From hit to lead: De novo design based on virtual screening hits of inhibitors of tRNA-guanine transglycosylase, a putative target of shigellosis therapy

Article abstract of DOI:10.1002/hlca.200390128

Shigellosis, a bacterial disease, causes the death of more than one million people per year. Extensive studies of Shigella flexneri have recognized tRNA-guanine transglycosylase (TGT, EC 2.4.2.29) as one of the key enzymes involved in the regulation of bacterial virulence. Based on the crystal structure of the Zymomonas mobilis enzyme, we have embarked on the rational design of TGT inhibitors. Herein, we describe the structure-based optimization of hits previously found by virtual screening (see Tables 1-3). For the pteridines, the most potent compound class discovered in a previous virtual screening run, a versatile synthesis could be established giving access to a broad range of substituted derivatives (see Scheme 5). The best ligand in this series, 14, exhibits a K_i = 0.45 μM.

Full text of DOI:10.1002/hlca.200390128

Helvetica Chimica Acta Vol. 86(2003)
1435
From Hit to Lead: De Novo Design Based on Virtual Screening Hits of
Inhibitors of tRNA-Guanine Transglycosylase, a Putative Target of Shigellosis
Therapy
by Ruth Brenka), Hans-Dieter Gerbera), Jeffrey D. Kittendorfb), George A. Garciab), Klaus Reutera), and
Gerhard Klebe*a)
a) Institut f¸r Pharmazeutische Chemie, Philipps-Universit‰t Marburg, Marbacher Weg 6, D-35032 Marburg
b) Interdepartmental Program in Medicinal Chemistry, College of Pharmacy, University of Michigan, Ann
Arbor MI 48109-1065, USA
This paper is dedicated to Professor Jack D. Dunitz on the occasion of his 80th anniversary
Shigellosis, a bacterial disease, causes the death of more than one million people per year. Extensive studies
of Shigella flexneri have recognized tRNA-guanine transglycosylase (TGT, EC 2.4.2.29) as one of the key
enzymes involved in the regulation of bacterial virulence. Based on the crystal structure of the Zymomonas
mobilis enzyme, we have embarked on the rational design of TGT inhibitors. Herein, we describe the structure-
based optimization of hits previously found by virtual screening (see Tables 1 3). For the pteridines, the most
potent compound class discovered in a previous virtual screening run, a versatile synthesis could be established
giving access to a broad range of substituted derivatives (see Scheme 5). The best ligand in this series, 14,
exhibits a Ki ˆ 0.45 mm.
Introduction. Shigellosis, a bacterial disease, causes death of more than one
million people per year [1], mainly in the developing countries with substandard
hygiene and water supplies. Usually, this disease is treated by antibiotics [2][3], but
increasingly multidrug-resistant strains are reported. Accordingly, there is urgent
demand for the development of new and selective antibiotics against shigellosis [4].
Extensive studies of Shigella flexneri have recognized tRNA-guanine transglyco-
sylase (TGT, EC 2.4.2.29) as one of the key enzymes involved in the regulation of
bacterial virulence [5][6]. In eubacteria, it catalyzes the incipient step in biosynthesis of
the highly modified nucleoside queuine (see Fig. 1), which is inserted in the anticodon
loop of certain tRNAs. The exchange of guanine in the unmodified tRNA by the
queuine precursor preQ1 (ˆ 7-(methylamino)-7-deazaguanine; see Scheme 1) is
performed via an associative mechanism [7][8]. Once bound to tRNA, preQ1 is
further chemically modified through at least two subsequent enzymatic steps to result
in the final queuine modification [9][10]. The presence of such modified bases in the
anticodon loop supposedly modulates the efficiency and fidelity of tRNAs during
translation [11 13]. Inhibiting the biological function of TGT is expected to result in
tRNAs lacking the modified base in the anticodon. This leads to a significantly reduced
expression of the virF gene resulting in a dramatic loss of virulence [5][14].
Based on the crystal structure of the Zymomonas mobilis enzyme [15], which
exhibits an active site of identical composition apart from a Tyr-106Phe replacement
[16], we have embarked on the rational design of TGT inhibitors. In a recent study, we

1436
Helvetica Chimica Acta Vol. 86(2003)
Fig. 1. Chemical structures of preQ1 and queuine
Scheme 1. Catalytic Mechanism of TGT. TGT catalyzes the base exchange of guanine in the wobble position of
certain tRNAs against preQ1. The reaction follows an associate mechanism via a covalent intermediate.
performed a virtual screening [17] applying a protocol of several consecutive
hierarchical filters, involving a preselection based on functional-group requirements
and fast pharmacophore matching. In a first step, a suitable pharmacophore has been
derived applying a sophisticated −hot-spot× analysis of the binding site to detect regions
of favorable proteinÀligand interactions. After a crude screening for compounds in
agreement with this protein-derived pharmacophore, flexible docking of the best
candidate molecules was performed. A remarkable variety of structurally deviating hits
has been detected by this approach. Experimental testing of the most-promising hits
demonstrated several of the test compounds as micromolar to sub-micromolar level
inhibitors. In a following step, described in the present paper, these hits have to be
validated and assessed whether they are suitable as potential leads for further
development and optimization. In this context, the synthetic accessibility and the
versatility of the functional decoration for chemical modifications of the discovered hits
has to be explored. Furthermore, the chemical modifications planned by rational
concepts are used to map different subsites of the binding pocket of TGT. Preliminary
structureÀactivity relationships help to rank the constitutions of different binding
subsites to elucidate their relevance to enhance binding affinity. In a recent study [18],

Helvetica Chimica Acta Vol. 86(2003)
1437
we started with 2-aminoquinazolin-4(3H)-one as a promising scaffold of TGT inhibitor
design. In an iterative drug-design process involving crystallography, computer design,
synthesis, and enzyme testing, we succeeded to establish a first compound class binding
to the target enzyme with sub-micromolar affinity. In the present contribution, we
report on other compound classes possessing the potential to deliver an additional lead
structure for further optimization to result in high-affinity TGT inhibitors.
Results and Discussion.
Dihydro-pyridazino-quinoline-triones. In a virtual
screening run described previously, 1 and 2 (Table 1) have been discovered as low
micromolar inhibitors of TGT [17]. Probably due to their limited solubility under
aqueous-buffer conditions, we failed to soak these compounds into crystals of
uncomplexed TGT. However, soaking was successful for compound 3, which was
detected in the same screening run. Crystal-structure analysis of this compound in
complex with TGT (PDB-Code 1N2V) revealed an unexpected flip of a backbone
peptide group in the binding pocket (Fig. 2). The backbone carbonyl group of Leu-231,
originally exposed towards the binding site in all previously studied complexes,
becomes buried, whereas the donor functionality of the adjacent backbone NH is now
available for ligand binding. In addition, an interstitial H2O molecule is incorporated,
thus mediating a favorable proteinÀligand interaction. Based on the complex structure
with 3, the binding modes of 1 and 2 could be modeled accordingly (Fig. 2). Therefore,
a plausible binding mode can be assumed for 1, which exhibits binding affinity to TGT
in the low micromolar range. Furthermore, a three-step synthesis of this compound has
been published [19]. These aspects strongly argue to consider 1 as a prospective lead
structure for new TGT inhibitors.
To discover suitable regions in the binding pocket where additional molecular
portions and functional groups hooked-up to 1 could possibly interact with the protein
to further improve binding affinity, a −hot-spot× analysis (see Exper. Part) was
performed. The result of such calculations by means of a H-bond-donor probe (N.3) is
Table 1. Three Pyridazinediones Retrieved in the Previous Virtual Screening and Tested for TGT Inhibition
Structure
Ki [mm]a)
Structure
Ki [mm]a)
5.0 Æ 1.2
3
83 Æ 18
1
2
4.7 Æ 1.0
a) The error given is the standard deviation of two independent measurements with different substrate
concentrations. Due to the elaborate determination of the Ki values, the actual error is assumed to be ca. 20
30% [34].

1438
Helvetica Chimica Acta Vol. 86(2003)
Fig. 2. Superposition of the crystallographically determined binding mode of 3 (orange) with the modeled
binding mode of 1 (green). Both ligands can form a H-bond to the interstitial H2O molecule W1.
shown in Fig. 3. The three N-atoms of 1 fall precisely into regions where the analysis
predicts favorable areas. Additionally, two so far unused −hot-spot× areas are found near
Asp-102 and Asp-280. The first design goal was, therefore, to decorate 1 accordingly
with further functional groups to address these additional regions and to hopefully gain
in affinity. Considering in detail the required topological connectivity, the core
structure of 1 does not appear ideally suited for this optimization due to the placement
of the S-atom in the thiophene moiety. For chemical reason, no further portions can be
linked via the S-atom. Focusing on the 2-position as substitution site appears rather
unreasonable due to steric reasons since substituents at this position could rather
unlikely address the indicated −hot spots× but would orient towards the rim of the
binding pocket. Therefore, in the first step, the thiophene ring of 1 was replaced
bioisosterically by a benzene ring (Scheme 2). Within the error limits of the
biochemical assay, the affinity of 4 (3.7 Æ 1.5 mm) remained unchanged compared to
1 (5.0 Æ 1.2 mm). For subsequent decoration of 4, positions 8 and 9 appear suitable. For
this purpose, an OH group was introduced into the new core structure that could serve
as a linker at C(8). A versatile synthesis concept for this compound class starts with
substituted anthranilic acid (Scheme 3). In a first trial, we selected the 2-amino-5-
hydroxy derivative (RˆOH). After esterification and addition to the triple bond of
but-2-ynedioic acid ester, the central pyridinone ring was closed. The dihydropyrida-
zine ring moiety was introduced by means of hydrazine. Due to synthetic reasons, any
substituent at the linker OH group had to be introduced into the hydroxyanthranilic
acid prior to the construction of the central ring moiety.

Helvetica Chimica Acta Vol. 86(2003)
1439
Fig. 3. Mapping of putative −hot spots× of binding in the active site of TGT for an N.3 probe. The −hot spots×,
calculated with DrugScore, are contoured at a 80 (cyan), 84 (blue), and 88% (magenta) level above the global
minimum. For orientation, the modeled-binding geometry of 1 is also shown.
Scheme 2. Optimization of 1 via Bioisosteric Replacement. The Ki values are given.
Already, the introduction of the OH group at C(8) improved binding affinity
leading to a Ki value of 0.7 mm for 5. Molecular modeling based on MOLOC [20] and
SYBYL [21] suggested that a CH2CH2NH chain added to 5 could successfully address
the −hot spot× near Asp-280. Directly in the neighborhood of this polar spot, a small
hyrophobic pocket opens, formed by the amino acids Val-45, Leu-68, Cys-281, and

1440
Helvetica Chimica Acta Vol. 86(2003)
Scheme 3. General Synthesis Scheme for Dihydro-pyridazino-quinoline-triones 4
7
Val-282. The −hot-spot× analysis based on a C.3 probe indicates favorable binding of
hydrophobic groups in this area (Fig. 4,b).
In a previous study, we successfully addressed this pocket during the optimization of
quinazolinone derivatives as TGT inhibitors [18]. In consequence, we tried to design a
compound that could occupy both the hydrophobic pocket and the donor spot near
Asp-280. A pyrrolidine moiety appeared to be a suitable candidate. The modeled
binding mode of 6 (Table 2) is shown in Fig. 4. In the computed binding mode, the
aliphatic C-atoms of the pyrrolidine ring fill the hydrophobic pocket (Fig. 4,b), and its
N-atom donates a H-bond to Asp-280 (Fig. 4,a). The pKa of pyrrolidine in H2O is
reported to be 11.35 [22]. We, therefore, assume that this group is positively charged in
the binding pocket leading to a charge-assisted H-bond with Asp-280. Depending on
the local surrounding of such H-bonds, an affinity increase of up to 3000-fold could be
expected compared to the unsubstituted derivative [23 25]. Furthermore, all torsion
angles of 6 in its bound conformation are indicated as energetically favorable [26]. To
our surprise, the Ki value of 6 did not improve compared to the unsubstituted parent
structure. Indeed, a Ki of 9.4 mm indicates even reduced binding compared to the core
structure 4. To investigate whether the aliphatic C-atoms of the pyrrolidine moiety are
responsible for this effect, we considered testing the structurally reduced analogue 7

Fig. 4. Modeled binding mode of 6: a) −Hot spots×, calculated with DrugScore for an N.3 probe contoured at a 80 (cyan), 84 (blue), and 88% (magenta) level above the
global minimum; all H-bond donors of the ligand fall into regions were their presence is predicted to be favorable; the N-atom of the pyrrolidine ring forms a H-bond to
Asp-280). b) − Hot spots×, calculated with DrugScore for a C.3 probe contoured at a 88 (magenta), 90 (blue), and 92% (yellow) level; the aliphatic C-atoms of the
pyrrolidine ring fill the hydrophobic pocket formed of Val-45, Leu-68, Cys-281, and Val-282).

1442
Helvetica Chimica Acta Vol. 86(2003)
Table 2. Dihydro-pyridazino-quinoline-triones and a Dihydro-tetraaza-anthracenedione Synthesized and Sub-
sequently Tested for TGT Inhibition
Structure
Ki [mm]a)
Structure
Ki [mm]a)
6
7
9.4 Æ 0.1
8
5.6 Æ 0.4
a) The error given is the standard deviation of two independent measurements with different substrate
concentrations. Due to the elaborate determination of the Ki values, the actual error is assumed to be ca. 20
30% [34].
lacking these C-atoms. However, compound 7 polymerized, probably, because its
terminal amino group reacts with the carbonyl group of the pyridinone ring.
Unfortunately, it could not be tested for biological activity.
Analyzing in detail the suggested binding mode of 1 and its dihydro-pyridazino-
quinoline-trione analogues 4 7 (Figs. 3 and 4), the above-mentioned possibly reactive
carbonyl group is not involved in any favorable interactions with the surrounding
amino acids. As a consequence, it should be possible to avoid this for the stability of the
compound class unfortunate functional group without loosing affinity. Accordingly, we
considered dihydro-tetraaza-anthracenes (ˆdihydro-pyridazino-quinoxalines) such as
8 as possible bioisosteric analogues of 4. In comparison to 4, the pyridinone ring is
replaced by a pyrazine ring in 8. This exchange corresponds to the replacement of a H-
bond donor facility of the pyridinone by a H-bond acceptor in the pyrazine ring. The
inhibitor presumably interacts via this functional group with the interstitial H2O
molecule W1. However, due to its bifunctionality, this H2O molecule can act either as a
H-bond donor or acceptor with respect to the ligand functional group [17]. Therefore, it
can be expected that this exchange does not affect binding affinity. This new compound
class of dihydro-tetraaza-anthracene-diones is synthetically accessible via a route
described by Chattaway and Humphrey [27]. Interestingly, the Ki value of 8 is 5.6 mm,
and falls indeed into the same range as that of 4. Unfortunately, the yield of the
followed synthetic route is very low; thus, 8 does not appear to be a suitable choice for a
prospective lead structure.
Pteridines. For both compound classes, dihydro-pyridazino-quinoline-triones and
dihydro-tetraaza-anthracene-diones described above, unexpected difficulties occurred
during synthesis. This prompted us to search for a better-suited lead structure that could
be more easily modified. In a virtual screening run, we discovered pteridines to be

Helvetica Chimica Acta Vol. 86(2003)
1443
potent inhibitors of Z. mobilis TGT. Experimental verification of this suggested hit
demonstrated 2-aminopteridin-4(3H)-one to be a 0.6 mm inhibitor [17]. This compound
is synthetically accessible in a one-step reaction starting from 2,4,5-triaminopyrimidin-
6-ol and glyoxal (ˆethanedial; Scheme 4) [28]. If instead of glyoxal (RˆH) substituted
keto aldehydes are used, pteridines substituted at the 6- or 7-position are obtained [29].
Obviously, pteridines are a well-studied class of compounds, readily derivatized: more
than 60 2-aminopteridin-4(3H)-ones with varying substituents at the 7-position are
stored in the Beilstein Database (status October 2002). This makes 2-aminopteridin-
4(3H)-one an interesting suggestion as a lead structure for TGT inhibitors.
Scheme 4. General Synthesis Scheme for 2-Aminopteridin-4(3H)-ones
Presumably, 2-aminopteridin-4(3H)-one adopts a similar binding mode as 3. The
modeled binding geometry is shown in Fig. 5. In this orientation, it places its H-bond
donors in regions indicated as favorable in the −hot-spot× analysis. However, again, it
leaves the −hot spot× near Asp-280 as well as the hydrophobic pocket formed by Val-45,
Leu-68, Cys-281, and Val-282 unoccupied. Accordingly, we decided to address these
areas of the binding pocket by substituents attached to C(7) of the pteridine skeleton.
Simple access to pteridines follows the reaction of substituted aldehydes with 2,4,5-
triaminopyrimidinol (Scheme 4). Starting with substituted ketoaldehydes, constitu-
tional isomers have to be expected. To yield the desired isomer, the reaction conditions
have to be optimized for each starting material individually. They depend on the
ketoaldehyde used and must be re-established for any new reagent. To circumvent this
obvious disadvantage, we selected a pteridine derivative as core structure that already
incorporated an appropriate substituent at C(7) as precursor for a linker functional
group. After functionalizing this linker, differently substituted compounds could be
obtained avoiding possible constitutional isomers.
A pteridine with suitable precursor for a potential linker group is 2-amino-7-
methylpteridin-4(3H)-one (9; Table 3); the Me group can be easily functionalized by
bromination. Subsequently, the Br-atom can be substituted by different S- and O-
nucleophiles to introduce suitable side chains. For all respective steps, related syntheses
have been described in the literature [29 31], the best yield for the last step of the
reaction being obtained for S-nucleophiles [31]. Following the three subsequent
reaction steps shown in Scheme 5, different pteridine derivatives should be easily
available. According to the expected yield in the first design cycle, only S-nucleophiles
were considered.
To select appropriate thiols, the Available Chemicals Directory (ACD) was
screened as an idea generator following a hierarchical protocol. First, the database
was screened with SELECTOR [32] for aliphatic thiols. Out of 215212 compounds,
only 701 passed this criterion. In a next step, the remaining hitlist was filtered for
compounds that contained exactly one thiol group along with a molecular mass of less
than 300 D and up to six rotatable bonds. These criteria were chosen to reveal unique

1444
Helvetica Chimica Acta Vol. 86(2003)
Fig. 5. Modeled binding mode of 2-aminopteridin-4(3H)-one together with −hot spots×, calculated with
DrugScore for an N.3 probe. − Hot spots× contoured at a 80 (cyan), 84 (blue), and 88% (magenta) level. The
ligand can form a H-bond to the interstitial H2O molecule W1. The unoccupied −hot spot× near Asp-280 as well
as the hydrophobic pocket formed by Val-45, Leu-68, Cys-281, and Val-282 can be addressed via substitutions at
C(7).
product formation during synthesis and to avoid compounds with high molecular mass
and pronounced flexibility. After this step, 204 compounds remained. Filtering the
hitlist for duplicates reduced the number of prospective candidates to 148 compounds.
Scheme 5. Synthesis of Derivates of 2-Aminopteridin-4(3H)-ones with Substituents at C(7)

Helvetica Chimica Acta Vol. 86(2003)
1445
Table 3. Pteridinones that were Synthesized and Subsequently Tested for TGT Inhibition
Structure
Ki [mm]a)
Structure
Ki [mm]a)
9
5.6 Æ 1.7
12
5.2 Æ 0.8
10
11
3.4 Æ 0.4
7.9 Æ 1.9
13
14
6.7 Æ 0.9
0.45 Æ 0.05
a) The error given is the standard deviation of two independent measurements with different substrate
concentrations. Due to the elaborate determination of the Ki values, the actual error is assumed to be ca. 20
30% [34].
By using a SPL script in SYBYL [21], these fragments were assembled to putative
pteridine-type inhibitors. Subsequently, they were docked with FlexX [33] into the
binding pocket of the Z. mobilis TGT. For 144 of the candidate molecules, a docking
solution was generated. Subsequently, the obtained hitlist was inspected visually.
Prospective-looking hits were minimized by using the MAB force field [20] simulta-
neously keeping the binding pocket rigid. A final selection for synthesis and enzyme
testing regarded in particular the following criteria: a) synthetic accessibility, b)
complementarity between ligand and protein surface in terms of spatial occupancy and
matching contacts in hydrophobic/hydrophilic surface patches, c) absence of unfavor-
able Van der Waals interactions after minimization, and d) diversity of the attached
fragments within the considered series. The compounds 10 12 finally selected are
shown in Table 3. Compound 13 was obtained as an O-analog and tested for
comparison. Derivative 12 with a terminal cyclopentyl moiety was expected to occupy
the small hydrophobic binding niche composed by Val-45, Leu-68, Cys-281, and Val-
282. Since the −hot-spot× analysis (see below, Fig. 8) indicated favorable binding of
hydrophobic moieties to this subsite, the analogous derivative 14 with a terminal
thiophene group was selected likewise.
Surprisingly, the Ki value of 2-amino-7-methylpteridin-4(3H)-one (9) is by about
one order of magnitude higher than that of the parent pteridine (5.6 vs. 0.6 mm).
Consulting force-field calculations with MOLOC did not reveal any unfavorable
interactions for the Me group of 9 with the protein environment. Thus, a simple

1446
Helvetica Chimica Acta Vol. 86(2003)
explanation for this finding cannot be provided. Possibly, the additional Me group
ruptures a H-bond network mediated by interstitial H2O molecules to polar residues in
the binding pocket. Such interactions have been observed in the crystal structures of
3,5-diaminophthalhydrazide with the enzyme [34].
The best docking solution of 10 suggested by FlexX is shown in Fig. 6. According to
this binding geometry, the terminal NH2 group forms a H-bond to Asp-280. Due to
chemical reasons, a similar H-bond cannot be established by 11 and 12. Nevertheless,
the Ki values of all three compounds fall, within the error limits of the assay, into the
same range as the lead compound 9 (Ki ˆ 5.6( 9), 3.4 (10), 7.9 (11), 5.2 (12) mm).
Obviously, and in agreement with the results described for the pyridazinedione series,
the assumed charge-assisted H-bond, likely to be formed by 10, cannot improve affinity
significantly compared to 9. Furthermore, the purely lipophilic moieties of 11 and 12 do
not improve binding affinity either.
Fig. 6. Best-ranked docking solution for 10 together with −hot spots×, calculated with DrugScore for an N.3 probe,
− Hot spots× contoured at a 80 (cyan), 84 (blue), and 88% (magenta) level. In the suggested binding mode, the
terminal NH2 group forms
a
H-bond to Asp-280. The conformational angle N(ar-
om.)ÀC(arom.)ÀC(sp3)ÀS(sp3) (circle) adopts a value of ca. 308 which does not correspond to the most
favorable arrangement of this torsional fragment.
Consulting conformational preferences as observed in small-molecule crystal
structures (see, e.g., the MIMUMBA approach [26]), energetically favorable con-
formers of the torsional fragment N(arom.)ÀC(arom.)ÀC(sp3)ÀS(sp3) (Fig. 7)
correspond to values of 08, 908, and 1808. The conformer at 908 is the most favorable
conformation followed by the arrangement at 08. But the first conformation cannot be
adopted by the pteridine derivatives 10 12 at the binding site due to unfavorable steric

Helvetica Chimica Acta Vol. 86(2003)
1447
interactions with Tyr-106and Gly-216 on either sides of the binding pocket. All
derivatives could be accommodated with the less-favorable value of 08 for this
torsional fragment. However, if this angle is adopted by 10, the assumed H-bond to
Asp-280 could not be formed. To achieve this interaction, the torsion angle must be
shifted to ca. 308, thus driving this fragment significantly out of a local minimum.
Possibly, this slightly unfavorable arrangement could be one explanation for why the
assumed charge-assisted H-bond to Asp-280 does not increase binding affinity.
Furthermore, there are indications from enzyme kinetics that Asp-280 acts as a general
acid/base during the enzyme reaction (Scheme 1) [35]. In such a case, its functional role
would be to protonate the leaving guanine moiety and to deprotonate the incoming
preQ1. As a consequence, Asp-280 would be protonated most of the time. Accordingly,
its pKa value is expected to exhibit a higher value than measured for the free acid.
Therefore, much less-favorable H-bond-acceptor properties are expected for Asp-280.
This assumption is supported by our observation that the placement of positively
charged donor groups (e.g., in 6 or 10) in the neighborhood of Asp-280 does not
improve binding affinity. Nevertheless, it also has to be kept in mind that the enzymatic
reaction used to record inhibition data involves tRNA binding to TGT. Thus, possibly
through tRNA complexation, the conformation and local properties of Asp-280 could
be affected.
Fig. 7. Torsion Angle N(arom.)ÀC(arom.)ÀC(sp3)ÀX (XˆS(sp3), O(sp3)).
Favorable angles for XˆS are 08, 908, and 1808 with the arrangement at 908 as
most favorable. For XˆO, conformers at 08, 908, and 1808 are energetically well
disposed, the arrangement at 08 is most preferred, with a broad distribution around
the maxima. These considerations are based on statistical preferences observed
in torsional libraries retrieved from small-molecule crystal structures [26].
The ethoxy derivative 13 shows a Ki of 6.7 mm. It falls into the same range as the
unsubstituted lead 9. In directly comparing its properties to those of 11, it has to be
considered that the side chain of the latter is extended by one CH2 group. Cases have
been reported where an optimally placed additional Me group increases binding
affinity by a factor of 10 [23 25]. Nevertheless, it seems likely in the present case that
the substitution of S for O has no dramatic influence on affinity. The conformational
preferences of an N(arom.)ÀC(arom.)ÀC(sp3)ÀO(sp3) fragment differ from those of
the previously described N(arom.)ÀC(arom.)ÀC(sp3)ÀS(sp3) chain (Fig. 7). In the O-
case, favorable angles are 08, 908, and 1808, and the 08 conformer corresponds to the
global minimum. Furthermore, a CÀO bond length is by 0.4 ä shorter than a CÀS
bond. Therefore, 7-(oxymethyl)pteridines such as 13, in contrast to 7-(thiomethyl)p-
teridines, can adopt a torsion angle of 908 in the binding pocket of TGT without
creating steric conflicts with the surrounding amino acids. Additionally, the distribution
of torsion angles around 908 is quite broad, thus, slight distorsions to lower values
appear energetically equivalent. Accordingly, 7-(oxymethyl)pteridines possibly exhibit
better-suited conformational properties compared to the 7-(thiomethyl)pteridines to
properly accommodate the binding pocket. Moreover, data for a large variety of
alcohols are stored in the ACD. In a subsequent design cycle, we want to investigate

1448
Helvetica Chimica Acta Vol. 86(2003)
whether such derivatives can further improve the affinity of the lead structure by using
such side chains.
The most-potent inhibitor in the discussed series is 14. With 0.45 mm, its Ki value is
more than one order of magnitude lower compared to the original lead structure 9. The
modeled binding mode is shown in Fig. 8 together with −hot spots× calculated with
DrugScore using a C.ar probe. The thiophene ring fully accommodates the binding
niche formed by Val-45, Leu-68, Cys-281, and Val-282. This area is predicted to be
favorable for aromatic C-atoms. In a following design cycle, we will further map the
molecular-recognition properties exhibited by this subpocket using other aromatic
portions. Interestingly, in a previous optimization study based on quinazolinones as the
core skeleton, we also found that an aromatic moiety ideally fits into this hydrophobic
binding niche simultaneously increasing the binding affinity [18].
Fig. 8. Modeled binding mode of 14 together with −hot spots×, calculated with DrugScore for a C.ar-probe. − Hot
spots× contoured at a 89 (purple), 91 (orange), and 95% (yellow) level. The thiophene moiety falls exactly into a
region predicted as favorable for aromatic C-atoms.
Conclusions. Starting with various hits from virtual screening, we tried to improve
binding affinity focusing mainly on two lead structures. Two compound classes, the
dihydropyridazino-quinoline-triones and dihydro-tetraaza-anthracene-diones, resulted
in unexpected problems during synthesis. For the pteridines, the most-potent
compound class discovered in a previous virtual screening run, a versatile synthesis
could be established giving access to a broad range of substituted derivatives. Based on
structural information present in the ACD and guided by the results of −hot-spot×
analyses, several new inhibitors were synthesized and tested. The best ligand, 14, in this

Helvetica Chimica Acta Vol. 86(2003)
1449
series exhibits a Ki ˆ 0.45 mm. The thiophene moiety of this compound accommodates
exactly a region in the binding pocket that is predicted to be favorable for aromatic C-
atoms. In a following design cycle, we will try to further experience the molecular-
recognition properties exhibited by this subpocket using different aromatic portions.
Experimental Part
1. Molecular Modeling. The modeling was performed with SYBYL 6.8 [21] and MAB/Moloc 01/05/08 [20]
running on a Silicon-Graphics-O2 (R10000) workstation.
2. − Hot-Spot× Analysis of the Binding Pocket. − Hot spots× were calculated with DrugScore [36] by using the
unoccupied binding pocket of TGT in its conformation adopted with the bound ligand 3 (PDP-code 1N2V). In
these calculations, the position of the interstitial H2O molecule W1, assigned to atom-type O.3, was considered.
The N.3 probe was selected to analyze H-bond-donor interactions, and the C.ar probe as well as the C.3 probe
were used to map hydrophobic interactions. − Hot spots× are contoured on percentage levels with respect to the
global minimum.
3. Construction of 3D Database. The 2D connection tables of entries in the Available Chemicals Directory
(ACD) were converted as described in [17].
4. Hierarchical Filtering. The hierarchical filtering was performed by using SELECTOR [32]. The SLN
code (Sybyl line notation) for the required aliphatic thiol fragments was defined as SHCH2.
5. Docking. With an in-house script, exocyclic guanidino and amidino groups or primary and secondary
aliphatic amino groups were protonated; similarly, phosphoric, sulfonic, and carboxylic acid groups were
deprotonated. Docking calculations were performed with FlexX, version 1.102. To speed up calculations, the
−mapref× mode was activated, and all pteridines were matched on a preoriented core structure. The FlexX
scoring function was used during the complex-construction phase. The obtained solutions were reranked by
DrugScore [37] as implemented into FlexX.
6. Force-Field Minimization. Ligands were minimized in the binding pocket exhibiting the protein
conformation as complexed with 3 by using the MAB force field [20]. The default parameters were applied, and
the binding pocket was kept rigid.
7. Kinetic Analysis. The apparent Ki values were measured as described elsewhere [34].
8. Syntheses. 8.1. General. Starting compounds were purchased from Sigma-Aldrich Chemie GmbH. Unless
stated otherwise, commercially available solvents were used without further purification. Column chromatog-
raphy (CC): silica gel 60 (0.063 0.200 mm), Merck, solvents for CC were distilled before use. NMR Spectra:
Jeol JNM-GX 400 (1H at 399.79 MHz, −400 MHz×; 13C at 100.54 MHz, −100.5 MHz×) or Jeol Lambda 500 Delta
(1H at 500.00 MHz, −500 MHz×; 13C at 125.65 MHz, −125.7 MHz×); d in ppm referenced to SiMe4 (ˆ0.00 ppm) for
1H and to CDCl3 (ˆ 77.00 ppm) or DMSO (ˆ 39.50 ppm) for 13C, J in Hz; due to poor solubility, only 1H-NMR
spectra were obtained for all pteridine compounds, recorded in 20% DCl/D2O (referenced to 5.48 ppm);
13C-NMR assignments are in some cases subject to change. MS: double-focussing sectorfield mass spectrometer
VG 7070 H by Vacuum Generators or VG-AutoSpec by Micromass applying electron impact (EI) ionization, in
m/z (rel. %).
8.2. Dihydro-thieno-pyrido-pyridazine-trione 1 and Dihydro-pyridazino-quinoline-triones 4 6. 6,7-Dihy-
drothieno[2'3':5,6]pyrido[2,3-d]pyridazine-5,8,9(4H)-trione (1) was prepared according to [19].
2,3-Dihydropyridazino[4,5-b]quinoline-1,4,10(5H)-trione (4). Analogously to the three-step synthesis of 1,
from methyl 2-aminobenzoate (5.0 mmol) (see also [38]): 0.75 g (66%) of 4. Light-yellow solid. 1H-NMR
(400 MHz, (D6)DMSO): 13.14, 12.51, 12.31 (3s, 3 NH); 8.29 (d, 3J ˆ 8.1, HÀC(9)); 8.16( d, 3J ˆ 8.2, HÀC(6));
7.93 (dd, 3J ˆ 7.8, 7.2, HÀC(8)); 7.57 (dd, 3J ˆ 7.8, 7.5, HÀC(7)). 13C-NMR (100.5 MHz, (D6)DMSO): 178.70
(C(10)); 154.01, 152.46(C(1), C(4)); 139.67, 137.86(C(4a), C(5a)); 134.08 (C(9)); 124.59, 124.36(C(7), C(8));
‡
‡
123.59 (C(9a)); 119.68 (C(6)); 104.74 (C(10a)). EI-MS: 229 (100, M , C11H7N3O3 ).
2,3-Dihydro-8-hydroxypyridazino[4,5-b]quinoline-1,4,19(5H)-trione (5). Analogously to the three-step
synthesis of 4, from methyl 5-hydroxy-2-aminobenzoate (1.6mmol). The latter was obtained by the following
procedure: To a soln. of 5-hydroxy-2-aminobenzoic acid (6.13 g, 40.0 mmoles) in MeOH (60 ml), conc. H2SO4
soln. (10 ml) was added cautiously. The mixture was refluxed for 20 h and the major excess of MeOH
evaporated. The residual dark suspension was poured into ice, the mixture adjusted to pH 8 with NaHCO3 soln.,
and the resulting precipitate filtered. After washing with H2O, the product was dried in vacuo: methyl ester
(5.60 g, 84%). Grey solid, which was recrystallized from EtOH for further purification. 1H-NMR (400 MHz,

1450
Helvetica Chimica Acta Vol. 86(2003)
(D6)acetone): 7.64 (s, OH); 7.27 (d, 4J ˆ 2.9, HÀC(6)); 6.88 (dd, 3J ˆ 8.8, 4J ˆ 2.9 Hz, HÀC(4)); 6.70 (d, 3J ˆ 8.8,
HÀC(3)); 5.93 (br. s, 2 NH2); 3.81 (s, MeO).
Data of 5: 320 mg (80%). Brownish solid. 1H-NMR (500 MHz, (D6)DMSO): 8.30 (br. s, OH, 3 NH); 7.82
(d, 3J ˆ 8.8, HÀC(6)); 7.51 (d, 4J ˆ 2.7, HÀC(9)); 7.27 (dd, 3J ˆ 9.0, 4J ˆ 2.8, HÀC(7)). 13C-NMR (125.7 MHz,
(D6)DMSO): 175.60 (C(10)); 159.07, 156.59, 154.00 (C(1), C(4), C(8)); 142.23, 141.84 (C(5a), C(4a)); 128.26
‡
‡
(C(9a)); 126.46, 123.78 (C(6), C(9)); 106.08, 104.55 (C(7), C(10a)). EI-MS: 245 (100, M , C11H7N3O4 ).
2,3-Dihydro-8-{[(2S)-pyrrolidin-2-yl]methoxy}-pyridazino[4,5-b]quinoline-1,4,10(5H)-trione (6). Analo-
gously to the three-step synthesis of 4 and 5 via the following isolated and characterized intermediates A D.
Methyl 2-Amino-5-{{(2S)-1-[(tert-butoxy)carbonyl]pyrrolidin-2-yl}methoxy}benzoate (A). To a soln. of the
sodium salt of methyl 2-amino-5-hydroxybenzoate, freshly prepared from methyl 2-amino-5-hydroxybenzoate
(1.00 g, 6.0 mmol) and 60% NaH (0.250 g, 6.25 mmol) in anh. DMSO (15 ml), tert-butyl (2S)-2-{{[(4-
methylphenyl)sulfonyl]oxy}methyl}pyrrolidine-1-carboxylate (1.78 g, 5.0 mmol); prepared from tert-butyl (2S)-
2-(hydroxymethyl)pyrrolidine-1-carboxylate according to [39] prior to use), in anh. DMSO (10 ml) was added
dropwise, and the mixture was stirred for 1 h. Stirring was continued for another 18 h at 758. The mixture was
poured into H2O (100 ml), and after addition of brine (100 ml), it was extracted thoroughly with AcOEt. The
combined org. extract was washed with NaOH soln. and then with NaH2PO4 soln., dried (Na2SO4), and
evaporated and the obtained dark-red oil (1.50 g) submitted to CC (silica gel, hexane/AcOEt 3 :1): A (1.33 g,
63%). Light-yellow oil. 1H-NMR (500 MHz, CDCl3): 7.37 (br. d, HÀC(6)); 6.96 (d, HÀC(4)); 6.62
(d, HÀC(3)); 5.40 (br. s, NH2); 4.10 3.30 (m, 5 H, CH2O, pyrr.); 3.87 (s, MeO); 2.10 1.80 (m, 4 H, pyrr.);
1.46( s, t-Bu).
Dimethyl (2E)/(2Z)-2-{{4-{{2S)-1-[(tert-Butoxy)carbonyl]pyrrolidin-2-yl}methoxy}-2-(methoxycarbonyl)-
phenyl}amino}but-2-enedioate (B). From A (5.0 mmol): B (2.17 g, 86%), (E)/(Z) 95 :5 (by 1H-NMR), after CC
(hexane/AcOEt 2 :1). Colorless oil. 1H-NMR (500 MHz, CDCl3; (E)-isomer): 11.21 (s, NH); 7.48 (br. d,
HÀC(3')); 6.98 (br. s, HÀC(5')); 6.62 (d, HÀC(6')); 5.48 (br. s, HÀC(3)); 4.20 3.30 (m, 5 H, CH2O, pyrr.);
‡
‡
3.95, 3.78, 3.72 (3s, 3 MeO); 2.05 1.85 (m, 4 H, pyrr.); 1.47 (s, t-Bu). EI-MS: 492 (38, M , C24H32N2O9 ).
Dimethyl 6-{{(2S)-1-[(tert-Butoxy)carbonyl]pyrrolidin-2-yl}methoxy}-1,4-dihydro-4-oxo-quinoline-2,3-di-
carboxylate (C). From B (4.4 mmol): C (1.31 g, 65%), after CC (hexane/AcOEt 1:3). Light-yellow solid.
1H-NMR (500 MHz, CDCl3): 12.41 (br. d, NH); 7.95 (br. s, HÀC(8)); 7.60 (−t×, HÀC(5)); 7.47 (dd, 3J ˆ 9.2, 2.7,
‡
HÀC(7)); 4.30 3.30 (m, 5 H, CH2O, pyrr.); 2.1 1.9 (m, 4 H, pyrr.); 1.48 (s, t-Bu). EI-MS: 460 (34, M ,
‡
C23H28N2O8 ).
tert-Butyl (2S)-2-{[(1,2,3,4,5,10-hexahydro-1,4,10-trioxopyridazino[4,5-b]quinolin-8-yl)oxy]methyl}pyrro-
lidine-1-carboxylate (D). From C (2.3 mmol): D (0.84 g, 85%). Yellow solid. 1H-NMR (500 MHz, (D6)DMSO):
13.25, 12.70, 12.28 (3 br. s, 3 NH); 8.11 (d, 3J ˆ 8.9, HÀC(6)); 7.63, 7.61 (−dd×, HÀC(7), HÀC(9)); 4.19, 4.09 (2m,
‡
CH2O, CHN (pyrr.)); 3.28 (m, CH2N (pyrr.)); 2.05 1.75 (m, 2 CH2 (pyrr.)); 1.40 (s, t-Bu). EI-MS: 428 (25, M ,
‡
‡
C21H24N4O6 : 245 (78, [M À CH2Àpyrr.ÀBoc) ‡ 1] ).
Deprotection of D: 6. To a suspension of the above obtained Boc-protected D (429 mg, 1.0 mmol) in DMF
(15 ml), 32% HCl soln. (2 ml) was added. The mixture was heated to 1008 thus resulting in a clear soln. which
was stirred for another hour at 1008. After cooling to 208, H2O (150 ml) was added, and the soln. was neutralized
with NaHCO3 (2.5 g, 30 mmol). Stirring was continued for 10 min, the resulting precipitate filtered, thoroughly
washed with H2O, and dried in vacuo: 6 (285 mg, 87%). Light-yellow solid. 1H-NMR (500 MHz, (D6)DMSO):
11.40 (br. s, NH); 7.85 (d, 3J ˆ 8.7, HÀC(6)); 7.56 (br. s, HÀC(9)); 7.33 (d, 3J ˆ 8.0, HÀC(7)); 4.25, 4.17, 3.87, 3.20
(4m, 5 H, CH2O, pyrr.); 2.11, 1.98, 1.89, 1.78 (4 m, 4 H, pyrr.). 13C-NMR (125.7 MHz, (D6)DMSO): 175.53
(C(10)); 159.52, 157.01, 154.80 (C(1), C(4), C(8)); 143.92, 142.65 (C(5a), C(4a)); 128.46 (C(9a)); 125.90, 123.75
(C(6), C(9)); 105.41, 103.93 (C(10a), C(7)); 68.54 (CH2O); 58.20 (CHN); 46.04 (CH2N); 27.35, 24.28 (2 CH2).
‡
‡
‡
EI-MS: 328 (7, M , C16H16N4O4 ): 245 (100, [M À CH2Àpyrr.) ‡ 1] ).
2,3-Dihydropyridazino[4,5-b]quinoxaline-1,4-dione (8). Diethyl quinoxaline-2,3-dicarboxylate (274 mg,
1.0 mmol; obtained by esterification of quinoxaline-dicarboxylic acid [27]) and 85% hydrazine hydrate (0.5 g,
15.6mmol) in EtOH (15 ml) were refluxed for 6h. The resulting precipitate was filtered and washed with 2 n
1
HCl, followed by H2O: 8 (170 mg, 79%). Yellow solid. H-NMR (400 MHz, (D6)DMSO): 11.80 (br. s, 2 NH);
8.38 (non-resolved AA' of AA'BB', HÀC(6), HÀC(9)); 8.13 (unresolved BB' of AA'BB', HÀC(7), HÀC(8)).
‡
‡
EI-MS: 214 (100, M , C10H6N4O2 ).
8.3. 2-Amino-7-(thiomethyl)pteridin-4(3H)-ones 10 12 and 14. General Procedure. To a soln. of sodium
ethanolate (5.0 mmoles) in abs. EtOH, freshly prepared from Na (115 mg, 5.0 mmol) in abs. EtOH (15 ml), the
appropriate thiol (8.0 mmol) was added and the soln. stirred for 30 min. Then, 2-amino-7-(bromomethyl)pter-
idin-4(3H)-one hydrobromide (1.0 mmol) [30], prepared from 2-amino-7-methylpteridin-4(3H)-one [29], was

Helvetica Chimica Acta Vol. 86(2003)
1451
added and the mixture refluxed for 6h. After adjusting to pH 7, the resulting precipitate was filtered, thoroughly
washed with H2O and then with acetone, and finally dried.
2-Amino-7-{[(2-aminoethyl)thio]methyl}pteridin-4(3H)-one (10): Only 2 mmol of 2-aminoethanethiol
hydrochloride was added: 215 mg (85%) of 10. Brown solid. 1H-NMR (500 MHz, DCl in D2O): 6.94
‡
‡
(s, HÀC(6)); 2.21 (s, CH2ÀC(7)); 1.42 (t, CH2N); 1.04 (t, CH2S). EI-MS: 252 (0.2, M , C9H12N6OS ), 177 (100,
‡
[M À H2N(CH2)2S ‡ 1] ).
2-Amino-7-[(propylthio)methyl]pteridin-4(3H)-one (11). Yield 225 mg (90%) of 11. Orange solid.
1H-NMR (500 MHZ, DCl in D2O): 6.92 (s, HÀC(6)); 2.12 (s, CH2ÀC(7)); 0.62 (t, CH2S); À0.38 (sext.,
‡
‡
‡
CH2); À1.04 (s, Me). EI-MS: 251 (1, M , C10H13N5OS ), 177 (100, [M À Me(CH2)2S ‡ 1] ).
2-Amino-7-[(cyclopentylthio)methyl]pteridin-4(3H)-one (12). Yield 270 mg (97%) of 12. Brownish solid.
1H-NMR (500 MHz, DCl in D2O): 6.84 (s, HÀC(6)); 2.08 (s, CH2ÀC(7)); 1.11 (q, CHS); À0.08, À0.37, À0.48,
‡
‡
‡
À0.59 (4m, 4 CH2 (cyclopentyl)). EI-MS: 277 (1, M , C12H15N5OS ), 177 (100, [M À C5H9S ‡ 1] ).
2-Amino-7-[(2-thienylthio)methyl]pteridin-4(3H)-one (14). Yield 265 mg (91%) of 14. Brownish solid.
1H-NMR (500 MHz, DCl in D2O): 8.47 (s, HÀC(5')); 7.91 (d, 3J ˆ 5.2, HÀC(3')); 7.41 (s, HÀC(6)); 7.40 (d, 3J ˆ
‡
‡
5.2, HÀC(4')); 3.03 (s, 1 CH2ÀC(7)); 4.22, 4.14 (2m, 1 H, CH2ÀC(7)). EI-MS: 291 (76, M , C11H9N5OS2 ), 177
‡
(100, [M À (C4H3S)S ‡ 1] ).
2-Amino-7-(ethoxymethyl)pteridin-4(3H)-one (13). As described for 10 12 and 14, without addition of a
thiol: 170 mg (77%) of 13. Brown solid. 1H-NMR (400 MHz, DCl in D2O): 6.88 (s, HÀC(6)); 3.00
‡
‡
(s, CH2ÀC(7)); 1.80 (q, MeCH2O); À0.70 (t, MeCH2O). EI-MS: (221 (1, M , C9H11N5O2 , 177 (100, [M À
‡
C2H5O ‡ 1] ).
This work was supported by the Deutsche Forschungsgemeinschaft (Grant KL-1204/1).
REFERENCES
[1] K. Kotloff, J. Winickoff, B. Ivanoff, J. Clemens, D. Swerdlow, P. Sansonetti, G. Adak, M. Levine, Bull. World
Health Organiz. (WHO Bull.) 1999, 77, 651.
[2] W. A. Khan, C. Seas, U. Dhar, M. A. Salam, M. L. Bennish, Ann. Intern. Med. 1997, 1, 697.
[3] M. L. Replogle, D. W. Fleming, P. R. Cieslak, Clin. Infect. Dis. 2000, 30, 515.
[4] WHO, in −Overcoming Antimicrobial Resistance×, http://www.who.int/infectious-disease-report/2000/
index.html, 2000.
[5] J. M. Durand, B. Dagberg, B. E. Uhlin, G. R. Bjork, Mol. Microbiol. 2000, 35, 924.
[6] N. Okada, S. Nishimura, J. Biol. Chem. 1979, 254, 3061.
[7] J. D. Kittendorf, L. M. Barcomb, S. T. Nonekowski, G. A. Garcia, Biochemistry 2001, 40, 14123.
[8] C. Romier, K. Reuter, D. Suck, R. Ficner, Biochemistry 1996, 35, 15734.
[9] B. Frey, J. McCloskey, W. Kersten, H. Kersten, J. Bacteriol. 1988, 170, 2078.
[10] R. K. Slany, H. Kersten, Biochimie 1994, 76, 1178.
[11] L. Curran, in −Modification and Editing of RNA×, Eds. H. Grosjean and B. Benne, American Society for
Microbiology Press, Washington, DC, 1998, p. 493.
[12] G. R. Bjˆrk, in −tRNA: Structure, Biosynthesis, and Function×, Eds. D. Sˆll and U. L. Rajbhandary,
American Society for Microbiology Press, Washington, DC, 1995, p. 165.
[13] G. Bjˆrk, in −Escherichia coli and Salmonella: Cellular and Molecular Biology×, Eds. F. C. Neidhardt, I. R.
Curtiss III, J. L. Ingraham, E. C. C. Lin, K. B. Low, B. Magasanik, W. S. Reznikoff, M. Riley, M. Schaechter,
H. E. Umbarger, American Society for Microbiology Press, Washington, DC, 1996, p. 861.
[14] J. M. Durand, N. Okada, T. Tobe, M. Watarai, I. Fukuda. T. Suzuki, N. Nakata, K. Komatsu, M. Yoshikawa,
C. Sasakawa, J. Bacteriol. 1994, 176, 4627.
[15] C. Romier, K. Reuter, D. Suck, R. Ficner, EMBO. J. 1996, 15, 2850.
[16] C. Romier, J. E. Meyer, D. Suck, FEBS Lett. 1997, 416, 93.
[17] R. Brenk, L. Naerum, U. Gr‰dler, H.-D. Gerber, G. A. Garcia, K. Reuter, M. T. Stubbs, G. Klebe, J. Med.
Chem. 2003, 46, 1133.
[18] E. A. Meyer, R. Brenk, R. K. Castellano, M. Furler, G. Klebe, F. Diederich, ChemBioChem 2002, 3, 250.
[19] J. K. Luo, R. N. Castle, J. Heterocycl. Chem. 1991, 28, 205.
[20] P. R. Gerber, K. M¸ller, J. Comput. Aided Mol. Des. 1995, 9, 251.
[21] −SYBYL Molecular Modeling Software×, 6.8 edn., Tripos, Inc., St. Louis, MO.
[22] D. D. Perrin −Dissociation Constants of Organic Bases in Aqueous Solution×, Butterworths, London, 1965.
[23] H. J. Bˆhm, G. Klebe, Angew. Chem., Int. Ed. 1996, 35, 2588.

1452
Helvetica Chimica Acta Vol. 86(2003)
[24] A. Davis, S. Teague, Angew. Chem., Int. Ed. 1999, 38, 736.
[25] H. Gohlke, G. Klebe, Angew. Chem., Int. Ed. 2002, 41, 2644.
[26] G. Klebe, T. Mietzner, J. Comput. Aided Mol. Des. 1994, 8, 583.
[27] F. D. Chattaway, W. S. Humphrey, J. Chem. Soc. 1929, 645.
[28] J. H. Mowat, J. H. Boothe, B. L. Hutchings, E. L. R. Stokstad, C. W. Waller, R. B. Angier, J. Semb, D. B.
Cosulich, Y. SubbaRow, J. Am. Chem. Soc. 1948, 70, 14.
[29] C. B. Storm, R. Shiman, S. Kaufman, J. Org. Chem. 1971, 36, 3925.
[30] C. W. Waller, M. J. Fahrenbach, J. H. Boothe, R. B. Angier, B. L. Hutchings, J. H. Mowat, P. J. F. Poletto, J.
Semb, J. Am. Chem. Soc. 1952, 74, 5405.
[31] G. Frˆhlich, P. Kotsonis, H. Traub, S. Taghavi-Moghadam, N. Al-Masoudi, H. Hofmann, H. Strobel, H.
Matter, W. Pfleiderer, H. H. Schmidt, J. Med. Chem. 1999, 42, 4108.
[32] −Selector×, Tripos Inc., St. Louis, MO 63144, 1996.
[33] M. Rarey, B. Kramer, T. Lengauer, G. Klebe, J. Mol. Biol. 1996, 261, 470.
[34] U. Gr‰dler, H. D. Gerber, D. M. Goodenough-Lashua, G. A. Garcia, R. Ficner, K. Reuter, M. T. Stubbs, G.
Klebe, J. Mol. Biol. 2001, 306, 455.
[35] J. Kittendorf, T. Sgraja, G. Klebe, K. Reuter, G. A. Garcia, J. Biol. Chem., submitted.
[36] H. Gohlke, M. Hendlich, G. Klebe, Persp. Drug Discov. Design 2000, 20, 115.
[37] H. Gohlke, M. Hendlich, G. Klebe, J. Mol. Biol. 2000, 295, 337.
[38] H. Biere, W. Seelen, Liebigs Ann. Chem. 1976, 1972.
[39] W. Szeja, Synthesis 1979, 822.
Received February 10, 2003