Theoretical and experimental studies of biotin analogues that bind almost as tightly to streptavidin as biotin

Article abstract of DOI:10.1021/jo991846s

We have used a newly developed qualitative computational approach, PROFEC (Pictorial Representation of Free Energy Changes), to visualize the areas of the ligand biotin where modifications of its structure might lead to tighter binding to the protein streptavidin. The PROFEC analysis, which includes protein flexibility and ligand sotvation/desolvation, led to the suggestion that the pro-9R hydrogen atom of biotin, which is in α-position to the CO₂^- group, might be changed to a larger group and lead to better binding with streptavidin and avidin. Free energy calculations supported this suggestion and predicted that the methyl analogue should bind ≈3 kcal/mol more tightly to streptavidin, with this difference coming exclusively from the relative desolvation free energy of the ligand. The PROFEC analysis further suggested little or no improvement for changing the pro-9S hydrogen atom to a methyl group, and great reduction in changing the ureido N-H groups to N-CH₃. Stimulated by these results, we synthesized 9R-methylbiotin and 9S-methylbiotin, and their binding free energies and enthalpies were measured for interaction with streptavidin and avidin, respectively. In contrast to the calculated results, experiments found both 9-methylbiotin isomers to bind more weakly to streptavidin than biotin. The calculated preference for the binding of the 9R- over the 9S-stereoisomer was observed. In addition, 9-methylbiotin is considerably less soluble in water than biotin, as predicted by the calculation, and the 9R isomer is, to our knowledge, thus far the tightest binding analogue of biotin to streptavidin. Subsequently, X-ray structures of the complexes between streptavidin and both 9R- and 9S-methylbiotin were determined, and the structures were consistent with those used in the free energy calculations. Thus, the reason for the discrepancy between the calculated and experimental binding free energy does not lie in unusual binding modes for the 9-methylbiotins.

Full text of DOI:10.1021/jo991846s

J . Org. Chem. 2002, 67, 1827-1837
1827
Th eor etica l a n d Exp er im en ta l Stu d ies of Biotin An a logu es Th a t
Bin d Alm ost a s Tigh tly to Str ep ta vid in a s Biotin
Richard W. Dixon,* Randall J . Radmer, Bernd Kuhn, and Peter A. Kollman
Department of Pharmaceutical Chemistry, University of California, San Francisco,
San Francisco, California 94143
J aemoon Yang, Cesar Raposo, and Craig S. Wilcox
Department of Chemistry, University of Pittsburgh, Pittsburgh, Pennsylvania 15260
Lisa A. Klumb and Patrick S. Stayton
Department of Bioengineering, University of Washington, Seattle, Washington 98195
Craig Behnke, Isolde Le Trong, and Ronald Stenkamp
Department of Biological Structure, University of Washington, Seattle, Washington 98195
richard.dixon@abbott.com
Received December 1, 1999 (Revised Manuscript Received November 28, 2001)
We have used a newly developed qualitative computational approach, PROFEC (Pictorial
Representation of Free Energy Changes), to visualize the areas of the ligand biotin where
modifications of its structure might lead to tighter binding to the protein streptavidin. The PROFEC
analysis, which includes protein flexibility and ligand solvation/desolvation, led to the suggestion
that the pro-9R hydrogen atom of biotin, which is in R-position to the CO2- group, might be changed
to a larger group and lead to better binding with streptavidin and avidin. Free energy calculations
supported this suggestion and predicted that the methyl analogue should bind ≈3 kcal/mol more
tightly to streptavidin, with this difference coming exclusively from the relative desolvation free
energy of the ligand. The PROFEC analysis further suggested little or no improvement for changing
the pro-9S hydrogen atom to a methyl group, and great reduction in changing the ureido N-H
groups to N-CH3. Stimulated by these results, we synthesized 9R-methylbiotin and 9S-methylbiotin,
and their binding free energies and enthalpies were measured for interaction with streptavidin
and avidin, respectively. In contrast to the calculated results, experiments found both 9-methylbiotin
isomers to bind more weakly to streptavidin than biotin. The calculated preference for the binding
of the 9R- over the 9S-stereoisomer was observed. In addition, 9-methylbiotin is considerably less
soluble in water than biotin, as predicted by the calculation, and the 9R isomer is, to our knowledge,
thus far the tightest binding analogue of biotin to streptavidin. Subsequently, X-ray structures of
the complexes between streptavidin and both 9R- and 9S-methylbiotin were determined, and the
structures were consistent with those used in the free energy calculations. Thus, the reason for
the discrepancy between the calculated and experimental binding free energy does not lie in unusual
binding modes for the 9-methylbiotins.
In tr od u ction
tremendous computational demand. That is why one of
us (R.J .R.) has developed a more approximate method,
PROFEC (Pictorial Representation Of Free Energy
Changes), for ligand design that can use modest length
molecular dynamics trajectories on the solvated ligand
and the solvated ligand-macromolecule complex to es-
timate where a “lead” compound can best be modified to
bind more tightly to the macromolecule. This method
looks quite promising, having qualitatively ranked the
observed relative free energies of binding of benzamidine
analogues to trypsin.4
We decided to try PROFEC on a more challenging
problem such as the biotin-streptavidin system. The
complexes of biotin with streptavidin and its structural
homologue avidin are known to be among the strongest
ligand-protein complexes,5-10 with measured binding
Given a structure of a macromolecule, the goal of
structure-based design is to find ligands that bind to it
tightly and specifically. The computational methodologies
designed to do so vary widely,1 from methods that analyze
databases with many thousands of ligands,2 to methods
that focus on single, small changes such as free energy
perturbation calculations.3 Whereas the simplest meth-
ods must use very approximate scoring schemes to filter
good from bad candidate ligands, free energy approaches
can, in principle, consider such critical aspects as ligand
solvation/desolvation and (limited) binding site flexibility.
The difficulty with a free energy calculation is its
(1) Kuntz, I. D. Science 1992, 257, 1078-1082.
(2) Kuntz, I. D.; Meng, E. C.; Shoichet, B. K. Acc. Chem. Res. 1994,
27, 117-123.
(4) Radmer, R. J .; Kollman, P. A. J . Comput.-Aided Mol. Des. 1998,
12, 215-227.
(3) Kollman, P. Chem. Rev. 1993, 93, 2395-2417.
10.1021/jo991846s CCC: $22.00 © 2002 American Chemical Society
Published on Web 02/22/2002

1828 J . Org. Chem., Vol. 67, No. 6, 2002
Dixon et al.
constants of 1.7 × 1015 and 2.5 × 1013, respectively.5-8,11
Because of the very high binding affinities of biotin to
these two proteins, which are certainly the highest known
per number of atoms in the ligand (excluding metal
complexes), many studies have been directed to under-
stand and elucidate the mechanism of binding at the
molecular level.9,10,12,13 Given this strong binding, it is a
significant challenge to predict new biotin analogues that
bind even more tightly to streptavidin than biotin itself.
The reference ligand D-(+)-biotin 1, vitamin H, serves
as a CO2 carrier14 and was first isolated as a growth
factor for yeast in 1936 in its methyl ester form.15
Unlike other biotin homologues such as (+)-norbiotin
(4a ), (+)-homobiotin (4b), (+)-bis-homobiotin (4c), and
(+)-tris-homobiotin (4d ), 9-methylbiotin keeps the valeric
acid side chain and has an additional methyl substituent
in R-position to the carboxylic acid group.
The absolute configuration of D-(+)-biotin was deter-
mined by measurement of the anomalous dispersion of
X-ray data on bis-p-bromoanilide carbon dioxide biotin
2.16 Subsequently, X-ray crystallographic data on biotin
itself, and its analogues have been published.17
Streptavidin and avidin have very similar primary
structures and can be isolated from the bacterium
Streptomyces avidinii and from hen egg white, respec-
tively. They are both tetrameric proteins, but differ in
their overall charge, with streptavidin being highly acidic
and avidin being highly basic. X-ray structures of biotin
cocrystallized with both streptavidin22-24 and avidin13
revealed the protein residues involved in binding.
In the following, we describe a combined theoretical
and experimental approach to investigate the binding
affinities of different methylbiotin analogues to strepta-
vidin and avidin. This involves the initial PROFEC
analysis, subsequent binding free energy and solvation
free energy calculations, organic synthesis of the two
9-methylbiotin stereoisomers, thermodynamic measure-
ments of the binding affinities of the 9-methylbiotins to
streptavidin and avidin, and X-ray analysis of the binding
modes of these compounds in streptavidin.
The proposed structure for the D-(+)-biotin was con-
firmed by the first total synthesis by Merck18 and later
numerous organic syntheses employing different meth-
odologies have appeared.19
9R- and 9S-methylbiotin, 3a /3b, are structural ana-
logues of D-(+)-biotin, first isolated in 1971 from the
fermentation of a strain of S. lydicus on Saccharomyces
pastorianus.20 Studies of 9-methylbiotin on the antimi-
crobial activity were performed in detail to show that it
could not substitute for the biotin requirement for the
growth of Saccharomyces cerevisiae although it has
affinity for avidin.21
Meth od s
A. Th eor etica l Ca lcu la tion s. The quantitative determi-
nation of free energy changes is perhaps the most powerful
tool available to the computational chemist. In principle, given
two states A and B, the free energy associated with converting
the system in question from A to B can be calculated by
(5) Green, N. M. Biochem. J . 1963, 89, 585-591.
(6) Melamed, M. D.; Green, N. M. Biochem. J . 1963, 89, 591-599.
(7) Green, N. M. Adv. Protein Chem. 1975, 29, 85-133.
(8) Green, N. M. Methods Enzymol. 1990, 184, 51-67.
(9) Miyamoto, S.; Kollman, P. A. Proteins: Struct., Funct., Genet.
1993, 16, 226-245.
∆G ) GB - GA ) -RT ln e[V (x) - V (x)]/RT
(1)
B
A
A
(10) Miyamoto, S.; Kollman, P. A. Proc. Natl. Acad. Sci. U.S.A. 1993,
90, 8402-8406.
In this case, sampling of configuration space is performed
at state A according to a potential energy function VA. The
accuracy of the calculated ∆G in eq 1 is determined by the
quality of the potential function VA and the completeness of
the sampling of configuration space during the course of the
simulation. In practice, it has been found that the highest
quality results can be obtained when states A and B are
(11) Wilchek, M.; Bayer, E. A. Methods Enzymol. 1990, 184, 5-13.
(12) Livnah, O.; Bayer, E. A.; Wilchek, M.; Sussman, J . L. Proc. Natl.
Acad. Sci. U.S.A. 1993, 90, 5076-5080.
(13) Pugliese, L.; Coda, A.; Malcovati, M.; Bolognesi, M. J . Mol. Biol.
1993, 231, 698-710.
(14) Stryer, L. Biochemistry, 2nd ed.; W. H. Freeman: San Fran-
cisco, 1981.
(15) Kogl, F.; Tonnis, B. Z. Z. Physiol. Chemie 1936, 242, 43.
(16) Trotter, J .; Hamilton, J . A. Biochemistry 1966, 5, 713.
(17) Stallings, W.; Detitta, G. T. Ann. N. Y. Acad. Sci. 1985, 447,
152-168.
(21) Hanka, L. J .; Martin, D. G.; Reineke, L. M. Antimicrob. Agents
Chemother. 1972, 1, 135-138.
(22) Weber, P. C.; Cox, M. J .; Salemme, F. R.; Ohlendorf, D. H. J .
Biol. Chem. 1987, 262, 12728-12729.
(18) Harris, S. A.; Wolf, D. E.; Mozingo, R.; Folkers, K. Science 1943,
97, 447.
(23) Pahler, A.; Hendrickson, W. A.; Kolks, M. A.; Argarana, C. E.;
Cantor, C. R. J . Biol. Chem. 1987, 262, 13933-13937.
(24) Hendrickson, W. A.; Pahler, A.; Smith, J . L.; Satow, Y.; Merritt,
E. A.; Phizackerley, R. P. Proc. Natl. Acad. Sci. U.S.A. 1989, 86, 2190-
2194.
(19) Deroose, F. D.; Declercq, P. J . J . Org. Chem. 1995, 60, 321-
330.
(20) Martin, D. G.; Hanka, L. J .; Reineke, L. M. Tetrahedron Lett.
1970, 3791-3794.

Biotin Analogues as Strepavidin Substrates
J . Org. Chem., Vol. 67, No. 6, 2002 1829
similar. Therefore, a number of innovations have become
common practice in the determination of free energy changes
through simulation methodology. The first is that the path
between states A and B is usually divided into a number of
discrete steps, with a coupling parameter λ defining this path,
where λ ) 0 at A and λ ) 1 at B. This allows a conversion
between two states to be represented as a collection of
arbitrarily small subdivisions. The free energy change at any
point along the simulation path is defined as
proposed change. This derivative is calculated by means of
umbrella sampling
dG(i,j,k)
)
[
]
dq
LJ (i,j,k)
dV(i,j,k)
dq
e-∆V(i,j,k)/RT
Φ(i,j,k)e-∆V(i,j,k)/RT
A
A
)
(6)
e-∆V(i,j,k)/RT
e-∆V(i,j,k)/RT
A
A
∆Gλ ) -RT ln e[V
(2)
i+1(x)-Vλi(x)]RT
λ
λi
i
displayed on the van der Waals free energy contours as color.
The resulting image can be used to identify regions of the
ligand susceptible to derivatization and to predict the effect
of the charge distribution on the proposed physical changes.
Once a broad scan of the free energy “surface” has been
completed, and candidates chosen, the more CPU-intensive
free energy simulation techniques can be used to evaluate each
candidate individually. This prescription has been followed for
the ligand biotin in the protein streptavidin.
The method described in eqs 1 and 2 is called free energy
perturbation. An alternative way to calculate the free energy
by slowly varying the Hamiltonian with the coupling param-
eter λ is called thermodynamic integration (eq 3).
1
∂H
∂λ
∆G )
dλ
λ
(3)
∫
0
The trajectory information necessary to calculate the free
energies needed for PROFEC was collected from a 500 ps
simulation of biotin in tetrameric streptavidin using the
Cornell et al. force field.25 An 18 Å belly of 231 TIP3P26 water
molecules was centered on one biotin in the system, and a
small harmonic restraint force was applied to the backbone
atoms of the mobile portions of the protein. This last restraint
was found to play an important role in the free energy
simulations and so was included in these calculations for
consistency. In addition, SHAKE27,28 was applied to all bonds,
and an 8.0 Å cutoff for nonbonded interactions was imposed.
The system temperature was maintained at 300 K by means
of the Berendsen temperature coupling algorithm. All simula-
tions began with at least 40 ps of equilibration. Trajectory
information was collected every 0.5 ps, resulting in 1000
coordinate sets, which were used to calculate the appropriate
ensemble averages. A second simulation of biotin solvated in
a box of 555 TIP3P water molecules, approximately 25 ×
25 × 30 Å3, was performed with the same total time and
trajectory collection history. A 0.5 Å grid was placed around
the biotin molecule, and a united atom carbon atom (R* ) 2.00
Å, ꢀ ) 0.15 kcal/mol) was inserted at each grid point. A series
of grids, each requiring a separate PROFEC calculation, was
chosen due to the flexible nature of the ligand.
The free energy simulations were carried out in both the
forward and reverse direction between biotin and the biotin
analogues reported in the next section. Simulation conditions
were the same as those described above. The simulation time
for each leg of the calculation was 200 ps with a time step of
1 fs. Two hundred windows, each composed of 500 steps of
equilibration and 500 steps of data collection, were employed
in total. Twenty ps of equilibration was performed prior to each
reverse simulation. The reported errors in the calculated free
energy values represent the difference between the forward
and reverse simulations. Thermodynamic integration29 was
used to calculate the free energy changes.
Both of these methods have been extensively used in free
energy calculations.3
Another conceptual change made in free energy calculations
is to simulate “nonphysical” paths and take advantage of the
state function nature of the free energy to relate the results
to physical processes. Thus, thermodynamic cycles are invoked,
such as the one in eq 4, for the relative binding free energy of
two ligands to a protein.
∆Gprot - ∆Gsolv ) ∆G4 - ∆G3 ) ∆Gtot
(4)
The free energy calculations, which have been performed
as a part of this effort, have all relied on the above methodol-
ogy. The details of these calculations will be described below.
A major drawback of these calculations is that they carry
the price of significant computational burden. If one wishes
to examine the effects of a relatively large number of changes
between ligand systems, the computational needs may soon
outstrip available resources. For this reason, a qualitative
methodology has been developed which allows relatively rapid
assessment of a broad range of free energy changes. In this
approach, called PROFEC (Pictorial Representation Of Free
Energy Changes), eq 5, in analogy to eq 1, is used to calculate
the free energy cost of inserting a particle at a point space
(i,j,k).
∆G(i,j,k) ) -RT ln e-∆V(i,j,k)/RT
(5)
∆V(i,j,k) is the van der Waals interaction energy between a
Lennard-J ones particle and the surrounding atoms. The
The atomic partial charges for biotin were those used
previously9,10 and were calculated by means of the RESP
methodology30 at the HF/6-31G* level31 of theory. Charges for
ensemble average ... is calculated from stored trajectories
A
collected from a single simulation involving the reference
ligand, state A, only. This exercise is carried out for a grid of
insertion points surrounding the reference ligand. Given a set
of grid points, each with an associated van der Waals free
energy value, contours can be drawn with a suitable graphics
program corresponding to different values of free energy. For
example, the contour of 0.0 kcal/mol represents a boundary
for free energy changes. Particles that fall “inside” this contour
are predicted to lead to improved interaction free energies,
while those outside are predicted to have a deleterious effect
on the interactions.
The electrostatic interactions can be examined by calculat-
ing the derivative of the binding free energy at each grid point
once the particle has been inserted. This derivative can also
be calculated in the absence of the Lennard-J ones particle;
however, it has been found to be more useful to weight the
electrostatic calculation with the steric favorability of the
(25) Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Gould, I. R.; Merz, K.
M., J r.; Ferguson, D. M.; Spellmeyer, D. C.; Fox, T.; Caldwell, J . W.;
Kollman, P. A. J . Am. Chem. Soc. 1995, 117, 5179-5197.
(26) J orgensen, W. L.; Chandrasekhar, J .; Madura, J . D.; Impey, R.
W.; Klein, M. L. J . Chem. Phys. 1983, 79, 926-935.
(27) Van Gunsteren, W. F.; Berendsen, H. J . C. Mol. Phys. 1977,
34, 1311-1327.
(28) Ryckaert, J . P.; Ciccotti, G.; Berendsen, H. J . C. J . Comput.
Phys. 1977, 23, 327-341.
(29) van Gunsteren, W. F.; Weiner, P. K.; Wilkinson, A. J .; Eds.
Computer Simulation of Biomolecular Systems, Volume 3: Theoretical
and Experimental Applications; Kluwer: Dordrecht, Netherlands,
1997.
(30) Bayly, C. I.; Cieplak, P.; Cornell, W.; Kollman, P. A. J . Phys.
Chem. 1993, 97, 10269-10280.
(31) Hariharan, P. C.; Pople, J . A. Theor. Chim. Acta 1973, 28, 213-
222.

1830 J . Org. Chem., Vol. 67, No. 6, 2002
Dixon et al.
the methylbiotin analogues were obtained by using partial
charges from a calculation on methane for the methyl group
and adjusting the charge of the attachment point atom to
maintain correct overall charge.
We also carried out MM-PBSA (molecular mechanics Pois-
son Boltzmann Surface Area) binding free energy calculations32
on biotin and its 9R- and 9S-methyl analogues. We first used
the crystal structure of the biotin-streptavidin complex to
model the methyl analogues; subsequently, we used the
experimental X-ray structures of the 9-methylbiotin-strepta-
vidin complexes. The method of analyzing the binding free
energies is described in ref 32.
resulting solution was heated at 95-100 °C for 12 h. The
mixture was cooled, and volatile components were removed
by rotary evaporation. The residual solid was recrystallized
from 20 mL of water. After 12 h, the crystalline solid was
filtered, washed with water, and vacuum-dried at 65-70 °C
for 12 h to afford 0.3992 g (62%) of 3a and 3b, (9R and 9S)-
methylbiotin: [R]23 +109 (c 0.010, pH ) 9 NH4OAc buffer);
D
mp 185-188 °C. This product was not further characterized
but was immediately converted to the amides.
[9R a n d 9S]-Meth ylbiotin (N-(S-r-p h en eth yl))a m id es
(9a a n d 9b). To an ice-cooled mixture of the (9R and 9S)-
methylbiotin carboxylic acids 3a /3b (258 mg, 1.0 mmol), CH2-
Cl2 (10 mL), and HMPA (2.5 mL), under argon, was added 486
mg of BOP reagent (1.1 mmol) followed by 0.42 mL (3.3 mmol)
of 9S-methylbenzylamine. After 10 min, Et3N (0.14 mL, 1.0
mmol) was added, and the reaction mixture was stirred for
3.5 h. Water (20 mL) was added, the organic phase was
separated, and the aqueous layer was extracted with (6 × 2
mL, 5% MeOH-CH2Cl2). The combined organic layers were
washed with brine and dried (MgSO4). After filtration, the
filtrate was concentrated, and residual HMPA was removed
by Kugelrohr vacuum distillation. Purification of the residue
by flash chromatography (SiO2, 5% MeOH-CH2Cl2) afforded
0.37 g of material which was a 1:1 mixture of two diastereo-
mers as shown by 1H NMR and by HPLC. Separation of these
stereoisomers was achieved by preparative HPLC using a C18
silica column (10 mm ID × 25 cm) and 0.5% aqueous TFA:
CH3CN (75:25 v:v) at a flow rate of 2 mL/min. (The pH of the
TFA solution was adjusted to 2.5-3.0 by using 33% KOH.)
(9S)-Meth ylbiotin (N-(S-r-p h en eth yl))a m id e (9b): yield
108 mg (60%); [R]23D +13 (c 0.60, MeOH); Rf 0.12 (5% MeOH-
CH2Cl2); mp 179-181 °C; IR (KBr) 3283, 2923, 1690, 1678,
1630, 1534 cm-1; 1H NMR (300 MHz, CD3OD) δ 7.40-7.20 (m,
5 H), 5.00 (q, J ) 7.0 Hz, 1 H), 4.45 (dd, J ) 7.6, 5.0 Hz, 1 H),
4.16 (dd, J ) 7.7, 4.4 Hz, 1 H), 3.08 (m, 3 H), 2.90 (dd, J )
12.8, 4.8 Hz, 1 H), 2.68 (d, J ) 12.7 Hz, 1 H), 2.35 (m, 1 H),
1.80-1.20 (m, 9 H), 1.11 (t, J ) 6.8 Hz, 3 H); 13C NMR (75
MHz, CD3OD) δ 178.58, 166.21, 145.62, 129.67, 128.14, 127.32,
63.45, 61.67, 57.04, 41.98, 41.22, 35.34, 29.67, 28.37, 22.45,
18.53; 13C NMR (75 MHz, DMSO-d6) δ 174.67, 162.70, 145.06,
128.21, 126.47, 125.88, 61.09, 59.15, 55.55, 47.56, 33.92, 28.25,
26.73, 22.51, 17.89; HRMS calcd for C19H27N3O2S 361.1824,
found 361.1842.
B. Exp er im en ta l Da ta for Syn th etic Wor k . Biotin
Meth yl Ester (7). A mixture composed of d-(+)-biotin 1 (2.90
g, 11.9 mmol), MeOH (60 mL), and sulfuric acid (6 g) was
refluxed under N2 for 5 h, and then 10 mL of MeOH was
removed by distillation. Heating at reflux was continued for
an additional 12 h, and then all MeOH was removed by rotary
evaporation. The residue was added to a mixture of ice and
water, and the pH of the resulting solution was adjusted (final
pH ∼ 6) with 25% NaOH. The precipitated solid was extracted
with EtOAc and washed with brine. The organic layer was
dried (MgSO4) and filtered. The filtrate was concentrated in
vacuo to give d-(+)-biotin methyl ester 7 (2.31 g, 75%) as a
white solid: [R]23 +52 (c 0.21, CHCl3) ([28] [R]22 +57 (c 1,
D
D
CHCl3)); Rf 0.24 (5% MeOH-CH2Cl2); mp 160-161 °C; IR
(KBr) 3252, 2913, 1734, 1698, 1462 cm-1; 1H NMR (300 MHz,
CDCl3) δ 5.24 (br s, 1 H), 4.91 (br s, 1 H), 4.53 (m, 1 H), 4.34
(m, 1 H), 3.68 (s, 3 H), 3.17 (m, 1 H), 2.94 (dd, J ) 12.8, 5.1
Hz, 1 H), 2.75 (d, J ) 12.8 Hz, 1 H), 2.35 (t, J ) 7.3 Hz, 2 H),
1.80-1.40 (m, 6 H); HRMS calcd for C11H18N2O3S 258.1038,
found 258.1048.
[9R a n d 9S]-Meth ylbiotin Meth yl Ester (8a ). To THF
(75 mL) under an argon atmosphere at -78 °C were added 40
mL (20 mmol) of a 0.5 M solution of KHDMS in toluene
(dropwise, over 5 min) and 3.0 mL (20 mmol) of TMEDA. The
mixture was stirred for an additional 5 min, and then a
solution of (+)-biotin methyl ester 7 (1.29 g, 5 mmol) in a
mixture of HMPA/THF (10 mL/7.5 mL) was added via syringe
over 25 min. Stirring was continued for 1 h at -78 °C, the
color of the solution changed to light brown, and then 2.50
mL of MeI (40 mmol) was added dropwise. The reaction
mixture was stirred for 1.5 h at -78 °C, 25 mL of 10% aqueous
HCl was added to the cold mixture, cooling was discontinued,
and the mixture was warmed to 25 °C. Volatile components
were removed by rotary evaporation, and the product was
extracted with CH2Cl2 (10 × 20 mL), washed with brine (1 ×
20 mL), and dried (MgSO4). After filtration, the filtrate was
concentrated, and residual HMPA was removed by Kugelrohr
vacuum distillation. Purification of the residue by flash
chromatography (SiO2, 5% MeOH-CH2Cl2) afforded 1.95 g of
material, which was a mixture (1:6 by 1H NMR) of starting
material and R-methylated product. Further purification was
achieved by preparative HPLC using a C18 silica column (10
mm ID × 25 cm) and 0.5% aqueous TFA:CH3CN (75:25 v:v)
at a flow rate of 2 mL/min. (The pH of the TFA solution was
adjusted to 2.5-3.0 by using 33% KOH.)
(9R)-Meth ylbiotin (N-(S-r-p h en eth yl))a m id e (9a ): yield
157 mg (87%); [R]23D -30 (c 0.73, MeOH); Rf 0.12 (5% MeOH-
CH2Cl2); mp 209-210 °C; IR (KBr) 3281, 2917, 1694, 1630,
1534 cm-1; 1H NMR (300 MHz, CD3OD) δ 7.40-7.20 (m, 5 H),
5.00 (q, J ) 7.0 Hz, 1 H), 4.49 (dd, J ) 7.6, 4.7 Hz, 1 H), 4.29
(dd, J ) 7.8, 4.4 Hz, 1 H), 3.19 (m, 3 H), 2.93 (dd, J ) 12.8, 4.9
Hz, 1 H), 2.71 (d, J ) 12.7 Hz, 1 H), 2.38 (m, 1 H), 1.80-1.20
(m, 9 H), 1.11 (t, J ) 6.9 Hz, 3 H); 13C NMR (75 MHz, CD3OD)
δ 178.51, 166.25, 145.46, 129.63, 128.13, 127.20, 63.58, 61.76,
57.34, 41.88, 41.19, 35.34, 29.99, 28.58, 22.72, 18.77; 13C NMR
(75 MHz, DMSO-d6) δ 174.67, 162.85, 144.97, 128.32, 126.60,
125.95, 61.31, 59.28, 55.72, 47.53, 33.81, 28.47, 26.80, 22.74,
18.18; HRMS calcd for C19H27N3O2S 361.1824, found 361.1836.
(9R)-Meth ylbiotin (3a ). A mixture of 1 mL of 48% HBr
and amide 9a (20.0 mg, 55 µmol) was heated at100-105 °C
(oil bath) for 3 h. The mixture was evaporated to dryness in
vacuo, the brownish solid was suspended in water, and the
mixture was basified to pH ) 9.5-10 with 0.5% NaOH. Basic
byproducts were extracted with Et2O, and the aqueous layer
was acidified to pH 2 by 0.6 N HCl. Upon standing for 12 h,
the mixture deposited a white solid which was removed by
filtration, washed with a small amount of cold water, and
vacuum-dried for 12 h to give 10.6 mg (75%) of pure 3a , (9R)-
In this manner we obtained 0.88 g (64%) of pure 9R- and
9S-methylbiotin methyl ester 8a : [R]23 +39 (c 0.19, CHCl3);
D
Rf 0.24 (5% MeOH-CH2Cl2); mp 116-117 °C; IR (KBr) 3245,
2926, 1725, 1698, 1462 cm-1 1H NMR (300 MHz, CDCl3) δ
;
4.85 and 4.79 (br m, 1 H), 4.68 (br m, 1 H), 4.54 (m, 1 H), 4.34
(m, 1 H), 3.68 (s, 3 H), 3.17 (m, 1 H), 2.95 (dd, J ) 12.7, 5.0
Hz, 1 H), 2.74 (d, J ) 12.7 Hz, 1 H), 2.49 (m, 1 H), 1.80-1.30
(m, 6 H), 1.17 (t, J ) 7.0 Hz, 3 H); 13C NMR (75 MHz,
CDCl3 + CD3OD) δ 177.41, 177.35, 164.05, 61.86, 61.76, 59.92,
55.48, 55.32, 51.50, 40.22, 39.15, 39.00, 33.36, 33.20, 28.31,
28.06, 26.70, 26.41, 16.95, 16.79; HRMS calcd for C12H20N2-
O3S 272.1195, found 272.1191.
methylbiotin: [R]23 +80 (c 0.013, pH ) 9 NH4OAc buffer),
D
[R]23 +14 (c 0.20, MeOH); mp 201-205 °C; IR (KBr) 3362,
D
2923, 1680 cm-1; 1H NMR (300 MHz, CD3OD) δ 4.49 (dd, J )
7.7, 4.7 Hz, 1 H), 4.30 (dd, J ) 7.9, 4.5 Hz, 1 H), 3.21 (m, 1 H),
2.93 (dd, J ) 12.8, 5.0 Hz, 1 H), 2.70 (d, J ) 12.7 Hz, 1 H),
2.42 (m, 1 H), 1.80-1.40 (m, 6 H), 1.15 (t, J ) 7.0 Hz, 3 H);
13C NMR (125 MHz, CD3OD) δ 180.61, 166.15, 63.43, 61.62,
57.06, 41.04, 40.53, 34.82, 29.78, 28.07, 17.65; HRMS calcd
for C11H18N2O3S 258.1038, found 258.1034.
[9R a n d 9S]-Meth ylbiotin (3a a n d 3b). To an ice cold
mixture of water (4 g) and TFA (12 g) was added 0.6801 g (2.50
mmol) of (9R- and 9S-)methylbiotin methyl ester 8a . The
(32) Kuhn, B.; Kollman, P. A. J . Med. Chem. 2000, 43, 3786-3791.

Biotin Analogues as Strepavidin Substrates
J . Org. Chem., Vol. 67, No. 6, 2002 1831
(9S)-Meth ylbiotin (3b). By the above procedure used for
the 9R isomer, (9S)-methylbiotin (N-(S-R-phenethyl))amide, 9b
(20.0 mg, 55 µmol) afforded 10.1 mg (67%) of pure 3b, (9S)-
ing the amount of unbound biotin. Streptavidin was added to
a final concentration of 9 nM to mixtures of H-biotin (Amer-
3
sham, Chicago, IL) at 10 nM and the competing ligand in
phosphate buffered saline (50 mM NaH2PO4, 100 mM NaCl,
pH 7). The concentration of the competing ligand was varied
from 0 nM to 50 × ∆Kd × 10 nM, where ∆Kd is the ratio of the
streptavidin equilibrium dissociation constants for competing
ligand to biotin. The solution containing the protein and
ligands was equilibrated over several half-lives for the wild-
type streptavidin-biotin dissociation rate (t1/2 is 4.7 h at 37
°C) with typical incubation times of 24 h at 37 °C. After
equilibration, 50 µL aliquots were added to 200 µL of 0.2 M
ZnSO4 followed by addition of 200 µL of 0.2 M NaOH to
precipitate the protein bound ligands. The supernatant was
assayed by for radioactivity in a liquid scintillation counter
(LS-7000, Beckman, Fullerton, CA).
methylbiotin: [R]23 +127 (c 0.010, pH ) 9 NH4OAc buffer),
D
[R]23 +76 (c 0.11, MeOH); mp 177-180 °C; IR (KBr) 3291,
D
2913, 1705, 1688, 1676 cm-1
;
1H NMR (300 MHz, CDCl3) δ
4.51 (dd, J ) 7.6, 4.9 Hz, 1 H), 4.31 (dd, J ) 7.8, 4.4 Hz, 1 H),
3.20 (m, 1 H), 2.93 (dd, J ) 12.8, 4.9 Hz, 1 H), 2.70 (d, J )
12.7 Hz, 1 H), 2.42 (m, 1 H), 1.80-1.40 (m, 6 H), 1.15 (t, J )
6.9 Hz, 3 H); 13C NMR (125 MHz, CD3OD) δ 180.68, 166.12,
63.34, 61.59, 56.93, 41.01, 40.49, 34.70, 29.58, 27.93, 17.57;
HRMS calcd for C11H18N2O3S 258.1038, found 258.1033.
P r op er ties for Ad d ition a l [9R a n d 9S]-Alk yla ted Bi-
otin Meth yl Ester s. The method used for R-methylation of
biotin methyl ester was applied to prepare the following
products, which were not further investigated.
3
[9R a n d 9S]-Eth ylbiotin Meth yl Ester (8b). Prepared
from biotin methyl ester by use of KHDMS and ethyl iodide.
The amount of unbound H-biotin was plotted as a function
of the total concentration of competing ligand, with the
streptavidin and total 3H-biotin concentrations kept constant.
The solution to the quadratic expression shown in eq 7 was
applied to the data with the ratio of the dissociation constants
(∆Kd ) (KC/KL)) as a fitting parameter using nonlinear chi
square minimization (Igor Pro, Wavemetrics, Inc., Lake Os-
wego, OR).
Yield 49%; [R]23 +53 (c 0.15, CHCl3); Rf 0.24 (5% MeOH-
D
CH2Cl2); mp 141 °C; IR (KBr) 3235, 2921, 1728, 1698, 1455
1
cm-1; H NMR (300 MHz, CDCl3) δ 5.59-5.39 (m, 2 H), 4.51
(m, 1 H), 4.31 (m, 1 H), 3.69 (s, 3 H), 3.16 (m, 1 H), 2.92 (dd,
J ) 12.7, 4.8 Hz, 1 H), 2.74 (d, J ) 12.8 Hz, 1 H), 2.33 (m, 1
H), 1.80-1.40 (m, 8 H), 0.89 (t, J ) 7.3 Hz, 3 H); 13C NMR (75
MHz, CDCl3 δ 176.60, 176.56, 163.40, 61.97, 61.73, 60.08,
55.42, 55.16, 51.45, 46.99, 46.77, 40.52, 31.88, 31.66, 28.64,
28.22, 27.03, 26.56, 25.49, 25.38, 11.79; HRMS calcd for
KC
KL
K
- 1 [L]2 + C(PT - LT) + CT - PT + 2LT [L] +
(
)
(
)
KL
C
13H22N2O3S 286.1358, found 286.1354.
(PT - CT - LT)LT ) 0 (7)
[9R a n d 9S]-Bu t ylb iot in Met h yl E st er (8c). Prepared
from biotin methyl ester by use of KHDMS and butyl iodide.
Yield 19%; [R]23 +45 (c 0.14, CHCl3); Rf 0.24 (5% MeOH-
where
D
CH2Cl2); mp 122-123 °C; IR (KBr) 3370, 2917, 1728, 1696,
1
[P][C]
[PC]
1455 cm-1; H NMR (300 MHz, CDCl3) δ 5.48-5.31 (m, 2 H),
KC )
) equilibrium dissociation constant for
4.51 (m, 1 H), 4.31 (m, 1 H), 3.69 (s, 3 H), 3.14 (m, 1 H), 2.92
(dd, J ) 12.7, 4.9 Hz, 1 H), 2.74 (d, J ) 12.8 Hz, 1 H), 2.35 (m,
1 H), 1.90-1.20 (m, 12 H), 0.89 (t, J ) 6.9 Hz, 3 H); 13C NMR
(75 MHz, CDCl3) δ 176.83, 176.72, 163.31, 61.96, 61.69, 60.08,
55.39, 55.13, 51.46, 45.45, 45.21, 40.53, 32.30, 32.10, 29.57,
28.65, 28.21, 27.05, 22.58, 13.93; HRMS calcd for C15H26N2O3S
314.1664, found 314.1679.
streptavidin-competing ligand complex
[P][L]
[PL]
KL )
) equilibrium dissociation constant for
streptavidin-biotin complex
[9R a n d 9S]-Ben zylbiotin Meth yl Ester (8d ). Prepared
from biotin methyl ester by use of KHDMS and benzyl
[C] ) concentration of competing ligand
bromide. Yield 23%; [R]23 +50 (c 0.17, CHCl3); Rf 0.28 (5%
D
[L] ) concentration of 3H-biotin
MeOH-CH2Cl2); mp 101-102 °C; IR (KBr) 3409, 2899, 1719,
1711, 1694, 1663, 1445 cm-1 1H NMR (300 MHz, CDCl3) δ
;
[P] ) concentration of wild-type streptavidin
7.31-7.14 (m, 5 H), 5.41 (br m, 1 H), 5.19 (br m, 1 H), 4.49
(m, 1 H), 4.29 (m, 1 H), 3.62 (s, 3 H), 3.13-2.69 (m, 6 H), 1.90-
1.40 (m, 6 H); 13C NMR (75 MHz, CDCl3) δ 175.91, 175.82,
163.21, 139.12, 128.83, 128.41, 126.38, 61.97, 61.63, 60.06,
55.30, 55.09, 51.58, 47.32, 47.12, 40.59, 40.49, 38.53, 38.42,
31.74, 31.59, 28.52, 28.13, 26.88, 26.47; HRMS calcd for
Equation 7 was derived from the ratio of the equilibrium
dissociation constants of the ligand and biotin-streptavidin
complexes, the mass balances for both ligands, and the protein.
The derivation assumes that there is negligible free strepta-
vidin, which is true under these experimental conditions.
C
18H24N2O3S 348.1508, found 348.1504.
[9R a n d 9S]-Allylbiotin Meth yl Ester (8e). Prepared
Resu lts
from biotin methyl ester by use of KHDMS and allyl bromide.
Yield 35%; [R]23 +55 (c 0.19, CHCl3); Rf 0.24 (5% MeOH-
D
A. Th eor etica l Ca lcu la tion s. The type of information
typical of a PROFEC analysis is displayed in Figure 1.
In this case, the three atoms of the R-methylene group
have been used to anchor the set of grid points. The
surface displayed is the 0.0 kcal/mol free energy contour.
As can be seen from this image, one of the two methylene
hydrogen atoms penetrates the surface. This leads to the
rejection of this site as candidate for substitution. The
other hydrogen atom, however, is enclosed in a small
region of “favorable” interaction free energies. This site
may admit substitution by a small functional group. The
electrostatic character of the site, indicated by the color
of the contour surface, suggests that a neutral functional
group would be the best choice. The surface colors can
range from red at the negative end of the charge
spectrum to blue at the positive end. The colors represent
the receptor charge distribution and so should be matched
by complementary charge distributions on the added
CH2Cl2); mp 122-123 °C; IR (KBr) 3403, 2905, 1725, 1692,
1
1676, 1449 cm-1; H NMR (300 MHz, CDCl3) δ 5.72 (m, 1 H),
5.55 and 5.48 (2 s, 1 H), 5.31 (m, 1 H), 5.04 (m, 2 H), 4.51 (m,
1 H), 4.31 (m, 1 H), 3.68 (s, 3 H), 3.14 (m, 1 H), 2.95 (dd, J )
12.8, 4.9 Hz, 1 H), 2.74 (d, J ) 12.8 Hz, 1 H), 2.60-2.20 (m, 3
H), 1.80-1.30 (m, 6 H); 13C NMR (75 MHz, CDCl3) δ 175.94,
175.83, 163.31, 135.24, 116.90, 61.96, 61.72, 60.07, 55.36,
55.12, 51.55, 45.05, 44.82, 40.57, 40.51, 36.46, 36.36, 31.58,
31.37, 28.59, 28.19, 26.88, 26.43; HRMS calcd for C14H22N2O3S
298.1351, found 298.1347.
C. Det er m in in g R ela t ive E q u ilib r iu m Dissocia t ion
Con st a n t s. A recently developed assay to measure the
streptavidin binding free energy of a ligand relative to biotin
was used to characterize the binding affinities of the methyl-
biotin derivatives. In a mixture of the competing ligand and
tritiated biotin, the amount of unbound biotin in solution at
equilibrium can be determined by precipitating the protein
bound ligands. A competition curve was generated by varying
the concentration of the competing ligand, keeping the tritiated
biotin and streptavidin concentrations constant, and determin-

1832 J . Org. Chem., Vol. 67, No. 6, 2002
Dixon et al.
Ta ble 1. Ca lcu la ted Bin d in g F r ee En er gy Ch a n ges Rela tive to Biotin (k ca l/m ol)
∆Gprot (∆G2) ∆Gsolv (∆G1) ∆Gtot (∆G2 - ∆G1)
Streptavidin 100 ps forward
final structure
9R-methylbiotin
9S-methylbiotin
8S-methylbiotin
N1-methylbiotin
+0.2
+2.9
+0.9
+9.8
+3.0
+3.2
+3.3
-0.4
-2.8
-0.3
-2.4
+10.2
Streptavidin 200 psa
9R-methylbiotin
9S-methylbiotin
8S-methylbiotin
+0.05 ( 0.1 (2.2)b
+3.18 ( 0.6
+1.31 ( 0.3
+2.95 ( 0.3 (2.3)b
+2.85 ( 0.5
-2.9 ( 0.4 (-0.1)b
+0.3 ( 1.1
-1.5 ( 0.6
+2.85 ( 0.3
avidin 100 ps forward
9R-methylbiotin
9R-methylbiotin
-0.6
+3.0
Avidin 200 psa
-3.6
-1.33 ( 0.3
+2.95 ( 0.3
-4.3 ( 0.6
a
b
Average is reported with the forward and backward values described by the ( entry. Calculation repeated after experiment was
done using the same conformation (t,t,t,g-) for the side chain of biotin. These were 150 ps simulations run only in the forward direction.
than expected given the size of the free energy change
observed for the proposed modifications.
Given the results of the PROFEC analysis, the follow-
ing free energy calculations were undertaken. First, we
studied methyl substitution of the pro-9R hydrogen and
of the pro-8S hydrogen of biotin. The introduction of
chiral centers can often increase the complexity of the
synthesis of a target compound, so the 9S diastereomer
of methylbiotin was also included in these studies. A final
target compound, N-methylbiotin, was chosen as an
example of a clearly unfavorable analogue. In addition
to the PROFEC analysis, which led to the rejection of
this site, this atom is involved in a hydrogen bond that
will be lost upon methylation. The increased size of the
ligand may also disrupt the important hydrogen bond
network to the ureido oxygen. And finally, the X-ray
conformation of the ligand has the valeric acid “tail” of
the biotin near the area the added methyl group will
occupy. There may well be an unfavorable steric effect
upon addition of the methyl group in the binding pocket.
The calculated binding free energies of these compounds
relative to biotin are reported in Table 1. Also reported
in this table is the relative binding free energy of the 9R-
methylbiotin analogue to the protein avidin.
-
F igu r e 1. PROFEC contour around biotin. Note that the CO2
As can be seen from the free energy data of Table 1,
the 9R methyl analogue of biotin is predicted to bind
more strongly to the protein streptavidin than biotin. In
addition, 9R-methylbiotin is also predicted to bind more
tightly to the protein avidin than biotin. This increase
in affinity, particularly for the protein streptavidin, is
almost completely due to changes in the solvation prop-
erties of the ligands. It is therefore unlikely that these
improvements could have been predicted simply from
visual inspection of the binding pocket.
Are the methyl-substituted biotins significantly less
soluble than biotin as suggested by the ∆Gsolv values in
Table 1? Simple continuum approaches to solvation free
energy calculations support this claim. Results of calcula-
tions with the semiempirical solvation model SM233 are
reported in Table 2 for the molecules considered in our
free energy studies. As can be seen from this table, the
continuum methods reproduce the trends observed in our
simulations.
group is somewhat hidden and that the pro-9R hydrogen atom
is within a favorable (green) contour while the pro-9S hydrogen
atom is protruding into an unfavorable (red) region.
functionality, red matched with positive and blue with
negative. The basically green contour surface in the
PROFEC image, consistent with the predominantly
hydrophobic character of the streptavidin binding pocket,
leads to the proposal of a methyl group as the added
functionality at this point. This type of analysis was
performed for all regions of the biotin molecule. The only
other possible site for adding functionality was the pro-S
hydrogen of the â-methylene group. All other possible
choices were already bordering the region of “unfavor-
able” interaction free energies. This was not a surprising
result. In many ways, the surprise was in finding any
indication of favorable changes at all. It is for this reason
that we are also content to base further free energy
simulations on what appear at first glance to be fairly
small cavities. The raw simulation data used to generate
the PROFEC images are not biased toward any particu-
lar changes in the ligand. For a “cavity” to be present in
the image, it must form naturally during the course of
the simulation, without any help from knowledge of the
proposed end state. This may lead to smaller cavities
B. Syn th esis of 9R-Meth ylbiotin 3a a n d 9S-Meth -
ylbiotin 3b. The first synthesis of R-methylbiotin (RMB)
in racemic form was done by Hanka et al.21 Treatment
(33) Cramer, C. J .; Truhlar, D. G. J . Comput.-Aided Mol. Des. 1992,
6, 629-666.

Biotin Analogues as Strepavidin Substrates
J . Org. Chem., Vol. 67, No. 6, 2002 1833
Ta ble 2. Rela tive Solva tion F r ee En er gies for Biotin
a n d An a logu es
generation in mind, excess (4 equiv) LDA was used, but
R-methylbiotin methyl ester was obtained in a relatively
low yield (Table 3, entry B). When TMEDA was employed
as an additive with the bulky base LiHMDS, however,
R-methylation occurred smoothly at -74 °C after stirring
for 2 h (Table 3, entry C). Better results were obtained
when the countercations of HMDS bases were varied,38-40
and KHMDS was found to be the most effective base in
R-methylation (Table 3, entries D, E). R-Methylbiotin
methyl ester was separated from the starting material
by preparative HPLC on C18 column.41
molecule
biotin
9R-methylbiotin
9S-methylbiotin
∆Gsolv (SM2)a
∆Gsolv (TI)b
0.0
+1.9
+1.6
0.0
+2.95
+2.85
a
b
Using the SM2 solvation model.33 Free energy calculations
using thermodynamic integration (this study).
Sch em e 1
The reaction conditions (KHMDS/TMEDA) used in
R-methylation were found to be effective with other
electrophiles. R-C-alkylation with different alkyl halides
and benzaldehyde was tried to make various R-alkylated
D-(+)-biotin methyl esters 8b-f (Table 4). Primary alkyl
iodides reacted slowly to give R-ethyl D-(+)-biotin methyl
ester (8b) and R-butyl D-(+)-biotin methyl ester (8c) in
49% and 19% respectively (Table 4, entries A, B). For
the R-ethylation, longer reaction times (up to 6 h at -74
°C) and use of more HMPA (up to 0.6 mL in 0.1 mmol
scale) were tried, but the reaction conversion was not
improved. The enolate of D-(+)-biotin methyl ester was
found to be relatively unreactive toward C-alkylation.
Alkylation with sec-alkyl iodide such as isopropyl iodide
was also tried; however, the reaction was sluggish and
the conversion was about 5%. For the reactive electro-
philes such as allylic halides or benzaldehyde, the yield
was relatively low because of the formation of N-alkylated
side product (Table 4, entries D-F).
Sch em e 2
Sch em e 3
R-Methylbiotin methyl ester was hydrolyzed with a
mixture of TFA/H2O (3:1) to give R-methylbiotin 3a /3b
in 56-62% yield.42
Diastereomers of R-methylbiotin methyl ester or R-
methylbiotin were not separable on TLC. Although
diastereomers of R-methylbiotin methyl ester were sepa-
rable on a chiral HPLC column (Cyclobond I 2000
column, Astec Inc.), the separation was not good enough
to allow us to perform preparative HPLC. Therefore
further derivatization43,44 was carried out with R-meth-
ylbiotin. Because DCC/DMAP coupling45 did not work
well, R-methylbiotin 3a /3b was coupled with 1S-(-)-R-
methylbenzylamine using BOP-reagent46 to give (9R)-
methyl amide 9a and (9S)-methyl amide 9b in 87 and
60% yield, respectively, after preparative HPLC purifica-
tion.
of racemic thionium salt 5 with the thallium salt of ethyl
methylacetoacetate at room temperature in DMF af-
forded 6, the precursor of R-methylbiotin 3 (Scheme 1).
Alcoholysis of the acetyl group followed by hydrolysis of
the resulting ester provided racemic methylbiotin in 72%
overall yield.
Our synthesis started from readily available D-(+)-
biotin 1. D-(+)-Biotin methyl ester 7 was prepared by
Fischer esterification (Scheme 2).34
The next step was C-alkylation of D-(+)-biotin methyl
ester with MeI. We envisioned that treatment of D-(+)-
biotin methyl ester with 2.2 equiv of LDA followed by
trapping the resultant dianion with 2.2 equiv of MeI
would give diastereomeric mixture of R-methylbiotin
methyl ester 8a in one step (Scheme 3).35-37
1
Specific rotation values and H NMR spectra provided
evidence necessary for assigning structures to these
In our first trial, when D-(+)-biotin methyl ester was
treated with 2.2 equiv of LDA in THF at -70 °C and
quenched with 10 equiv of CH3I, only N-methylated
D-(+)-biotin methyl ester was obtained from the 1H NMR
(Table 3, entry A). With the possibility of trianion
(38) Evans, D. A.; Britton, T. C. J . Am. Chem. Soc. 1987, 109, 6881-
6883.
(39) Humphrey, J . M.; Bridges, R. J .; Hart, J . A.; Chamberlin, A.
R. J . Org. Chem. 1994, 59, 2467-2472.
(40) Differding, E.; Ruegg, G. M. Tetrahedron Lett. 1991, 32, 3815-
3818.
(41) Bowers-Komro, D. M.; Chastain, J . L.; McCormick, D. B.
Methods Enzymol. 1986, 122, 63-67.
(34) Corey, E. J .; Mehrotra, M. M. Tetrahedron Lett. 1988, 29, 57-
60.
(42) Volkmann, R. A.; Davis, J . T.; Meltz, C. N. J . Am. Chem. Soc.
1983, 105, 5946-5948.
(35) Seebach, D.; Estermann, H. Tetrahedron Lett. 1987, 28, 3103-
3106.
(36) Seebach, D.; Bossler, H.; Gruendler, H.; Shoda, S.; Wenger, R.
Helv. Chim. Acta 1991, 74, 197-224.
(37) Hanessian, S.; Schaum, R. Tetrahedron Lett. 1997, 38, 163-
166.
(43) Helmchen, G.; Nill, G. Angew. Chem. 1979, 91, 66-68.
(44) Helmchen, G.; Nill, G.; Flockerzi, D.; Youssef, M. S. K. Angew.
Chem. 1979, 91, 65-66.
(45) Kaneda, T. J . Chromatogr. 1986, 366, 217-224.
(46) Castro, B.; Evin, G.; Selve, C.; Seyer, R. Synthesis 1977, 413.

1834 J . Org. Chem., Vol. 67, No. 6, 2002
Ta ble 3. Effect of TMEDA a n d Ba se on Alk yla tion of Biotin Meth yl Ester
Dixon et al.
entry
basea
TMEDA
Mel
% yield/(convnb)
A
B
C
D
E
LDA, -78 °C/0.7 h
LDA, -78 °C/0.5 h
LiHMDS, -78 °C/1 h
NaHMDS, -78 °C/1 h
KHMDS, -78 °C/1 h
0 equiv
0 equiv
4 equiv
4 equiv
4 equiv
10 equiv, -60 °C f -50 °C/3.5 h
1 equiv, -60 °C f -30 °C/5 h
8 equiv, -78 °C f -60 °C/2 h
8 equiv, -78 °C/1 h
only N-methylation occurred.
12c
26/(50:50)
59d/(80:20)
64e/(87/13)
8 equiv, -78 °C/1.5 h
a
b
2 equiv (entry A) and 4 equiv (entries B-E) of base was used. Conversion (convn) is expressed as the ratio of product to starting
material, as determined by integration of crude reaction mixture. c The product contained some Claisen condensation adduct. Referees
d
to crude yield. e Isolated yield at 5 mmol (1.25 g) scale. The ratio of diastereomers was 1:1 by 1H NMR.
Ta ble 4. C-Alk yla tion s of D-(+)-Biotin Meth yl Ester 7a
Ta ble 5. Op tica l Rota tion s of Som e Rela ted Com p ou n d s
entry
electrophile
CH3I
CH3CH2I
n-C4HgI
C6H5CH2Br
CH2dCHCH2Br
C6H5CHO
product
% yield/(convnb)
A
B
C
D
E
F
8a
8b
8c
8d
8e
8f
64/6:1
46/(2:1)
19/(1:2)
23/(NDc)
35/(NDc)
39/(NDc)
a
All reactions were carried out at -78 °C with 4 equiv of
KHMDS, 4 equiv of TMEDA and 8 equiv of electrophile. b The ratio
of product to starting material. c Conversion ratio was not deter-
mined.
isomeric molecules. We observed that the C-9 methyl
group in amide 9b is more downfield than the C-9 methyl
group in 9a . Helmchen’s method47-49 predicts that the
absolute stereochemistry at C-9 should therefore be (R)
for 9a and (S) for 9b. Furthermore, the specific rotation
of 9a (-27.3) is more negative than that of 9b (+12.8).
Optical rotation data for model compounds 10-13 (Table
5) show that these R-2-methyl acids and amides should
have more negative specific rotations the S-isomers. The
sum of the optical rotations of 1 and 10 is more negative
than the sum of the optical rotations of 1 and 11. We
conclude that the absolute stereochemistry at C-9 is (R)
for 9a and (S) for 9b.
the optical rotation of acid 3a , the 9-(R)-methyl isomer,
would be more negative than that of acid 3b, the 9-(S)-
methyl isomer. In other words, the sum (+73.3) of the
optical rotations of compounds 1 and 12 is more negative
than the sum (+110.5) of the optical rotations of com-
pounds 1 and 12. By this reasoning we assigned the
absolute stereochemistry at C-9 as (R) for acid 3a and
(S) for acid 3b. This has been verified by the X-ray
structures for the two molecules complexed to strepta-
vidin.
Each isomeric amide was hydrolyzed with 48% HBr.50
There was no epimerization at C-9 under these conditions
as confirmed by HPLC injection of amide, which was
made from the coupling of isolated R-methylbiotin with
(S)-(-)-R-methylbenzylamine using BOP-reagent. The
absolute stereochemistry at C-9 was assigned as (R) for
the acid 3a and (S) for the acid 3b from the measurement
of optical rotation. The optical rotation values ([R]D) for
acid 3a is +80 and that for acid 3b is +127. From the
argument of the rule of the sum of the optical rotations,
C. Ca lor im et r ic Mea su r em en t s of ∆H of Biot in
Bin d in g to Avid in . The calorimetric experiments were
performed with the OMEGA titration calorimeter from
MicroCal, Inc., Northampton, MA.51 The titration of
avidin with D-(+)-biotin 1, 9R-methylbiotin 3a , and 9S-
methylbiotin 3b was done isothermally at 25 °C. The
reference cell was charged with pure distilled water, and
avidin solution of known concentration (typically about
1.5 µM) was transferred to the reaction cell. A 250 µL
(47) Helmchen, G. Tetrahedron Lett. 1974, 1527.
(48) Helmchen, G.; Volter, H.; Schule, W. Tetrahedron Lett. 1977,
1417.
(49) Rettinger, K.; Burschka, C.; Scheeben, P.; Fuchs, H.; Mosandl,
A. Tetrahedron: Asymmetry 1991, 2, 965-968.
(50) Liwschitz, Y.; Singerman, A. J . Chem. Soc., Perkin 1 1966, 13,
1200-1202.
(51) Connelly, P. R.; Varadarajan, R.; Sturtevant, J . M.; Richards,
F. M. Biochemistry 1990, 29, 6108-6114.

Biotin Analogues as Strepavidin Substrates
J . Org. Chem., Vol. 67, No. 6, 2002 1835
F igu r e 3. Competition curves at 37 °C for cold biotin and
the two 9-methylbiotins competing with 3H-biotin for strepta-
vidin.
Ta ble 7. Str ep ta vid in Bin d in g Affin ities of
9-Meth ylbiotin s Rela tive to Biotin a t 37 °C
∆∆G° b
F igu r e 2. Typical titration curve for calorimetry experiment.
a
competing ligand
∆Kd
(kcal/mol)
Ta ble 6. Ca lor im etr ic Mea su r em en ts of ∆H on Titr a tin g
Avid in w ith Biotin An a logu esa ,b
biotin
trial 1:c 0.99 ( 0.07
trial 2: 1.06 ( 0.07
methylbiotin isomer A(R), 3a trial 1: 7.80( 0.60
trial 2: 8.00 ( 0.50
methylbiotin isomer B(S), 3b trial 1: 34.00 ( 2.00
trial 2: 38.00 ( 2.00
0.0
0.0
1.3
1.3
2.2
2.2
∆H (kcal/mol)
at pH ) 8.91, 25 °C
∆H (kcal/mol)
at pH ) 7.00, 25 °C
biotin analogues
biotin
9R-methylbiotin
9S-methylbiotin
-23.1 ( 1.1
-17.6 ( 0.4
-16.7 ( 0.4
-23.6 ( 0.4
-14.5 ( 0.2
-14.6 ( 0.2
a
∆Kd ) (KC/KL) The standard deviations are estimated from
the propagation of errors in multiple sample measurements.
a
Values are the average of three measurements. b ∆H for biotin
has been measured to be: (1) -20.3 kcal/mol;55 (2) -22.5 ( 0.1
kcal/mol.56
b
∆∆G° ) ∆G° (ligand) - ∆G° (biotin) ) RT ln ∆Kd. Characteriza-
tion was performed at 37 °C. c Two independent protein and ligand
preparations were tested with multiple samples for each trial.
injection syringe was filled with the solutions (typically
50 µM) of biotin analogues. After thermal equilibrium
was reached, the sample was injected as 10-25 µL
aliquots at intervals of 3 min. A control experiment with
pH ) 9 buffer solution or pH ) 7 100 mM KCl solution
was done to measure the heat of dilution. The heat of
binding was determined by best-fit curve. A typical
titration diagram is shown in Figure 2, and the results
are summarized in Table 6. The ∆H of binding of the
R-methyl biotins are considerably less exothermic than
is biotin.
D. Com p etitive Bin d in g of Biotin , 9R-Meth ylbi-
otin , a n d 9S-Meth ylbiotin to Str ep ta vid in . We used
a competition assay to quantitate the equilibrium ∆∆G°
of the biotin analogues, relative to biotin at 37 °C. The
competitive binding isotherms are shown in Figure 3 and
Table 7 summarizes the relative binding affinities and
∆∆G°’s. The 9R-methylbiotin displays a decrease in
binding free energy of 1.3 kcal/mol relative to biotin,
while the 9S ligand displays a larger decrease of 2.2 kcal/
mol.
that binds more tightly to streptavidin/avidin. This study
has emphasized the potentially important role of improv-
ing the binding of ligands to their targets by reducing
their solvation free energy. Such desolvation contribu-
tions to binding would not be obvious from simple
visualization of macromolecular target structures. We
should also note that to our knowledge 9R-methylbiotin
is the tightest binding analogue of biotin to streptavidin,
binding more tightly than the more than 20 analogues
previously studied.5-8
The calculations predicted that 9R-methylbiotin would
bind ∼3 kcal/mol more tightly to streptavidin than biotin
with this preference coming exclusively from the relative
solvation free energies of the ligands. SM2 quantum
mechanical calculations also find 9-methylbiotin to be
1.6-1.9 kcal/mol less water soluble than biotin whereas
the ∆Gsolv from the free energy calculations is ∼2.9 kcal/
mol. Experimentally, the 9-methylbiotin was indeed
found to be significantly less soluble than biotin, although
given the limited amounts of material this difference was
difficult to quantify. The calculations further predicted
that 9R-methylbiotin would bind ∼3 kcal/mol more
tightly to streptavidin than the 9S isomer.While the
preference of the 9R- over the 9S-stereoisomer was
Discu ssion a n d Con clu sion
We have carried out a collaborative theoretical/
experimental study, attempting to find a biotin analogue

1836 J . Org. Chem., Vol. 67, No. 6, 2002
Dixon et al.
F igu r e 4. Stereoviews of the complexes of (a) 9R- (3a ) and (b) 9S- (3b) methylbiotin with streptavidin. The structures are those
at the final step of the free energy calculations. C22 corresponds to the R methyl group in both cases.
confirmed by our experiments we find that 9R-methyl-
biotin binds ∼1 kcal/mol less tightly than biotin, not ∼3
kcal/mol more tightly.
9S-methylbiotin to streptavidin. Using 750 ps molecular
dynamics trajectories and 100 snapshots (see ref 32 for
details), we found free energies of binding of -21.6 ( 0.4,
-21.9 ( 0.5, and -19.9 ( 0.4 kcal/mol, respectively,
which corresponds to a ∆∆G of -0.3 kcal/mol for 9R-
methylbiotin relative to biotin. Finally, the X-ray struc-
tures of the 9R- and the 9S-methylbiotin/streptavidin
complexes were solved by Behnke and Stenkamp to 1.7
Å resolution.52 We used these new structures in MM-
PBSA calculations to be compared with MM-PBSA
calculations using the biotin-streptavidin structure.
These calculations led to binding free energies of
-20.2 ( 1.1, -21.5 ( 0.8, and -19.7 ( 1.2 kcal/mol for
Since the experiments showed a discrepancy with the
calculated results, we revisited the calculations on the
free energy difference between biotin and 9R-methylbi-
otin using different free energy calculation protocols. We
first repeated the calculations described in Table 1,
running the simulations for 150 ps only in the forward
direction and using the same t,t,t,g- conformation of the
biotin side chain as used previously. Instead of 0.05 and
2.95 kcal/mol, we calculated 2.2 and 2.3 kcal/mol, respec-
tively, for ∆G2 and ∆G1, which led to a ∆∆G ) -0.1 kcal/
mol, closer to the experimental value of +1.3 kcal/mol.
We then used a new approach, MM-PBSA, to estimate
the binding free energies of biotin, 9R-methylbiotin and
(52) Behnke, C.; Stenkamp, R., Unpublished crystal structures of
9R- and 9S-methylbiotin complexed with streptavidin.

Biotin Analogues as Strepavidin Substrates
J . Org. Chem., Vol. 67, No. 6, 2002 1837
biotin, 9R- and 9S-methylbiotin, and thus to ∆∆G ) -1.3
kcal/mol for biotin f 9R and ∆∆G ) 1.8 kcal/mol for
9R f 9S, which is similar to the results obtained with
the biotin/streptavidin structure. Thus, one consistently
predicts 9R-methylbiotin to bind more tightly to strepta-
vidin than biotin, mainly due to its less favorable
solvation free energy, although their binding free energy
difference varies from -0.1 kcal/mol to -2.9 kcal/mol. On
the other hand, the calculated stereoselectivity (R vs S)
ranges from 1.8 to 3.2 kcal/mol compared to the experi-
mental value of 0.9 kcal/mol.
Calorimetry experiments on the two 9-methylbiotins
binding to avidin also find a significantly less exothermic
∆H of binding than found for biotin (∆∆H ) 6-9 kcal/
mol [Table 6]). Assuming a similar ∆∆G ) 1-2 kcal/mol
in avidin and streptavidin, there is clearly significant
enthalpy/entropy compensation in the relative binding
of biotin and its 9-methyl analogues to (strept)avidin.
At this point, it is not clear why the calculations
overestimated the relative binding free energy of 9R-
methylbiotin and biotin. The conformation around the
R-â bond of biotin in streptavidin appears to have the
R-methyl group of 9R-methylbiotin trans to the γ-CH2
unit, with the carboxylate in gauche position (Figure 4).
This should be a low energy conformer, in contrast to the
R-methyl group of 9S-methylbiotin where both CO2- and
R-CH3 are gauche with respect to the γ-CH2 unit. These
conformations were assumed in the model calculations
and supported by the X-ray crystal structures.
to have a ∆Gsolv of ∼0.4 kcal/mol, compared to 0.2 kcal/
mol experimentally, but perhaps that the R-CH3 disrupts
the solvation of the CO2- group. Such R-CH3 substitutions
would also disrupt the solvation of CO2H groups, given
the small ∆pKa effect observed in comparing primary and
secondary carboxylic acids.
Given the fact that the X-ray structure of the 9R-
methylbiotin/streptavidin complex confirms that the
conformations used in the free energy and MM-PBSA
calculations are correct, the most plausible explanation
for the discrepancy in the calculated ∆∆G is that there
is a subtle overestimate of the ∆Gsolv calculated for adding
the 9-methyl group, which is the dominant cause of the
negative ∆∆G bind between biotin and 9R-methylbiotin.
-
Since the total free energy of solvation of a CO2 group
is ∼ -70 kcal/mol, it is perhaps not surprising that
accurately calculating the ∆Gsolv for putting a methyl
group R to the carboxylate tail is challenging. Future
studies are clearly needed to assess this issue. Nonethe-
less it is possible that a modification of the -(CH2)4- side
chain of biotin that binds more tightly to streptavidin or
avidin than biotin may still be found.54
Ack n ow led gm en t. We acknowledge help from the
UCSF Computer Graphics Lab (T. Ferrin, P. I., sup-
ported by grant NIH P41-RR01081), which provided
visualization facilities. We are also grateful for support
from the NIH (Grant GM-29072 to P.A.K., Grant
DK49655 to P. S. Stayton).
The length of the free energy calculations is relatively
short, but one might expect with limited sampling that
the free energy of growing in a methyl group to be
overestimated rather than underestimated. It is clear
from the rather large size of ∆Gsolv that the addition of
the R-methyl group is not only a simple “hydrophobic
interaction”, where Sun et al.53 found ethane f propane
J O991846S
(53) Sun, Y.; Spellmeyer, D.; Pearlman, D. A.; Kollman, P. J . Am.
Chem. Soc. 1992, 114, 6798-801.
(54) Kuhn, B.; Kollman, P. A. J . Am. Chem. Soc. 2000, 122, 3909-
3916.
(55) Green, N. M. Biochem. J . 1966, 101, 774-80.
(56) Suurkuusk, J .; Wadso, I. Eur. J . Biochem. 1972, 28, 438-441.