl-d-ribityl segment of riboflavin tet-
raacetate[13] by addition of a benzyl moiety at N3. This starting
geometry of (S)-3 was subjected to a series of Monte Carlo
(MC) conformational search algorithms in vacuo (see the Com-
putational Section). The MC search provided a pool of approxi-
mately 3000 conformers, which was narrowed down to 25 dis-
tinct representative structures based on the redundant-confor-
mer elimination method (see the Computational Section for
details). Among the representative 25 structures, 12 conform-
ers (henceforth indicated as “closed” structures, (S)-3C) showed
intramolecular hydrogen bonding between the 4a-OH moiety
and one of the C=O groups attached to the d-ribityl unit at
positions 2’, 4’, or 5’. The remaining 13 conformers showed no
intramolecular hydrogen bonding (indicated as “open” struc-
tures; i.e.,
Figure 1. a,b) CD and c,d) absorption spectra of (+)-3 (a,c) and (À)-3 (b,d) in
ACN at approximately 258C. The concentrations of (+)- and (À)-3 were
0.5 mm and their de values were 94%.
The absorption spectra of the two diastereomers were almost
identical and showed a manifold of bands in the region of l=
200–450 nm allied with the 4a-hydroxyflavin and benzene
chromophores. In the CD spectra, (+)-3 and (À)-3 (both with
94% de) showed intense bands of alternating sign centered at
around l=348, 285, and 240 nm and weaker bands at shorter
wavelengths. Very interestingly, the two diastereomers were
associated with almost mirror-image CD patterns. The mirror-
image relationship was not perfect, as evidenced by slightly
different absolute values of the CD intensities and by the ob-
servation that the two spectra do not cross exactly at zero
values. Still, the spectra clearly suggested that the chiroptical
properties of the diastereomers are significantly governed by
the chirality at the 4a position and that the contribution from
the tetraacetyl-d-ribityl group is negligibly small. In other
words, the two diastereomers behave almost as pseudoenan-
tiomers from the viewpoint of their electronic CD spectra. We
measured concentration-dependent absorption and CD spectra
(S)-3O), and the majority of these conformers had the 4a-OH
group and the tetraacetyl-d-ribityl side chain on opposite sides
of the flavin ring. The 25 unique conformations identified by
the MC search were submitted to geometry optimization and
frequency calculation by using DFT methods at the B3LYP/6-
31G(d) level of theory in the gas phase. Surprisingly, DFT-opti-
mized hydrogen-bonded (S)-3C conformers were much more
stable than the (S)-3O conformers. Based on the relative free
energies, the overall Boltzmann population at 300 K for the hy-
drogen-bonded (S)-3C conformers was as high as 94.4% (see
Table S1 in the Supporting Information for a list of the calculat-
ed free energies and populations). To ascertain possible errors
related to the employed functional and basis set, the two most
stable structures belonging to the “open” and “closed” families
Chem. Eur. J. 2014, 20, 4386 – 4395
www.chemeurj.org
4388
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
(indicated as C12O and C2C in Table S1, respectively) were reop-
timized at the M06-2X/6-311+G(d,p) level. The reoptimization
did not substantially change the free-energy difference be-
tween the two structures (see Table S2 in the Supporting Infor-
mation). This finding that the most stable conformations calcu-
lated for (S)-3 possessed an intramolecular hydrogen bond not
only contradicted the NMR spectroscopic data discussed
above, but was also hard to reconcile with the experimental
CD data. In fact, any strong interaction between the tetraace-
tyl-d-ribityl and 4a-hydroxyflavin moieties would be expected
to produce sizable differences in the CD profiles for the two
diastereomers, which was not the case.
Next, we explored the possibility that the theoretically over-
predicted stability of the hydrogen-bonded (S)-3C conformers
could be a computational artifact due to the absence of a sol-
vent environment. Therefore, we first attempted to use an im-
plicit solvent model by running a single-point-energy calcula-
tion on the pool of 25 conformers at the M06-2X/6-311+
G(d,p) level of theory by employing the integral equation for-
malism polarizable continuum model (IEF-PCM) for ACN.[14]
Once again, however, the overall population calculated with
M06-2X/6-311+G(d,p)[IEF-PCM]//B3LYP/6-31G(d) for the family
of “closed” (S)-3C conformers was much higher than expected
(91.3%; see Table 1). Apparently, the problem is that the con-
Table 1. Single-point M06-2X/6-311+G(d,p) energy calculations on (S)-3
obtained from an implicit solvent model (IEF-PCM) of ACN (geometry opti-
mized at the B3LYP/6-31G(d) level of theory).
Figure 2. The three starting conformations of (S)-3·ACN that underwent
a MC search with an explicit solvent model for ACN. For clarity of represen-
tation, hydrogen atoms have been omitted from the tetraacetyl-d-ribityl
backbone. A color version of this figure is given in the Supporting Informa-
tion.
Conformer[a] DE[b]
Population Conformer[a] DE[b]
[kcalmolÀ1] [%]
Population
[kcalmolÀ1] [%]
C1C
C2C
C3O
C4C
C5C
C6O
C7O
C8C
C9O
C10O
C11O
C12O
C13C
7.5
0.0
C14O
C15C
C16O
C17O
C18C
C19O
C20O
C21C
C22C
C23O
C24C
C25C
3.9
2.1
8.5
3.4
0.9
1.8
5.7
1.1
0.0
7.7
2.8
0.9
0.1
1.6
0.0
0.2
12.8
2.8
0.0
8.4
55.7
0.0
0.5
17.7 0.0
7.6
6.3
5.9
2.3
2.4
4.3
3.5
5.0
1.9
2.2
3.0
0.0
0.0
0.0
1.2
0.9
0.0
0.1
0.0
2.2
1.3
0.3
13 non-hydrogen-bonded (S)-3O “open” conformers and the 12
hydrogen-bonded (S)-3C “closed” conformers obtained from
the IEF-PCM calculations (Table 1), the two most stable con-
formers that belong to each family, namely C19O and C22C in
Table 1, were selected as starting structures for subsequent ex-
plicit solvent MC searches. By starting from the non-hydrogen-
bonded geometry C19O, a structure for the adduct (S)-3O·ACN
was obtained by placing one ACN molecule in the vicinity of
the 4a-OH group with a 4a-OH···NCCH3 distance of 2.1 ꢁ
(Figure 2). From the hydrogen-bonded geometry C22C, two
starting structures were generated by positioning ACN at two
different orientations designated as (S)-3C·ACN/A and
(S)-3C·ACN/B (Figure 2). After an iterative MC search, a total of
147 conformations resulted within an energy window of
20 kJmolÀ1. More specifically, the initially “open” (S)-3O·ACN re-
sulted in 37 structures, most of which showed a hydrogen-
bonding interaction between the 4a-OH group and ACN. A
few conformers lacked any inter- and intramolecular hydrogen
bonds, whereas only one conformer showed an intramolecular
OH···3’-Ac hydrogen bond.
11.9
[a] o=“open” conformation with no intramolecular hydrogen bonding, c=
“closed” conformation with intramolecular hydrogen bonding. [b] Internal
energy relative to the lowest-energy conformer.
tinuum solvation model accounts for bulk solvent effects, but
is unsuited to describe the ability of the solvent to directly par-
ticipate as a hydrogen-bond acceptor/donor, thus leading to
a realistic competition between intra- and intermolecular hy-
drogen bonds.[14] A better description of such a kind of short-
range solute–solvent interaction might be achieved by inclu-
sion of explicit solvent molecules, although one may expect
that entropy (conformational versus solvent release/engage-
ment) would also play a role in this balance.
Among the 50 output structures obtained from the starting
“closed” geometry (S)-3C·ACN/A, most displayed an intermolec-
ular 4a-OH···ACN hydrogen bond, whereas some exhibited an
intramolecular 4a-OH···2’-Ac hydrogen bond. On the other
Therefore, further molecular modeling was performed by in-
cluding one ACN molecule in the input structures. Among the
Chem. Eur. J. 2014, 20, 4386 – 4395
www.chemeurj.org
4389
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Figure 3. Nine representative conformers with an explicit ACN, obtained from the MC pool of 147, selected based on hydrogen-bonding structural criteria to
undergo an initial geometry optimization at the DFT/M06-2X/6-311+G(d,p) level of theory. In addition, the mutual disposition of the OH, ribityl, and benzyl
moieties was taken into account. For clarity of representation, the hydrogen atoms have been omitted from the tetraacetyl-d-ribityl backbone. A color version
of this figure is given in the Supporting Information.
hand, among the 60 output structures from the starting
“closed” geometry (S)-3C·ACN/B, most retained their intramo-
lecular 4a-OH···4’-Ac hydrogen bonds.
Because it would be impractical to submit the whole ensem-
ble of 147 conformers generated for the (S)-3·ACN adduct to
DFT optimizations, we analyzed these MC conformations in
order of increasing energy and selected a few structures that
represented the various possible hydrogen-bonding interac-
tions of the 4a-OH group with ACN or an acetyl group (2’-Ac
or 4’-Ac), as well as the ones without any hydrogen bonding at
all. Additional attention was paid to the relative disposition of
the 4a-OH and tetraacetyl-d-ribityl groups and to the orienta-
tion of the benzyl group. As a result, we obtained nine repre-
sentative conformers for the (S)-3·ACN adducts displayed in
Figure 3, which were geometry optimized at the M06-2X/6-
311+G(d,p) level to result in the energies listed in Table 2. In-
Chem. Eur. J. 2014, 20, 4386 – 4395
www.chemeurj.org
4390
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
3O·ACN(8), which cover all the relevant possible hydrogen
bonding situations and orientations of the various groups. By
considering the chemical structure of (S)-3 and its experimen-
tal CD data, it was expected that the conformation of the 4a-
hydroxyflavin moiety would be of key importance in determin-
ing the calculated CD spectra, which would in turn largely
depend on the presence or absence of an intramolecular hy-
drogen bond between the 4a-OH moiety and one of the
acetyl groups. In fact, the three structures (S)-3O·ACN(1),
(S)-3O·ACN(3), and (S)-3O·ACN(8) with no intramolecular hydro-
gen bonds show an essentially identical puckering of the 4a-
hydroxyflavin skeleton, whose three rings lie almost coplanar
to each other (see Figure 3). Conversely, the two conformers
(S)-3C·ACN(4) and (S)-3C·ACN(5), both with 4a-OH···2’-Ac intramo-
lecular hydrogen bond, feature a more distorted 4a-hydroxy-
flavin moiety.
Table 2. Comparison of the electronic energy for the nine representative
conformers shown in Figure 3 (geometry optimized at the M06-2X/6-
311+G(d,p) level of theory with explicit solvent model for ACN).
Conformer[a]
3O·ACN(1) 3O·ACN(2) 3O·ACN(3) 3C·ACN(4) 3C·ACN(5)
0.00 5.28 1.62 2.33 4.50
DE[b] [kcalmolÀ1
]
]
Conformer[a]
DE[b] [kcalmolÀ1
3C·ACN(6) 3C·ACN(7) 3O·ACN(8) 3O·ACN(9)
6.90 4.53 1.83 7.77
[a] o=“open” conformation with no intramolecular hydrogen bonding,
c=“closed” conformation with intramolecular hydrogen bonding. [b] In-
ternal energy relative to the first listed conformer.
terestingly enough, the most stable structure was an “open”
conformer, namely (S)-3O·ACN(1), which lacked any intramolecu-
lar hydrogen bonding with the acetyl groups and exhibited in-
stead an intermolecular 4a-OH···ACN hydrogen bond with an
H···N distance of 1.87 ꢁ. In this structure, the tetraacetyl-d-ribi-
tyl chain and the benzyl group were both oriented on the
same side as the 4a-OH group. The second most stable struc-
ture was (S)-3O·ACN(3), which also displayed intermolecular 4a-
OH···ACN hydrogen bonding, yet with an opposite orientation
of the 4a-OH and tetraacetyl-d-ribityl groups. Notably, in both
these structures the benzyl moiety was in close vicinity to the
methyl group of the ACN molecule, perhaps due to a stabilizing
non-bonding interaction between the benzyl and methyl
groups. The third structure in order of stability was
(S)-3O·ACN(8), without any hydrogen bonding. Among the
“closed” conformers, the most stable structure was (S)-
3C·ACN(4), with an intramolecular 4a-OH···2’-Ac hydrogen bond,
which was higher in energy by 2.33 kcalmolÀ1 than (S)-
3O·ACN(1). Although the set of nine selected structures does
not necessarily cover the whole conformational ensemble, it
clearly indicated a strong preference for structures devoid of
intramolecular hydrogen bonds, which was eventually in
accord with the experimental results.
The CD spectra calculated at the CAM-B3LYP/TZVP level for
the five structures are shown in Figure 4. The three structures
(S)-3O·ACN(1), (S)-3O·ACN(3), and (S)-3O·ACN(8) gave similar calcu-
TDDFT CD calculations on representative structures of (S)-3
The selection of the best functional/basis-set combination for
the TDDFT calculations was based on a preliminary screening
run on an essential molecular model that represented the 4a-
hydroxyflavin chromophore (see the Computational Section).
On this model, optimized at the B3LYP/6-31G(d) level, the
hybrid functionals B3LYP CAM-B3LYP and M06-2X and the
basis sets TZVP, ma-TZVP, and aug-TZVP were tested (see the
Computational Section for details). All the combinations gave
consistent results. The best functional in terms of agreement
with the experimental absorption and CD spectra was CAM-
B3LYP; the basis set with the best cost/efficiency compromise
was ma-TZVP. The combination CAM-B3LYP/TZVP was selected
for the calculations on the relatively large structures of
(S)-3·ACN adducts, whereas the larger basis set ma-TZVP was
employed for the simplified model (S)-4 discussed below.
TDDFT calculations were run on the five representative
stable structures of (S)-3·ACN discussed above, namely
(S)-3O·ACN(1), (S)-3O·ACN(3), (S)-3C·ACN(4), (S)-3C·ACN(5), and (S)-
Figure 4. The TDDFT-calculated CD spectra at the CAM-B3LYP/TZVP level for
the five representative structures of the (S)-3·ACN adducts (the geometry
was optimized at the M06-2X/6-311+G(d,p) level; see the structures in
Figure 3). The rotational strengths computed in the dipole-length gauge
were each associated with a Gaussian band-shape with an exponential half-
width of 0.4 eV.
Chem. Eur. J. 2014, 20, 4386 – 4395
www.chemeurj.org
4391
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
lated CD spectra (Figure 4a). The CD profile calculated for the
lowest-energy structure (S)-3O·ACN(1) is in especially good
agreement with the experimental CD spectrum for the (À)-3
isomer (Figure 1b) in terms of sign, relative intensities, and po-
sitions of the three major bands, despite a systematic wave-
length blue-shift of about Dl=40 nm; a direct comparison, in-
cluding absorption spectra, is shown in Figure S4 (see the Sup-
porting Information). On the contrary, the CD spectra calculat-
ed for the two structures (S)-3C·ACN(4) and (S)-3C·ACN(5) deviate
substantially from the experimental CD profiles (Figure 4b).
These results clearly indicated that the conformation assumed
by the 4a-hydroxyflavin moiety is a major determinant of the
observed and calculated CD spectra. Furthermore, we also con-
sidered two simplified structures obtained from the lowest-
energy geometry (S)-3O·ACN(1) by subsequent removal of the
tetraacetyl-d-ribityl moiety and the ACN molecule. In both
cases, the calculated CD spectra were almost superimposable
to that of the original (S)-3O·ACN(1) (see Figure S5 in the Sup-
porting Information). This observation demonstrated that both
the tetraacetyl-d-ribityl moiety and the ACN molecule have no
direct impact on the excited-state calculations. Finally, we no-
ticed that the CD spectra calculated for (S)-3O·ACN(1) and the
other representative structures (S)-3O·ACN(3) and (S)-3O·ACN(8)
reproduced well the sequence of signs of CD bands observed
for the (À)-3 isomer (compare Figures 1 and 4 or see Figure S4
in the Supporting Information). Therefore, we conclude that
the absolute configurations at the 4a position of the diastereo-
meric 4a-hydroxyflavins (+)- and (À)-3 should be assigned as
the R and S isomers, respectively.
Figure 5. The TDDFT-calculated CD spectrum at the CAM-B3LYP/ma-TZVP
level on the model compound (S)-4 (see inset) obtained as a Boltzmann-
weighted average at 300 K for the two minima optimized at the B3LYP/6-
31G(d) level. The rotational strengths computed in the dipole-length gauge
were each associated with a Gaussian band-shape with an exponential half-
width of 0.3 eV.
resulted for (S)-4 (see Figure S6 in the Supporting Information).
These structures differed only in the orientation of the benzyl
group, whereas the conformation assumed by the 4a-hydroxy-
flavin ring was almost planar, similar to the lowest-energy
structure (S)-3O·ACN(1) found for the (S)-3·ACN adduct. The CD
spectra calculated at the CAM-B3LYP/ma-TZVP level for the
two conformations were very similar to each other. The Boltz-
mann-weighted average of these conformers at 300 K
(Figure 5) revealed a very close similarity to the CD spectrum
calculated for (S)-3O·ACN(1) (Figure 4) and to that observed for
(À)-3 (Figure 1; a direct comparison is shown in Figure S7 in
see the Supporting Information). In practice, consideration of
model (S)-4 would lead to the same configurational assign-
ment reached above, but with a much reduced computational
effort.
TDDFT CD calculations on model (S)-4
The whole set of the experimental data for (+)-3 and (À)-3
and the theoretical data collected for (S)-3 presented above
clearly demonstrated a hypothesis, which we anticipated on
the basis of stereochemical and spectroscopic reasoning; that
is, the CD spectra of these compounds are not affected by the
tetraacetyl-d-ribityl moiety attached at the N10 position. First
of all, because the tetraacetyl-d-ribityl chain does not contain
any strong chromophoric group, it cannot perturb the transi-
tions of the 4a-hydroxyflavin chromophore. Second, the tetra-
acetyl-d-ribityl chain is unlikely to impact the conformation as-
sumed by the 4a-hydroxyflavin ring because any specific inter-
action (for example, an intramolecular hydrogen bond) would
lead to a thermodynamically unstable medium-size cycle (see
the discussion below). As a consequence, the hypothetical
compound (S)-4 (Figure 5), in which the tetraacetyl-d-ribityl
group has been cut away and replaced by a “capping” methyl
group, should represent a valid candidate model to interpret
the stereochemical information contained in the CD spectra of
(+)-3 and (À)-3. Because most of the molecular flexibility of
(+)-3 and (À)-3 is confined to the tetraacetyl-d-ribityl side
chain, the model dramatically simplified the molecular-model-
ing and CD-calculation steps.[15] In fact, as a result of MMFF
conformational searches followed by B3LYP/6-31G(d) geometry
optimizations in vacuo (see the Computational Section for de-
tails), only two populated conformations at room temperature
On this model, it was also easier to assign the main ob-
served CD bands by means of orbital and population analysis.
It turned out that the two major bands in the region of l=
260–400 nm are allied to two p–p* transitions both localized
on the 4a-hydroxyflavin chromophore and endowed with
some charge-transfer character from the benzene ring to the
uracil-like ring.
Discussion
Although the main target of our investigation was the abso-
lute configurational assignment of 4a-hydroxyflavin diastereo-
mers 3 mentioned above, there are some important additional
aspects that we would like to discuss shortly in the present
section.
The foremost problem concerns the evaluation of the most
representative conformers that one should consider for pre-
dicting CD or other chiroptical properties by means of quan-
tum-mechanics methods such as TDDFT. As mentioned earlier,
Chem. Eur. J. 2014, 20, 4386 – 4395
www.chemeurj.org
4392
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
preliminary computational data obtained by using the intact
molecule (S)-3 in the gas phase followed by an implicit PCM
solvent model resulted in the presence of an intramolecular
hydrogen bond between the 4a-OH moiety and one of the
acetyl groups. This outcome was at odds with simple chemical
reasoning and the experimental VT-NMR and CD spectroscopic
data in ACN. More specifically, an intramolecular hydrogen
bond between the 4a-OH and the acetyl groups at the 2’ or 4’
position of the tetraacetyl-d-ribityl side chain leads to the for-
mation of 10- or 12-membered cycles, respectively, which are
energetically disfavored medium-size rings. The conformational
entropy associated with these ring-closure processes may not
be entirely captured by frequency calculations run on single
structures because a correct evaluation would require evalua-
tion of a large statistical ensemble. Moreover, the formation of
the hydrogen-bonded cycles also forces the 4a-hydroxyflavin
moiety to adopt a distorted “unnatural” conformation (see, for
example, structures for (S)-3C·ACN(4) and (S)-3C·ACN(5) in
Figure 3), which is likely to occur only in the artificial environ-
ment of computational simulations. This apparent computa-
tional artifact was not unveiled even with the use of a new-
generation functional M06-2X, which is thought to be more ac-
curate than the classical B3LYP in describing hydrogen-bond-
ing interactions.[16] As implied earlier, although an implicit PCM
solvent model is adequate for describing bulk solvent ef-
fects,[17] it may be ineffective when specific short-range solute–
solvent interactions have a relevant impact.[18] In the present
investigation, the apparent stabilization of (S)-3C conformers
held together by intramolecular hydrogen bonds warned
against the inadequacy of the continuum solvation model and
called for further computational analysis. In fact, only the use
of an explicit solvent model resolved the discrepancy associat-
ed with intramolecular hydrogen bonding by favoring intermo-
lecular hydrogen bonding between the 4a-OH group and
ACN.
tra. In this approach, the molecule under investigation is
“pruned” at each “branch”, which is envisaged not to affect in
any sizable way the observed chiroptical response, and the CD
prediction is then run on the simplified truncated model. The
method is particularly well suited for electronic CD spectrosco-
py, which responds primarily to chromophores and their imme-
diate environment and is less sensitive to spectroscopically
silent molecular portions.[19] The truncated analogue 4 incorpo-
rates the prior knowledge that intramolecular hydrogen bond-
ing does not exist and gives rise to a straightforward confor-
mation analysis, which is focused on the spectroscopically rele-
vant molecular portion. The use of a truncated model may
greatly simplify the analysis, but it represents a double-edged
sword because it is based on the assumption, which needs to
be verified, that the “pruned branches” do not contribute to
the CD spectrum in any way, either directly, by perturbing
chromophore transitions, or indirectly, by impacting on its
structure. An example that makes evident the caution to be
paid in the use of truncated models is the study on gymno-
cin B, in which the attempt to truncate a molecular “tail” led to
erroneous conformational results.[20] On the other hand, when
it is successful,[21] this double-edged sword allows one to
doubly reduce computational effort, that is, for the conforma-
tion analysis and for the excited-state calculations necessary to
predict electronic CD. In the current case, the use of truncated
model (S)-4 would apparently provide the right answer for the
right reason, that is, the neglect of any effect exerted by the
tetraacetyl-d-ribityl side chain on the 4a-hydroxyflavin core.
Conclusion
Determination of the absolute configuration of an organic
compound by means of quantum-mechanical calculations of
chiroptical spectra, such as electronic CD spectroscopy, neces-
sarily relies on the detailed knowledge of a molecular geome-
try. Because CD spectroscopic measurements are most often
run in solution, the resulting conformational ensemble ob-
served in the presence of a particular solvent must be known
with accuracy. Herein, we have reported a thorough CD study
on the 4a-hydroxyflavin derivative 3 of flavin analogue 1,
which was previously obtained as a diastereomeric mixture
and separated into diastereomers (+)-3 and (À)-3 by HPLC.
The molecular structure of 3 includes an intrinsically chiral
chromophore, the 4a-hydroxyflavin, attached at the N10 posi-
tion to a tetraacetyl-d-ribityl substituent, which itself is chiral,
non-chromophoric, and very flexible. The collection of experi-
mental and computational data demonstrated that the tetraa-
cetyl-d-ribityl chain does not affect the conformation and elec-
tronic transitions of the 4a-hydroxyflavin chromophore to any
extent. We have also demonstrated that two alternative ap-
proaches could be used to reproduce the CD spectrum of
(À)-3 by TDDFT calculations and to assign its configuration. In
the first approach, a thorough conformation analysis was run
on the whole molecular structure. Importantly, the results of
the conformation analysis in the gas phase or implicit solvent
environment were at odds with the experimental observations.
Further analysis by including one explicit solvent molecule
A discussion worth mentioning concerns the compromise
between completeness and simplicity, which is well represent-
ed herein by the problem of solvation. In this respect, we have
considered that a truly complete computational model would
consist of the solute surrounded by a certain number of sol-
vent molecules, yet this protocol is still out of the reach of
most ab initio calculations as far as excited states are con-
cerned. Luckily, in the present case, a single explicit solvent
molecule reconciled the difference between the molecule in
the gas phase (or in polarizable continuum) and in the real sol-
vent. In many cases, as in the present one, electronic CD spec-
troscopic analysis may offer itself a proof of internal consisten-
cy because the CD spectrum consists of several bands; there-
fore, the experimental and calculated absorption and CD spec-
tra should be compared in terms of global appearance, thus
not limited to the sign of the first or few CD bands. In fact, the
CD spectra calculated for the two “suspicious” structures
(S)-3C·ACN(4) and (S)-3C·ACN(5) (Figure 4B) had overall profiles
very different from the experimental one (i.e., the CD spectrum
of (À)-3 in Figure 1).
The simplification from compound 3 to 4 corresponds to
a so-called “truncation approach” for the analysis of CD spec-
Chem. Eur. J. 2014, 20, 4386 – 4395
www.chemeurj.org
4393
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
calculations were carried out under an implicit (IEF-PCM) ACN sol-
vent environment, with Gaussian09 default parameters. Due to the
size of the system and the level of the theory, we used a more
cost-effective approach by submitting all 25 conformers from the
M06–2X/6–311+G(d,p) geometry optimization in the gas phase to
a single-point energy calculation, rather than the full geometry op-
timization. The calculation was performed under an implicit ac-
count for the solvent presence by using the IEF-PCM method.
(i.e., ACN) in the calculations proved to be necessary to obtain
good agreement between the theoretical and experimental CD
spectroscopic data. In the second approach, the tetraacetyl-d-
ribityl chain was truncated and the simplified model 4 was
considered. In both cases, calculated CD spectra by using
TDDFT led to a reliable assignment of the absolute configura-
tions of (+)-3 and (À)-3 as (R)-3 and (S)-3 isomers, respectively.
The present configurational assignment of a 4a-hydroxyfla-
vin may be valid for the analogue 4a-hydroperoxyflavin, the
key intermediate responsible for diverse transformation reac-
tions catalyzed by a large family of flavoenzymes and synthetic
flavin derivatives. The results will not only provide new insight
into the molecular basis mechanism of flavin-catalyzed reac-
tions but also open possibilities for the design and synthesis of
novel flavin-based asymmetric organocatalysts.
Excited-state calculations: The TDDFT calculations were run with
Gaussian09 with default grids and without any implicit solvent
model. Several hybrid DFT functionals (B3LYP, CAM-B3LYP, M06–2X)
and basis sets (TZVP, ma-TZVP, aug-TZVP)[24] were initially tested on
an essential molecular model that represented the 4a-hydroxyfla-
vin chromophore, which was obtained by replacing both tetraace-
tyl-d-ribityl and benzyl groups in (S)-3 with two methyl groups.
The two augmented basis sets were constructed by adding to the
standard Ahlrichs TZVP basis set, a set of (1p1d) diffuse functions
(ma-TZVP)[25] or a set of (1s1p1d/1s1p) diffuse functions taken from
the most diffuse functions of aug-cc-pVDZ (aug-TZVP),[26] respec-
tively. The final calculations were run with CAM-B3LYP/TZVP on (S)-
3·ACN adducts, including 48 excited states, and with CAM-B3LYP/
ma-TZVP on the truncated model (S)-4, including 36 excited states.
Electronic CD spectra were generated with the program Specdis[27]
by applying a Gaussian band shape with an exponential half-width
of 0.3–0.4 eV by using dipole-length rotational strengths; the dif-
ference in the dipole–velocity values was checked to be minimal
for all the relevant transitions.
Experimental Section
Separation of (+)-3 and (À)-3
The 4a-hydroxyflavin 3 was separated into diastereomers (+)-3
and (À)-3 on a chiral HPLC column (Chiralpak IB, 2 (i.d.)ꢂ25 cm, n-
hexane/CHCl3 =1:1 (v/v), 5.5 mLminÀ1) or an HPLC column (Cosmo-
sil 5C18-MS-II, 2.8 (i.d.)ꢂ25 cm, ACN/H2O=11:9 (v/v), 12 mLminÀ1).
The prefixes + and À denote the signs of the optical rotation at
the sodium d-line. The 1H NMR spectra of 3 showed two sets of
peaks attributed to the diastereomeric pair (+)-3 and (À)-3, and
the diastereomeric excess values of (+)-3 and (À)-3 were estimated
to be 94% de (see Figures S1B–D in the Supporting Information).
The molar ellipticity De values at the first, second, and third Cotton
effects were De=9.55, À13.8, and 14.25 (l=348, 284, and
240 nm), respectively, for (+)-3 and De=À11.07, 12.94, and À17.00
(l=346, 285, and 241 nm), respectively, for (À)-3 (Figure 1).
Acknowledgements
The authors thank Prof. Koji Nakanishi for his helpful discussion
and suggestions. This research was supported in part by
a Grant-in-Aid for Young Scientists (B) (H.I.) from the Japan So-
ciety for the Promotion of Science (JSPS) and in part by the
Italian Ministero dell’Istruzione, dell’Universitꢃ e della Ricerca
(MIUR), “Progetto PRIN 2009 prot. 2009PRAM8L”. S.I. is grateful
for a JSPS Research Fellowship for Young Scientists (no. 4919),
the Institutional Program for Young Researcher Overseas Visits
from JSPS, and to Prof. K. Nakanishi for his hospitality during
his stay at Columbia University during 2012–2013. N.B. appreci-
ates the instrumental and technical support of JASCO Corp.,
(Japan) and JASCO, Inc. (USA).
Computational Section
Molecular modeling: The initial survey of the potential-energy sur-
face (PES) was performed by using a molecular-mechanics Monte
Carlo (MC) algorithm in the gas phase and employing three confor-
mational search methods within the MacroModel applet of Schro-
dinger software:[22] 1) Monte Carlo multiple minimum (MCMM),
2) systematic pseudo-Monte Carlo (SPMC), and 3) mixed torsional/
Larges-scale low-mode sampling. Each search was set to 10000
steps for (S)-3 treated in vacuo (gas phase). The minimization
method used was the polar-Ribiere conjugate gradient (PRCG),
with 1000 iteration steps to reach the minimum. All three of the
MC searches were performed at both OPLS-2005 and MMFFs force
fields to ensure a thorough PES search, regardless of the parame-
terization. The search was considered to be converged, and hence
complete, once iterative MC submission provided no further lower-
energy conformations. Independently, each of the six MC searches
resulted in approximately 3000 conformers. The redundant con-
formers from the six composite searches were eliminated on this
basis: 1) a distance cutoff of 7 ꢁ for all atoms and 2) an energy
window of 20 kJmolÀ1. As a result of the redundant-conformer
elimination, 25 conformers were identified. DFT calculations were
run with the Gaussian09 program[23] with default grids and conver-
gence criteria. DFT geometry optimization and frequency calcula-
tions were performed under the following two levels of theory:[24]
B3LYP/6–31G(d), M06–2X/6–311+G(d,p) geometry optimization in
the gas phase and M06–2X/6–311+G(d,p) geometry optimization
with explicit ACN. The M06–2X/6–311+G(d,p) single-point energy
Keywords: circular dichroism
·
conformation analysis
·
hydrogen bonds · solvent effects · structure elucidation
[1] a) T. C. Bruice, Acc. Chem. Res. 1980, 13, 256–262; b) V. Massey, Biochem.
Soc. Trans. 2000, 28, 283–296; c) F. Mꢄller, Chemistry and Biochemistry
of Flavoenzymes, CRC Press, Boston, 1991; d) C. Walsh, Acc. Chem. Res.
1980, 13, 148–155.
[2] a) N. B. Beaty, D. P. Ballou, J. Biol. Chem. 1981, 256, 4619–4625; b) J. R.
Cashman, Chem. Res. Toxicol. 1995, 8, 165–181; c) L. L. Poulsen, D. M.
Ziegler, J. Biol. Chem. 1979, 254, 6449–6455; d) D. M. Ziegler, Drug
Metab. Rev. 2002, 34, 503–511; e) K. C. Jones, D. P. Ballou, J. Biol. Chem.
1986, 261, 2553–2559.
[3] The half-life of FlOOH has been estimated to be as low as 2.5 ms:
a) R. F. Anderson, in Flavins and Flavoproteins (Eds.: R. C. Bray, P. C. Engel
and S. E. Mayhew), DeGruyter, Berlin, 1984, ch. 57–60. Recently, 4a–
oxygen adducts (either FlOOH or FlOH) were reported to be present as
an intermediate in choline oxidase and pyranose dehydrogenase by
crystallographic analysis: b) A. M. Orville, G. T. Lountos, S. Finnegan, G.
Gadda, R. Prabhakar, Biochemistry 2009, 48, 720–728; c) T. C. Tan, O.
Chem. Eur. J. 2014, 20, 4386 – 4395
www.chemeurj.org
4394
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Spadiut, T. Wongnate, J. Sucharitakul, I. Krondorfer, C. Sygmund, D. Hal-
trich, P. Chaiyen, C. K. Peterbauer, C. Divne, PLoS One 2013, 8, e53567.
[4] a) G. de Gonzalo, M. W. Fraaije, ChemCatChem 2013, 5, 403–415; b) F. G.
Gelalcha, Chem. Rev. 2007, 107, 3338–3361; c) Y. Imada, T. Naota, Chem.
Rec. 2007, 7, 354–361.
[5] C. Kemal, T. C. Bruice, Proc. Natl. Acad. Sci. USA 1976, 73, 995–999.
[6] S.-I. Murahashi, T. Oda, Y. Masui, J. Am. Chem. Soc. 1989, 111, 5002–
5003.
[14] B. Mennucci, R. Cammi, Continuum Solvation Models in Chemical Physics:
From Theory to Applications, Wiley, Chichester, 2007.
[15] It must be stressed that the following analysis is based on model (S)-4
treated as an independent molecule; thus, the structures of this com-
pound were not derived by the truncation of optimized structures of
(S)-3, but found by independent molecular modeling and DFT calcula-
tions.
[16] Y. Zhao, D. Truhlar, Theor. Chem. Acc. 2008, 119, 525–525.
[17] B. Mennucci, C. Cappelli, R. Cammi, J. Tomasi, Chirality 2011, 23, 717–
729.
[18] a) V. P. Nicu, E. J. Baerends, P. L. Polavarapu, J. Phys. Chem. A 2012, 116,
8366–8373; b) F. Lipparini, F. Egidi, C. Cappelli, V. Barone, J. Chem.
Theory Comput. 2013, 9, 1880–1884.
[19] a) N. Berova, L. Di Bari, G. Pescitelli, Chem. Soc. Rev. 2007, 36, 914–931;
b) G. Snatzke, Angew. Chem. 1979, 91, 380–393; Angew. Chem. Int. Ed.
Eng. 1979, 18, 363–377.
[7] For examples of 1- or 5-alkylated flavin catalysts, see: a) K. Bergstad, J.-
E. Bꢅckvall, J. Org. Chem. 1998, 63, 6650–6655; b) S. Chen, M. S. Hos-
sain, F. W. Foss, Org. Lett. 2012, 14, 2806–2809; c) Y. Imada, H. Iida, T. Ki-
tagawa, T. Naota, Chem. Eur. J. 2011, 17, 5908–5920; d) Y. Imada, H. Iida,
S.-I. Murahashi, T. Naota, Angew. Chem. 2005, 117, 1732–1734; Angew.
Chem. Int. Ed. 2005, 44, 1704–1706; e) Y. Imada, H. Iida, T. Naota, J. Am.
Chem. Soc. 2005, 127, 14544–14545; f) Y. Imada, H. Iida, S. Ono, S.-I.
Murahashi, J. Am. Chem. Soc. 2003, 125, 2868–2869; g) A. A. Linden, L.
Krueger, J.-E. Bꢅckvall, J. Org. Chem. 2003, 68, 5890–5896; h) B. J.
Marsh, E. L. Heath, D. R. Carbery, Chem. Commun. 2011, 47, 280–282;
i) C. Mazzini, J. Lebreton, R. Furstoss, J. Org. Chem. 1996, 61, 8–9;
j) A. B. E. Minidis, J.-E. Bꢅckvall, Chem. Eur. J. 2001, 7, 297–302. For ex-
amples of chiral flavin-catalyzed asymmetric oxidations, see: k) R. Jurok,
R. Cibulka, H. Dvorakova, F. Hampl, J. Hodacova, Eur. J. Org. Chem. 2010,
5217–5224; l) V. Mojr, M. Budesinsky, R. Cibulka, T. Kraus, Org. Biomol.
Chem. 2011, 9, 7318–7326; m) S.-I. Murahashi, Angew. Chem. 1995, 107,
2670–2693; Angew. Chem. Int. Ed. Eng. 1995, 34, 2443–2465; n) S.-I.
Murahashi, S. Ono, Y. Imada, Angew. Chem. 2002, 114, 2472–2474;
Angew. Chem. Int. Ed. 2002, 41, 2366–2368; o) S. Shinkai, T. Yamaguchi,
A. Kawase, O. Manabe, R. M. Kellogg, J. Am. Chem. Soc. 1989, 111,
4935–4940; p) S. Shinkai, T. Yamaguchi, O. Manabe, F. Toda, J. Chem.
Soc. Chem. Commun. 1988, 1399–1401.
[20] K. Tanaka, Y. Itagaki, M. Satake, H. Naoki, T. Yasumoto, K. Nakanishi, N.
Berova, J. Am. Chem. Soc. 2005, 127, 9561–9570.
[21] a) X.-W. Shi, A.-L. Zhang, G. Pescitelli, J.-M. Gao, Chirality 2012, 24, 386–
390; b) J. M. Dragna, G. Pescitelli, L. Tran, V. M. Lynch, E. V. Anslyn, L.
Di Bari, J. Am. Chem. Soc. 2012, 134, 4398–4407.
[22] Schrçdinger Suite 2012: Macromodel, version 9.9. New York NY, 2012.
[23] Gaussian 09, Revision C.01. M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E.
Scuseria, M. A. Robb, J. R. Cheeseman, G. Scalmani, V. Barone, B. Men-
nucci, G. A. Petersson, H. Nakatsuji, M. Caricato, X. Li, H. P. Hratchian,
A. F. Izmaylov, J. Bloino, G. Zheng, J. L. Sonnenberg, M. Hada, M. Ehara,
K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O.
Kitao, H. Nakai, T. Vreven, J. A. Montgomery, J. E. Peralta, F. Ogliaro, M.
Bearpark, J. J. Heyd, E. Brothers, K. N. Kudin, V. N. Staroverov, R. Kobaya-
shi, J. Normand, K. Raghavachari, A. Rendell, J. C. Burant, S. S. Iyengar, J.
Tomasi, M. Cossi, N. Rega, J. M. Millam, M. Klene, J. E. Knox, J. B. Cross, V.
Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Stratmann, O. Yazyev,
A. J. Austin, R. Cammi, C. Pomelli, J. W. Ochterski, R. L. Martin, K. Moro-
kuma, V. G. Zakrzewski, G. A. Voth, P. Salvador, J. J. Dannenberg, S. Dap-
prich, A. D. Daniels, Farkas, J. B. Foresman, J. V. Ortiz, J. Cioslowski, D. J.
Fox. Wallingford CT, 2009.
[8] H. Iida, S. Iwahana, T. Mizoguchi, E. Yashima, J. Am. Chem. Soc. 2012,
134, 15103–15113.
[9] A close inspection of the reported crystallographic data of 4a–oxygen
adducts (either FlOOH or FlOH) in choline oxidase (see ref. [3b]) and
pyranose dehydrogenase (see ref. [3c]) indicates that the 4a–oxygen
adducts may have the R configurations, although the uncertainty of the
crystal structures could not discriminate between the FlOOH and FlOH
structures.
[10] a) B. R. Leeflang, J. F. G. Vliegenthart, L. M. J. Kroonbatenburg, B. P. Va-
neijck, J. Kroon, Carbohydr. Res. 1992, 230, 41–61; b) K. D. Zhang, R. B.
Teklebrhan, G. Schreckenbach, S. Wetmore, F. Schweizer, J. Org. Chem.
2009, 74, 3735–3743.
[24] All references about DFT functionals and basis sets can be found in the
online documentation for Gaussian09 at www.gaussian.com/g tech/
g ur/g09 help.htm.
[25] J. Zheng, X. Xu, D. Truhlar, Theor. Chem. Acc. 2011, 128, 295–305.
[26] R. A. Kendall, J. T. H. Dunning, R. J. Harrison, J. Chem. Phys. 1992, 96,
6796–6806.
[11] C. A. Hunter, Angew. Chem. 2004, 116, 5424–5439; Angew. Chem. Int. Ed.
2004, 43, 5310–5324.
[12] G. Pescitelli, L. Di Bari, N. Berova, Chem. Soc. Rev. 2011, 40, 4603–4625.
[13] a) M. Ebitani, Y. In, T. Ishida, K.-i. Sakaguchi, J. L. Flippen-Anderson, I. L.
Karle, Acta Crystallogr. Sect. B 1993, 49, 136–144; b) P. Mꢆnovꢇ, V.
[27] a) Specdis v. 1.60. T. Bruhn, Y. Hemberger, A. Schaumlçffel, G. Bring-
mann. University of Wꢄrzburg (Germany), 2012; b) T. Bruhn, A. Schaum-
lçffel, Y. Hemberger, G. Bringmann, Chirality 2013, 25, 243–249.
ˇ
ˇ
ˇ
Received: November 8, 2013
Eigner, J. Cejka, H. Dvorꢇkovꢇ, M. Sanda, R. Cibulka, J. Mol. Struct. 2011,
1004, 178–187.
Published online on March 3, 2014
Chem. Eur. J. 2014, 20, 4386 – 4395
www.chemeurj.org
4395
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim