An N1-hydrogen bonding model for flavin coenzyme

Article abstract of DOI:10.1016/S0960-894X(01)00712-0

A model flavin possessing a specific hydrogen bond to the N1-position has been synthesized. The redox potential has been measured in aqueous buffer and found to be shifted +21 mV as compared to a similar flavin lacking this hydrogen bond. The reaction of the N1-hydrogen-bonding model with sulfite and 1-benzyl-dihydronicotinamide were examined and compared with the non-hydrogen-bonded flavin. The N1-hydrogen bond did not accelerate the rate of sulfite ion or hydride addition to N5, however the N5-sulfite complex was stabilized by nearly 4-fold over a non-hydrogen-bonding model. The model flavins were also studied computationally.

Full text of DOI:10.1016/S0960-894X(01)00712-0

Bioorganic & Medicinal Chemistry Letters 12 (2002) 151–154
An N1–Hydrogen Bonding Model for Flavin Coenzyme
Fengli Guo, Bryan H. Chang and Carmelo J. Rizzo*
Department of Chemistry and Center for Molecular Toxicology, Vanderbilt University, VU Station B 351822,
Nashville, TN 37235-1822, USA
Received 14 May 2001;accepted 15 October 2001
Abstract—A model flavin possessing a specific hydrogen bond to the N1-position has been synthesized. The redox potential has
been measured in aqueous buffer and found to be shifted +21 mV as compared to a similar flavin lacking this hydrogen bond. The
reaction of the N1–hydrogen-bonding model with sulfite and 1-benzyl-dihydronicotinamide were examined and compared with the
non-hydrogen-bonded flavin. The N1-hydrogen bond did not accelerate the rate of sulfite ion or hydride addition to N5, however
the N5–sulfite complex was stabilized by nearly 4-fold over a non-hydrogen-bonding model. The model flavins were also studied
computationally. # 2002 Elsevier Science Ltd. All rights reserved.
The mechanism by which the protein environment
influences the redox and catalytic properties of flavin
coenzyme are poorly understood. Specific interactions
Flavins 3a–c and 4 were previously developed as N1–
hydrogen bonding models (Fig. 1). Muller examined 3a
using vapor phase UV photoelectron spectroscopy and
¨
1
between the protein and the flavin such as hydrogen
bonding are believed to be important in defining the
reactivity of a particular flavoenzyme. Massey and
Hemmerich suggested one criterion by which flavo-
enzymes could be classified pertained to the presence of
a specific hydrogen bond to the N1- or N5-position.
Hydrogen bonding to N1 activates the flavin cofactor to
addition reactions directly to N5 while a hydrogen bond
showed that the p- and s-orbital framework of 3a are
significantly altered by N1–hydrogen bonding. Shinkai
7
subsequently examined related models 3b and c in solu-
tion and found that the appended hydrogen-bonding
8
group had no effect on their chemistry. Model 4, also
studied by Shinkai, was only mildly effective as an N1–
8
hydrogen bonding model. Molecular modeling and X-
ray crystallographic analysis showed that the N10-aryl
group is orthogonal to the flavin ring system, and is
therefore not in the optimal geometry for hydrogen
bonding to N1 in the oxidized state.
2
to N5 activates the 4a-position.
A number of flavin models have been developed to
study how specific hydrogen bonds to the isoalloxazine
ring system influences its chemistry. These include ele-
gant host–guest systems to examine the influence of
hydrogen bonding to the pyrimidine portion of the iso-
alloxazine ring system as well as models designed to
3
,4
have a specific hydrogen bond to N5. We have pre-
pared flavin model 1 to probe the influence of a specific
hydrogen bond to the N1-position (Fig. 1). Flavin 2
served as a reference compound, since it will have simi-
5
6
lar electronic and conformational effects as 1, however
is not capable of forming an intramolecular hydrogen
bond.
*Corresponding author. Fax: +1-615-343-1234;e-mail: c.j.rizzo@
vanderbilt.edu
Figure 1. N1–Hydrogen bonding models of flavin coenzyme.
0
960-894X/02/$ - see front matter # 2002 Elsevier Science Ltd. All rights reserved.
PII: S0960-894X(01)00712-0

152
F. Guo et al. / Bioorg. Med. Chem. Lett. 12 (2002) 151–154
Figure 2. Crystal structure of the FAD cofactor of an electron-trans-
˚
fer flavoprotein (ETF) at 2.6 A resolution.
0
chain to the flavin N1-position (Fig. 2). The ETF was
reconstituted with an unnatural 4 -deoxy-FAD, which
gen bond from the C4 hydroxyl group of the ribityl
1
3
0
presumably eliminated the critical hydrogen-bonding
interaction. As a result of this modification, the oxi-
dized-semiquinone couple was destabilized by 116 mV
(2.7 kcal/mol), over 5-fold greater than we observed for
the two electron reduction of model 1. Zhou and Swen-
son examined the role of electrostatic interactions on
the redox potential of a bacterial flavodoxin from
Scheme 1. Synthesis of N1–hydrogen-bonding flavin model.
14
Desulfovibrio vulgaris.
Mutant proteins in which
aspartate and glutamate residues near the FMN binding
site were systematically neutralized resulted in less
negative midpoint potentials for the semiquinone/
hydroquinone couple;the oxidize/semiquinone couple
was not significantly influenced by the changes. The
calculated electrostatic surface potential of the ETF
The synthesis of 1 is outlined in Scheme 1 and begins
with the bromination of 8-nitroquinaldine (5) to give 6.
Hydrolysis and esterification provided methyl ester 7 in
9
5
0% overall yield. The nitro group was then reduced
and protected as the BOC-derivative (8). At this stage,
the quinoline ring was reduced under 45 psi hydrogen
pressure over Pd. When the reduction of the nitro group
and quinoline was attempted in a single step, low yields
of the desired product were obtained, possibly due to
poisoning of the catalyst by the 8-amino-tetra-
hydroquinoline product. Reduction of 9 with lithium
borohydride in THF at reflux gave alcohol 10 in high
yield. Optimal conditions for the removal of the BOC-
group involved treatment of 10 with boron trifluoride
1
3
from P. denitrificans was found to be highly negative.
Eliminating the N1–hydrogen bond would greater
expose the one-electron reduced semiquinone anion to
the electrostatic field. Thus, electrostatic may play a
significant role in the destabilization of the semiquinone
in the reconstituted ETF lacking the N1–hydrogen
bond. It was also reported that the long wavelength
absorbance of the ETF reconstituted with the unnatural
cofactor was red shifted 7 nm versus that of the native
protein (443 vs 436 nm). However, we observed the
opposite behavior for model 1 where the presence of the
hydrogen bond caused a 4 nm red shift relative to 2.
1
0
etherate in CH Cl . The product from the deprotec-
2
2
tion was condensed without purification with alloxan in
1
1
glacial acetic acid to give 1 in 86% overall yield.
The redox potentials of flavins 1 and 2 (Table 1) were
measured by cyclic voltammetry in 100 mM, pH 7.4
HEPES buffer using a standard three-electrode cell with
a glassy carbon working electrode. The redox potential
of 1 was found to be shifted +21 mV (0.5 kcal/mol)
from that of 2. We attribute this shift to specific hydro-
gen bonding to the N1-position. The redox potential of
We have examined two reactions of flavins 1 and 2, the
addition of sulfite and the oxidation of the NAD(P)H
analogue 1-benzyl-dihydronicotinamide (Tables 2 and
3). These reactions involve a direct nucleophilic addition
to N5 and should be accelerated by hydrogen bonding
to the N1-position. We observed that the formation of
the sulfite–flavin complex was approximately four times
more favorable for N1–hydrogen bonding model 1 than
1
under these conditions, was nearly that of the unsub-
6
15
stituted flavin 3d. Our data should be contrasted to
that of Frerman and co-workers who examined the
redox chemistry of a bacterial electron-transfer flavo-
for 2 which lacks this hydrogen bonding interaction.
Table 2. Reaction of flavins 1 and 2 with sulfite in 100 mM, pH 7.4
ꢀ
1
2
protein (ETF) from Paracoccus denitrificans. This
particular ETF possesses a rare intramolecular hydro-
phosphate buffer at 30 C
Table 1. Midpoint potential (mV) and UV data of flavin models in
ꢀ
0
1
00 mM, pH 7.4 HEPES buffer at room temperature. E versus Ag/
AgCl (ꢁ197 mV vs NHE), scan rate=20 mV/s
a
p
c
p
ꢀ0
E
Flavin
E
E
l (e)
ꢁ
1
ꢁ1
ꢁ1
)
Flavin
K
d
(M)
k
1
(min
M
)
k
ꢁ1 (min
1
2
3
ꢁ446 ꢁ376 ꢁ411 440 (8400), 372 (9100), 265 (30,000)
ꢁ450 ꢁ414 ꢁ432 436 (6300), 370 (7000), 265 (23,300)
ꢁ428 ꢁ386 ꢁ407 429 (6500), 346 (4800), 261 (21,300)
1
2
0.371ꢂ0.019
1.47ꢂ0.173
0.493ꢂ0.012
0.443ꢂ0.343
0.183ꢂ0.007
0.651ꢂ0.504
d

F. Guo et al. / Bioorg. Med. Chem. Lett. 12 (2002) 151–154
153
Gaussian 98W;17 the optimized conformations are
shown in Figure 3. This analysis predicts the hydrogen-
bonding conformation to be favored by greater than
Table 3. Oxidation of 1-benzyldihydronicotinamide by flavins 1 and
in 100 mM, pH 7.4 phosphate buffer
2
5
The calculated N1–H and N1–O distances are 1.98 and
kcal/mol. These calculations did not include solvent.
˚
ꢀ
.78 A, respectively, and the N1–H–O angle is 136.6 .
2
The N1–H and N1–O distances are well within hydro-
gen-bonding range and compare favorably with the N1–
0
structure (2.89 A). The N1–H–O angle of 1, however, is
significantly smaller than optimal. In addition, the pro-
ton involved in the critical hydrogen bonding interac-
˚
tion is displaced 0.58 A from the best plane comprised
of the six-ring atoms of the pyrimidine ring.
O4 distances found in P. denitrificans ETF crystal
˚
(min 1
ꢁ
M
ꢁ1
)
Flavin
k
1
ꢀ
ꢀ
3
0 C
15 C
1
2
1494ꢂ47
1541ꢂ58
2576ꢂ109
646ꢂ32
612ꢂ34
928ꢂ34
3d
A recent computational study of some 27 oxidized flavin
models showed the two-electron reduction potential is
linearly correlated with the energy of their LUMO
Interestingly, the rate of the sulfite addition reaction
was not accelerated by the presence of the N1– hydro-
gen bond.
1
8
(R=0.96); the experimentally determined reduction
potentials of the flavins varied over a 300 mV range
while the computed LUMO energies spanned about 1
eV. The computed FMO energies for our flavins are
The oxidation of 1-benzyl-dihydronicotinamide was
1
6
examined under aerobic conditions. The produced
flavin hydroquinone is quickly reoxidized under these
conditions thus satisfying a first order equation;
dihydronicotinamides are not oxidized under these con-
ditions. Again, no significant increase in rate was
observed for the oxidation of 1-benzyl-dihydronicotin-
1
9
listed in Table 4. Although we can not determine the
redox potential of the non-hydrogen-bonding con-
formation 1b, we would expect it to be similar to that of
2. We find that the calculated LUMO energies of the
non-hydrogen-bonding conformation (1b) and 2 are in
reasonable agreement. The experimentally determined
reduction potential of 1 is similar to that of flavin model
3d (ꢁ411 vs ꢁ407 mV vs Ag/AgCl). The calculated
LUMO energies of the hydrogen-bonding conformation
1a is very close to that of 3d in agreement with similar
ꢀ
amide by 1 as compared to 2 at 30 or 15 C. We also
examined flavin model 3d, which has a similar redox
potential as 1. This ensured that flavins with relatively
small differences in redox potential would give measurable
differences in reaction rates.
ꢀ
0
E
values. The reduction potential of 1 would be
Hydrogen bonding effects should be enhanced in
organic solvents. Shinkai reported the reaction of N1–
hydrogen bonding model 4 with 1-benzyl-dihydronico-
tinamide in acetonitrile, although at a considerably
slower rate than in aqueous buffer. Reactions with non-
hydrogen bonding substrates were not observed.
Unfortunately, we were unable to observe a satisfactory
decay curve for the reaction of 1 or 2 with 1-benzyl-
dihydronicotinamide in acetonitrile.
dependent on the equilibrium of conformations 1a and
1b. We have not been able to determine this value at this
time, however the computational results lead us to
believe that the hydrogen-bonding conformation 1a is
the predominant form.
The geometry of reference flavin 2 with a molecule of
water hydrogen bonded to the N1-position was also
computed (Fig. 4). The geometry of the hydrogen-bon-
ded state is not as favorable as that in 1a due to steric
We have examined the conformation of 1 using a vari-
ety of computational methods. Molecular mechanics
and AM1 calculations were used to obtain starting
geometries for a hydrogen-bonding (1a) and non-
hydrogen-bonding conformation (1b). These conforma-
tions were then optimized using the B3LYP/6-31G*
hybrid density functional method as implemented in
Table 4. B3LYP/6-31G* calculated FMO energies of flavin models
Flavin
HOMO (eV)
LUMO (eV)
1a
1b
ꢁ6.606
ꢁ6.410
ꢁ6.382
ꢁ6.506
ꢁ6.508
ꢁ3.095
ꢁ2.967
ꢁ2.890
ꢁ3.004
ꢁ3.012
2
.
2
3
2
H O
d
Figure 3. Hydrogen-bonded and non-hydrogen-bonded conforma-
tions of flavin 1.
Figure 4. B3LYP/6-31G* optimized geometry of model 2 hydrogen
bonded to a water at N1.

154
F. Guo et al. / Bioorg. Med. Chem. Lett. 12 (2002) 151–154
interactions with the rigidly held N10-methylene group.
The calculated N1–H and N1–O distances are 2.16 and
1998, 2727. (c) Cuello, A. O.;McIntosh, C. M.;Rotello, V. M.
J. Am. Chem. Soc. 2000, 122, 3517.
˚
ꢀ
The calculated LUMO energy of 2 H O is intermediate
4. (a) Shinkai, S.;Honda, N.;Ishikawa, Y.;Manabe, O.
Am. Chem. Soc. 1985, 107, 6286. (b) Akiyama, T.;Simeno, F.;
Murakami, M.;Yoneda, F. J. Am. Chem. Soc. 1992, 114,
J.
3
.03 A, respectively, and the N1–H–O angle is 147.9 .
ꢃ
2
to that of 2 and 1a. Since solvent may be hydrogen
bonded to 2, the effect of hydrogen bonding in 1 may be
underestimated when compared to 2.
6
5
2
6
1
613.
. Hasford, J. J.;Rizzo, C. J. J. Am. Chem. Soc. 1998, 120,
251.
. Hasford, J. J.;Kemnitzer, W.;Rizzo, C. J. J. Org. Chem.
997, 62, 5244.
In summary, we have synthesized N1–hydrogen-bond-
ing model 1 and compared its redox chemistry with
related model 2 which lacks the N1–hydrogen-bonding
interaction. A shift in reduction potential and a sig-
nificant stabilization of the N5–sulfite complex indicates
the presence of the desired hydrogen bond. Contrary to
previous predictions, no measurable rate enhancement
for the addition of nucleophile to N5 was observed.
Deprotonated dihydroflavins are often drawn as the
N1-anion, however this charge is actually delocalized
into both pyrimidine carbonyls. The B3LYP/6-31+G*
optimized geometry of dihydro-lumiflavin anion pre-
dicts nearly equal distribution of the negative charge on
¨
7. Muller, F.;Eweg, J. K.;Szczesna, V. In Flavin and Flavo-
proteins;Brey, R. C., Engel, P. C., Meyhew, S. G., Eds.;W.
De Gruyter: Berlin, 1984;p 3.
8. Shinkai, S.;Kawanabe, S.;Kawase, A.;Yamaguchi, T.;
Manabe, O.;Harada, S.;Nakamura, H.;Kasai, N.
Chem. Soc. Jpn. 1988, 61, 2095.
Bull.
9
1
. Roth, R.;Erlenmeyer, H. Helv. Chim. Acta 1954, 37, 1064.
0. Evans, E. F.;Lewis, N. J.;Kapfer, I.;Macdonald, G.;
Taylor, R. J. K. Synth. Commun. 1999, 27, 1819.
1
1
1. Lambooy, J. P. Heterocycl. Comp. 1967, 9, 118.
2. Dweyer, T. M.;Mortl, S.;Kemter, K.;Bacher, A.;Fauq,
A.;Frerman, F. E. Biochemistry 1999, 38, 9735.
13. Roberts, D. L.;Salazar, D.;Fulmer, J. P.;Frerman, F. E.;
Kim, J.-J. Biochemistry 1999, 38, 1977.
2
0
N1 and both carbonyl oxygens. Flavoproteins often
show hydrogen bonds to these carbonyl groups. A
bifurcated hydrogen bond to N1 and the C2–carbonyl is
common. Hydrogen bonding to the C2 and C4 carbonyl
groups may play a significant role in activation of N5
toward nucleophilic addition.
1
1
1
1
1
4. Zhou, Z.;Swenson, R. P. Biochemistry 1995, 34, 3182.
5. Muller, F.;Massey, V. J. Biol. Chem. 1969, 244, 4007.
Macromolecules
978, 11, 65.
7. Frisch, M. J.;Trucks, G.W.;Schlegel, H. B.;Scuseria, G.
¨
6. Shinkai, S.;Yamada, S.;Kunitake, T.
E.;Robb, M. A.;Cheeseman, J. R.;Zakrzewski, V. G.;
Montgomery, J. A.;Stratmann, R. E.;Burant, J. C.;Dap-
prich, S.;Millam, J. M.;Daniels, A. D.;Kudin, K. N.;Strain,
M. C.;Farkas, O.;Tomasi, J.;Barone, V.;Cossi, M.;Cammi,
R.;Mennucci, B.;Pomelli, C.;Adamo, C.;Clifford, S.;Och-
terski, J.;Petersson, G. A.;Ayala, P. Y.;Cui, Q.;Morokuma,
K.;Malick, D. K.;Rabuck, A. D.;Raghavachari, K.;Fores-
man, J. B.;Cioslowski, J.;Ortiz, J. V.;Baboul, A. G.;Stefa-
nov, B. B.;Liu G.;Liashenko, A.;Piskorz, P.;Komaromi, I.;
Gomperts, R.;Martin, R. L.;Fox, D. J.;Keith, T.;Al-Laham,
M. A.;Peng, C. Y.;Nanayakkara, A.;Challacombe, M.;Gill,
P. M. W.;Johnson, B.;Chen, W.;Wong, M. W.;Andres, J.
L.;Gonzalez, C.;Head-Gordon, M.;Replogle, E. S.;Pople, J.
A., Gaussian 98, Revision A.9;Gaussian, Inc.;Pittsburgh, PA,
Acknowledgements
This work was supported by the National Institutes of
Health (GM56460). B.H.C. acknowledges fellowships
form the Vanderbilt Undergraduate Summer Research
Program and the Howard Hughes Medical Institute.
References and Notes
1
1
998.
8. Ridder, L.;Zuilhof, H.;Vervoot, J.;Rietjens, I. M. C. M.
1
. (a) Mu
voenzymes;Muller, F., Ed.;CRC: Boca Raton;1991;Vol. 1, p 1.
b) Ghisla, S.;Massey, V. Eur. J. Biochem. 1989, 181, 1.
¨
ller, F. In Chemistry and Biochemistry of Fla-
Methods Mol. Biol. 1999, 131, 207.
9. In addition to 1, 2 and 3d, the optimized geometries of
nine other oxidized flavin analogues were calculated (B3LYP/
-31G*). The calculated LUMO energies of these eleven flavins
1
(
2
8
3
. (a) Massey, V.;Hemmerich, P. Biochem. Soc. Trans. 1980,
, 246. (b) Ghisla, S.;Massey, V. Biochem. J. 1986, 239, 1.
. (a) Breinlinger, E.;Niemz, A.;Rotello, V. M. J. Am. Chem.
Soc. 1995, 117, 5379. (b) Kajiki, T.;Moriya, H.;Kondo, S.;
Nabeshima, T.;Yano, Y. J. Chem. Soc., Chem. Commun.
6
was found to be linearly correlated with the published two-
electron reduction potentials (R=0.977) as previously
observed. Details of our study will be reported elsewhere.
2
0. Rizzo, C. J. Antioxid. Redox Signal. 2001, 3, 737.