A fluorine scan of the phenylamidinium needle of tricyclic thrombin inhibitors: Effects of fluorine substitution on pKa and binding affinity and evidence for intermolecular C-F...CN interactions

Article abstract of DOI:10.1039/b402515f

The H-atoms of the phenylamidinium needle of tricyclic thrombin inhibitors, which interacts with Asp189 at the bottom of the selectivity pocket S1 of the enzyme, were systematically exchanged with F-atoms in an attempt to improve the pharmacokinetic prope

Full text of DOI:10.1039/b402515f

View Article Online / Journal Homepage / Table of Contents for this issue
A fluorine scan of the phenylamidinium needle of tricyclic
thrombin inhibitors: effects of fluorine substitution on pKa and
binding affinity and evidence for intermolecular C–F ؒ ؒ ؒ CN
interactions
Jacob Olsen,a Paul Seiler,a Björn Wagner,b Holger Fischer,b Thomas Tschopp,b
Ulrike Obst-Sander,b David W. Banner,b Manfred Kansy,b Klaus Müllerb and
François Diederich*a
a Laboratorium für Organische Chemie, ETH-Hönggerberg, HCI, CH-8093 Zürich,
Switzerland. E-mail: diederich@org.chem.ethz.ch
b Pharma Division, Präklinische Forschung, F. Hoffmann-La Roche AG, CH-4002 Basel,
Switzerland
Received 19th February 2004, Accepted 18th March 2004
First published as an Advance Article on the web 14th April 2004
The H-atoms of the phenylamidinium needle of tricyclic thrombin inhibitors, which interacts with Asp189 at the
bottom of the selectivity pocket S1 of the enzyme, were systematically exchanged with F-atoms in an attempt to
improve the pharmacokinetic properties by lowering the pKa value. Both the pKa values and the inhibitory constants
Ki against thrombin and trypsin were decreased upon F-substitution. Interestingly, linear free energy relationships
(LFERs) revealed that binding affinity against thrombin is much more affected by a decrease in pKa than the affinity
against trypsin. Surprising effects of F-substitutions in the phenylamidinium needle on the pKa value of the tertiary
amine centre in the tricyclic scaffold of the inhibitors were observed and subsequently rationalised by X-ray
crystallographic analysis and ab initio calculations. Evidence for highly directional intermolecular C–F ؒ ؒ ؒ CN
interactions was obtained by analysis of small-molecule X-ray crystal structures and investigations in the Cambridge
Structural Database (CSD).
reported the F-scan of the benzyl substituent in the tricyclic
Introduction
thrombin inhibitors, pointing into the hydrophobic D-pocket
The importance of organofluorine in medicinal chemistry has
been increasing ever since the seminal discovery in 1954 by
Fried and Sabo that introduction of a single F-atom in the
(Fig. 1).9 A single H/F mutation in position 4 of the benzylic
substituent in 1 resulted in a favourable F-interaction with the
H–C –C᎐O moiety of Asn98 lining the D-pocket and was
᎐
α
9α-position of cortisone acetate led to an increase in biological
activity by a factor of 7–10.1,2 This historical case illus-
trates well the complexity of the factors that may account for
favourable/unfavourable effects of F-substitution. Initially, the
enhanced biological activity of 9α-fluorocorticosteroids was
ascribed to an enhanced polarisation of the neighbouring 11β-
OH group by the strongly electron-withdrawing F-atom. Later,
it was attributed to conformational changes in the steroidal
A-ring induced by the F-substituent (van der Waals radius: 1.47
Å) that is slightly larger than a H-substituent (1.20 Å).3,4 Even
after the crystal structure analysis revealed the binding mode of
9α-fluorocortisol, complexed to the ligand-binding domain
of a mutant human androgen receptor,5 a good explanation of
the positive effect of 9α-F-substitution on biological activity
remained elusive. Most F-effects on protein–ligand binding
affinity/selectivity have indeed been discovered on a trial-and-
error basis.2 In contrast, modulation of the physico-chemical
properties (such as lipophilicity, metabolic stability, acid/base
properties) of drug candidates by introduction of F-substit-
uents is more predictable.6
found to increase the binding free enthalpy by ∆(∆G) ≈ 1 kcal
We have initiated a program to increase the knowledge of
how the thermodynamics of protein–ligand interactions are
affected by H- to F-mutations. In an experimental fluorine scan,
we undertook a systematic exchange of H- for F-substituents
in a rigid tricyclic thrombin inhibitor developed by de novo
design.7 With a conserved binding mode of the ligands in the
active site of thrombin, we hoped to obtain valuable structure–
activity relationships (SARs), thereby making thermodynamic
F-effects more predictable and helping to improve compu-
tational algorithms for predicting them.8 Thus, we recently
Fig. 1 Schematic representation of the binding mode of the tricyclic
inhibitor 1 in the active site of thrombin. Only the (3aS,4R,8aS,8bR)-
configured enantiomer is bound, according to X-ray crystallo-
graphy.7a,b,9 In addition to the catalytic triad flanked by the oxyanion
hole, the active site is described in terms of the selectivity pocket S1,
a small hydrophobic proximal pocket P and a large hydrophobic pocket
D.
T h i s j o u r n a l i s © T h e R o y a l S o c i e t y o f C h e m i s t r y 2 0 0 4
O r g . B i o m o l . C h e m . , 2 0 0 4 , 2, 1 3 3 9 – 1 3 5 2
1339

View Article Online
molϪ1. This study revealed that peptidic H–C –C᎐O fragments
substituents decrease the binding affinity.11,12 Monte-Carlo
᎐
α
provide a pronounced fluorophilic environment. Previously
simulations in conjunction with free-energy perturbation calcu-
lations by Jorgensen and co-workers nicely reproduced some of
the relative experimental binding free energies.15 According to
these calculations, different affinities are best rationalised by
bulk solvation effects while other computational studies also
invoke differences in the interactions with the protein environ-
ment.16 Isothermal calorimetric studies revealed that the bind-
ing of 4-substituted benzamidinium ligands to trypsin at room
temperature is driven by a large favorable enthalpic term and
a small favorable entropic term; however, as a result of a
large negative change in heat capacity, the net thermodynamic
driving force is highly temperature-dependent.11
The work presented herein contributes to the foundation of
the design of thrombin-selective inhibitors and resolves the
complexity of some of the effects of fluorine substitution on
protein binding affinity and selectivity, on physical properties
such as pKa and log D and on molecular conformation.
unrecognised favorable C–F ؒ ؒ ؒ C᎐O contacts are best
᎐
described in terms of multipolar interactions between the
intrinsically polar C–F and C᎐O units. Database mining
᎐
showed that, as the F-atom approaches the carbonyl C-atom,
it is preferentially located at or near the pseudo-trigonal axis
of the carbonyl system.
Here, we report a fluorine scan on the phenylamidinium side
chain (the so-called “needle”) of the tricyclic thrombin inhibi-
tors bound in the narrow selectivity pocket S1. A large number
of thrombin inhibitors feature as a crucial recognition element
such a needle that is a surrogate of the Arg side chain in the
natural substrate.10 Benzamidinium itself is an inhibitor of
thrombin (inhibitory constant Ki = 320 µM) and the related
serine protease trypsin (Ki = 20 µM).11,12 It binds into the S1-
pockets of the two enzymes, that differ only by one amino acid,
Ser190 in thrombin and Ala190 in trypsin. The phenyl-
amidinium needle of 1 adopts a similar geometry in the S1-
pocket of thrombin, forming a bidentate salt bridge with
Asp189 (Figs. 1 and 2). Its high basicity (pKa of benzamidinium
= 11.4112c) prevents good oral bioavailability. One approach to
overcome this problem could be a lowering of the basicity,13
and this could possibly be achieved by fluorination of the
needle. To the best of our knowledge, the modulation of the
phenylamidinium pKa through systematic F-substitution has
not been previously described, although mono-substitution of
H with F has recently been reported.14
Results and discussion
Synthesis
The construction of the tricyclic skeleton of the 2-fluorinated
(ortho to the amidinium residue) inhibitor ( )-2 required the
preparation of 2-fluoro-4-formylbenzonitrile (3) (Scheme 1). Br
CN exchange transformed commercial bromide 4 into the
benzonitrile 5 which was brominated (NBS) to give the benzyl
bromide 6.17 Finally, 3 was obtained by oxidation of 6 with
TMAO (trimethylamine-N-oxide).18
Scheme 1 Synthesis of inhibitor ( )-2. Reagents and conditions: i,
CuCN, DMF, ∆; quant.; ii, NBS, AIBN, CCl4, ∆; 55%; iii, TMAOؒ2
H2O, Me2SO, CH2Cl2, 0 ЊC; 74%; iv, N-(4-fluorobenzyl)maleimide,
-proline, MeCN, ∆; 14%; v, AcCl, MeOH, CH2Cl2, 5 ЊC; then NH3,
MeOH, 65 ЊC; 56%. NBS = N-bromosuccinimide; AIBN = azobis-
(isobutyronitrile); TMAO = trimethylamine N-oxide.
Fig. 2 Overlay of the crystal structures of 1 (protein data bank (pdb)
code: 1OYT) and benzamidinium (1C5O) bound in the active site of
thrombin. Color code: C-skeletons of protein and ligand in the complex
of 1: cyan; C-skeleton of benzamidinium: golden; C-skeleton of
thrombin in the benzamidinium complex: grey; O-atoms: red, S-atoms:
yellow, N-atoms: blue.
1,3-Dipolar cycloaddition7,19 of 3, -proline and N-(4-fluoro-
benzyl)maleimide9b gave a mixture of diastereoisomers that
was separated by column chromatography to afford the desired
benzonitrile ( )-7. The configuration of the tricyclic skeleton
1
In this paper, we analyse how F-substitution of the phenyl-
amidinium needle in the tricyclic thrombin inhibitors affects
the pKa of this binding element and the inhibitory constants
Ki against thrombin and trypsin. QSAR (quantitative struc-
ture–activity relationship) analysis of 3- and 4-substituted
benzamidinium ligands has previously revealed a significant
negative correlation between the binding affinity to both
enzymes and the Hammett substituent σ: electron-withdrawing
was initially assigned by H NMRand NOESY spectroscopy
and later confirmed by X-ray crystallography (Fig. 3). Finally,
Pinner reaction provided inhibitor ( )-2.
For the synthesis of inhibitor ( )-8 with the F-substituent in
position 3 of the phenylamidinium needle, 3-fluoro-4-formyl-
benzonitrile (9) was prepared from 10 by benzylic bromination
(
11) and oxidation, as described for 3 (Scheme 2).20 1,3-Di-
polar cycloaddition followed by column chromatography
O r g . B i o m o l . C h e m . , 2 0 0 4 , 2, 1 3 3 9 – 1 3 5 2
1340

View Article Online
Fig. 4 ORTEP representation of ( )-12 with vibrational ellipsoids
obtained at 223 K and shown at the 30% probability level.
Fig. 3 ORTEP representation of ( )-7 with vibrational ellipsoids
obtained at 223 K and shown at the 30% probability level.
Fig. 5 ORTEP representation of ( )-22 with vibrational ellipsoids
obtained at 243 K and shown at the 30% probability level.
into 4-formylbenzonitrile 15. 1,3-Dipolar cycloaddition
afforded an unusually complex mixture of diastereoisomers,
and the desired tricyclic product ( )-26 was only obtained after
tedious column chromatography. Attempts to convert ( )-26
into inhibitor ( )-18 by the Pinner reaction failed and gave only
starting material, even after one week. This transformation is
clearly sensitive to the presence of two substituents ortho to the
CN group. Instead, the desired conversion of ( )-26 into ( )-18
was possible by reaction with hydroxylamine, acetylation and
hydrogenation over Pd/C.23
Scheme 2 Synthesis of inhibitor ( )-8. Reagents and conditions: i,
NBS, AIBN, CCl4, ∆; 61%; ii, TMAOؒ2 H2O, Me2SO, CH2Cl2, 0 ЊC;
73%; iv, N-(4-fluorobenzyl)maleimide, -proline, MeCN, ∆; 22%; v,
AcCl, MeOH, CH2Cl2, 5 ЊC; then NH3, MeOH, 65 ЊC; 79%.
Small molecule crystal packing analysis
provided nitrile ( )-12, and its configuration was proven by
X-ray crystallography (Fig. 4). Subsequent Pinner reaction
yielded inhibitor ( )-8.
The crystal packings of the solved small molecule structures
were analysed24–26 to obtain additional information on the pre-
ferred environment for the small, hard and strongly electro-
negative F-substituents and for gaining further insight into
non-covalent interactions involving C–F residues. Such analysis
had previously been extremely useful in establishing weak
C–F ؒ ؒ ؒ H–C –C᎐O and C–F ؒ ؒ ؒ C᎐O interactions.9
The difluorinated benzonitriles 13–15 for the synthesis of
inhibitors ( )-16–( )-18, respectively, (Scheme 3) could be
effectively prepared from commercially available difluorinated
starting materials (19–21) by utilising fluorine-directed ortho-
metallation as the key transformation.21 Compound 13 was
obtained in high yield by ortho-lithiation of 19 with LDA at
Ϫ78 ЊC, followed by quenching with DMF. 1,3-Dipolar cyclo-
addition provided ( )-22, and its configuration was proven by
X-ray crystallography (Fig. 5).
ortho-Lithiated 20 was highly unstable even at Ϫ78 ЊC and
decomposed rapidly to an unidentified polar by-product (TLC).
Quenching with DMF immediately after lithiation afforded
benzonitrile 14 that was converted via ( )-23 into inhibitor
( )-17.
᎐
᎐
α
Close C–F ؒ ؒ ؒ H–C contacts. Several investigations on the
H-bonding potential of organic C–F residues have appeared.25
Whereas the FϪ anion is the strongest known H-bond acceptor,
it has become clear from crystallographic and computational
studies that organic C–F is a considerably weaker H-bond
acceptor than C–N or C–O groups. Thus, Dunitz and Taylor
found only very few cases in the Cambridge Structural Data-
base (CSD) of what, by geometrical considerations, could be
classified as moderate H–bonds with C–F as the acceptor.25a
Whether short contacts between C–F residues and strong
H-bond donors such as O–H, reported in the literature,27 are
true H-bonds or should rather be seen as multipolar inter-
For the synthesis of ( )-18, aldehyde 21 was transformed
into 1,3-dioxolane 24 that was ortho-lithiated and transformed
with tosyl cyanide22 into nitrile 25 and finally, by deprotection,
O r g . B i o m o l . C h e m . , 2 0 0 4 , 2, 1 3 3 9 – 1 3 5 2
1341

View Article Online
Scheme 3 Synthesis of inhibitors ( )-16–( )-18. Reagents and conditions: i, LDA, THF, hexane, Ϫ78 ЊC; then DMF; 82% (13), 35% (14);
ii, N-(4-fluorobenzyl) maleimide, -proline, MeCN, ∆; 29% (( )-22), 37% (( )-23), 15% (( )-26); iii, AcCl, MeOH, CH2Cl2, 5 ЊC; then NH3, MeOH,
65 ЊC; 58% (( )-16), 43% (( )-17); iv, HOCH2CH2OH, TsOHؒH2O, ∆; v) LDA, THF, hexane, Ϫ78 ЊC; then tosyl cyanide; vi, HCl, H2O, THF, ∆; 48%
(over 3 steps); vii, NH2OH, MeOH, THF, then Ac2O, then H2, Pd/C; 39%. LDA = lithium diisopropylamide; DMF = dimethylformamide, THF =
tetrahydrofuran.
actions remains a matter of ongoing discussion. This is even
more true for the increasing number of C–H ؒ ؒ ؒ F–C inter-
actions,9,28 involving very weak H-bond donors, that are
observed in crystal structures. Independent of naming and
classification, these contacts could make small yet selective
energetic contributions to protein–ligand binding.
For the packing analysis described in the following, idealised
C–H distances of 1.08 Å were used. In the crystal packing of
nitrile ( )-7 (see Fig. 3), the shortest intermolecular contact of
the two F-atoms is between F(30) and the exo H-atom of C(20)
(distance d(F ؒ ؒ ؒ H–C) 2.32 Å, d(F ؒ ؒ ؒ C) 3.34, angle
α(F ؒ ؒ ؒ H–C) 157Њ). An interesting intramolecular contact
is seen in which the F(30)-atom of the benzonitrile moiety is
Fig. 6 Short intermolecular Caryl–F ؒ ؒ ؒ H–Caryl contact observed in
the crystal packing of ( )-12.
situated on top of the electrophilic C(12)-atom of the benzyl
C–F group (d(F ؒ ؒ ؒ C–F) 3.09 Å. This C–F ؒ ؒ ؒ C–F inter-
action bears resemblance to the C–F ؒ ؒ ؒ C᎐O dipolar
᎐
(d(O ؒ ؒ ؒ H–C) 2.19 Å, d(O ؒ ؒ ؒ C) 3.25 Å, α(O ؒ ؒ ؒ H–C) 165Њ),
thereby presenting another example for the weak, well-studied
O ؒ ؒ ؒ H–C hydrogen bond.26,32
interaction.9
The crystal packing of ( )-12 (Fig. 6) features an extraordin-
arily short intermolecular contact between F(15) of the benzyl
group and the polarised C(24)–H ortho to the CN group
(d(F ؒ ؒ ؒ H–C) 2.22 Å, d(F ؒ ؒ ؒ C) 3.04 Å, α(F ؒ ؒ ؒ H–C) 131Њ).
This F ؒ ؒ ؒ H distance is significantly shorter than the sum of
the van der Waals radii (2.67 Å) and represents the short-
est Caryl–F ؒ ؒ ؒ H–Caryl contact observed in small molecule
crystal structures.28,29 C–F ؒ ؒ ؒ H–C contacts have not only
been recognised in protein–ligand interactions30 and in small
molecule crystal packings9,28 but have also been implied in
catalytic processes.29,31 The second polarised C(26)–H, ortho to
both CN- and F-substituents, contacts the imide carbonyl O(7)
Close C–F ؒ ؒ ؒ CN contacts. Intramolecular C–F ؒ ؒ ؒ CN
interactions have recently been reported.33 In this investigation,
we also observed several short intermolecular C–F ؒ ؒ ؒ CN con-
tacts below the sum of the van der Waals radii (ca. 3.3 Å). Thus,
in the crystal packing of difluorinated 4-formylbenzonitrile 13
(Fig. 7A), F(11) points directly toward the positively polarised
C(9)-atoms of the cyano groups of two molecules in a neigh-
boring stack (d (F ؒ ؒ ؒ CN) 3.11 Å, d(F ؒ ؒ ؒ NC) 3.36 Å,
α(F ؒ ؒ ؒ CN) 93Њ and d(F ؒ ؒ ؒ CN) 3.21 Å, d(F ؒ ؒ ؒ NC) 3.21 Å,
α(F ؒ ؒ ؒ CN) 80Њ). These polar C–F ؒ ؒ ؒ CN interactions
O r g . B i o m o l . C h e m . , 2 0 0 4 , 2, 1 3 3 9 – 1 3 5 2
1342

View Article Online
Fig. 7 Short intermolecular Caryl–F ؒ ؒ ؒ CN contacts observed in the crystal packings of 13 (A) and ( )-22 (B).
Table 1 Biological activities and physicochemical properties of the
fluorinated tricyclic thrombin inhibitors
probably stabilise the parallel-slipped stacks observed in the
crystal structure. Also, in the crystal packing of ( )-22, a very
short C–F ؒ ؒ ؒ CN contact is found (d(F ؒ ؒ ؒ CN) 3.01 Å,
d(F ؒ ؒ ؒ NC) 3.35 Å, α(F ؒ ؒ ؒ CN) 97Њ; Fig. 7B). These
C–F ؒ ؒ ؒ CN contacts bear strong resemblance to the multi-
c
c
Inhibitor
Ki [µM]a
Selectivityb
pKa1
pKa2
log D
( )-1
0.057
0.63
0.61
1.7
5.0
9.0
67
22
11
7.6
4.6
7.6
4.47
4.21
4.10
4.21
3.73
3.73
11.14
10.69
10.83
10.36
10.28
10.06
Ϫ1.16
polar C–F ؒ ؒ ؒ C᎐O interactions we have previously described.9
᎐
( )-2
Ϫ1.08
These findings prompted us to search the CSD34 for inter-
molecular interactions between C–F and CN moieties in crys-
tals of fluorine-containing organic nitriles. The search resulted
in 91 hits with a total of 261 occurrences of close F ؒ ؒ ؒ CN
contacts below 4.0 Å. There are quite a number of cases (44
occurrences, Fig. 8) with contact distances below the sum of the
van der Waals radii (ca. 3.3 Å), indicating an attractive inter-
action between the two polar groups. An analysis of the relative
orientation of the two interacting groups in terms of the non-
bonded distance d1(F ؒ ؒ ؒ CN) and the angles α1(F ؒ ؒ ؒ CN)
and α2(C–F ؒ ؒ ؒ N) reveals that, at short contact distances, the
F-atoms prefer a position perpendicular to the cyano group
(Fig. 8). The C–F bond is generally inclined to the F ؒ ؒ ؒ C axis
with angles α1 adopting values typically between 100Њ–170Њ
at short contact distances. Thus, the results of the CSD search
nicely parallel those of our previous investigations on
d
( )-8
d
d
( )-16
( )-17
( )-18
Ϫ0.35
a The uncertainty of the measured Ki values is
20%. b Ki(trypsin)/
Ki(thrombin). c pKa1: tertiary amine in tricyclic skeleton, pKa2: phenyl-
amidinium. Reproducibility of pKa values: 0.05 units. d Not deter-
mined.
measured for each compound, depending on whether the
potentiometric titration was conducted from pH ≈2 to pH ≈13
or in the reverse way from pH ≈13 to pH ≈2. The effect was
particularly pronounced for the tertiary amino group (pKa1),
yielding a higher apparent pKa value (by ca. 4 units) in the
titration starting from basic solution. The effect on the pKa of
the amidinium residue (pKa2) was smaller, with a discrepancy of
only ca. 0.2 units. Subsequent chemical analysis revealed that
this discrepancy resulted from the instability of the inhibitors in
basic solution and that the lower values included in Table 1,
obtained by titrations starting from acidic solution, are the cor-
rect ones. In basic solution (0.05 M NaOH in THF/H2O 2 : 1),
compounds such as ( )-277d undergo a surprisingly rapid,
regioselective opening of the succinimide ring in the tricyclic
skeleton to afford the corresponding sodium carboxylate ( )-
C–F ؒ ؒ ؒ C᎐O interactions.
᎐
1
28, as revealed by H and 13C NMR as well as high-resolution
mass spectrometry (HR-MS) (Scheme 4). The proposed con-
figuration of ( )-28 was confirmed by a combination of 2D
NMR sequences, including COSY (correlation spectroscopy),
HSQC (heteronuclear single-quantum correlation) and HMBC
(heteronuclear multiple bond correlation).35 In ( )-28, the pKa
value of the tertiary amine centre is raised by ca. 4 units, as
compared to ( )-27. Base-induced ring opening in 0.05 M
1
Fig. 8 Scatterplots of α1 vs. d1 (left) and α2 vs. d1 (right) of 261
occurrences of close C–F ؒ ؒ ؒ CN contacts with d1 < 4.0 Å obtained
from 91 hits of the CSD search of fluorine-containing organic nitriles.
NaOD in D2O was also confirmed by H NMR spectroscopy
for inhibitor ( )-29 (Scheme 4).9a Derivatisation of the carb-
oxylate functionality in ( )-28 and analogs with an appropriate
P-pocket binding element promises to generate a new class of
bicyclic thrombin inhibitors; this perspective is currently being
further explored.
Influence of F-substitution on the pKa values of the inhibitors
The influence of F-substitution on the pKa values of the ter-
tiary amino group in the tricyclic skeleton (pKa1) and the
amidinium residue (pKa2) of the inhibitors was investigated
(Table 1). Surprisingly, two different sets of pKa values were
The surprisingly low pKa value (5.2) of the tertiary amino
group in the tricyclic skeleton of ( )-27 was further investigated
in a series of related compounds, ( )-30a/b and ( )-31. The
O r g . B i o m o l . C h e m . , 2 0 0 4 , 2, 1 3 3 9 – 1 3 5 2
1343

View Article Online
Thus, starting from the typical pKa value of a trialkylamine
around 10, consideration of the two σ-inductive effects (∆pKa:
2 × 1.5 units) of the β-carbonyl groups leads to the pKa value
(7.04) of ( )-31. Taking further into account the α-phenyl effect
(∆pKa ca. 1 unit) gives the pKa values measured for ( )-30a
(5.63) and ( )-30b (5.87). Consideration of the larger σ-induct-
ive effects of 4-bromo- or 4-cyano-substituted (see below)
α-phenyl rings rationalises the further decrease in pKa to a value
of 5.2 in ( )-27.
Although ( )-28 features two C᎐O groups in the β-position
᎐
too, their σ-effects are reduced since they are less electro-
philic (compared to the imide function) and since such
effects are better transmitted across rigid rather than flexible
σ-scaffolds. Also, we do not wish to exclude some contribu-
tions from the carboxylate to the higher pKa. Clearly, this dis-
cussion shows that inductive effects of functional groups on
amine centres in bi- and tricyclic scaffolds can have a dramatic
effect on their pKa values, which is worth considering in the
optimisation of the pharmacokinetic properties of a lead
compound.
Following full clarification of the initially surprising dis-
crepancies obtained during pH-titrations of acidic and basic
solutions, we investigated the effects of fluorine substitution on
the pKa values. Exchanging hydrogen with fluorine in the
phenylamidinium needle reduces the basicity of the amidine
group. Specifically, ortho-fluorination decreases the pKa2 value
more than meta-substitution (( )-2 vs. ( )-8, Table 1). Not
surprisingly, the pKa2 values of difluorinated inhibitors ( )-16,
( )-17 and ( )-18 are further reduced compared to those of
the monosubstituted derivatives, with the pKa2 value of ( )-18
(10.06) being one magnitude lower than that of non-fluorin-
ated ( )-1 (11.14). In parallel, the lipophilicity improves as
measured by log D at pH 7.4 from Ϫ1.16 (( )-1) to Ϫ0.35
(( )-18).
The same trend (reduced pKa1 with increasing F-substitution)
was observed for the basicity of the tertiary amine centre in
the tricyclic skeleton, with one notable exception (Table 1).
Introduction of one F-atom meta to the amidinium residue
in ( )-8 expectedly decreases the pKa1 value to 4.10 (from
4.47 in ( )-1). In sharp contrast, introduction of a second meta-
Scheme 4 Regioselective ring opening of the tricyclic scaffold in
( )-27 and ( )-29 under basic conditions: i, 0.05 M NaOH in THF/H2O
2 : 1. Shown are additional comparison compounds together with the
pKa values of their tertiary amine centres. Precipitation occurred during
the titration of ( )-27.
endo- (( )-30a) and exo- (( )-30b) phenyl derivatives have pKa
values of 5.63 and 5.87, respectively, reflecting a minor but dis-
tinct configurational dependency. Omitting the phenyl ring in
the α-position to the tertiary amine centre has a strong effect on
the pKa value: it increases to 7.04 in compound ( )-31, lacking
such substituent.
The interesting pKa values for the tertiary amino groups
shown in Table 1 and Scheme 4 warrant further discussion. Our
initial hypothesis to explain the high pKa value of ( )-28 (com-
pared to ( )-27) mainly involved the presence of a negatively
charged carboxylate residue that would stabilise the protonated
amine by charge attraction. However, a more detailed analysis
suggests that large charge effects are not necessary to explain
the measured value. Tertiary amines typically have pKa values
around 10–10.3, and a search in the MedChem02 database36
shows that the introduction of a phenyl group in the α-position
to the N-atom reduces the pKa value by typically ca. one unit
(sometimes even slightly more), which would rationalize the
pKa value of 9.3 measured for ( )-28. Indeed, an α-phenyl effect
in this order of magnitude (and even slightly larger) is evident
from the comparison of the pKa values of ( )-31 vs. ( )-30a
and ( )-30b.
F-substituent in ( )-16 increases the basicity again to pKa1 =
4.21. This interesting effect was fully confirmed in the corre-
sponding series of nitrile precursors: pKa = 4.12 (( )-7), 4.16
(( )-12), 4.36 (( )-22), 3.58 (( )-23) and ( )-26).
The basicity of amines is influenced by several factors:
inductive and resonance effects (through bond), field effects
(through space), solvation, steric hindrance and internal
H-bonding.37 Equilibrium geometry determinations by density
functional theory (DFT) calculations (RB3LYP, 6–31G**)38
were conducted on the simple analogs 32a–d and 33a,b for the
nitrile precursors ( )-12 and ( )-22, respectively, to shed fur-
ther light onto the surprising pKa effect. In agreement with the
crystal structure analysis, it was found that conformer 32a with
the exo-F-substituent (exo with respect to the dioxoperhydro-
pyrrolo[3,4-c]pyrrole skeleton) is more stable than the endo con-
former 32b by 14.2 kJ molϪ1 (difference in heat of formation
However, it remains to rationalise the unusually low pKa
values of the tertiary amine centre in the tricyclic scaffolds such
as ( )-27 and ( )-30a/b. An important pKa-lowering contri-
bution arises from through-bond σ-effects of the two imide
carbonyl units on the amine. A search in the MedChem02
Database revealed substantial effects on amine basicities: intro-
ducing a keto function in the β-position to a saturated amine
∆Ho , Table 2). The exo-conformer observed in the crystal
f
structure (Fig. 4) minimises 1,3-allylic-like strain and unfavor-
able interactions between the C–F dipole and the nitrogen lone
pair.39 N-Protonation reverses the stability, with the endo-
conformer 32d now being stabilised by 17.1 kJ molϪ1 over the
exo-conformer 32c as a result of favorable C–F ؒ ؒ ؒ H–Nϩ
contacts.
The acid–base equilibrium of difluorinated model compound
33 was also evaluated by calculating the heats of formation of
free base 33a and protonated 33b, and the data compared to the
corresponding equilibrium between monofluorinated free base
32a (exo-F) and protonated 32d (endo-F; Table 2). These
calculations revealed that the acid–base equilibrium of the
difluorinated compound is shifted by Ϫ20.4 kJ molϪ1 toward
(structural element N–C(sp3)–C(sp3)–C᎐O) results in a lowering
᎐
of the amine pKa by ca. 1.5 units! Examples are the change from
piperidine (pKa 11.1) to 2-(2Ј-oxopropyl)piperidine (pKa 9.45)
or from 2,2,6,6-tetramethylpiperidine (pKa 11.1) to 4-oxo-
2,2,6,6-tetramethylpiperidine (pKa 7.9). Note that in the latter
case, there are two σ-paths; therefore the σ-inductive effect is
operating twice and hence the pKa shift is 2 × 1.5 = 3 pKa units.
O r g . B i o m o l . C h e m . , 2 0 0 4 , 2, 1 3 3 9 – 1 3 5 2
1344

View Article Online
Table 2 Heats of formation ∆Ho for unprotonated (32a,b, 33a) and
Table 3 Dihedral angles describing the orientation of the phenyl-
f
protonated (32c,d, 33b) model compounds, obtained by DFT calcu-
amidinium needle of inhibitor 1 bound in the active site of thrombin
(PDB-code: 1OYT) and of benzonitrile rings as seen in crystal struc-
tures of nitrile precursors
lations (RB3LYP, 6-31G**)
Compound
X
Y
Hof [Hartree]a
Compound
ω
ꢀ
32a
32b
32c
32d
33a
33b
H
F
H
F
F
F
F
H
F
H
F
F
Ϫ1072.0210798
Ϫ1072.0156243
Ϫ1072.3934282
Ϫ1072.4002484
Ϫ1072.2422662
Ϫ1072.6293077
1a
29Њ
40Њ
24Њ
9Њ
54Њ
43Њ
56Њ
Nitrile precursor to ( )-1a
23Њ
27Њ
33Њ
31Њ
Ϫ8Њ
( )-7
( )-12
( )-22
( )-22ؒHCl
a 1 Hartree = 2625 kJ molϪ1
.
a From ref. 9.
Influence of F-substitution on conformational properties
Exchanging H with F in the meta positions to the amidinium
moiety in the needle influences the conformational behaviour
of the inhibitor. In the 1H NMR spectra of ( )-16 and its nitrile
precursor ( )-22 at room temperature, the two protons ortho to
the functional group appear as two distinct broad doublets and
three fluorine signals are observed in the 19F NMR spectrum. In
contrast, the two protons of ( )-18 and ( )-26 in the meta pos-
ition to the amidinium/nitrile groups give one singlet and only
two peaks are observed in the 19F NMR spectrum. Obviously,
the barrier to rotation around the C–C bond connecting the
phenyl ring to the tricyclic skeleton is increased by F-substi-
tution in the meta positions, and rotation at room temperature
becomes slow on the NMR time scale. By variable-temperature
1H NMR spectroscopy, the kinetic parameters for the rotation
in nitrile ( )-22 were determined as ∆H≠ = 57.0 kJ molϪ1 and
the protonated form 33b compared with the monofluorinated
system. These computational results are in agreement with the
observed pKa effects.
Finally, the crystal structure of the HCl salt of ( )-22 was
solved (Fig. 9). The endo-F(30)-atom is found in close
intramolecular contacts with the H–Nϩ unit and the C-atom
of the imide carbonyl group C(1)᎐O(6) (d(F ؒ ؒ ؒ Nϩ) 2.83 Å,
᎐
d(F ؒ ؒ ؒ C᎐O) 3.11 Å). Similar favourable polar contacts should
᎐
stabilise the endo-conformation of monoprotonated, mono-
fluorinated ( )-12 (and of model compound 32d). In the crystal
structure of free amine ( )-22 (Fig. 5), however, the endo-F-
atom is twisted away from the nitrogen lone pair as clearly indi-
cated by the observed dihedral angle ω (C(23)–C(22)–C(16)–
N(17)) of Ϫ54.3Њ, the largest in all crystal structures (see Table 3
below). F(30) is in very close intramolecular contact with the
∆S≠ = Ϫ37.7 kJ molϪ1
.
Fig. 10 displays a succinimide ring-based overlay of the
conformation of 1 bound in the active site of thrombin9 with
the conformations of various benzonitrile precursors (nitrile
precursor of ( )-19 and nitriles ( )-7, ( )-12, ( )-22 and
( )-22ؒHCl) in crystal structures. The enzyme-bound phenyl-
aminidium needle adopts a different orientation than the benzo-
imide carbonyl group C(3)᎐O(7) (d(F ؒ ؒ ؒ C᎐O) 2.74 Å) and the
᎐
᎐
imide N-atom (d(F ؒ ؒ ؒ N) 3.09 Å). These results indicate that
the increase in basicity upon changing from ( )-12 to ( )-22
and from ( )-8 to ( )-16, respectively, results mainly from
intramolecular electrostatic field effects, i.e. unfavourable inter-
actions of the C–F dipole and the nitrogen lone pair in the free
base forms of ( )-22 and ( )-16. In the free base forms of ( )-8
and ( )-12, these interactions can be avoided by rotation of the
phenyl ring.
Fig. 10 Succinimide ring-based overlay of 1 (color code: yellow)
bound in the active site of thrombin9 with the crystal structures of
various phenylnitrile precursors, the nitrile precursor to ( )-1 (brown),9
( )-7 (red), ( )-12 (dark green), ( )-22 (blue) and ( )-22ؒHCl (cyan).
Fig.
9
ORTEP representation of the HCl salt of ( )-22 with
vibrational ellipsoids obtained at 223 K and shown at the 30%
probability level.
O r g . B i o m o l . C h e m . , 2 0 0 4 , 2, 1 3 3 9 – 1 3 5 2
1345

View Article Online
nitrile residues in the crystals; whether energetic costs are
associated with the re-orientation, which is possibly required
for binding into the S1-pocket, remains unclear. We defined two
torsional angles, ω (Caryl–Caryl–C(sp3)–N)) and ꢀ (Caryl–C(sp3)–
C(sp3)–CO)) to further analyse conformational preferences of
the benzonitrile rings (Table 3); however, apart from the specific
geometry adopted by ( )-22 to avoid repulsions between one
C–F dipole and the nitrogen lone pair (see above), no particular
effects resulting from F-substitution could be readily identified.
It also remains unclear how the introduction of F-substituents
ortho to the amidinium moiety affects the barrier for rotation of
this group about the bond connecting it to the phenyl ring.40
Biological results
The Ki values for the complexes of the fluorinated tricyclic
inhibitors with thrombin and trypsin were determined using a
chromogenic substrate (Table 1).41 The effect of F-substitution
is in all cases detrimental for inhibitory activity. Introduction of
one fluorine substituent in the 2- or 3-position of the phenyl-
amidinium needle reduces the inhibitory activity by one order
of magnitude. The difluorinated inhibitors are even less potent,
with activities reduced up to two orders of magnitude. It is
interesting to note that host–guest complexation studies with
synthetic receptors came to different conclusions, namely that
the strength of phenylamidinium ؒ ؒ ؒ carboxylate salt bridges
is slightly increased upon para-substitution of the phenyl-
amidinium moiety with electron-withdrawing groups.42
The detrimental effect of fluorine substitution of the phenyl-
amidinium needle on binding affinity is of complex origin.
Several characteristics of the inhibitors, such as acid–base
properties (pKa1 and pKa2), log D (Table 1) and conformational
preferences are affected in addition to the direct interactions
with the protein environment. The results are, however, in full
agreement with the previously reported “σ-effect” for substi-
tuted benzamidinium substrates,11–13 stating that electron-with-
drawing substituents decrease biological activity. Molecular
modeling using the MOLOC system43 showed that the intro-
duction of the F-substituents did not generate any repulsive van
der Waals contacts, however, potentially unfavourable electro-
static contacts between 2-F (ortho to the amidinium) with the
crystal-bound water and the carbonyl O-atom of Gly219 were
identified.
Strong linear free energy relationships (LFERs) between the
binding affinities (Ϫ∆G) to both enzymes and the pKa of the
amidinium residue of the inhibitors were found (Fig. 11). At
the pH (8.00) of the assay, the amidines in all inhibitors are
fully protonated. Interestingly, the dependence of the binding
affinity from the basicity of the amidine residues is much larger
for thrombin (slope of the linear regression: 2.18) than for
trypsin (slope 0.88). This means that the pKa of the phenyl-
amidinium residue not only affects binding affinity but also
selectivity. More basic inhibitors form more stable complexes
and are more selective for thrombin over trypsin and, with
decreasing pKa, much of the selectivity is lost. The results of the
LFERs presented here therefore provide a useful guideline in
the design of efficient and selective thrombin inhibitors.
Fig. 11 Plots of the binding free enthalpies Ϫ∆G for the complexes of
the fluorinated tricyclic inhibitors with thrombin (᭺) and trypsin (×)
against the pKa2 values (Table 1) of the amidinium group. The data were
fitted using a least-squares regression analysis. The slope of the linear
free energy relationships (LFERs) is 2.18 (R2 = 0.97) for thrombin and
0.88 (R2 = 0.86) for trypsin. The Ki value for the complex of trypsin with
( )-18 was above the threshold of the assay (>69 µM) and is therefore
not included in the regression analysis.
slopes for thrombin and trypsin: decreasing the pKa by intro-
duction of F-substituents affects complexation to thrombin
more than to trypsin, and, as a result, binding selectivity is lost.
Our data show that the “σ-effect” of substituents on the affinity
of benzamidinium needles should be rather described as a
“pKa-effect”.
A surprising increase in basicity of the tertiary amine centre
in the tricyclic skeleton of inhibitor ( )-16 (Table 1) could be
explained by DFT calculations and crystal structure analysis:
unfavourable interactions between a C–F dipole and the nitro-
gen lone pair in the free base are eliminated upon protonation.
Analysis of crystal packing environments for F-contacts
revealed an extraordinarily short Caryl–F ؒ ؒ ؒ H–Caryl inter-
action (d(F ؒ ؒ ؒ H–C) 2.22 Å), comparable to the well-
described, weak O ؒ ؒ ؒ H–C hydrogen bonds. Furthermore,
several examples of intermolecular C–F ؒ ؒ ؒ CN contacts
significantly below the sum of the van der Waals radii were
observed. Additional evidence for such interactions and their
orientational preferences—with the F-atom approaching the
electrophilic carbon atom of the cyano group in a perpendicu-
lar fashion—was obtained by searching the CSD. This inter-
action bears resemblance to the multipolar C–F ؒ ؒ ؒ C᎐O inter-
᎐
action that we have previously reported.9
Experimental
General details
Solvents and reagents were purchased reagent-grade and used
without further purification. 3-Fluoro-4-methylbenzonitrile
and 4-bromo-3-fluorotoluene were purchased from ABCR, 3,5-
difluorobenzaldehyde from Lancaster and 2,3-difluoro- and
3,5-difluorobenzonitrile from Acros. CH2Cl2 was freshly dis-
tilled from CaH2; THF was distilled form Na/benzophenone.
Concentration in vacuo is done at a pressure of ca. 20 Torr.
Column chromatography (CC) on SiO2 60 (230–400 mesh,
0.040–0.063 mm) from Fluka. TLC on SiO2 60 F245 from
Merck; visualisation by UV light at 245 nm. Melting points
(mp) were determined with a Büchi SMP-20 apparatus and are
uncorrected. In several cases, compounds decomposed (dec.)
upon mp determination. IR spectra were obtained on a Perkin-
Conclusions
The fluorine scan of the phenylamidinium needle in a series of
tricyclic thrombin inhibitors revealed that the introduction of
σ-accepting F-atoms is detrimental to binding affinity, in full
agreement with the previously reported “σ-effect” of substit-
uents in simple benzamidinium ligands.11–13 Improvement of
the pharmacokinetic properties of thrombin inhibitors through
F-substitution of the phenylamidinium needle seems therefore
not to be a viable strategy despite the desirable, significant low-
ering of the pKa of the amidinium moiety. Moreover, the strong
LFERs between binding affinity and pKa values have different
1
Elmer 1600-FT IR instrument. H, 13C and 19F NMR spectra
were obtained on a Varian Gemini 300 or a 500 GRX Bruker
spectrometer. The solvent peak was used as internal reference;
for 19F NMR spectra, CFCl3 was used as internal reference. In
the 13C NMR spectra of ( )-12 and ( )-23, one aliphatic signal
O r g . B i o m o l . C h e m . , 2 0 0 4 , 2, 1 3 3 9 – 1 3 5 2
1346

View Article Online
is missing and in the spectrum of ( )-17, one aromatic signal is
missing, as a result of signal overlap. The exchangeable amidin-
ium protons were not observed in 1H NMR spectra obtained in
CD3OD. MALDI-high-resolution-mass-spectrometry (HRMS)
was performed with 2,5-dihydroxybenzoic acid as matrix on a
IonSpec Ultima spectrometer. Molecular ions (Mϩ) reported
for phenylamidinium salts refer to the corresponding phenyl-
amidine derivatives. Elemental analyses were performed by the
Mikrolabor at the Laboratorium für Organische Chemie, ETH
Zürich.
H 4.70, F 9.33, N 10.31; found: C 67.93, H 4.94, F 9.33, N
10.25%.
2-Fluoro-4-[(3aSR,4RS,8aSR,8bRS )-2-(4-fluorobenzyl)-1,3-
dioxodecahydropyrrolo[3,4-a]pyrrolizin-4-yl]benzamidine
hydrochloride (( )-2)
Compound ( )-2 was synthesised according to the procedure
described below for ( )-8 except that the mixture was allowed
to stand for 3 d at 5 ЊC before precipitation with Et2O. CC
(CH2Cl2/CH3OH 99 : 10
90 : 10), followed by precipitation
from EtOH/Et2O gave ( )-2 (65 mg, 56%) as colourless solid,
mp > 212 ЊC (dec.); νmax(KBr)/cmϪ1 3376, 3109, 2962, 1776,
1709, 1674, 1627, 1613, 1487, 1471, 1402, 1348, 1237, 1180;
δH(300 MHz; CD3OD) 1.70–1.90 (2 H, m), 2.01–2.16 (2 H, m),
2.56–2.66 (1 H, m), 2.80–2.94 (1 H, m), 3.47 (1 H, dd, J 7.8,
0.6), 3.69 (1 H, dd, J 9.3, 7.2), 3.78 (1 H, t, J 8.3), 4.31 (1 H, d,
J 8.7), 4.47, 4.53 (2 H, AB, J 14.2), 6.99–7.09 (2 H, m), 7.21 (1
H, dd, J 11.7, 1.2), 7.24–7.30 (2 H, m), 7.32 (1 H, dd, J 8.1, 1.2),
7.56 (1 H, t, J 7.7); δC(75 MHz; CD3OD) 24.2, 30.5, 42.4, 50.2,
51.8, 52.0, 68.9, 69.5, 116.1 (d, J 21.9), 117.0 (d, J 9.8), 117.3 (d,
J 18.8), 126.1 (d, J 3.0), 130.3, 131.2 (d, J 8.0), 133.2 (d, J 3.0),
149.2 (d, J 7.9), 160.5 (d, J 251.3), 163.5 (d, J 242.9), 164.3,
176.9, 179.7; δF(282 MHz; CD3OD ϩ CFCl3) Ϫ112.9 (1 F, dd,
4-Methyl-2-fluorobenzonitrile (5)17
To 4 (2.46 g, 14.0 mmol) in DMF (40 ml), CuCN (2.50 g,
28 mmol) was added. The solution was degassed and heated to
reflux under Ar for 24 h. After cooling to room temperature
(rt), conc. NH3 (30 cm3) and Et2O (50 cm3) were added and the
mixture stirred for 1 h. After the addition of saturated (sat.)
aqueous (aq.) NaCl solution (100 cm3) and Et2O (100 cm3), the
formed phases were separated and the aqueous phase extracted
with Et2O (2 × 50 cm3). The combined organic phases were
washed with sat. aq. NaCl solution (2 × 100 cm3), dried
(Na2SO4) and concentrated in vacuo. CC (Et2O/pentane 1 : 5)
afforded 5 (1.89 g, ≈ quant.) as colourless crystals, mp 52–
53 ЊC; δH(300 MHz; CDCl3) 2.43 (3 H, s), 7.03 (1 H, d, J 9.6),
7.06 (1 H, d, J 9.3), 7.50 (1 H, t, J 7.5).
ϩ
J 11.1, 6.9), Ϫ114.0 (1 F, m); HRMS calcd for C23H23F2N4O2
(MHϩ): 425.1789; found 425.1780.
4-Bromomethyl-2-fluorobenzonitrile (6)17
4-Bromomethyl-3-fluorobenzonitrile (11)20
To 5 (1.35 g, 10 mmol), NBS (1.87 g, 10.5 mmol) and AIBN
(164 mg, 1.00 mmol) in CCl4 (45 cm3) were added. After heating
to reflux for 8 h, the solution was filtered and the mother liquor
concentrated in vacuo. CC (hexane/EtOAc 6 : 1) afforded 6
(1.18 g, 55%) as colourless needles, mp 44–45 ЊC; δH(300 MHz;
CDCl3) 4.45 (2 H, s), 7.24–7.31 (2 H, m), 7.61 (1 H, dd, J 8.1,
6.6).
A mixture of 10 (1.15 g, 8.5 mmol), NBS (1.51 g, 8.5 mmol) and
AIBN (140 mg, 0.85 mmol) in CCl4 (40 cm3) was heated to
reflux for 8 h. Filtration and evaporation of the filtrate in vacuo,
followed by CC (hexane/EtOAc 9 : 1) provided 11 (1.11 g, 61%)
as colourless needles, mp 75–76 ЊC; δH(300 MHz; CDCl3) 4.50
(2 H, s), 7.38 (1 H, dd, J 9.3, 1.5), 7.46 (1 H, dd, J 7.8, 1.5), 7.53
(1 H, t, J = 7.8).
2-Fluoro-4-formylbenzonitrile (3)17
3-Fluoro-4-formylbenzonitrile (9)
A solution of 6 (0.81 g, 2.8 mmol) in CH2Cl2 (2.3 cm3) and
Me2SO (4.7 cm3) was cooled to 0 ЊC, and TMAOؒ2 H2O (1.24 g,
11.2 mmol) was added. After stirring for 5 h at 0 ЊC, the mixture
was poured onto ice (25 cm3) and sat. aq. NaCl solution
(25 cm3) and the aqueous phase extracted with Et2O (4 ×
35 cm3). The combined organic phases were washed with sat.
aq. NaCl solution (2 × 35 cm3), dried (Na2SO4) and concen-
trated in vacuo. CC (CH2Cl2/hexane 4 : 6) afforded 3 (0.31 g,
74%) as colourless needles, mp 87–88 ЊC; δH(300 MHz; CDCl3)
7.73 (1 H, dd, J 8.4, 1.2), 7.80 (1 H, dd, J 7.8, 1.2), 7.85 (1 H, dd,
J 7.8, 5.7), 10.06 (1 H, d, J 1.8).
A solution of 11 (0.81 g, 3.8 mmol) in CH2Cl2 (3 cm3)
and Me2SO (6 cm3) was cooled to 0 ЊC, and TMAOؒ2 H2O
(1.69 g, 15.2 mmol) was added. After stirring for 5 h at 0 ЊC,
the mixture was poured onto ice (30 cm3) and sat. aq. NaCl
solution (30 cm3) and the aqueous phase extracted with Et2O
(4 × 45 cm3). The combined organic phases were washed with
sat. aq. NaCl solution (2 × 45 cm3), dried (Na2SO4) and
concentrated in vacuo to give 9 (0.41 g, 73%) as colourless
needles, mp 76–77 ЊC; δH(300 MHz; CDCl3) 7.53 (1 H, dd, J 9.3,
1.2), 7.59 (1 H, dd, J 8.1, 0.6), 8.00 (1 H, dd, J 7.8, 7.5), 10.41
(1 H, s).
2-Fluoro-4-[(3aSR,4RS,8aSR,8bRS )-2-(4-fluorobenzyl)-1,3-
dioxodecahydropyrrolo[3,4-a]pyrrolizin-4-yl]benzonitrile (( )-7)
3-Fluoro-4-[(3aSR,4RS,8aSR,8bRS )-2-(4-fluorobenzyl)-1,3-
dioxodecahydropyrrolo[3,4-a]pyrrolizin-4-yl]benzonitrile
(( )-12)
Compound ( )-7 was synthesised from 3 according to the pro-
cedure described below for ( )-12. CC (2 ×; EtOAc/hexane
55 : 45 and EtOAc/CH2Cl2 3 : 97) followed by recrystallisation
(EtOAc/hexane) afforded ( )-7 (102 mg, 14%) as colourless
crystals, mp 155–156 ЊC; νmax(KBr)/cmϪ1 2234, 1772, 1708,
1621, 1604, 1570, 1512, 1430, 1397, 1339, 1228, 1172, 1160;
δH(300 MHz; CDCl3) 1.60–1.88 (2 H, m), 1.95–2.22 (2 H, m),
2.52–2.62 (1 H, m), 2.81–2.93 (1 H, m), 3.31 (1 H, dd, J 8.1,
0.6), 3.53 (1 H, t, J 8.3), 3.75 (1 H, dd, J 10.1, 7.4), 4.10 (1 H,
d, J 8.7), 4.45, 4.51 (2 H, AB, J 14.1), 6.93–7.02 (2 H, m), 7.05–
7.14 (2 H, m), 7.22–7.30 (2 H, m), 7.48 (1 H, dd, J 8.1, 6.6);
δC(75 MHz; CDCl3) 23.4, 29.7, 41.8, 48.9, 48.9, 50.4, 50.8,
67.9, 100.3 (d, J 15.8), 113.9, 115.3 (d, J 21.8), 115.8 (d, J 20.6),
124.4 (d, J 3.0), 130.5 (d, J 8.5), 131.2 (d, J 3.7), 132.8, 147.2 (d,
J 7.3), 162.2 (d, J 245.4), 162.8 (d, J 256.9), 174.4, 177.2; δF(282
MHz; CDCl3 ϩ CFCl3) Ϫ105.9 (1 F, dd, J 9.6, 6.5), Ϫ113.5
A mixture of 9 (358 mg, 2.40 mmol), N-(4-fluorobenzyl)-
maleimide (492 mg, 2.40 mmol) and -proline (276 mg, 2.40
mmol) in MeCN (7.4 cm3) was stirred under reflux for 10 h.
After cooling to rt and filtration, the filtrate was concentrated
in vacuo and CC (EtOAc/hexane) (6 : 4)) followed by recrystal-
lisation (EtOAc/hexane) afforded ( )-12 (211 mg, 22%) as
colourless crystals, mp 159–160 ЊC; νmax(KBr)/cmϪ1 2230, 1778,
1707, 1607, 1573, 1515, 1498, 1429, 1418, 1400, 1342, 1232,
1181, 1160; δH(300 MHz; CDCl3) 1.63–1.88 (2 H, m), 1.96–2.22
(2 H, m), 2.56–2.66 (1 H, m), 2.86–2.98 (1 H, m), 3.33 (1 H,
d, J 7.8), 3.67 (1 H, t, J 8.3), 3.76 (1 H, dd, J 10.1, 7.4), 4.30
(1 H, d, J 8.4), 4.45 (2 H, s), 6.90–6.99 (2 H, m), 7.17–7.24 (2 H,
m), 7.26–7.34 (2 H, m), 7.40 (1 H, t, J 7.5); δC(75 MHz; CDCl3)
23.5, 29.8, 41.7, 48.6, 51.0, 61.8, 67.8, 112.3 (d, J 9.7), 115.2
(d, J 21.2), 117.6 (d, J 3.0), 118.0 (d, J 24.9), 128.0 (d, J 3.7),
129.2 (d, J 4.9), 130.5 (d, J 8.5), 131.3 (d, J 3.1), 131.8 (d,
J 13.4), 160.6 (d, J 24.0), 162.2 (d, J 245.0), 174.5, 177.3; δF(282
ϩ
(1 F, m); HRMS calcd for C23H20F2N3O2 (MHϩ): 408.1524;
found 408.1518; EA calcd for C23H19F2N3O2 (407.41): C 67.81,
O r g . B i o m o l . C h e m . , 2 0 0 4 , 2, 1 3 3 9 – 1 3 5 2
1347

View Article Online
MHz; CDCl3 ϩ CFCl3) Ϫ113.6 (1 F, m), Ϫ114.4 (1 F, s, br);
(1 H, m), 3.32 (1 H, dd, J 8.4, 1.5), 3.58 (1 H, t, J 8.4), 3.83 (1 H,
ϩ
HRMS calcd for C23H20F2N3O2 (MH)ϩ: 408.1524; found
t, J 8.4), 4.48 (1 H, d, J 8.7), 4.54, 4.60 (2 H, AB, J 13.8), 6.92–
7.01 (2 H, m), 7.04 (1 H, d, br, J 10.5), 7.20 (1 H, d, br, J 9.0),
7.28–7.36 (2 H, m); δC(75 MHz; CDCl3) 23.8, 30.1, 42.1, 49.0,
49.6, 51.9, 60.3, 68.1, 113.0 (t, J 12.4), 114.8 (dd, J 27.3, 3.6),
115.4 (d, J 21.2), 116.3 (t, J 2.9), 116.3 (dd, J 26.6, 3.3), 120.4
(t, J 13.7), 130.7 (d, J 8.5), 131.0 (d, J 3.0), 161.0 (dd, J 248.0,
9.1), 161.8 (dd, J 256.1, 6.7), 162.2 (d, J 244.4), 175.1, 177.0;
δF(282 MHz; CDCl3 ϩ CFCl3) Ϫ102.3 (1 F, t, J 7.6), Ϫ110.7
408.1515.
3-Fluoro-4-[(3aSR,4RS,8aSR,8bRS )-2-(4-fluorobenzyl)-1,3-
dioxodecahydropyrrolo[3,4-a]pyrrolizin-4-yl]benzamidine hydro-
chloride (( )-8)
To a 10 cm3 oven-dried flask, CH2Cl2 (1.0 cm3) and MeOH
(1.0 cm3) were added under N2 and the solution cooled to 0 ЊC.
AcCl (1.0 cm3) was added dropwise and, after 5 min, ( )-11
(101 mg, 0.25 mmol) in CH2Cl2 (0.5 cm3) was added and the
mixture allowed to stand for 36 h at 5 ЊC. The formed imidate
precipitated out as a white solid upon addition of Et2O (15 cm3)
and was washed with Et2O (2 × 15 cm3) and dried in vacuo. To
the imidate, MeOH (0.5 cm3) and a 2.0 M methanolic solution
of NH3 (0.5 cm3, 1 mmol) were added and the mixture was
stirred for 3.5 h at 65 ЊC. After cooling, NH4Cl was precipitated
out by addition of acetone (15 cm3). The filtrate was concen-
trated in vacuo and the residue purified by CC (CH2Cl2/MeOH
ϩ
(1 F, t, J 7.6), Ϫ113.6 (1 F, m); HRMS calcd for C23H19F3N3O2
(MHϩ): 426.1429; found 426.1420; EA calcd for C23H18F3N3O2
(425.40): C 64.94, H 4.26, F 9.88, N 13.40; found: C 65.07, H
4.39, F 9.65, N 13.39%.
3,5-Difluoro-4-[(3aSR,4RS,8aSR,8bRS )-2-(4-fluorobenzyl)-1,3-
dioxodecahydropyrrolo[3,4-a]pyrrolizin-4-yl]benzamidine
hydrochloride (( )-16)
Compound ( )-16 was synthesised from ( )-22 according to
the procedure described above for ( )-8. The crude product was
99 : 1
90 : 10). The product was dissolved in EtOH (0.5 cm3)
purified by CC (CH2Cl2/MeOH 99 : 1
90 : 10), and precipit-
and re-precipitation with Et2O yielded ( )-8 (91 mg, 79%) as a
colourless solid, mp >160 ЊC (dec.); νmax(KBr)/cmϪ1 3380, 3067,
2960, 1772, 1705, 1606, 1512, 1484, 1433, 1400, 1344, 1223,
1176; δH(300 MHz; CD3OD) 1.76–1.92 (2 H, m), 2.01–2.21
(2 H, m), 2.60–2.71 (1 H, m), 2.89–3.01 (1 H, m), 3.49 (1 H,
dd, J 8.0, 0.8), 3.73 (1 H, dd, J 9.6, 7.2), 3.85 (1 H, t, J 8.1),
4.40–4.52 (3 H, m), 6.98–7.07 (2 H, m), 7.16–7.25 (2 H, m), 7.46
(1 H, dd, J 8.1, 2.0), 7.51–7.59 (2 H, m); δC(75 MHz; CD3OD)
24.4, 30.6, 42.3, 50.0, 50.3, 52.2, 63.2, 69.3, 115.3 (d, J 24.9),
116.0 (d, J 21.8), 124.5 (d, J 3.0), 129.8 (d, J 8.5), 130.6
(d, J 4.8), 131.0 (d, J 8.5), 133.1 (d, J 3.0), 134.1 (d, J 14.0),
162.4 (d, J 245.9), 163.4 (d, J 242.8), 166.9, 176.7, 179.7; δF(282
MHz; CD3OD ϩ CFCl3) Ϫ113.7 (1 F, s, br), Ϫ114.2 (1 F, m);
ation from EtOH/Et2O gave ( )-16 (70 mg, 58%) as a colourless
solid, mp >160 ЊC (dec.); νmax(KBr)/cmϪ1 3381, 3062, 2960,
1772, 1700, 1604, 1576, 1511, 1429, 1400, 1345, 1305, 1222,
1177; δH(300 MHz; CD3OD) 1.72–1.91 (2 H, m), 2.02–2.21
(2 H, m), 2.64–2.76 (1 H, m), 2.82–2.94 (1 H, m), 3.50 (1 H, dd,
J 8.4, 1.8), 3.73–3.86 (2 H, m, 2H), 4.53, 4.58 (2 H, AB, J 14.7),
4.65 (1 H, d, J 9.0), 6.98–7.08 (2 H, m), 7.24 (1 H, d, br, J 11.1),
7.28–7.36 (2 H, m), 7.49 (1 H, d, br, J 10.2); δC(75 MHz;
CD3OD) 24.7, 31.0, 42.7, 50.5, 51.2, 53.1, 61.5, 69.4, 112.0
(d, J 26.7), 113.2 (d, J 24.3), 116.0 (d, J 21.9), 121.7 (t, J 14.3),
130.9 (t, J 11.0), 131.5 (d, J 8.5), 132.9 (d, J 3.1), 162.5 (dd,
J 247.2, 9.1), 163.1 (dd, J 253.9, 6.7), 163.5 (d, J 242.9), 165.8,
177.3, 179.4; δF(282 MHz; CD3OD ϩ CFCl3) Ϫ101.4 (1 F, dd,
J 10.0, 5.8), Ϫ110.4 (1 F, dd, J 9.6, 6.5), Ϫ114.3 (1 F, m);
ϩ
HRMS calcd for C23H23F2N4O2 (MHϩ): 425.1789; found
ϩ
425.1787.
HRMS calcd for C23H22F3N4O2 (MHϩ): 443.1695; found
443.1694.
3,5-Difluoro-4-formylbenzonitrile (13)
2,3-Difluoro-4-formylbenzonitrile (14)
i
To Pr2NH (0.392 cm3, 2.8 mmol) in THF (5.0 cm3) at 0 ЊC
under Ar, 1.6 M n-BuLi in hexane (1.56 cm3, 2.5 mmol) was
added dropwise over 5 min. After 10 min, the mixture was
cooled to Ϫ78 ЊC and 19 (348 mg, 2.5 mmol) in THF (2 cm3)
added dropwise over 5 min. After stirring for 1 h at Ϫ78 ЊC,
DMF (0.232 cm3, 3.0 mmol) was added dropwise over 5 min
and the mixture stirred for 45 min. AcOH (0.5 cm3) and H2O
(13 cm3) were added and the cold solution extracted with Et2O
(4 × 10 cm3). The combined organic phases were washed with
0.2 M HCl (10 cm3), H2O (10 cm3) and sat. aq. NaCl solution
(10 cm3), followed by drying (Na2SO4). Concentration in vacuo
and CC (CH2Cl2/hexane 6 : 4) afforded 13 (0.34 g, 82%) as
colourless crystals, mp 97–98 ЊC; νmax(KBr)/cmϪ1 2248, 1694,
1686, 1627, 1568, 1429, 1417, 1324, 1209, 1196, 1181; δH(300
MHz; CDCl3) 7.33 (2 H, ddd, J 13.2, 3.6, 1.8), 10.35 (1 H, s);
δC(75 MHz; CDCl3) 115.4, 116.6 (dd, J 266.1, 6.1), 117.4 (t,
J 11.3), 118.8 (t, J 12.2), 162.7 (dd, J 20.1, 9.1), 182.9 (t, J 4.2);
δF(282 MHz; CDCl3 ϩ CFCl3) Ϫ110.7 (2 F, dd, J 11.6, 4.2);
HR-EI-MS for C8H3F2NOϩ (Mϩ):167.0183; found 166.9772,
165.9725.
To Pr2NH (1.08 cm3, 7.7 mmol) in THF (12.0 cm3) at 0 ЊC
i
under Ar, 1.6 M n-BuLi in hexane (4.40 cm3, 7.0 mmol) was
added dropwise over 5 min. After 10 min, the mixture was
cooled to Ϫ78 ЊC and 20 (970 mg, 7.0 mmol) in THF (5 cm3)
added dropwise over 5 min. After stirring for 1 min at Ϫ78 ЊC,
DMF (0.812 cm3, 10.5 mmol) was added dropwise over 5 min
and the mixture stirred for 45 min. AcOH (2.0 cm3) and H2O
(50 cm3) were added and the cold solution extracted with Et2O
(3 × 30 cm3). The combined organic phases were washed with
0.2 M HCl (30 cm3), H2O (30 cm3) and sat. aq. NaCl solution
(30 cm3), followed by drying (Na2SO4). Concentration in vacuo
and CC (Et2O/pentane 3 : 7), followed by recrystallisation
(CH2Cl2/pentane) afforded 14 (0.41 g, 35%) as colourless
needles, mp 52–53 ЊC; δH(300 MHz; CDCl3) 7.54 (1 H, dddd,
J 8.4, 5.1, 1.8, 0.9), 7.76 (1 H, ddd, J = 8.1, 5.4, 1.5), 10.39 (1 H,
s); δC(75 MHz; CDCl3) 108.2 (d, J 11.6), 111.8 (d, J 3.7), 123.6
(d, J, 3.7), 128.1 (d, J 4.9), 128.7 (d, J 5.5), 152.0 (dd, J 263.4,
14.0), 152.2 (dd, J 262.3, 11.5), 184.1 (dd, J 6.0 2.5); δF(282
MHz; CDCl3 ϩ CFCl3) Ϫ128.1 (1 F, ddd, J = 20.3, 5.4, 2.0),
Ϫ142.9 (1 F, ddd, J = 20.3, 5.8, 2.0); HRMS calcd for
C8H3F2NOϩ (Mϩ): 167.0183; found 167.0170.
3,5-Difluoro-4-[(3aSR,4RS,8aSR,8bRS )-2-(4-fluorobenzyl)-1,3-
dioxodecahydropyrrolo[3,4-a]pyrrolizin-4-yl]benzonitrile
(( )-22)
2,3-Difluoro-4-[(3aSR,4RS,8aSR,8bRS )-2-(4-fluorobenzyl)-1,3-
dioxodecahydropyrrolo[3,4-a]pyrrolizin-4-yl]benzonitrile
(( )-23)
Compound ( )-22 was synthesised from 13 according to the
procedure described above for ( )-12. CC (2 ×; EtOAc/hexane
(1 : 1) and CH2Cl2), followed by recrystallisation (EtOAc/
hexane) afforded ( )-22 as a colourless solid (300 mg, 29%), mp
165–166 ЊC; νmax(KBr)/cmϪ1 2237, 1768, 1704, 1628, 1604, 1576,
1509, 1429, 1398, 1341, 1222; δH(300 MHz; CDCl3) 1.60–1.88
(2 H, m), 1.98–2.23 (2 H, m), 2.58–2.70 (1 H, m), 2.82–2.88
Compound ( )-23 was synthesised from 14 according to the
procedure described above for ( )-12. CC (EtOAc/hexane 1 : 1),
followed by recrystallisation (EtOAc/hexane) afforded ( )-23
(377 mg, 37%) as colourless crystals, mp 134–135 ЊC; νmax(KBr)/
cmϪ1 2239, 1776, 1710, 1631, 1604, 1512, 1472, 1432, 1399,
O r g . B i o m o l . C h e m . , 2 0 0 4 , 2, 1 3 3 9 – 1 3 5 2
1348

View Article Online
1345, 1300, 1223, 1175; δH(300 MHz; CDCl3) 1.60–1.88 (2 H,
m), 1.96–2.24 (2 H, m), 2.54–2.64 (1 H, m), 2.86–2.98 (1 H, m),
3.34 (1 H, d, J 8.4), 3.67 (1 H, t, J 8.1), 3.76 (1 H, dd, J 10.1,
7.4), 4.31 (1 H, d, J 8.4), 4.43, 4.49 (2 H, AB, J 14.4), 6.90–7.01
(2 H, m), 7.13–7.30 (4 H, m); δC(75 MHz; CDCl3) 23.4, 29.8,
41.8, 48.6, 51.1, 61.8, 67.9, 101.9 (d, J 12.2), 112.9 (d, J 3.6),
115.3 (d, J 21.2), 123.4 (t, J 3.6), 127.4 (d, J 4.2), 130.5 (t, J 7.9),
131.2 (d, J 3.1), 134.5 (d, J 10.4), 148.9 (dd, J 249.0, 11.0), 150.7
(dd, J 258.8, 14.6), 162.2 (d, J 244.7), 174.3, 177.1; δF(282 MHz;
CDCl3 ϩ CFCl3) Ϫ113.5 (1 F, m), Ϫ131.7 (1 F, dd, J 21.2, 5.4),
2,6-Difluoro-4-[(3aSR,4RS,8aSR,8bRS )-2-(4-fluorobenzyl)-1,3-
dioxodecahydropyrrolo[3,4-a]pyrrolizin-4-yl]benzonitrile
(( )-26)
Compound ( )-26 was synthesised according to the procedure
described above for ( )-12. CC (2 ×; EtOAc/hexane 40 : 60 and
EtOAc/CH2Cl2 10 : 90) followed by recrystallisation (EtOAc/
hexane) provided ( )-26 (154 mg, 15%) as colourless crystals,
mp 169–170 ЊC; νmax(KBr)/cmϪ1 2238, 1773, 1706, 1632, 1628,
1605, 1576, 1572, 1511, 1441, 1400, 1344, 1230, 1172; δH(300
MHz; CDCl3) 1.58–1.71 (1 H, m), 1.72–1.88 (1 H, m), 1.95–2.23
(2 H, m), 2.50–2.61 (1 H, m), 2.82–2.94 (1 H, m), 3.32 (1 H, d,
J 8.1), 3.53 (1 H, t, J 8.4), 3.74 (1 H, dd, J 9.4, 7.5), 4.07 (1 H, d,
J 8.4), 4.47, 4.52 (2 H, AB, J 14.1), 6.92 (2 H, d, J 8.7), 6.94–
7.03 (2 H, m), 7.28–7.36 (2 H, m); δC(75 MHz; CDCl3) 23.5,
29.8, 41.9, 48.8, 50.4, 50.9, 68.0, 68.1, 91.4 (t, J 19.4), 109.3,
111.5 (dd, J 20.6, 3.4), 115.5 (d, J 21.2), 130.5 (t, J 8.6), 131.1 (d,
J 3.6), 148.7 (t, J 8.8), 162.3 (d, J 245.0), 162.7 (dd, J 259.0, 4.9),
174.2, 176.9; δF(282 MHz; CDCl3 ϩ CFCl3) Ϫ110.4 (2 F, d,
ϩ
Ϫ138.6 (1 F, d br, J 18.3); HRMS calcd for C23H19F3N3O2
(MHϩ): 426.1429; found 426.1426.
2,3-Difluoro-4-[(3aSR,4RS,8aSR,8bRS )-2-(4-fluorobenzyl)-1,3-
dioxodecahydropyrrolo[3,4-a]pyrrolizin-4-yl]benzamidine
hydrochloride (( )-17)
Compound ( )-17 was synthesised from ( )-23 according to
the procedure described above for ( )-8. CC (CH2Cl2/MeOH
ϩ
99 : 1
90 : 10) and precipitation from EtOH/Et2O yielded
J 8.5), Ϫ113.3 (1 F, m); HRMS calcd for C23H19F3N3O2
(MHϩ): 426.1429; found 426.1420.
( )-17 (51 mg, 43%) as a colourless solid, mp >152 ЊC (dec.);
νmax(KBr)/cmϪ1 3382, 2970, 1772, 1699, 1638, 1607, 1511, 1464,
1436, 1401, 1345, 1222, 1177; δH(300 MHz; CD3OD) 1.72–1.92
(2 H, m), 2.02–2.21 (2 H, m), 2.60–2.72 (1 H, m), 2.82–3.02
(1 H, m), 3.51 (1 H, dd, J 8.0, 0.8), 3.72 (1 H, dd, J 9.3, 7.2), 3.88
(1 H, t, J 8.3), 4.41–4.53 (3 H, m), 6.98–7.07 (2 H, m), 7.17–7.25
(2 H, m), 7.30 (1 H, d, J 8.7), 7.35 (1 H, d, J 8.7); δC(75 MHz;
CD3OD) 24.4, 30.7, 42.4, 50.0, 50.2, 52.2, 63.2, 69.4, 116.1 (d,
J 21.8), 118.9 (d, J 9.7), 124.8, 131.0 (d, J 8.5), 133.1 (d, J 3.1),
135.7 (d, J 10.4), 148.6 (dd, J 238.4, 15.9), 150.5 (dd, J 235.4,
12.9), 163.4, 163.5 (d, J 243.2), 176.7, 179.6; δF(282 MHz;
CD3OD ϩ CFCl3) Ϫ114.2 (1 F, m), Ϫ138.7 (1 F, s br), Ϫ139.2
2,6-Difluoro-4-[(3aSR,4RS,8aSR,8bRS )-2-(4-fluorobenzyl)-1,3-
dioxodecahydropyrrolo[3,4-a]pyrrolizin-4-yl]benzamidine
hydrochloride (( )-18)
A solution of ( )-26 (106 mg, 0.25 mmol), NH2OHؒHCl
(48 mg, 0.5 mmol) and Et3N (104 cm3, 0.75 mmol) in MeOH
(0.5 cm3) and THF (2 cm3) was stirred for 16 h after which it
was partitioned between CH2Cl2 (15 cm3) and H2O (15 cm3).
The organic phase was washed with H2O (15 cm3), dried
(Na2SO4) and concentrated in vacuo. The residual solid was
taken up in AcOH (2 cm3), and Ac2O (35 cm3, 0.38 mmol) was
added. After stirring for 5 min, Pd/C (10%, 15 mg) was added
and hydrogenation (H2 balloon) conducted for 5 h. Filtration
through Celite, concentration and CC (CH2Cl2/MeOH 99 : 1
90 : 10), followed by precipitation from HCl/EtOH/Et2O
afforded ( )-18 (47 mg, 39%) as colourless solid, mp >171 ЊC
(dec.); νmax(KBr)/cmϪ1 3389, 3048, 2973, 1773, 1705, 1692,
1641, 1604, 1512, 1467, 1435, 1402, 1346, 1224, 1177; δH(500
MHz; CD3OD ϩ TMS) 1.72–1.88 (2 H, m), 2.02–2.14 (2 H),
2.54–2.63 (1 H, m), 2.83–2.93 (1 H, m), 3.46 (1 H, d, J 8.0), 3.69
(1 H, dd, J 9.6, 7.4), 3.74 (1 H, t, J 8.3), 4.27 (1 H, d, J 8.7), 4.47,
4.55 (2 H, d, J 14.5), 7.00–7.07 (2 H, m), 7.10 (2 H, d br, J 9.8),
7.23–7.32 (2 H, m); δC(125 MHz; CD3OD ϩ TMS) 24.2, 30.5,
42.5, 50.1, 51.8, 51.9, 68.9, 69.6, 107.7 (t, J 19.3), 113.1 (dd,
J 22.1, 2.9), 116.3 (d, J 21.8), 131.3 (d, J 7.5), 133.4 (d, J 3.2),
149.9 (t, J 9.2), 160.1, 160.4 (dd, J 253.4, 5.9), 163.8 (d, J 245.0),
177.2, 179.8; δF(282 MHz; CD3OD ϩ CFCl3) Ϫ111.7 (2 F, d,
(1 F, d, J 19.2); HRMS calcd for C23H22F3N4O2 (MHϩ):
ϩ
443.1695; found 443.1691.
2,6-Difluoro-4-formylbenzonitrile (15)
To a mixture of 21 (2.56 g, 18 mmol), TsOHؒH2O (17 mg,
0.09 mmol) and ethyleneglycol (2.5 cm3, 45 mmol), benzene
(30 cm3) was added. After heating to reflux for 16 h using a
Dean-Stark trap, the mixture was partitioned between EtOAc
(30 cm3) and sat. aq. NaHCO3 solution (30 cm3). The organic
phase was washed with H2O, dried (Mg2SO4) and concentrated
in vacuo to give acetal 24 as a colourless oil that was used with-
i
out further purification. To Pr2NH (2.78 cm3, 19.8 mmol)
in THF (36 cm3) at 0 ЊC under Ar, 1.6 M n-BuLi in hexane
(11.3 cm3, 18.0 mmol) was added dropwise over 5 min. After
10 min, the mixture was cooled to Ϫ78 ЊC and 24 in THF
(12 cm3) added dropwise over 5 min. After stirring for 2 h at
Ϫ78 ЊC, TsCN (3.91 g, 21.6 mmol) in THF (15 cm3) was
added dropwise over 10 min and the mixture stirred for 1 h
at Ϫ78 ЊC. The mixture was allowed to warm to rt and then
partitioned between Et2O (60 cm3) and H2O (60 cm3). The
aqueous phase was extracted with Et2O (3 × 40 cm3), and
the combined organic phases were washed with 1 M HCl (80
cm3), H2O (80 cm3) and sat. aq. NaCl solution (80 cm3). Con-
centration in vacuo afforded crude 25, to which THF (60 cm3)
and 2 M HCl (30 cm3) were added. The solution was heated
to reflux for 6 h and, after cooling to rt, Et2O (60 cm3)
was added and the two phases were separated. The aqueous
phase was extracted with E2O (2 × 60 cm3), and the combined
organic phases were washed with sat. aq. NaHCO3 solution
(80 cm3), H2O (80 cm3) and sat. aq. NaCl solution (80 cm3).
Drying (Na2SO4), concentration in vacuo and CC (Et2O/
pentane 3 : 7) afforded 15 (1.43 g, 48%) as colourless needles,
mp 48–50 ЊC; δH(300 MHz; CDCl3) 7.58 (2 H, ddd, J 9.0,
1.5, 0.9), 10.00 (1 H, t, J 1.8); δC(75 MHz; CDCl3) 97.5 (t,
J 16.7), 108.2, 112.4 (dd, J 20.7, 3.6), 141.4 (t, J 7.3), 163.5 (dd,
J 264.1, 3.6), 187.5; δF(282 MHz; CDCl3 ϩ CFCl3) Ϫ100.0 (2 F,
t, J 7.3); HRMS calcd for C8H3F2NOϩ (Mϩ): 167.0183; found
167.0142.
ϩ
J 9.6), Ϫ114.1 (1 F, m); HRMS calcd for C23H22ClF3N4O2
(MHϩ): 443.1695; found 443.1694.
(3aSR,4RS,8aSR,8bRS )-2-(4-Fluorobenzyl)-4-phenylhexa-
hydropyrrolo[3,4-a]pyrrolizine-1,3(2H,4H)-dione (( )-30a) and
(3aSR,4SR,8aRS,8bRS )-2-(4-Fluorobenzyl)-4-phenylhexa-
hydropyrrolo[3,4-a]pyrrolizine-1,3(2H,4H)-dione (( )-30b)
Compounds ( )-30a and ( )-30b were synthesised according to
the procedure described above for ( )-12 from N-(4-fluoro-
benzyl)maleimide (307 mg, 1.5 mmol), -proline (173 mg, 1.5
mmol) and benzaldehyde (159 mg, 1.5 mmol). CC (CH2Cl2/
EtOAc 100 : 0
70 : 30) afforded ( )-30a (220 mg, 40%) as
colourless needles, mp 109–110 ЊC and ( )-30b (200 mg, 37%)
as a yellow oil. ( )-30a: νmax(neat)/cmϪ1 1772, 1698, 1603, 1510,
1430, 1397, 1340, 1221, 1169. δH(300 MHz; CDCl3) 1.60–1.85
(2 H, m), 1.96–2.19 (2 H, m), 2.62–2.77 (1 H, m), 2.80–2.94
(1 H, m), 3.32 (1 H, dd, J 7.8, 0.6), 3.49 (1 H, t, J 8.4), 3.74 (1 H,
dd, J 9.6, 6.9), 4.08 (1 H, d, J 8.7), 4.50 (2 H, s), 6.94–7.03 (2 H,
m), 7.18–7.33 (7 H, m); δC(75 MHz; CDCl3) 23.4, 29.6, 41.7,
49.0, 50.6, 50.7, 67.8, 68.7, 115.2 (d, J 21.4), 127.7, 127.8, 128.0,
130.7 (t, J 8.5), 131.5 (d, J 3.0), 137.6, 162.2 (d, J 246.0), 175.0,
O r g . B i o m o l . C h e m . , 2 0 0 4 , 2, 1 3 3 9 – 1 3 5 2
1349

View Article Online
177.8; δF(282 MHz; CDCl3 ϩ CFCl3) Ϫ114.1 (1 F, m); HRMS
UV absorption. The obtained optical density (reference) is
ϩ
calcd for C22H22FN2O2 (MHϩ): 365.1660; found 365.1655.
equal to the concentration of the substance before partitioning.
An exact amount of 1-octanol is added and the mixture incu-
bated by quiet shaking (2 h). The emulsion is allowed to stand
overnight to be sure that the partition equilibrium is reached.
The next day, the layers need thorough centrifugation at 3000
rpm for 10 min and the concentration of the substance in the
aqueous phase is determined again by measuring the UV
absorption. The procedure above is carried out at four
different octanol/water ratios, two with a large volume of octa-
nol for hydrophilic compounds (log D < 1) and two with a low
volume of octanol for the lipophilic compounds (log D > 1).
( )-30b: νmax(neat)/cmϪ1 1771, 1697, 1603, 1509, 1431, 1393,
1339, 1222, 1158; δH(300 MHz; CDCl3) 1.55–1.78 (3 H, m),
1.88–2.00 (1 H, m), 2.38–2.47 (1 H, m), 2.88–2.98 (1 H, m), 3.35
(1 H, dd, J 9.0, 5.1), 3.51 (1 H, t, J 9.0), 3.84–3.94 (1 H, m), 4.13
(1 H, d, J 5.1), 4.61 (2 H, s), 6.95–7.04 (2 H, m), 7.22–7.48 (7 H,
m); δC(75 MHz; CDCl3) 24.2, 26.2, 41.8, 48.0, 52.1, 55.6, 66.3,
69.5, 115.4 (d, J 21.4), 126.6, 127.3, 128.5, 130.8 (t, J 8.5), 131.2
(d, J 3.0), 141.9, 162.2 (d, J 246.0), 176.5, 177.6; δF(282
MHz; CDCl3 ϩ CFCl3) Ϫ113.3 (1 F, m); HRMS calcd for
C22H22FN2O2ϩ (MHϩ): 365.1660; found 365.1656.
(3aSR,8aSR,8bRS )-2-(4-Fluorobenzyl)hexahydropyrrolo-
[3,4-a]pyrrolizin-1,3(2H,4H)-dione (( )-31)
pKa Determination by potentiometric titration44
The pKa values (reproducibility: 0.05 units) were determined
by potentiometric titrations using GLpKa from Sirius Analyti-
cal Instruments Ltd. In the applied pH-metric method, a solu-
tion of the sample is titrated over a pH range in which it passes
from being fully protonated to fully deprotonated. The pKa
values are calculated by analysing the shape of the titration
curve. General titration procedure: An exact amount of the
sample is placed directly into the vial. The instrument adds
0.1 M KNO3 (20 cm3) as background electrolyte. It then checks
the pH of the sample solution and adds 0.5 M HCl to bring the
pH down to the initial pH 2. It then titrates with standardised
base solution (0.5 M KOH) until the final pH value of 12 has
been reached, at which point data collection is terminated. The
titration is running under argon atmosphere to minimise
absorption of atmospheric CO2 because of the pKa values of
carbonate. This procedure results in a data set containing a
single potentiometric titration curve. The data set is down-
loaded to the computer, and pKa results are calculated by pKa-
LOGP V5.2a. In comparison with a blank titration (without
sample), the shape of such a curve indicates the amount of
substance present and its characteristic acid–base ionisation
properties.
Compound ( )-31 was synthesised according to the procedure
described above for ( )-12 from N-(4-fluorobenzyl)maleimide
(307 mg, 1.5 mmol), -proline (173 mg, 1.5 mmol) and para-
formaldehyde (45 mg, 1.5 mmol). CC (EtOAc/Et3N 97.5 : 2.5)
afforded ( )-31 (116 mg, 27%) as a slightly yellow oil;
νmax(neat)/cmϪ1 1772, 1696, 1603, 1510, 1430, 1397, 1340,
1220, 1166; δH(300 MHz; CDCl3) 1.50–1.80 (2 H, m), 1.90–
2.10 (2 H, m), 2.70–2.92 (3 H, m), 3.11 (1 H, dd, J 8.1, 2.1), 3.23
(1 H, dd, J 10.2, 2.1), 3.31 (1 H, td, J 8.1, 2.1), 4.08 (1 H, m),
4.57 (2 H, s), 6.89–6.98 (2 H, m), 7.25–7.34 (2 H, m); δC(75
MHz; CDCl3) 23.6, 29.6, 41.7, 45.9, 50.0, 51.8, 54.5, 68.7, 115.2
(d, J 21.4), 130.2 (d, J 7.5), 131.4 (d, J 3.0), 137.6, 162.0 (d,
J 246.0), 177.7, 178.2; δF(282 MHz; CDCl3 ϩ CFCl3) Ϫ114.0
ϩ
(1 F, m); HRMS calcd for C16H18FN2O2 (MHϩ): 289.1347;
found 289.1350.
Imide opening under basic conditions
Solutions of ( )-277d in THF (1.22 cm3) and NaOH (16 mg,
0.4 mmol) in H2O (0.66 cm3) were stirred together at rt. After
1 h, only traces of starting material were detected by TLC.
After 2 h, the solution was concentrated in vacuo and re-
concentrated after addition of toluene (3 ×) to give ( )-28 in a
quantitative conversion (1H NMR); νmax(neat)/cmϪ1 1622, 1578
(s), 1603, 1446, 1416, 1356, 1227, 1110; δH(300 MHz; CD3OD)
1.80–2.12 (2 H, m), 2.20–2.34 (1 H, m), 2.50–2.60 (1 H, m),
2.94–3.05 (2 H, m), 3.13 (1 H, dd, J 9.0, 6.9), 3.74 (1 H, t, J 5.7),
4.33 (1 H, d, J 5.7), 4.38–4.48 (1 H, m), 4.07, 4.53 (2 H, AB,
J 13.3), 6.89–6.98 (5 H, m), 7.25–7.34 (4 H, m); δC(75 MHz;
D2O) 33.3, 33.4, 42.8, 54.3, 56.7, 57.6, 57.7, 72.8, 120.3, 126.4,
127.7, 128.2, 130.2, 131.2, 139.3, 140.7, 172.4, 178.7; HRMS
Crystal structures
The crystal structures were solved by direct methods45 and
refined by full-matrix least-squares analysis46 (heavy atoms
anisotropic, H-atoms isotropic, whereby H-positions are based
on stereochemical considerations).†
Crystal data of ( )-7 at 223 K: C23H19F2N3O2, (Mr = 407.41);
triclinic, space group P1, Dc= 1.404 g cmϪ3, Z = 2, a = 9.4926(2),
¯
b = 10.3004(3), c = 10.9831(3) Å, α = 93.680 (1)Њ, β = 94.130(1)Њ,
γ = 115.220(1)Њ, V = 963.64(4) Å3. Linear crystal dimensions
ca. 0.3 × 0.3 × 0.25 mm. Bruker-Nonius Kappa-CCD, MoKα
radiation, λ = 0.7107 Å. Final R(F) = 0.0437, wR(F2) = 0.1161
for 291 parameters and 3458 reflections with I > 2σ(I ) and
θ < 27.49Њ. CCDC 231344.
ϩ
calcd for C22H24BrN2O3 (MHϩ): 443.0970, 445.0950; found
443.0960, 445.0944. The imide ring opening of ( )-29 (5 mg,
0.012 mmol) in D2O (0.9 cm3) and 0.5 M NaOD (0.1 cm3) was
also monitored and confirmed by 1H NMR.
Computational investigation
Crystal data of ( )-12 at 223 K: C23H19F2N3O2, (Mr
=
407.41); triclinic, space group P1, Dc= 1.373 g cmϪ3, Z = 2,
a = 7.673(1), b = 11.394(2), c = 11.894(2) Å, α = 97.22(1)Њ, β =
106.92(1)Њ, γ = 90.85(1)Њ, V = 985.5(3) Å3. Linear crystal dimen-
sions ca. 0.22 × 0.2 × 0.18 mm. Nonius CAD4 diffractometer,
CuKα radiation, λ = 1.5418 Å. Final R(F) = 0.0449, wR(F2)=
0.1256 for 291 parameters and 3272 reflections with I > 2σ(I )
and θ < 74.9Њ. CCDC 231343.
¯
The enthalpy of formation ∆Hf of the different conformers of
32 and 33 in unprotonated and protonated forms was obtained
by calculating the equilibrium geometry in Spartan’0438 utilis-
ing density functional theory and a 6-31G** basis set.
High-throughput log D screening
The applied high-throughput (HT) method for the determin-
ation of the distribution coefficient HT-log D is based on
microplate technique and derived from the conventional ‘shake
flask’ method. The compound is distributed between H2O
buffered at a specific pH and 1-octanol. The distribution co-
efficient is then calculated from the difference in concentration
in the aqueous phase before and after partitioning and the ratio
of the two phases. The “one phase-analysed” experiment starts
with a pure Me2SO solution of the lead molecule of interest
which is dispensed in aqueous buffer with c (compound) = 0.5
mM. A part of this solution is then analysed by measuring the
Crystal data of 13 at 100 K: C8H3F2NO, (Mr = 167.11);
orthorhombic, space group P212121, Dc= 1.660 g cmϪ3, Z = 4,
a = 3.7369(2), b = 6.2036(3), c = 28.8526(2) Å, V = 668.87(6) Å3.
Linear crystal dimensions ca. 0.1 × 0.1 × 0.07 mm. Bruker-
Nonius Kappa-CCD, MoKα radiation, λ = 0.7107 Å. Final
R(F) = 0.0327, wR(F2) = 0.0677 for 113 parameters and 1267
reflections with I > 2σ(I ) and θ < 27.45Њ. CCDC 231345.
† CCDC reference numbers 231343–231347. See http://www.rsc.org/
suppdata/ob/b4/b402515f/ for crystallographic data in.cif or other
electronic format.
O r g . B i o m o l . C h e m . , 2 0 0 4 , 2, 1 3 3 9 – 1 3 5 2
1350

View Article Online
13 (a) P. E. J. Sanderson, M. G. Stanton, B. D. Dorsey, T. A. Lyle,
Crystal data of ( )-22 at 243 K: C23H18F3N3O2, (Mr
=
425.40); monoclinic, space group P21/n, Dc= 1.452 g cmϪ3, Z =
4, a = 4.8890(7), b = 20.180(2), c = 19.844(2) Å, β = 96.211(5)Њ,
V = 1946.3(4) Å3. Linear crystal dimensions ca. 0.09 × 0.07 ×
0.05 mm. Bruker-Nonius Kappa-CCD, MoKα radiation, λ =
0.7107 Å. Final R(F) = 0.1252, wR(F2) = 0.3214 for 299 param-
eters and 2225 reflections with I > 2σ(I ) and θ < 24.12. CCDC
231347.
C. McDonough, W. M. Sanders, K. L. Savage, A. M. Naylor-Olsen,
J. A. Krueger, S. D. Lewis, B. J. Lucas, J. J. Lynch and Y. Yan, Bioorg.
Med. Chem. Lett., 2003, 13, 795–798; (b) T. J. Tucker, S. F. Brady,
W. C. Lumma, S. D. Lewis, S. J. Gardell, A. M. Naylor-Olsen,
Y. Yan, J. T. Sisko, K. J. Stauffer, B. J. Lucas, J. J. Lynch, J. J. Cook,
M. T. Stranieri, M. A. Holahan, E. A. Lyle, E. P. Baskin, I.-W. Chen,
K. B. Dancheck, J. A. Krueger, C. M. Cooper and J. P. Vacca,
J. Med. Chem., 1998, 41, 3210–3219; (c) Y. St-Denis, S. Lévesque,
B. Bachand, J. J. Edmunds, L. Leblond, P. Préville, M. Tarazi, P. D.
Winocour and M. A. Siddiqui, Bioorg. Med. Chem. Lett., 2002, 12,
1181–1184.
Crystal data of ( )-22ؒHCl at 223 K: C23H19ClF3N3O2
ϩ
1H2O (Mr = 479.88); monoclinic, space group P21/c, Dc= 1.496 g
cmϪ3, Z = 4, a = 11.6360(6), b = 6.9444(4), c = 26.6949(16) Å,
β = 98.983(2)Њ, V = 2130.63(17) Å3. Linear crystal dimensions
ca. 0.1 × 0.1 × 0.03 mm. Bruker-Nonius Kappa-CCD, MoKα
radiation, λ = 0.7107 Å. Final R(F) = 0.0664, wR(F2) = 0.1639
for 321 parameters and 2354 reflections with I > 2σ(I ) and
θ < 24.11Њ. CCDC 231346.
14 (a) B. A. Katz, P. A. Sprengeler, C. Luong, E. Verner, K. Elrod,
M. Kirtley, J. Janc, J. R. Spencer, J. G. Breitenbucher, H. Hui,
D. McGee, D. Allen, A. Martelli and R. L. Mackman, Chem. Biol.,
2001, 8, 1107–1121; (b) K. Lee, C. W. Park, W.-H Jung, H. D. Park,
S. H. Lee, K. H. Chung, S. K. Park, O. H. Kwon, M. Kang, D.-H.
Park, S. K. Lee, E. E. Kim, S. K. Yoon and A. Kim, J. Med. Chem.,
2003, 46, 3612–3622.
15 J. W. Essex, D. L. Severance, J. Tirado-Rives and W. L. Jorgensen,
J. Phys. Chem. B., 1997, 101, 9663–9669.
16 (a) C. R. W. Guimaraes and R. B. de Alencastro, J. Phys. Chem. B.,
2002, 106, 466–476; (b) S. M. Schwarzl, T. M. Tschopp, J. C. Smith
and S. Fischer, J. Comput. Chem., 2002, 23, 1143–1149; (c) C. F.
Wong and J. A. McCammon, J. Am. Chem. Soc., 1986, 108, 3830–
3832; (d ) G. Náray-Szabó, J. Am. Chem. Soc., 1984, 106, 4584–4589;
(e) H.-J. Henkel, D. Labes and V. Hagen, Pharmazie, 1983, 38, 342–
346.
17 C. J. Dinsmore, M. J. Bogusky, J. C. Culberson, J. M. Bergman,
C. F. Homnick, C. B. Zartman, S. D. Mosser, M. D. Schaber,
R. G. Robinson, K. S. Koblan, H. E. Huber, S. L. Graham,
G. D. Hartman, J. R. Huff and T. M. Williams, J. Am. Chem. Soc.,
2001, 123, 2107–2108.
Acknowledgements
We thank Dr. F. Hof (ETH Zürich) for assistance with the
DFT calculations and manuscript corrections, Prof. B. Jaun
(ETH Zürich) for advice with the NMR experiments and
Dr. C. Thilgen (ETH Zürich) for help with the nomenclature.
The research was supported by the ETH research council,
F. Hoffmann-La Roche Ltd and the Carlsberg Foundation
(postdoctoral fellowship to J.A.O.).
18 I. M. Bell, S. N. Gallicchio, M. Abrams, L. S. Beese, D. C. Beshore,
H. Bhimnathwala, M. J. Bogusky, C. A. Buser, J. C. Culberson,
J. Davide, M. Ellis-Hutchings, C. Fernandes, J. B. Gibbs, S. L.
Graham, K. A. Hamilton, G. D. Hartman, D. C. Heimbrook, C. F.
Homnick, H. E. Huber, J. R. Huff, K. Kassahun, K. S. Koblan,
N. E. Kohl, R. B. Lobell, J. J. Lynch Jr., R. Robinson, A. D.
Rodrigues, J. S. Taylor, E. S. Walsh, T. M. Williams and C. B.
Zartman, J. Med. Chem., 2002, 45, 2388–2409.
19 (a) R. Grigg, J. Idle, P. McMeekin and D. Vipond, J. Chem. Soc.,
Chem. Commun., 1987, 49–51; (b) O. Tsuge and S. Kanemasa, Adv.
Heterocycl. Chem., 1989, 45, 231–349.
20 C.-B. Xue, J. Wityak, T. M. Sielecki, D. J. Pinto, D. G. Batt, G. A.
Cain, M. Sworin, A. L. Rockwell, J. F. Roderick, S. Wang, M. J.
Orwat, W. E. Frietze, L. L. Bostrom, J. Liu, C. A. Higley, F. W.
Rankin, A. E. Tobin, G. Emmett, G. K. Lalka, J. Y. Sze, S. V. Di
Meo, S. A. Mousa, M. J. Thoolen, A. L. Racanelli, E. A. Hausner,
T. M. Reilly, W. F. DeGrado, R. R. Wexler and R. E. Olson, J. Med.
Chem., 1997, 40, 2064–2084.
21 A. J. Bridges, A. Lee, E. C. Maduakor and C. E. Schwartz,
Tetrahedron Lett., 1992, 33, 7499–7502.
22 K. J. Rutan, F. J. Heldrich and L. F. Borges, J. Org. Chem., 1995, 60,
2948–2950.
23 B. D. Judkins, D. G. Allen, T. A. Cook, B. Evans and T. E.
Sardharwala, Synth. Commun., 1996, 26, 4351–4367.
24 E. A. Meyer, R. K. Castellano and F. Diederich, Angew. Chem., Int.
Ed., 2003, 42, 1210–1250.
References
1 J. Fried and E. F. Sabo, J. Am. Chem. Soc., 1954, 76, 1455–1456.
2 F. M. D. Ismael, J. Fluorine Chem., 2002, 118, 27–33.
3 M. Schlosser, in Enantiocontrolled Synthesis of Fluoro-Organic
Compounds: Stereochemical Challenges and Biomedicinal Targets,
ed. V. A. Soloshonok, Wiley, Chichester, 1999, pp 613–659.
4 C. M. Weeks, W. L. Duax and M. E. Wolff, J. Am. Chem. Soc., 1973,
95, 2865–2868.
5 P. M. Matias, M. A. Carrondo, R. Coelho, M. Thomaz, X.-Y. Zhao,
A. Wegg, K. Crusius, U. Egner and P. Donner, J. Med. Chem., 2002,
45, 1439–1446.
6 (a) B. K. Park, N. R. Kitteringham and P. M. O’Neill, Annu. Rev.
Pharmacol. Toxicol., 2001, 41, 443–470; (b) D. O’Hagan and H. S.
Rzepa, Chem. Commun., 1997, 645–652; (c) B. E. Smart, J. Fluorine
Chem., 2001, 109, 3–11.
7 (a) U. Obst, V. Gramlich, F. Diederich, L. Weber and D. W. Banner,
Angew. Chem., Int. Ed., 1995, 34, 1739–1742; (b) U. Obst, D. W.
Banner, L. Weber and F. Diederich, Chem. Biol., 1997, 4, 287–295;
(c) U. Obst, P. Betschmann, C. Lerner, P. Seiler, F. Diederich,
V. Gramlich, L. Weber, D. W. Banner and P. Schönholzer, Helv.
Chim. Acta, 2000, 83, 855–909; (d ) P. Betschmann, S. Sahli,
F. Diederich, U. Obst and V. Gramlich, Helv. Chim. Acta, 2002, 85,
1210–1245.
8 B. Kuhn and P. A. Kollman, J. Am. Chem. Soc., 2000, 122, 3909–
3916.
25 (a) J. D. Dunitz and R. Taylor, Chem. Eur. J., 1997, 3, 89–98;
(b) J. A. K. Howard, V. J. Hoy, D. O’Hagan and G. T. Smith,
Tetrahedron, 1996, 52, 12613–12622.
9 (a) J. A. Olsen, D. W. Banner, P. Seiler, U. Obst Sander, A. D’Arcy,
M. Stihle, K. Müller and F. Diederich, Angew. Chem., Int. Ed., 2003,
42, 2507–2511; (b) J. A. Olsen, D. W. Banner, P. Seiler, U. Obst-
Sander, A. D’Arcy, M. Stihle, K. Müller and F. Diederich,
ChemBioChem, in press.
10 (a) For examples of phenylamidinium-containing thrombin
inhibitors, see: J. W. Nilsson, I. Kvarnstrom, D. Musil, I. Nilsson and
B. Samulesson, J. Med. Chem., 2003, 46, 3985–4001; (b)
S. Hanessian, E. Therrien, K. Granberg and I. Nilsson, Bioorg. Med.
Chem. Lett., 2002, 12, 2907–2911; (c) N. H. Hauel, H. Nar,
H. Priepke, U. Ries, J.-M. Stassen and W. Wienen, J. Med. Chem.,
2002, 45, 1757–1766; (d ) for a review, see: J. B. M. Rewinkel and
A. E. P. Adang, Curr. Pharm. Des., 1999, 5, 1043–1075.
11 R. Talhout and J. B. F. N. Engberts, Eur. J. Biochem., 2001, 268,
1554–1560.
12 (a) M. Recanatini, T. Klein, C.-Z. Yang, J. McClarin, R. Langridge
and C. Hansch, Mol. Pharmacol., 1986, 29, 436–446; (b) D. Labes
and V. Hagen, Pharmazie, 1979, 34, 649–653; (c) M. Mares-Guia,
D. L. Nelson and E. Rogana, J. Am. Chem. Soc., 1977, 99, 2331–
2336; (d ) B. A. Katz, R. Mackman, C. Luong, K. Radika,
A. Martelli, P. A. Sprengeler, J. Wang, H. Chan and L. Wong, Chem.
Biol., 2000, 7, 299–312.
26 T. Steiner, Angew. Chem., Int. Ed., 2002, 41, 48–76.
27 (a) J. P. Snyder, N. S. Chandrakumar, H. Sato and D. C. Lankin,
J. Am. Chem. Soc., 2000, 122, 544–545; (b) T. J. Barbarich, C. D.
Rithner, S. M. Miller, O. P. Anderson and S. H. Strauss, J. Am.
Chem. Soc., 1999, 121, 4280–4281.
28 (a) G. R. Desiraju, Acc. Chem. Res., 2002, 35, 565–573;
(b) V. R. Thalladi, H.-C. Weiss, D. Bläser, R. Boese, A. Nangia
and G. R. Desiraju, J. Am. Chem. Soc., 1998, 120, 8702–
8710.
29 A. J. Mountford, D. L. Hughes and S. J. Lancaster, Chem. Commun.,
2003, 2148–2149.
30 (a) J. Parsch and J. W. Engels, J. Am. Chem. Soc., 2002, 124, 5664–
5672; (b) G. M. Dubowchik, V. M. Vrudhula, B. Dasgupta, J. Ditta,
T. Chen, S. Sheriff, K. Sipman, M. Witmer, J. Tredup, D. M. Vyas,
T. A. Verdoorn, S. Bollini and A. Vinitsky, Org. Lett., 2001, 3, 3987–
3990; (c) A. DerHovanessian, J. B. Doyon, A. Jain, P. R. Rablen and
A.-M. Sapse, Org. Lett., 1999, 1, 1359–1362.
31 (a) E. J. Corey and T. W. Lee, Chem. Commun., 2001, 1321–1329;
(b) V. V. Grushin and W. J. Marshall, Angew. Chem., Int. Ed., 2002,
41, 4476–4479.
O r g . B i o m o l . C h e m . , 2 0 0 4 , 2, 1 3 3 9 – 1 3 5 2
1351

View Article Online
32 P. W. Baures, A. M. Beatty, M. Dhanasekaran, B. A. Helfrich,
W. P érez-Segarra and J. Desper, J. Am. Chem. Soc., 2002, 124,
11315–11323.
33 K. Nishide, Y. Hagimoto, H. Hasegawa, M. Shiro and M. Node,
Chem. Commun., 2001, 2394–2395.
34 CSD Version 5.24 of November 2002, holding over 270000 small-
molecule crystal structures. The search was limited to well-resolved
(R-factor < 7.5%, no disorder) neutral organic molecules, excluding
metallorganic or polymeric compounds, with an intermolecular
(non-bonded) distance between a C-bound F-atom and a nitrile
C-atom in the range of 2.5 Å < d1 F. . .CN < 4.0 Å.
35 W. R. Croasmun and R. M. K. Carlson (eds), Two-Dimensional
NMR Spectroscopy, 2nd ed., VCH, Weinheim, 1994.
36 MedChem02 database, Daylight Chemical Information Systems,
Inc., 27401, Los Altos #360, Mission Viejo, CA 92691, USA.
37 J. Clark and D. D. Perrin, Q. Rev. Chem. Soc., 1964, 18, 295–320.
38 Spartan ’04 Windows, Copyright 2003, Wavefunction, Inc., 18401
Von Karman Ave., Irvine, CA 92612.
39 Unfavorable C–F dipole–lone pair interactions were elegantly taken
in consideration to explain the primary stereoelectronic effect in the
Baeyer–Villiger reaction; see: C. M. Crudden, A. C. Chen and
L. A. Calhoun, Angew. Chem., Int. Ed., 2000, 39, 2852–2855.
40 B. A. Katz, R. Mackman, C. Luong, K. Radika, A. Martelli,
P. A. Sprengeler, J. Wang, H. Chan and L. Wong, Chem. Biol., 2000,
7, 299–312.
41 K. Hilpert, J. Ackermann, D. W. Banner, A. Gast, K. Gubernator,
P. Hadváry, L. Labler, K. Müller, G. Schmid, T. B. Tschopp and
H. van de Waterbeemd, J. Med. Chem., 1994, 37, 3889–3901.
42 (a) L. Sebo, B. Schweizer and F. Diederich, Helv. Chim. Acta, 2000,
83, 80–92; (b) T. Grawe, G. Schäfer and T. Schrader, Org. Lett., 2003,
5, 1641–1644.
43 (a) P. R. Gerber and K. Müller, J. Comput.-Aided Des., 1995, 9, 251–
268; (b) Gerber Molecular Design (http://www.moloc.ch).
44 (a) A. Albert and E. P. Serjeant, The Determination of Ionization
Constants, T&A Constable Ltd., Edinburgh, 1971; (b) Sirius
Analytical Instruments Ltd.: Applications and Theory Guide to pH-
Metric pKa and log P Determination, Revision, 1.0, August 1993;
(c) Sirius Analytical Instruments Ltd.: GLpKa instruction manual,
January 1999.
45 A. Altomare, G. Cascarano, C. Giacovazzo, A. Guagliardi,
M. Burla, G. Polidori and M. Camalli, J. Appl. Crystallogr., 1994,
27, 435.
46 G. M. Sheldrick, SHELX-97 program for the refinement of crystal
structures, University of Göttingen, Germany, 1997.
O r g . B i o m o l . C h e m . , 2 0 0 4 , 2, 1 3 3 9 – 1 3 5 2
1352