Novel dihydrofolate reductase inhibitors. Structure-based versus diversity-based library design and high-throughput synthesis and screening

Article abstract of DOI:10.1021/jm020495y

Novel 2,4-diaminopyrimidines bearing N,N-disubstituted aminomethyl residues at the 5-position were designed as dihydrofolate reductase (DHFR) inhibitors. These compounds were obtained by treatment of 1-[(2,4-diamino-5-pyrimidinyl)methyl]pyridinium bromide

Full text of DOI:10.1021/jm020495y

2304
J . Med. Chem. 2003, 46, 2304-2312
Novel Dih yd r ofola te Red u cta se In h ibitor s. Str u ctu r e-Ba sed ver su s
Diver sity-Ba sed Libr a r y Design a n d High -Th r ou gh p u t Syn th esis a n d Scr een in g
Pierre C. Wyss,*,† Paul Gerber, Peter G. Hartman, Christian Hubschwerlen,† Hans Locher,†
Hans-Peter Marty, and Martin Stahl
Preclinical Research, F. Hoffmann-La Roche Ltd., CH-4070 Basle, Switzerland
Received November 1, 2002
Novel 2,4-diaminopyrimidines bearing N,N-disubstituted aminomethyl residues at the 5-position
were designed as dihydrofolate reductase (DHFR) inhibitors. These compounds were obtained
by treatment of 1-[(2,4-diamino-5-pyrimidinyl)methyl]pyridinium bromide with secondary
amines in a polar solvent and in the presence of triethylamine at room temperature. The
procedure was found to be very efficient and suitable for application in high-throughput
synthesis. In addition, we found that high-throughput screening for enzymatic and in vitro
antibacterial activity could be performed on crude reaction mixtures, thus avoiding any
purification step. Over 1200 proprietary secondary amines were selected for high-throughput
synthesis, based on structural and diversity-related criteria, and the resulting products were
submitted to high-throughput screening. A greater number of hits, and significantly more active
compounds, were obtained through structure-based library design than through diversity-based
library design. Different classes of inhibitors of DHFR were identified in this way, including
compounds derived from di-, tri-, and tetracyclic amines. In general, these products showed
high activity against the enzymes derived from both TMP-sensitive and TMP-resistant
Streptococcus pneumoniae. Some compounds possessed appreciable selectivity for the bacterial
over the human enzyme, whereas other compounds were not at all selective. In most cases,
active enzyme inhibitors also displayed antibacterial activity.
In tr od u ction
RO-64-5781, showed inhibitory activity that was up to
190 000 times greater than that of TMP, allowing the
coverage of TMP-resistant strains, and selectivity was
at least as good as that of TMP.4 These new compounds
were highly active in vitro against all staphylococci,
including TMP-resistant strains, and showed good an-
tistaphylococcal activity in a mouse septicaemia model
after parenteral application. However, the high level of
binding to plasma proteins exhibited by these com-
pounds, coupled with a low solubility, has caused
problems in achieving their formulation at the dose
estimated to be required for clinical use. The high
lipophilicity and molecular weight of these compounds
may be responsible for their unfavorable physicochem-
ical properties. In an attempt to solve these problems,
we decided to design inhibitors of different structural
classes and reduced molecular size.
Information available from crystal structures of DHFR
and complexes with inhibitors has shown that the 2,4-
diaminopyrimidine moiety of TMP fits nearly ideally
into a narrow pocket in the active site of the enzyme.5,6
As all previous efforts aimed at replacing this moiety
have so far failed, we decided to retain the 2,4-diami-
nopyrimidine “needle” fragment7 and to focus our efforts
to structure modifications of the side chain at the
5-position of the 2,4-diaminopyrimidine moiety.
The spread of antibiotic resistance has reached alarm-
ing proportions in some species, and one of the most
worrying trends is the increasing incidence of multire-
sistant Staphylococcus aureus (MRSA) in hospitals and
multiresistant Streptococcus pneumoniae in the com-
munity.1,2 There is therefore an urgent need for effective
antibacterial agents to treat infections caused by these
organisms.
The enzyme dihydrofolate reductase (DHFR) has been
established clinically as a proven target for chemo-
therapy. The DHFR inhibitor trimethoprim (TMP,
Figure 1), often used in combination with sulfamethox-
azole (SMZ), was introduced primarily for the treatment
of community-acquired infections and urinary tract
infections, with emphasis on Gram-negative pathogens.
Despite some recent work on Pneumocystis carinii,3 the
enzyme remains an underexploited target in the anti-
bacterial field, and no optimization of inhibitors against
Gram-positive pathogens has been performed.
As part of a previous program, we designed and
synthesized a large number of 5-benzyl-2,4-pyrim-
idinediamines incorporating lipophilic substituents in
the 3′-position of the benzene nucleus with the objective
of seeking derivatives with antibacterial activity against
highly resistant Gram-positive organisms, particularly
against S. aureus. The best compounds identified, e.g.,
Ra tion a le
We hypothesized that 2,4-diaminopyrimidine deriva-
tives incorporating basic N-disubstituted aminomethyl
residues at the 5-position (1) might have better solubil-
ity properties than TMP itself. Moreover, we anticipated
that an appropriate selection of candidates from the
* To whom correspondence should be addressed. Address: Morpho-
chem AG, Schwarzwaldallee 215, CH-4058 Basle, Switzerland. Phone:
++41-61-695 21 39. Fax: ++41-61-695 21 22. E-mail: pierre.wyss@
morphochem.ch.
† Present address: Morphochem AG, Schwarzwaldallee 215, CH-
4058 Basle, Switzerland.
10.1021/jm020495y CCC: $25.00 © 2003 American Chemical Society
Published on Web 05/08/2003

Dihydrofolate Reductase Inhibitors
J ournal of Medicinal Chemistry, 2003, Vol. 46, No. 12 2305
F igu r e 1. Inhibitors of DHFR containing 2,4-diaminopyrimidine moieties.
Ta ble 1. Inhibition of TMP-Resistant DHFR from S. aureus
and S. pneumoniae and Antibacterial Activity of Compound 4
Compared to Those of TMP
IC50 (µM)
S1a Sp1b
19 34
>100 0.21
MIC (µg/mL)
compd
Sa 157c
Spnd
Sp 1/1e
TMP
4
>32
>32
2
0.5
>64
4
a
S1, TMP-resistant DHFR from Staphylococcus aureus 157/
b
4696. Sp1, TMP-resistant DHFR from Streptococcus pneumoniae
1/1. c Sa 157, Staphylococcus aureus 157/4696 (TMP-resistant).
d
Spn, Streptococcus pneumoniae ATCC 49619 (TMP-susceptible).
e Sp 1/1, Streptococcus pneumoniae 1/1 (TMP-resistant).
F igu r e 2. General structure of designed inhibitors (1) and
examples of analogues known before initiating this work
(2-4).
amines were initially obtained by treating 5-(bromo-
methyl)-2,4-pyrimidinediamine dihydrobromide (5) with
a number of primary, secondary, and cyclic amines, as
shown in Scheme 1. The starting material 5 was
prepared in four steps following essentially the proce-
dures described in the literature.10-13 We soon realized
that the insolubility of compound 5 in most usual
solvents, including DMF and DMSO, and its high
reactivity, often leading to complex mixtures of products,
would probably limit the potential of the method. In
contrast, the crystalline pyridinium salt 6, obtained on
treatment of compound 5 with pyridine in DMF, was
found to be a very effective starting material for
performing the nucleophilic substitutions described
above. In contrast to most quaternary salts, this com-
pound was found to be nonhygroscopic. Moreover, it is
stable, very soluble in polar solvents, including DMF,
DMSO, and methanol, and reactions with amines are
much cleaner than the corresponding reactions previ-
ously performed using bromide 5 as a starting material.
High -Th r ou gh p u t Syn th esis a n d Scr een in g. On
the basis of the interesting biological results obtained
for 4, we decided to synthesize a large number of diverse
compounds of type 1 in order to identify a large variety
of chemical classes exhibiting biological activity in the
first hit. For this purpose, over 9000 secondary amines
were extracted from the Roche proprietary library. The
large pool of available amines would allow modulation
of the pKa and clogP properties of the resulting products
and thereby influence penetration into bacteria and the
antibacterial properties of the compounds.
Only a limited number of pyrimidines of type 1
(Figure 2) have been described so far in the literature.
For instance, Godel et al. have synthesized compounds
derived from substituted and unsubstituted piperazines
(e.g., 2) and piperidines as antagonists of the dopamine
D4 receptor.8 The unsubstituted 1,2,3,4-tetrahydroiso-
quinoline 3 is mentioned in a patent of the Wellcome
Foundation on antibacterial pyrimidine compounds.9
However, this compound was found to be inactive in our
DHFR and in vitro antibacterial assays. In the first step,
we prepared a limited number of compounds of type 1
and, interestingly, one of the first compounds synthe-
sized in this series (4) showed an appreciable inhibitory
activity against TMP-resistant DHFR from Streptococ-
cus pneumoniae but less activity against S. aureus
(Table 1).
Resu lts a n d Discu ssion
Ch em istr y. The first representatives of this new
class of N-substituted 5-(aminomethyl)-2,4-pyrimidinedi-
Sch em e 1. General Strategy for the Synthesis of 2,4-Diaminopyrimidinesa
a
Conditions: (a) NaOEt, EtOH, 20 °C, 18 h (100%); (b) Ni-Al (1:1), HCO2H, 110 °C, 2 h (59%); (c) NaBH4, H2O, 50 °C, 4 h (86%); (d)
HBr, AcOH, 95 °C, 2 h (93%); (e) pyridine, DMF, 20 °C, 2 h (96%); (f) RR′NH (1 equiv), Et3N (5 equiv), DMF, 20 °C, 18 h.

2306 J ournal of Medicinal Chemistry, 2003, Vol. 46, No. 12
Ta ble 2. Inhibitors of DHFR Identified in Libraries
Wyss et al.
based on docking calculations, and the other one was
based on diversity analysis.
library
no.
inhibitors
n (%)
selection procedure
size
(a ) Str u ctu r e-Ba sed Com p ou n d Selection . The
DHFR enzymes from different species (human, S. au-
reus, and S. pneumoniae) are highly homologous. The
main difference at the center of the active site is the
exchange of the three hydrophobic amino acids that are
in contact with the 2,4-diaminopyrimidine ring of TMP
on the side facing away from the cofactor NADPH. A
crystal structure of S. pneumoniae DHFR is not avail-
able, but homology modeling indicates that the active
site shape and polarity close to the 2,4-diaminopyrimi-
dine binding site is closely related to that of S. aureus
DHFR. The sequence alignment of human and S. aureus
DHFR, as defined by the overlay of the available crystal
structures, is shown in Figure 3. The sequence of S.
pneumoniae DHFR has been manually aligned to these
two sequences, and this alignment was used to build a
homology model. Side chains of the framed sets of
residues are close to the 2,4-diaminopyrimidine moiety.
1
2
structure-based, top-score
structure-based, low-score or
no docking solution
252
269
54 (21)
4 (1)
3
4
diversity-based
501
370
17 (3)
90 (24)
analogues of hitsa
a
Hits identified in library no. 1.
first selection of approximately 500 candidates was
made according to the two selection criteria discussed
below in the library design section. A number of pre-
liminary experiments, including TLC and HPLC con-
trols, showed that it is possible to perform the reaction
efficiently on a 0.35-0.7 µM scale. It was initially
planned to purify the crude reaction mixtures by HPLC
before submitting the compounds for measuring the
inhibition of DHFR enzymes and in vitro activity
against Gram-positive bacteria. Fortunately enough, we
found that this time-consuming operation was not
required. A number of control experiments showed that
neither starting materials (amines and pyridinium salt
6), released pyridine, triethylamine, nor solvent (MeOH)
interfered in any way with the biological assays under
the chosen test conditions. Eighty different reactions
were run in parallel on a single standard 96-well plate.
Each plate included samples of TMP and epiroprim as
biological references, and one reference reaction known
to give an active compound was performed as well, as
an internal control of the reaction. A total of 1392
compounds, included in libraries nos. 1-4 (Table 2),
were prepared according to this method and evaluated
for DHFR inhibition and antibacterial activity.
All compounds obtained by high-throughput synthesis
were evaluated for inhibition of human DHFR and the
bacterial enzymes from TMP-sensitive S. aureus (ATCC
25923) and TMP-resistant S. pneumoniae (1/1), which
are denoted human, Sa, and Sp1, respectively, in the
tables. Active compounds were also tested against TMP-
resistant S1 DHFR from S. aureus and TMP-sensitive
Spn DHFR from S. pneumoniae. Screening for antibac-
terial activity was performed using the following four
bacterial strains: TMP-susceptible (ATCC 25923) and
TMP-highly resistant S. aureus (157/4696), TMP-
susceptible (ATCC 49619) and TMP-resistant S. pneu-
moniae (1/1). In addition, TMP-susceptible S. aureus
was tested in a medium containing 10 µg/mL thymidine
to test for activity that is not caused by antifolates. It
is well-known that the growth inhibiting effect of
antifolates is reversed in rich medium containing thy-
midine, since some bacteria are not dependent on de
novo synthesis of thymidine but can take it up from the
medium.14 Finally, some subsets of highly hydrophobic
compounds were tested in the presence of 10% human
serum in order to discard highly lipophilic compounds
that may act nonspecifically by damaging bacterial
membranes.
Several structures of complexes between S. aureus
DHFR and various Roche inhibitors have been resolved
at Roche. No significant induced fit effects can be
observed in these structures such that the receptor can
be assumed to be rigid to a good approximation. Because
of the rigidity of the receptor and the high similarity
between the S. aureus and S. pneumoniae enzymes, we
assumed that the use of a single-crystal structure of S.
aureus DHFR should be sufficient for structure-based
selection of compounds against both targets. We also
reasoned that the difference between resistant and
sensitive enzymes should not be relevant for docking
studies, since TMP-resistance is not coupled with a
change of binding mode of TMP in the enzymes and the
shape of the binding site remains unchanged. For
instance, the main determinant of TMP-resistance in
S. aureus DHFR is a single mutation in the active site
that leads to the loss of a hydrogen bond between ligand
and receptor.15 While this mutation does lead to weaker
binding of the 2,4-diaminopyrimidine moiety, it does not
alter its binding mode.
A list of 9448 secondary amines available in the Roche
collection were retrieved as virtual reagents. 2,4-Diami-
nopyrimidine products 1 were generated with an in-
house library enumeration program. Single 3D confor-
mations of this library were generated with the program
Corina16,17 in mol2-format.18 The protonation states of
all compounds were adjusted with a rule-based in-house
program. The nitrogen atom of the aminomethyl ring
substituent common to all inhibitors (Nb in compound
1, Figure 2) was not protonated, since this atom must
be situated in a lipophilic environment within the DHFR
binding site. For docking, the crystal structure of DHFR
from TMP-sensitive S. aureus complexed with the Roche
inhibitor RO-62-6091 (Figure 4) was chosen. The library
members were docked with the constraint of a fixed
position of the 2,4-diaminopyrimidine fragment taken
from the crystal structure (schematic view in Figure 4).
The docking program FlexX,19-22 version 1.7.0, was used
for the calculations. Default parameter settings of FlexX
were used except for a modified scoring function that
penalizes hydrogen bonds formed at solvent-accessible
parts of the protein surface.23 The calculation was
Libr a r y Design . We tried to guide the synthesis by
covering representative compounds of the whole library
as well as specifically selecting subsets where binding
affinity to DHFR could be expected to be high. There-
fore, two alternative methods were applied to select
compounds out of the virtual library of over 9000
theoretical compounds. One selection strategy was

Dihydrofolate Reductase Inhibitors
J ournal of Medicinal Chemistry, 2003, Vol. 46, No. 12 2307
F igu r e 3. Sequence alignment of DHFR enzymes from three species (hum ) human; Sa ) S. aureus; Spn ) S. pneumoniae).
Crystal structures have been determined for the human and S. aureus enzymes, whereas the S. pneumoniae structure was modeled.
The 16 C-terminal residues of each sequence have been omitted. Red lines denote helical regions, and yellow lines indicate â-strands
as secondary structures. The three sets of highlighted residues are those mentioned in the text whose side chains are situated
directly next to the 2,4-diaminopyrimidine fragment.
of library molecules was superimposed at the newly
formed C-N bond (bond a-b in compound 1, Figure 2),
and the amine substituent was rotated around this bond
to generate conformers with maximum volume and
H-bond-donor and H-bond-acceptor overlap. The list of
pairwise similarity scores was used to cluster the
compounds in a binary tree (complete linkage algo-
rithm). It was found that approximately 500 compounds
were needed to adequately represent the chemical space
spanned by the library. A set of 501 compounds were
synthesized (library no. 3).
Selection a n d Va lid a tion of Scr een in g Hits.
Primary selection of hits was based on activity against
isolated DHFR from S. aureus or S. pneumoniae at 10
µM (>50% inhibition). Only candidates showing in
addition antibacterial activity at 25 µM were selected.
Further selection criteria were thymidine antagonism
and activity in the presence of 10% human serum. This
allowed elimination of highly hydrophobic compounds
(e.g., membrane-active compounds) and those having
modes of action other than DHFR inhibition. Applica-
tion of these selection criteria to compounds obtained
by high-throughput chemistry, and summarized in
F igu r e 4. Structure of the Roche inhibitor whose complex
library nos. 1-3, led to the identification of a number
of hits, as shown in Table 2. Most of the hit compounds
identified and mentioned in Table 2 were resynthesized
by classical methods, purified, characterized, and sub-
mitted to antibacterial testing. Results validated in most
cases the preliminary biological data obtained by high-
throughput synthesis and screening of all libraries of
compounds mentioned above. They also confirmed the
higher activity of this new class of inhibitors for DHFR
from S. pneumoniae vs S. aureus, initially observed for
the first interesting compound identified in this series
(4). Important quantitative and qualitative differences
were also observed between the different libraries. Thus,
structure-based library no. 1 led to a significantly higher
hit rate (21% hit rate) than diversity-based library no.
3 (3% hit rate). Library no. 2, containing compounds
that had received low ranks in the docking calculation
or could not be docked at all, led to only few weakly
active compounds. It is important to note that the hits
from library no. 1 were not biased toward activity
against S. aureus DHFR but rather showed higher
activity against S. pneumoniae DHFR, although a
crystal structure of S. aureus DHFR was used for
docking. This confirms our experience that the power
of structure-based virtual screening generally lies in its
structure with S. aureus DHFR was used for docking calcula-
tions (top) and binding mode of the 2,4-diaminopyrimidine
fragment of the inhibitor that was held fixed for library
docking (bottom). Details of the binding mode of the entire
compound are shown in Figure 8.
performed within 2 h of CPU time on six processors of
a R10k SGI Origin. FlexX produced docking solutions
for about half of the library. For each of the docked
compounds, the solution with the lowest score value
(highest calculated binding energy) was selected. Com-
pounds were then sorted according to their scores, and
252 out of the 300 top-ranking candidates were synthe-
sized (library no. 1; see Table 2). In addition, 150
candidates were randomly selected from those com-
pounds for which no docking solution could be found.
This list was combined with those 150 compounds that
had received the lowest docking ranks. A majority of
these molecules (269 representatives) could be synthe-
sized (library no. 2).
(b) Diver sity-Ba sed Com p ou n d Selection . In an
alternative selection process, the virtual library of 9448
2,4-diaminopyrimidines was clustered according to chemi-
cal similarity. The similarity measure applied was an
in-house method based on single conformers. Each pair

2308 J ournal of Medicinal Chemistry, 2003, Vol. 46, No. 12
Wyss et al.
and compounds with long flexible moieties. The latter
are penalized by the scoring function to account for the
entropic cost of conformational restriction upon binding.
In contrast, FlexX assigns high ranks to 2,4-diaminopy-
rimidines with flat, rigid, policyclic nitrogen substitu-
ents, many of which are highly potent compounds. For
example, a methylated derivative of compound 13 is
rank 6 and compound 14 is rank 30 in the docking
calculation (see discussion in the next section and
Figure 7 below). A moderate correlation between the
experimentally determined IC50 values against S. pneu-
moniae DHFR and the FlexX score values could be
observed (Figure 5), indicating that a more stringent
selection of compounds based on docking scores may
have been possible, e.g., with a score cutoff of about -20.
The cutoff for library no. 1, corresponding to the selec-
tion of the top 300 compounds, was approximately -13
(note that only relative score values are meaningful; the
absolute values strongly depend on the details of the
calculation setup and the protein structure used).
Nonetheless, library no. 1 contained not only more hits
but also significantly more potent hits than library no.
3. The activity ranges of resynthesized hits from library
nos. 1 and 3 are displayed in Figure 6. Median values
of activity between library no. 1 (shaded box) and library
no. 3 differ by 2 orders of magnitude for the DHFRs from
sensitive and resistant S. pneumoniae and by almost 1
order of magnitude for S. aureus. Activity against DHFR
from TMP-resistant S. aureus was weak for all libraries.
This result further underlines the efficiency of the
structure-based virtual screening process.
F igu r e 5. Correlation between FlexX score (arbitrary units)
and the IC50 values of individual library hits against S.
pneumoniae DHFR.
The best inhibitors identified in library no. 1 (Table
2) are derived from amines belonging to three different
chemical classes, namely, bi-, tri-, and tetracyclic amines.
Therefore, an additional library (no. 4, with 370 com-
pounds) was constructed from cyclic amines belonging
to these three classes and available in our corporate
collection. As expected, compounds included in library
no. 4 showed high activity vs DHFR enzymes.
F igu r e 6. Box plots illustrating the activity distribution of
hits from library no. 1 (shaded boxes) and library no. 3. The
median is depicted as a small square. The area in the box
contains all data between the 25% and 75% quantiles. Outliers
beyond 1.5 times this interval are shown as open circles.
Str u ctu r e-Activity Rela tion sh ip s a n d Str u c-
tu r a l Ba sis for DHF R In h ibition . A number of
compounds belonging mostly to the bi-, tri-, and tetra-
cyclic amine series showed a potent inhibitory activity
mainly against TMP-sensitive and TMP-resistant DHFR
from S. pneumoniae but less activity against S. aureus.
Some interesting features of representatives of these
classes are discussed below (Table 3 and Figure 7).
ability to filter out undesirable compounds, i.e., in
excluding many inactive compounds rather than in
identifying specific active ones. Specifically, FlexX does
not find docking solutions for compounds with more
than one bulky substituent attached to the aminomethyl
nitrogen because of the steric constraints of the active
site. FlexX also discards particularly small compounds
F igu r e 7. Individual hits from library no. 1. Activity data are summarized in Table 3.

Dihydrofolate Reductase Inhibitors
J ournal of Medicinal Chemistry, 2003, Vol. 46, No. 12 2309
Ta ble 3. IC50 Values, Selectivity over Human DHFR, and Antibacterial Activity of Selected Compounds
IC50 (µM)
MIC (µg/mL)
Sp 1/1
selectivity (-fold)
human/Spn
compd
Sa
Spn
Sp1
human
Spn
TMP
3
4
7
8
rac-9
(S)-9
(R)-9
10
11
12
0.027
40
30
18
2
9.8
5.4
1.3
3.2
0.075
1.1
0.01
3
5.5
0.21
900
>100
30
0.67
1.7
1.4
0.52
1.2
100
18
12000
>90
3000
15
243
117
2
8
>64
32
4
0.5
0.5
0.5
0.25
1
0.125
2
32
0.044
0.007
0.012
0.021
0.0098
0.006
0.45
0.002
0.004
0.004
0.053
0.031
0.024
0.029
0.0028
0.056
0.55
0.031
0.047
0.01
0.5
4
0.5
2
0.125
8
32
4
4
8
25
122
16667
40
2600
45000
21
0.042
0.073
7.8
5.2
180
0.085
2
1
8
13
14
6.8
F igu r e 8. Binding mode of RO-62-6091 as determined by crystal structure analysis (top structures) complexed with S. aureus
DHFR and the modeled binding mode of (R)-9 (bottom structures) complexed with S. pneumoniae DHFR. The cofactor NADPH
is partially visible at the bottom of the binding sites.
(a ) Bicyclic Am in es. Most compounds in this series
are derived from 1,2,3,4-tetrahydroisoquinolines. It ap-
pears that the substitution pattern of the tetrahydroiso-
quinoline moiety plays an essential role for both potency
and selectivity of the inhibitors. For instance, the
unsubstituted derivative 3 is a very weak inhibitor of
the DHFR from wild-type (Spn) and TMP-resistant S.
pneumoniae (Sp1). Introduction of methoxy substituents
at C-6′ and C-7′ (4) showed 100- and 25-fold enhance-
ment of DHFR inhibitory activity from Spn and Sp1,
respectively. Further introduction of a methyl substitu-
ent at C-1′ in compound 4 leads to a small 2- to 4-fold
increase of activity (7). However, the presence of larger
alkyl (e.g., cyclopropyl, isopropyl, and isobutyl) and aryl
substituents at this position gives nearly inactive
compounds. In contrast, aryl substituents at C-4′ (e.g.,
2,4-dichlorophenyl) are tolerated (8). Tetrasubstituted
derivative rac-9 was identified as one of the most
interesting compounds in this series. Resolution into
enantiomerically pure compounds was achieved by
HPLC on a chiral phase. One enantiomer showed a 10-
fold higher activity against S. pneumoniae DHFRs. A
detailed analysis of potential binding modes of both
(R)-9 and (S)-9 was carried out by molecular modeling.
It became clear that only the (R) isomer could fit into
the DHFR active site in a manner not requiring either
a strained conformation of the molecule or a slight
movement of amino acid side chains in the active site.
In addition, the predicted binding mode of (R)-9 in the
active site (Figure 8) points the methyl group in position
1 into a lipophilic pocket that is not filled by other
known DHFR inhibitors. This pocket exists in all three

2310 J ournal of Medicinal Chemistry, 2003, Vol. 46, No. 12
Wyss et al.
ing into DHFR from Streptococcus aureus) gave a higher
hit rate and led to the identification of more potent
inhibitors than diversity-based compound selection, thus
allowing us to focus on the more promising candidates
in a time- and cost-efficient manner.
Exp er im en ta l Section
Gen er a l Meth od s. Melting points were determined on a
Buchi 510 apparatus and are uncorrected. Column chroma-
tography was accomplished on Kieselgel 60 (70-230 mesh)
obtained from E. Merck, Darmstadt. Kieselgel 60 F254 plates,
also from E. Merck, were used for thin-layer chromatography,
F igu r e 9. Crystal structure conformation of (S)-9.
DHFR isoforms. The identity of the more active enan-
tiomer with (R)-9 was confirmed by X-ray structure
analysis of enantiomerically pure 9 (Figure 9).
1
and compounds were visualized with UV light and iodine. H
nuclear magnetic resonance (NMR) spectroscopic data were
recorded on a Bruker DRX 400 instrument; δ values in parts
per million relative to tetramethylsilane are given. Elemental
analyses are indicated by the symbol of the elements; analyti-
cal results were within 0.4% of the theoretical values.
1-[(2,4-Diam in o-5-pyr im idin yl)m eth yl]pyr idin iu m Br o-
m id e Mon oh yd r obr om id e (6). A solution of 20.6 g (56.4
mmol) of 5-(bromomethyl)-2,4-pyrimidinediamine dihydrobro-
mide in 170 mL of DMF was treated with 18.2 mL (225 mmol)
of pyridine, and the mixture was stirred at 20 °C for 2 h,
whereupon crystallization occurred. The reaction mixture was
treated with 200 mL of dichloromethane, and the solids were
filtered, washed with dichloromethane, and dried to give 19.7
g (96%) of 6 as a colorless solid, mp 224-226 °C. 1H NMR
(DMSO-d6): δ 12.35-12.15 (1H, broad s), 9.07 (2H, d), 8.64
(1H, t), 8.8-8.3 (2H, broad s), 8.2-8.1 (3H, m), 8.1-7.7 (2H,
broad s), 5.69 (2H, s). MS m/z: 202.2 (M + H)+. Anal.
(C10H12N5Br‚HBr) C, H, N.
5-[(3,4-Dih yd r o-6,7-d im et h oxy-2(1H )-isoq u in olin yl)-
m eth yl]-2,4-p yr im id in ed ia m in e (4). A mixture of 5-(bro-
momethyl)-2,4-pyrimidinediamine dihydrobromide (5) (365 mg,
1 mmol) and 1,2,3,4-tetrahydro-6,7-dimethoxyisoquinoline hy-
drochloride (690 mg, 3 mmol) in DMF (5 mL) was vigorously
stirred with triethylamine (1.38 mL, 10 mmol) for 18 h. The
reaction mixture was diluted with water (20 mL), and the
solids were filtered, washed with water, and dried to provide
4 (245 mg, 78%) as colorless crystals, mp 235-240 °C (dec).
1H NMR (DMSO-d6): δ 7.57 (1H, s), 6.66 (1H, s), 6.63 (1H, s),
6.30 (2H, s), 5.83 (2H, s), 3.69 (3H, s), 3.67 (3H, s), 3.38 (2H,
s), 3.35 (2H, signal collapsed by H2O peak), 2.72 (2H, m), 2.61
(2H, m). MS m/z: 316.2 (M + H)+. Anal. (C16H21N5O2) C,
H, N.
Gen er a l P r oced u r e for th e P r ep a r a tion of Com p ou n d s
of Typ e 1. Reactions were performed on a 0.7 mM scale
(occasionally 0.3-0.5 mM) by treating pyridinium salt 6 with
1 equiv of primary or secondary amine in 5 mL of DMF in the
presence of 5 equiv of triethylamine. After the mixture was
stirred at 20 °C for 20 h, a solution was usually obtained.
Evaporation of the reaction mixture to dryness, followed by
column chromatography on silica gel using a mixture of
dichloromethane, methanol, ammonium hydroxide and/or
preparative HPLC gave pure compounds (>95%, HPLC). For
some amines, 1 precipitated during the reaction and was
obtained in pure form after filtration.
5-[(3,4-Dih yd r o-6,7-d im eth oxy-1-m eth yl-2(1H)-isoqu i-
n olin yl)m eth yl]-2,4-p yr im id in ed ia m in e (7). Reaction of
pyridinium salt 6 with 1,2,3,4-tetrahydro-6,7-dimethoxy-1-
methylisoquinoline by the general procedure gave 7 as color-
less crystals, mp 203-206 °C. 1H NMR (DMSO-d6): δ 7.55 (1H,
s), 6.66 (1H, s), 6.63 (1H, s), 6.36 (2H, s), 5.81 (2H, s), 3.71
(1H, m, collapsed by methoxy peak), 3.70 (3H, s), 3.69 (3H, s),
3.52 (1H, d), 3.45 (1H, d), 2.95-2.45 (4H, m, partly collapsed
by DMSO peak), 1.23 (3H, d). MS m/z: 330.4 (M + H)+.
5-[[4-(2,4-Dich lor op h en yl)-3,4-d ih yd r o-6,7-d im eth oxy-
2(1H)-isoqu in olin yl]m eth yl]-2,4-p yr im id in ed ia m in e (8).
Reaction of pyridinium salt 6 with 4-(2,4-dichlorophenyl)-
1,2,3,4-tetrahydro-6,7-dimethoxyisoquinoline by the general
procedure gave 8 as colorless crystals, mp 118-124 °C. 1H
NMR (DMSO-d6): δ 7.60 (1H, d), 7.51 (1H, s), 7.27 (1H, dd),
6.84 (1H, d), 6.78 (1H, s), 6.36 (1H, s), 5.98 (2H, s), 5.82 (2H,
The experimentally determined conformation of the
fused ring system is virtually identical to the predicted
model of the complex in Figure 8. X-ray structure
analysis of racemic 9 complexed with S. aureus DHFR
also supports this binding mode, although a high-
resolution complex structure could not be obtained. A
comparison of this binding mode with that of RO-62-
6091 (Figure 8) shows that compound (R)-9 tightly
occupies the main lipophilic binding cleft extending
along the plain of the 2,4-diaminopyrimidine ring
system, whereas RO-62-6091 gains much of its binding
energy through additional lipophilic interactions in the
cleft oriented perpendicularly to the main cavity. The
side chain of Met 50 in S. pneumoniae DHFR is oriented
such that it interacts closely with the phenyl ring of the
tetrahydroisoquinoline system. This interaction might
be a reason for the selective inhibition of S. pneumoniae
DHFR through the tetrahydroisoquinoline derivatives,
since the human and S. aureus enzymes have a less
polarizable isoleucin residue in this position. Still, most
representatives of the tetrahydroisoquinoline series
showed lower selectivity than TMP for bacteria over
human DHFR, with the exception of 10, which had
comparable selectivity (16700-fold) to that of TMP
(12000-fold). Finally, a single compound (11) showed
appreciable inhibitory activity against DHFR from S.
aureus (IC50 ) 0.042 µM).
(b) Tr icyclic Am in es. A number of compounds in
this series were found to be very potent inhibitors of
both wild-type and TMP-resistant DHFR from S. pneu-
moniae. In addition, one of these representatives (12)
also strongly inhibited the enzyme from TMP-sensitive
S. aureus. Finally, compound 13 showed higher selectiv-
ity for the enzyme from S. pneumoniae against human
source (human/Spn ) 45000) than TMP itself (12000-
fold).
(c) Tetr a cyclic Am in es. Most compounds of this
type (e.g., 14) are strong inhibitors of DHFR from S.
pneumoniae sources, but their selectivity against human
DHFR is usually low.
Con clu sion
A series of novel and potent inhibitors of DHFR from
wild-type and TMP-resistant Streptococcus pneumoniae
have been identified using high-throughput synthesis
and screening methodology. The procedure developed
for the synthesis of the novel 2,4-diaminopyrimidines
is simple, efficient, and rapid. In addition, screening for
enzymatic and in vitro antibacterial activity could be
performed on crude reaction mixtures, thus avoiding
any tedious and time-consuming purification. The li-
brary design method based on structural criteria (dock-

Dihydrofolate Reductase Inhibitors
J ournal of Medicinal Chemistry, 2003, Vol. 46, No. 12 2311
s), 4.50 (1H, t), 3.71 (3H, s), 3.72-3.65 (2H, q), 3.56 (3H, s),
3.35-3.20 (2H, m), 2.77 (2H, m). MS m/z: 460.4 (M + H)+.
8.05 (1H, d), 7.90 (1H, d), 7.60 (1H, s), 7.55 (1H, d), 7.40 (1H,
t), 7.15 (1H, t), 6.93 (1H, d), 6.38 (2H, s), 5.82 (2H, s), 4.11
(3H, s), 3.97 (1H, q), 3.58 (1H, d), 3.52-3.40 (2H, m), 3.34 (1H,
m), 3.05 (1H, m), 2.80 (1H, m), 1.37 (3H, d). MS m/z: 373.4
(M + H)+.
5-[(6,9-Dih yd r o-4-m eth oxy-9-m eth yl-1,3-d ioxolo[4,5-h ]-
isoqu in olin -8(7H)-yl)m eth yl]-2,4-p yr im id in ed ia m in e (9).
Reaction of pyridinium salt 6 with 6,7,8,9-tetrahydro-4-meth-
oxy-9-methyl-1,3-dioxolo[4,5-h]isoquinoline by the general pro-
Cr yst a l Da t a for 5-[[(9R)-6,9-Dih yd r o-4-m et h oxy-9-
m et h yl-1,3-d ioxolo[4,5-h ]isoq u in olin -8(7H )-yl]m et h yl]-
2,4-p yr im id in ed ia m in e. The compound was crystallized
from an ethanol/water mixture. Crystals with dimensions
0.5 × 0.5 × 0.4 mm3 were in orthorhombic space group
P2(1)2(1)2(1) with a ) 7.2468(14) Å, b ) 14.658(3) Å, and c )
16.629(3) Å at 293 K. The 7073 reflections were measured on
a SIEMENS P3 diffractometer with graphite-monochromated
Cu KR radiation. The 2385 (Rint ) 0.0692) unique reflections
were used in all calculations, and the structure was resolved
by direct methods and was expanded using Fourier techniques
and the program SHELX-97.24 Non-hydrogen atoms were
refined anisotropically using full-matrix least-squares refine-
ment without restraints. The R factor and Rw factor for all
data were refined to 0.0396 and 0.1032, respectively. The
structure has been deposited at the Cambridge Crystal-
lographic Data Centre, deposition number CCDC 196327.
1
cedure gave (R,S)-9 as colorless crystals, mp 191-192 °C. H
NMR (DMSO-d6): δ 7.55 (1H, s), 6.39 (1H, s), 6.33 (2H, s),
5.98 (1H, s), 5.89 (1H, s), 5.80 (2H, s), 3.81-3.72 (1H, m,
collapsed by methoxy peak), 3.77 (3H, s), 3.46-3.35 (2H, AB),
2.87-2.45 (4H, m, partly collapsed by DMSO peak), 1.19 (3H,
d). MS m/z: 344.3 (M + H)+. Anal. (C17H21N5O3) C, H, N. A
530 mg sample of (R,S)-9 was separated by chiral HPLC
(Chiracel-OD CSP 20 µm, 50 mm i.d. × 250 mm, heptane-
2-propanol (1%), flow rate 60 mL/min) to yield 165 mg (S)-9
and 135 mg (R)-9 as colorless solids.
5-[[(9S)-6,9-Dih yd r o-4-m et h oxy-9-m et h yl-1,3-d ioxolo-
[4,5-h ]isoq u in olin -8(7H )-yl]m et h yl]-2,4-p yr im id in ed ia -
1
m in e: mp 205-210 °C. H NMR (DMSO-d6): δ 7.55 (1H, s),
6.39 (1H, s), 6.31 (2H, s), 5.98 (1H, s), 5.89 (1H, s), 5.80 (2H,
s), 3.81-3.72 (1H, m, collapsed by methoxy peak), 3.77 (3H,
s), 3.46-3.31 (2H, AB), 2.87-2.45 (4H, m, partly collapsed by
DMSO peak), 1.19 (3H, d). MS m/z: 344.4 (M + H)+. [R]D +2.8°
(c 1, CHCl3).
Ackn owledgm en t. The authors thank Andrea Arau-
jo Del Rosario, Heidi Schlunegger, Marie-Claude Vigliano,
Re´my Halm, Fabian Rufi, and David Wechsler for
excellent technical assistance, Michael Hennig and
Christian Oefner for X-ray structure analyses, and Paul
Hadva´ry for his encouragement and support.
5-[[(9R)-6,9-Dih yd r o-4-m eth oxy-9-m eth yl-1,3-d ioxolo-
[4,5-h ]isoq u in olin -8(7H )-yl]m et h yl]-2,4-p yr im id in ed ia -
m in e: 1H NMR (DMSO-d6): δ 7.55 (1H, s), 6.39 (1H, s), 6.31
(2H, s), 5.98 (1H, s), 5.89 (1H, s), 5.80 (2H, s), 3.81-3.75 (1H,
m, collapsed by methoxy peak), 3.77 (3H, s), 3.46-3.31 (2H,
AB), 2.87-2.45 (4H, m, partly collapsed by DMSO peak), 1.19
(3H, d). MS m/z: 344.3 (M + H)+. [R]D -1.9° (c 1, CHCl3).
5-[[3,4-Dih yd r o-7-m eth oxy-4-(3-n itr op h en yl)-2(1H)-iso-
qu in olin yl]m eth yl]-2,4-p yr im id in ed ia m in e (10). Reaction
of pyridinium salt 6 with 1,2,3,4-tetrahydro-7-methoxy-4-(3-
nitrophenyl)isoquinoline by the general procedure gave 10 as
yellowish crystals, mp 115 °C. 1H NMR (DMSO-d6): δ 8.06
(1H, d), 7.98 (1H, s), 7.61 (1H, d), 7.56 (1H, d), 7.53 (1H, s),
6.76 (2H, m), 6.70 (1H, dd), 5.98 (2H, s), 5.79 (2H, s), 4.37 (1H,
t), 3.75-3.71 (1H, d, partly collapsed by methoxy peak), 3.71
(3H, s), 3.46 (1H, d), 3.33-3.27 (2H, AB), 2.90-2.86 (1H, dd),
2.76-2.72 (1H, dd). MS m/z: 407.4 (M + H)+.
5-[[3,4-Dih yd r o-1-m et h yl-6-p h en yl-7-(p h en ylm et h yl)-
pyr r olo[1,2-a ]pyr a zin -2(1H)-yl]m eth yl]-2,4-p yr im idin ed i-
a m in e (11). Reaction of pyridinium salt 6 with 1,2,3,4-
tetrahydro-1-methyl-6-phenyl-7-(phenylmethyl)pyrrolo[1,2-
a]pyrazine by the general procedure gave 11 as a brownish
solid, mp 98-103 °C. 1H NMR (DMSO-d6): δ 7.56 (1H, s), 7.43
(2H, m), 7.31 (3H, m), 7.24 (2H, m), 7.12 (3H, m), 6.25 (2H, s),
5.82 (2H, s), 5.75 (1H, s), 3.81 (1H, d), 3.75-3.60 (5H, m), 3.07
(1H, d), 2.95-2.87,1H, m), 2.55-2.45 (1H, m, collapsed by
DMSO peak), 1.40 (3H, d). MS m/z: 425.5 (M + H)+.
5-[[7-Ch lor o-3,4-d ih yd r o-10-(m eth ylth io)p yr a zin o[1,2-
a ]in d ol-2(1H)-yl]m eth yl]-2,4-p yr im id in ed ia m in e (12). Re-
action of pyridinium salt 6 with 7-chloro-1,2,3,4-tetrahydro-
10-(methylthio)pyrazino[1,2-a]indole by the general procedure
gave 12 as a colorless solid, mp 210-218 °C. 1H NMR (DMSO-
d6): δ 7.62 (1H, s), 7.55 (2H, m), 7.14 (1H, d), 6.25 (2H, s),
5.86 (2H, s), 4.07 (2H, t), 3.79 (2H, s), 3.47 (2H, s), 2.91 (2H,
t), 2.18 (3H, s). MS m/z: 375.4 (M + H)+.
Refer en ces
(1) Houvinen, P.; Sundstro¨m, L.; Swedberg, G.; Sko¨ld, O. Trimeth-
oprim and sulfonamide resistance. Antimicrob. Agents Chemo-
ther. 1995, 39, 279-289.
(2) Then, R. L.; Kohl, I.; Burdeska, A. Frequency and transferability
of trimethoprim and sulfonamide resistance in methicillin-
resistant Staphylococcus aureus and Staphylococcus epidermi-
dis. J . Chemother. 1992, 4, 67-71.
(3) Durand, R.; Savel, J . Dihydrofolate reductase inhibitors: New
developments in antiparasitic chemotherapy. Expert Opin. Ther.
Pat. 2001, 11, 1285-1290.
(4) Wyss, P. C.; Guerry, P.; Hartman, P. G.; Hubschwerlen, C.;
J olidon, S.; Locher, H.; Specklin, J .-L.; Stalder, H. Anti-MRSA
Dihydrofolate Reductase Inhibitors: Synthesis and SAR. Pre-
sented at the 39th Interscience Conference on Antimicrobial
Agents and Chemotherapy, San Francisco, CA, September 26-
29, 1999; Abstract No. 1800.
(5) Matthews, D. A.; Bolin, J . T.; Burridge, J . M.; Filman, D. J .;
Volz, K. W.; Kaufman, B. T.; Beddell, C. R.; Champness, J . N.;
Stammers, D. K.; Kraut, J . Refined crystal structures of Es-
cherichia coli and chicken liver dihydrofolate reductase contain-
ing bound trimethoprim. J . Biol. Chem. 1985, 260, 381-391.
(6) Dauber-Osguthorpe, P.; Roberts, V. A.; Osguthorpe, D. J .; Wolff,
J .; Genest, M.; Hagler, A. T. Structure and energetics of ligand
binding to proteins: Escherichia coli dihydrofolate reductase-
trimethoprim, a drug-receptor system. Proteins 1988, 4, 31-
47.
(7) Boehm, H.-J .; Boehringer, M.; Bur, D.; Gmuender, H.; Huber,
W.; Klaus, W.; Kostrewa, D.; Kuehne, H.; Luebbers, T.; Meunier-
Keller, N.; Mueller, F. Novel inhibitors of DNA gyrase: 3D
structure based biased needle screening, hit validation by
biophysical methods, and 3D guided optimization. A promising
alternative to random screening. J . Med. Chem. 2000, 43, 2664-
2674.
5-[(1,3,4,9-Tetr a h yd r o-2H-p yr id o[3,4-b]in d ol-2-yl)m eth -
yl]-2,4-p yr im id in ed ia m in e (13). Reaction of pyridinium salt
6 with 2,3,4,9-tetrahydro-1H-pyrido[3,4-b]indole by the general
procedure gave 13 as a colorless solid, mp 205-210 °C. 1H
NMR (DMSO-d6): δ 10.7 (1H, s), 7.59 (1H, s), 7.35 (1H, d),
7.25 (1H, d), 7.05-6.90 (2H, m), 6.35 (2H, s), 5.84 (2H, s), 3.53
(2H, s), 3.47 (2H, s), 2.76 (2H, s), 2.69 (2H, s). MS m/z: 295.4
(M + H)+.
(8) Godel, T.; Riemer, C.; Edenhofer, A. Preparation of pyrimidines
for treatment and/or prevention of illnesses caused by disorders
of the dopamine system. Patent WO 9713759, 1997.
(9) Daluge, S. M.; Skonezny, P. M.; Roth, B.; Rauckman, B. S.
Antibacterial pyrimidine compounds. Patent EP 96214, 1983.
(10) Huber, W. 2,4-Diamino-5-(4-methyl-5-â-hydroxyethylthiazolium
chloride)methylpyrimidine hydrochloride, a new analog of thia-
min. J . Am. Chem. Soc. 1943, 65, 2222-2226.
(11) Tieckelmann, H.; Guthrie, R.; Nain, J . G. 2,4-Diamino-5-
formylpyrimidine and 2,4-diamino-5-hydroxymethylpyrimidine.
J . Org. Chem. 1960, 25, 1257-1259.
(12) Weinstock, L. T.; O’Brien, D. E.; Cheng, C. C. Folic acid analogs.
I. p{[(2,4-Diamino-5-pyrimidinyl)methyl]amino}benzoyl-L-glutam-
ic acid and related compounds. J . Med. Chem. 1968, 11, 1238-
1241.
5-[(1,2,4,11-Tetr a h yd r o-4,11-d im eth yl-3H-p yr id o[4,3-a ]-
ca r ba zol-3-yl)m eth yl]-2,4-p yr im id in ed ia m in e (14). Reac-
tion of pyridinium salt 6 with 2,3,4,11-tetrahydro-4,11-dimethyl-
1H-pyrido[4,3-a]carbazole by the general procedure gave 14
1
as a colorless solid, mp 202-208 °C. H NMR (DMSO-d6): δ

2312 J ournal of Medicinal Chemistry, 2003, Vol. 46, No. 12
Wyss et al.
(13) Stuart, A.; Paterson, T.; Roth, B.; Aig, E. 2,4-Diamino-5-
benzylpyrimidines and analogues as antibacterial agents. 6. A
one-step synthesis of new trimethoprim derivatives and activity
analysis by molecular modeling. J . Med. Chem. 1983, 26, 667-
673.
(14) Methods for Dilution Antimicrobial Susceptibility Tests for
Bacteria That Grow Aerobically, 3rd ed.; National Committee
for Clinical Laboratory Standards: Villanova, Pa, 1993; Vol. 13,
No. 25, Article M7-A3.
(15) Dale, G. E.; Broger, C.; D’Arcy, A.; Hartmann, P. G.; DeHoogt,
R.; J olidon, S.; Kompis, I.; Labhardt, A. M.; Langen, H.; Locher,
H.; Page, M. G. P.; Stu¨ber, D.; Then, R. L.; Wipf, B.; Oefner, C.
A single amino acid substitution in staphylococcus aureus
dihydrofolate reductase determines trimethoprim resistance. J .
Mol. Biol. 1997, 266, 23-30.
(16) Sadowski, J .; Rudolph, C.; Gasteiger, J . Automatic generation
of 3D atomic coordinates for organic molecules. Tetrahedron
Comput. Methodol. 1990, 3, 537-547.
(19) Rarey, M.; Kramer, B.; Lengauer, T.; Klebe, G. A fast flexible
docking method using an incremental construction algorithm.
J . Mol. Biol. 1996, 261, 470-489.
(20) Rarey, M.; Wefing, S.; Lengauer, T. Placement of medium-sized
molecular fragments into active sites of proteins. J . Comput.-
Aided Mol. Des. 1996, 10, 41-54.
(21) Rarey, M.; Kramer, B.; Lengauer, T. Multiple automatic base
selection: Protein-ligand docking based on incremental con-
struction without manual intervention. J . Comput.-Aided Mol.
Des. 1997, 11, 369-384.
(22) Rarey, M.; Kramer, B.; Lengauer, T. Docking of hydrophobic
ligands with interaction-based matching algorithms. Bioinfor-
matics 1999, 15, 243-250.
(23) Stahl, M. Modifications of the scoring function in FlexX for
virtual screening applications. Perspect. Drug Discovery Des.
2000, 20, 83-98.
(24) Sheldrick, G. M. Phase annealing in SHELX-90: direct methods
for larger structures. Acta Crystallogr. 1990, A46, 467-473.
(17) Sadowski, J .; Schwab, C. H.; Gasteiger, J . Corina; Molecular
Networks GmbH Computerchemie: Erlangen, Germany, 1998.
(18) Clark, M.; Cramer, R. D., III; Van Opdenbosch, N. Validation
of the general purpose Tripos 5.2 force field. J . Comput. Chem.
1989, 10, 982-1012.
J M020495Y