Quantitative structure-activity relationships of 2,4-diamino-5-(2-X- benzyl)pyrimidines versus bacterial and avian dihydrofolate reductase

Article abstract of DOI:10.1021/jm970776j

Quantitative structure-activity relationships (QSAR) have been formulated for a set of 15 2,4-diamino-5-(2-X-benzyl)pyrimidines versus dihydrofolate reductase from Lactobacillus casei and chicken liver. QSARs were also developed for comprehensive data sets containing mono-, di-, and trisubstituted benzyl derivatives. Particular emphasis was placed on the role played by ortho substituents in the overall binding process and subsequent inhibition of the catalytic process in both the prokaryotic and eucaryotic DHFRs. Comparisons between the two QSARs reveal subtle differences at specific positions which can be optimized to design more selective antibacterial agents.

Full text of DOI:10.1021/jm970776j

J . Med. Chem. 1998, 41, 4261-4272
4261
Qu a n tita tive Str u ctu r e-Activity Rela tion sh ip s of
2,4-Dia m in o-5-(2-X-ben zyl)p yr im id in es ver su s Ba cter ia l a n d Avia n
Dih yd r ofola te Red u cta se†
Cynthia Dias Selassie,* Wei-Xi Gan, Lara S. Kallander, and Teri E. Klein‡
Department of Chemistry, Pomona College, Claremont, California 91711, and Department of Pharmaceutical Chemistry,
University of California, San Francisco, California 94143
Received November 14, 1997
Quantitative structure-activity relationships (QSAR) have been formulated for a set of 15
2,4-diamino-5-(2-X-benzyl)pyrimidines versus dihydrofolate reductase from Lactobacillus casei
and chicken liver. QSARs were also developed for comprehensive data sets containing mono-,
di-, and trisubstituted benzyl derivatives. Particular emphasis was placed on the role played
by ortho substituents in the overall binding process and subsequent inhibition of the catalytic
process in both the prokaryotic and eucaryotic DHFRs. Comparisons between the two QSARs
reveal subtle differences at specific positions which can be optimized to design more selective
antibacterial agents.
Dihydrofolate reductase (DHFR) catalyzes the reduc-
tion of 7,8-dihydrofolate to 5,6,7,8-tetrahydrofolate using
NADPH as a cofactor (Scheme 1). Since it wields such
Earlier studies from our laboratory have focused on
the inhibition of chicken liver DHFR and Lactobacillus
casei DHFR by meta- and para-substituted 2,4-diami-
nobenzylpyrimidines. Our previous quantitative struc-
ture-activity relationship (QSAR) for benzylpyrim-
idines acting on L. casei DHFR is described in eq 1:5
Sch em e 1. Reduction of 7,8-Dihydrofolate to
5,6,7,8-Tetrahydrofolate
log(1/Ki) ) 1.24((0.21)MR4′ + 0.52((0.27)MR3′ +
0.42((0.45)MR5 - 0.13((0.26)MR2 +
0.46((0.21)π4 - 0.92((0.31) log(â4‚10π4 + 1) +
0.31(( 0.23)π3′ - 0.71((0.36) log(â3‚10π3′ + 1) +
5.45((0.17) (1)
an important role in the DNA synthetic pathway, its
prominence as a target in antibacterial, antiparasitic,
and antineoplastic chemotherapy is unsurpassed.1 In
recent years, enhanced structural characterization of
various DHFRs from prokaryotic, eucaryotic, and para-
sitic sources has accelerated the study of various
ligand-enzyme interactions as well as the design of
selective inhibitors. Relative high homology (75-90%)
is found between various vertebrate enzymes unlike
bacterial species where homology is around 25-40%.
Although the vertebral DHFR enzymes are generally
longer (around 25 residues), those insertions are located
at various points throughout the sequence and are not
clustered in one area. These residues also occur on the
protein surface, where they do not appear to affect the
overall geometry of the enzyme. Thus, the three-
dimensional structures of the vertebrate and bacterial
enzymes are similar despite the low degree of homology
between them.2 Previous efforts from this laboratory
have carefully examined the molecular interactions
between numerous antifolates and various DHFRs.3,4
In a continuation of our study to delineate the inter-
molecular forces at play in the binding of various
inhibitors to DHFR, we have carefully and systemati-
cally examined the effects of ortho substitution on the
inhibitory potencies of 2,4-diamino-5-(X-benzyl)pyrim-
idines, I.
n ) 65 r ) 0.894 s ) 0.245 F3,54 ) 6.14
π0 ) 0.49 log â4 ) -0.501 log â3 ) -1.431
4
MR0 ) 1.66 π0 ) 1.33
5
3
The subscripts 3, 4, and 5 refer to substituents in the
meta, para, and other meta positions, respectively.
Molecular graphics analysis in tandem with QSAR
analyses has established that with certain enzymes,
such as DHFR, porcine pancreatic elastase, and papain,
the more hydrophobic meta substituent will tend to bind
in hydrophobic space placing its counterpart in hydro-
philic or aqueous space.6 Using selectively 13C-enriched
trimethoprim, Cheung et. al. have also elegantly il-
lustrated the phenomenon of “ring flipping” via two-site
exchange between nuclei at the C3′ and C5′ positions.7
Molar refractivity (MR) is often used to quantify steric
effects in QSAR studies. It is defined by the Lorentz-
Lorenz equation: MR ) (n2 - 1)/(n2 + 2)(MW/δ). In
this expression, n represents the refractive index, MW
is the molecular weight, and δ is the density of the
compound. It is thus a crude means of characterizing
the bulk and polarizability of a substituent/compound.
Although it contains no information on shape, it has
found considerable usage in biological QSAR where
†Dedicated to Professor Corwin Hansch on the occasion of his 80th
birthday.
‡ University of California.
10.1021/jm970776j CCC: $15.00 © 1998 American Chemical Society
Published on Web 10/02/1998

4262 J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 22
Selassie et al.
intermolecular effects predominate. MR is usually
scaled by 0.1 to make it equiscalar with π. The prime
with MR indicates that the MR term is truncated at a
value of 0.79 (the MR value for the methoxy substitu-
ent), regardless of the size of the substituent. The occur-
rence of this feature in the model indicates that beyond
a fixed point, the substituents do not effectively contact
the enzyme. The positive coefficients of the MR terms
suggest that positive steric and/or dispersion effects are
operative, thus enhancing the value of log(1/Ki).
In this model, the steric terms as represented by MR
account for approximately 50% of the variance in the
data while the other 30% is explained by hydrophobicity.
The bilinear model of Kubinyi is used to model the hy-
drophobic characteristics of the substituents.8 The coef-
ficients with both π3 and π4 suggest that partial desol-
vation occurs at the binding site. Despite the presence
of the hydrophobic terms in eq 1, most of the variance
in the data is explained by the molar refraction terms
which suggests that polar/steric interactions are of pri-
mary importance in the bacterial enzyme. This result
is analogous to what has been observed in the case of
DHFR from Escherichia coli.9 In the 3 and 4 position
where any substituent larger than three or two carbon
atoms, respectively, encounters interference from the
active site residues, the activity falls off dramatically.
Hydrophilic substituents in the 5 position have more
polar space to maneuver in, around an area that in-
corporates portions of the sugar moiety of the NADPH
cofactor.
sites on the bacterial and avian enzymes. In the present
study, 15 new ortho, monosubstituted derivatives of I
were synthesized, and their inhibitory potencies versus
avian and bacterial DHFRs at pH 7.4 were assessed.
In addition, five multisubstituted analogues containing
substituents in the ortho position were also tested
versus these two diverse enzymes.
Resu lts
The following mathematical equations were developed
in a stepwise fashion for the inhibition of L. casei DHFR
at pH 7.20 by ortho-, meta-, and para-substituted 2,4-
diamino-5-(X-benzyl)pyrimidines:
log(1/Ki) ) 6.68((0.20) - 0.52((0.11)B56 (3)
n ) 88 r ) 0.700 s ) 0.545 F1,86 ) 82.47 q2 ) 0.464
log(1/Ki) ) 6.12((0.24) - 0.42((0.10)B56 +
1.00((0.30)MR′4 (4)
n ) 88 r ) 0.812 s ) 0.448 F1,85 ) 42.54 q2 ) 0.636
log(1/Ki) ) 5.72((0.26) - 0.34((0.09)B56 +
1.14((0.27)MR′4 + 0.67((0.28)MR′3 (5)
Equation 2 was based on the inhibition of chicken
liver DHFR by meta- and para-substituted I.5 Chicken
liver DHFR, because of its strong homology with mouse
and human DHFR, has been used as a surrogate for
the human enzyme.10,11
n ) 88 r ) 0.855 s ) 0.400 F1,84 ) 22.36 q2 ) 0.706
log(1/Ki) ) 5.57((0.26) - 0.31((0.09)B56 +
1.11((0.26)MR′4 + 0.65((0.27)MR′3 +
0.30((0.16)MR5 (6)
log(1/K ) ) 0.39π + 0.44π + 0.37π + 0.44 σ -
∑
i
3
4
5
0.75MR5 - 1.04 log(â3‚10π3 + 1) -
0.32 log(â4‚10π4 + 1) + 4.70 (2)
n ) 65 r ) 0.906 s ) 0.207
log â3 ) -2.69 log â4 ) -0.18 π0 ) 2.45 π0 = 3.00
n ) 88 r ) 0.876 s ) 0.375 F1,83 ) 12.88 q2 ) 0.742
3
4
log(1/Ki) ) 5.59((0.24) - 0.32((0.08)B56 +
1.26((0.24)MR′4 + 0.65((0.24)MR′3 +
0.28((0.15)MR5 + 0.20((0.18)π4 -
The σ (Hammett) term, albeit weak, was perhaps
indicative of dipolar interactions between the electron-
rich active site of the enzyme and the electron-deficient
phenyl ring of the antifolate. The coefficients with π3,
π4, and π5 suggested that surface binding of these
substituents was operative. The steep descending slope
with π3 (-0.65) indicated some obstruction to substitu-
ent binding; Val 115 appeared to be the culprit. The
flat descending slope with π4 (-0.10) suggested that
hydrophobic substituents larger than a hexyloxy chain
in the para position projected beyond the surface into
aqueous space. The relatively high negative coefficient
with the MR5 term revealed a constraint to binding in
5-space by large substituents in this area. Tyr 31 with
its ability to undergo a conformational change and
rotate 180° in the cleft was viewed as the source of this
unfavorable interaction.
0.84((0.39) log(â4‚10π4 + 1) (7)
n ) 88 r ) 0.907 s ) 0.332 F3,80 ) 8.47
q2 ) 0.800 π4,0 ) 0.82((0.72) log â4 ) -1.338
log(1/Ki) ) 5.61((0.23) - 0.32((0.08)B56 +
1.30((0.24)MR′4 + 0.62((0.28)MR′3 +
0.26((0.15)MR5 + 0.21((0.18)π4 -
0.79((0.37) log(â4‚10π4 + 1) + 0.34((0.30)π3 -
0.65((0.44) log(â3‚10π3 + 1) (8)
n ) 88 r ) 0.918 s ) 0.319 F3,77 ) 3.31
The two models, although complex in nature, clearly
delineated the interactions of the various substituents
in the meta and para positions with complementary
q2 ) 0.816 π4,0 ) 0.71((0.92) log â4 ) -1.159
π3,0 ) 1.10((0.76) log â3 ) -1.057

QSAR of 2,4-Diamino-5-(2-X-benzyl)pyrimidines
J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 22 4263
binding. Equation 3 suggests that the presence of any
substituent in the ortho position that is not small is
detrimental to overall binding in the bacterial enzyme.
Overall hydrophobicity of the substituents in the com-
bined data set only accounts for 10% of the variance in
the data. These results once again establish the im-
portance of polarity and/or steric effects in the interac-
tions of this class of antifolates with the bacterial
enzyme. Equations 3-9 are all significant at the 95%
level. Two points were not particularly well-predicted
in the final analysis; they were the 4-OH and 2-OH
derivatives which were underpredicted and overpre-
dicted, respectively. Excluding them from the analysis
resulted in the formulation of a slightly more rigorous
equation with the following statistics: n ) 86, r ) 0.970,
q2 ) 0.853, s ) 0.272.
The low coefficients with the hydrophobic terms in
eq 9 suggest that partial desolvation of the substituents
occurs at the enzyme surface. In the 3, 4, and 6
positions the optimum hydrophobicities (1.09, 0.78, and
0.66, respectively) are also relatively low which indicates
that added hydrophobicity beyond those cutoff values
results in a pronounced drop in inhibitory activity. This
is particularly apparent in the para position where the
descending slope of the bilinear equation is substantial
(0.78 - 0.17 ) 0.61); with moderately hydrophobic
substituents, the negative impact of hydrophobicity is
offset by the large positive contribution (+1.28) of the
steric term as represented by MR′4.
The two points that were not well-predicted in the
analysis, 4-OH and 2-OH, were off by more than two
standard deviations. The 2-OH derivative was 4 times
less active than predicted, while the 4-OH was under-
predicted by a factor of 7. The anomalous behavior of
the 2-OH analogue cannot be attributed to lack of
hydrophobicity. It has been suggested that the phenolic
hydroxy group may alter the positioning of the cofactor
and subsequently affect catalysis.13 Its placement in
6-space would place it in close proximity to the ribose
moiety of the cofactor. It is of interest to note that the
3-OH derivative is well-fit and falls within the standard
deviation of the equation.
log(1/Ki) )
5.87((0.30) - 0.27((0.12)B56 + 1.28((0.22)MR′4 +
0.64((0.27)MR′3 + 0.27((0.14)MR5 +
0.17((0.16)π4 - 0.78((0.36) log(â4‚10π4 + 1) +
0.34((0.29)π3 - 0.65((0.42) log(â3‚10π3 + 1) +
0.84((0.62)π6 - 1.02((0.95) log(â6‚10π6 + 1) (9)
n ) 88 r ) 0.929 s ) 0.304 F3,74 ) 3.59
q2 ) 0.821 π4,0 ) 0.78((0.96) log â4 ) -1.354
π3,0 ) 1.09((0.76) log â3 ) -1.037
π6,0 ) 0.66((0.85) log â6 ) 0.032
In these equations, n represents the number of data
points, r is the correlation coefficient, q2 is the cross-
validated r2, and s is the standard deviation from the
regression. F represents the F statistic for the signifi-
cance of each added variable. A sterimol parameter, B5,
is used for the first time with these compounds.12 B5
is defined as a maximum-width parameter. It repre-
sents a rough approximation since in the case of
unsymmetrical substituents, B5 can be defined in a few
different ways. As in our previous work, the more
hydrophobic meta substituent is placed in the 3 position,
while its more hydrophilic meta counterpart is located
in the 5 position. In the case of the ortho-substituted
I, these were mostly classified as being in the 6 position
where binding was favored over the 2 position as
delineated by a more favorable r and substantiated by
molecular graphics analysis. Exceptions to this rule
were the 2,3-Cl2 and 2,3-(OCH3)3 analogues. Placing
them in 6,5-space led to a slightly poorer overall
correlation for eq 9 (r ) 0.920, s ) 0.323). More than
that, they were outliers whose inhibitory activity was
off by 0.69 (2,3-(OCH3)2) and 0.67 (2,3-Cl2). Interest-
ingly, the activity of the former analogue (5.92) was very
similar to that of the 3-OCH3 (5.88), while the activity
of the latter analogue (6.37) was around that of the 3-Cl
analogue (5.90). This behavior suggested that the slight
hydrophobicity of the 3-m-methoxy group (0.02) in 2,3-
(OCH3)3 and the pronounced hydrophobicity of the
m-chloro (0.61) in 2,3-Cl2 analogues warranted their
placement in hydrophobic 3-space. Thus, the ortho
groups of these two analogues were forced into restricted
2-space unlike the other ortho-substituted I. With
pseudo-TMP (2,3,4-(OCH3)3) this would not occur be-
cause the hydrophilic methoxy group in the meta
position would favor binding in hydrophilic 5-space. (See
section on Molecular Graphics Analysis.)
An examination of the ortho-monosubstituted ana-
logues as a separate subset (Table 1) resulted in the
development of eqs 10 and 11. The 2-hydroxy analogue
was not included in the analysis.
log(1/Ki) ) 5.42((0.32) - 0.23((0.09)B56 (10)
n ) 15 r ) 0.827 s ) 0.229 F1,13 ) 28.02 q2 ) 0.607
The MR terms represent the bulk of the substituents
in the individual ortho, meta, or para positions. The
MR terms in the 3 and 4 positions are truncated at a
value of 0.79 which corresponds to the value of a
methoxy group, since earlier studies indicated that
substituents larger than a methoxy group did not
enhance binding to the receptor. The hydrophobicities
of the substituents in the 3, 4, and 6 positions are
represented by π3, π4, and π6, respectively. With the
bacterial enzyme, 80% of the variance in the data can
be explained by steric terms. In the 3, 4, and 5 positions
the presence of a bulky polarizable substituent tends
to enhance inhibitory potency, while the width of the
substituent in the ortho position appears to hinder
log(1/Ki) ) 5.30((0.21) + 0.50((0.23)π6 -
0.81((0.46) log(â6‚10π6 + 1) - 0.15((0.08)B56 (11)
n ) 15 r ) 0.958 s ) 0.133 F2,10 ) 9.46
q2 ) 0.824 π6,0 ) 0.78((0.58) logâ6 ) -0.559
Both eqs 10 and 11 are significant at the 99% level.
Approximately 70% of the variance in the monosubsti-
tuted ortho derivatives can be handled by the steric
term, while about 20% of the variance is defined by the
hydrophobic term. This is slightly higher than the value
obtained for the rest of the benzylpyridines (10%).

4264 J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 22
Selassie et al.
Ta ble 1. Physicochemical Parameters of 2,4-Diamino-5-(X-benzyl)pyrimidines I Used in the Derivation of Eqs 11 and 21
log(Ki)
chicken DHFR
L. casei DHFR
physicochemical parameters
X
obsd
preda
obsd
predb
SI ∆log(1/Ki) calcdc
π2(6)
B56
MR2
σ
2-H
2-F
2-Cl
2-Br
2-I
4.71
4.36
4.76
4.91
5.06
4.53
4.40
4.64
4.47
4.05
4.54
4.61
4.67
5.06
5.10
5.04
4.48
4.53
4.77
4.83
4.93
4.57
4.63
4.69
4.30
4.26
4.48
4.60
4.72
4.89
5.20
5.08
5.20
5.02
5.02
5.00
4.94
5.02
4.55
4.07
4.10
3.96
5.00
4.70
4.87
4.45
4.00
4.74
5.06
5.05
5.08
5.05
5.00
5.03
4.69
4.23
4.00
4.65
4.80
4.80
4.68
5.52
3.99
4.69
0.49
0.66
0.26
0.09
-0.12
0.49
0
1.00
1.35
1.80
1.95
2.15
2.04
3.40
4.10
3.98
1.93
3.07
3.36
4.42
4.79
6.23
3.50
0.1
0
0.06
0.23
0.23
0.18
-0.17
0.03
0.36
-0.33
-0.37
-0.27
-0.24
-0.25
-0.32
-0.34
-0.23
0.14
0.71
0.86
1.12
0.56
-0.03
-0.87
-1.37
-0.67
0.17
0.38
0.81
1.55
2.67
1.66
0.09
0.60
0.88
1.39
0.57
1.21
1.70
1.60
0.29
0.79
1.25
1.62
2.17
3.07
3.17
2-CH3
2-CH2OCH3
2-OSO2CH3
2-OCH2CONH2
2-OH
0.15
-0.57
-0.37
-0.09
0.46
0.09
0.20
-0.61
-1.10
-0.30
2-OCH3
2-OCH2CH3
2-OCH2CHdCH2
2-O(CH2)3CH3
2-O(CH2)5CH3
2-OCH2C6H5
a
b
Predicted using eq 21. Predicted using eq 11. c Selectivity index ) observed log(1/Ki (L. casei DHFR)) - log(1/Ki (chicken liver DHFR)).
Nevertheless, it is apparent that binding space is also
restricted in the 6 position and that substituents with
a wide girth are not conducive to effective binding in
this area. The activity of the 2-hydroxy analogue was
overpredicted by a factor of 5.
log(1/Ki) ) 4.73((0.10) + 0.45((0.10)π3 -
1.14((0.42) log(â3‚10π3 + 1) - 0.59((0.16)MR5 +
0.53((0.21)π4 - 0.49((0.29) log(â4‚10π4 + 1) +
0.24((0.11)π2 (15)
The results for the inhibition of chicken liver DHFR
by I are delineated in the following equations:
n ) 85 r ) 0.849 s ) 0.243 F1,76 ) 18.37
q2 ) 0.659 π3,0 ) 2.34((0.84) log â3 ) -2.529
π4,0 ) ∼3.0 log â4 ) -0.302
log(1/Ki) ) 4.53((0.08) + 0.40((0.15)π3 -
1.12((0.75) log(â3‚10π3 + 1) (12)
log(1/Ki) ) 4.74((0.10) + 0.43((0.09)π3 -
1.07((0.39) log(â3‚10π3 + 1) - 0.66((0.15)MR5 +
n ) 85 r ) 0.533 s ) 0.377 F3,81 ) 10.72
q2 ) 0.229 π3,0 ) 2.48((0.93) log â3 ) -2.735
0.50((0.21)π4 - 0.43((0.28) log(â4‚10π4 + 1) +
0.26((0.10)π2 + 0.32((0.17)Σσ (16)
log(1/Ki) ) 4.67((0.09) + 0.47((0.13)π3 -
1.13((0.51) log(â3‚10π3 + 1) - 0.60((0.21)MR5 (13)
n ) 85 r ) 0.874 s ) 0.225 F1,75 ) 13.91
q2 ) 0.708 π3,0 ) 2.35((0.91) log â3 ) -2.522
π4,0 ) ∼3.0 log â4 ) -0.218
n ) 85 r ) 0.695 s ) 0.322 F1,80 ) 30.76
q2 ) 0.434 π3,0 ) 2.26((1.04) log â3 ) -2.397
log(1/Ki) ) 4.77((0.11) + 0.39((0.09)π3 -
1.03((0.43) log(â3‚10π3 + 1) - 0.76((0.17)MR5 +
log(1/Ki) ) 4.82((0.12) + 0.45((0.11)π3 -
0.52((0.21)π4 - 0.43((0.28) log(â4‚10π4 + 1) +
0.26((0.10)π2 + 0.34((0.16)Σσ + 0.38((0.30)π5
(17)
1.11((0.44) log(â3‚10π3 + 1) - 0.63((0.18)MR5 +
0.60((0.25)π4 - 0.59((0.33) log(â4‚10π4 + 1) (14)
n ) 85 r ) 0.808 s ) 0.269 F3,77 ) 12.54
n ) 85 r ) 0.885 s ) 0.217 F1,74 ) 6.24
q2 ) 0.593 π3,0 ) 2.28((0.95) log â3 ) -2.442
π4,0 ) ∼3.0 log â4 ) -0.215
q2 ) 0.723 π3,0 ) 2.42((0.89) log â3 ) -2.646
π4,0 ) ∼3.0 log â4 ) -0.132

QSAR of 2,4-Diamino-5-(2-X-benzyl)pyrimidines
J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 22 4265
variation in the inhibitory activity of the ortho-substi-
tuted derivatives only extends from 4.03 to 5.06, despite
a range of 4 log units in hydrophobicity.
log(1/Ki) ) 4.68((0.10) + 0.39((0.09)π3 -
1.05((0.43) log(â3‚10π3 + 1) - 0.74((0.17)MR5 +
0.44((0.19)π4 - 0.34((0.26) log(â4‚10π4 + 1) +
0.17((0.13)π2 + 0.38((0.16)Σσ + 0.37((0.29)π5 +
0.11((0.10)MR2 (18)
Three data points were not included in the analysis.
The omitted compounds were the 3,4,5-(CH2CH3)3,
3-OH, and 4-OCF3 analogues. The first analogue was
approximately 4 times more active than predicted, while
the latter two were overpredicted by a factor of 4. In
the case of the 3,4,5-triethyl analogue which is an
isostere of trimethoprim, the low predicted activity may
be attributed to inadequate parametrization of the
hydrophobic contributions of each of the ethyl groups.
The partition coefficient of the molecule has been
measured, and the sum of the substituents’ hydropho-
bicity is deemed to be about 2.58; the contribution for
each ethyl group is ascertained to be the same at a value
of 0.86. Steric crowding around the p-ethyl substituent
subsequently alters its solvation by water, resulting in
unequal solvation of the adjacent ethyl groups. The
p-ethyl substituent would tend to be more hydrophilic
than the adjacent m-ethyl groups since it would be
forced out of the plane of the benzene ring. Greater
assignment of the hydrophobic contribution to the meta
positions would thus enhance the predictive inhibitory
activity of the 3,4,5-triethyl I.
n ) 85 r ) 0.892 s ) 0.212 F1,73 ) 4.62
q2 ) 0.737 π3,0 ) 2.46((0.85) log â3 ) -2.687
π4,0 ) ∼3.00 log â4 ) -0.331
Treatment of mono-ortho-substituted I as a subset leads
to the development of the following equations with
respect to the inhibition of chicken liver DHFR. (See
Table 1.)
log(1/Ki) ) 4.57((0.11) + 0.23((0.10)π2 (19)
n ) 16 r ) 0.785 s ) 0.191 F1,14 ) 22.44 q2 ) 0.473
log(1/Ki) ) 4.59((0.11) + 0.24((0.10)π2 +
0.32((0.42)Σσ (20)
n ) 16 r ) 0.825 s ) 0.180 F1,13 ) 2.64 q2 ) 0.428
An analysis of the mono-ortho-substituted I reveals
a similar dependence on hydrophobicity, size, and
electronic character of the substituents. Hydrophobicity
alone accounts for 62% of the variance in the data, while
Σσ only accounts for 6% of the variance in the data. The
95% confidence interval is also rather high. Further
inclusion of the MR2 term in eq 21 is significant at the
95% level of confidence as delineated by the F statistic.
The 95% confidence interval on the σ term is still high
but less than the coefficient of the electronic term. The
inadequacy of spanned space with regard to the σ
parameter (0.6) is responsible for the high confidence
intervals in eqs 20 and 21. It is of interest to note that
in the comprehensive data set, spanned space in σ )
1.61 and the confidence interval is quite satisfactory.
log(1/Ki) ) 4.46((0.15) + 0.18((0.10)π2 +
0.14((0.11)MR2 + 0.43((0.37)Σσ (21)
n ) 16 r ) 0.891 s ) 0.151 F1,12 ) 6.66 q2 ) 0.594
Equation 18 is very similar to eq 2 with the exception
of the inclusion of the parameters for the ortho position.
The coefficients of the various π terms (0.38 (π3), 0.45
(π4), and 0.38 (π5)) establish the hydrophobic homogene-
ity of the binding site and also indicate that partial
desolvation of the substituents provides the driving force
for these interactions. Using free energy simulation
methods, Fleischman and Brooks and McDonald and
Brooks have also concluded that entropic contributions
and desolvation thermodynamics play a critical role in
overall binding to the avian enzyme.14,15 The similarity
in the values, in particular, the two meta positions and
the para positions, suggests that these terms could be
incorporated into one parameter (π345) as was done in
one of our earlier studies.16 However this would be
misleading as the downward slope of the bilinear
equation in the 3 and 4 positions differs drastically and
the 5 position only calls for a linear dependence on
hydrophobicity. The strong negative slope in the 3
position (1.04 - 0.38 ) 0.66) is suggestive of either steric
hindrance encountered by a 3-substituent which may
be protruding into a bulky residue or a change in the
hydrophobic characteristics of the binding site beyond
five carbon atoms. The negative dependence on bulk
as denoted by the negative coefficient of the MR5 term
indicates that some amino acid residue presents a
strong deterrent to binding in this general vicinity. This
positioning also has an effect on the ortho-substituted
analogues which are all forced into the 2 position
irrespective of hydrophobicity or size. The coefficient
with the hydrophobic term in the 2 position is very small
and suggests that the ortho substituents barely make
hydrophobic contact with the enzyme. Indeed the
Molecu la r Gr a p h ics An a lysis of th e Bin d in g of
2,4-Dia m in o-5-(2′,3′,4′-t r im et h oxyb en zyl)p yr im i-
d in e (P seu d o-TMP ) to Ch ick en Liver DHF R a n d
L. ca sei DHF R. The molecular models of the binding
of avian and bacterial DHFR-benzylpyrimidine com-
plexes have been previously described.3 The structures
of pseudo-TMP 88 were generated using the program
LeAP from the Amber suite of programs.17 The inhibi-
tors were minimized in LeAP and then least-squared
fit to the inhibitors in the original crystallographic
structures. The solvent-accessible surface was calcu-
lated using the program dms contained in the UCSF
Midas Plus suite of programs. These images were
created using the program Neon contained in Midas
Plus.18 Hydrophobic surfaces are colored red, semihy-
drophobic surfaces are colored yellow, and hydrophilic
surfaces are visualized in blue.
1. Bin d in g to L. ca sei DHF R (F igu r e 1). The
benzyl side chain and its attendant phenyl ring avoid
coplanarity with the pyrimidine ring and are in a
perpendicular juxtaposition to it. It is also positioned
lower down in the active site pocket and oriented toward
the nicotinamide binding site. The binding area avail-
able to a substituent in the 2 position of the benzyl ring

4266 J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 22
Selassie et al.
F igu r e 1. Binding of pseudo-TMP to L. casei DHFR.
is constricted and limited by the presence of Phe 30.
More than that, there is also a steric interaction
between the C1′ and C2′ carbons and the H6 proton
which are in close van der Waals contact due to the
folded conformation adopted by the analogue when
bound to the enzyme. A 180° rotation of the benzyl ring
about the C7-C1′ bond places the methoxy substituent
in 6-space which is more extensive and hydrophilic and
away from the side chain of Leu 27 in 3-space. The low
optimum π6 (0.59) suggests that the binding area is
marginally hydrophilic and that desolvation of the ortho
substituent is limited such that binding occurs on the
surface. The weak negative coefficient of the B56 term
suggests that the binding of substituents with substan-
tial width is ill-favored in this region, which is defined
in part by the surface of the 2′-hydroxyl of the coenzyme
NADP. The methoxy substituent in the meta position
is also exposed to this hydrophilic surface where en-
hanced dispersion interactions occur as expressed by the
positive coefficient with the MR5 term. The methoxy
group in the 4 position is partially desolvated in this
extensive flat hydrophilic milieu delineated by residues
Ser 48, Phe 49, Pro 50, and Leu 19. Buttressing of the
4-substituent up against the Pro 50 residue results in
a substantial drop in activity with increasing hydro-
phobicity (i.e., beyond 0.8 unit).
2. Bin d in g to Ch ick en Liver DHF R (F igu r e 2).
This model was constructed based on the X-ray crystal-
lography coordinates of the ternary complex of tri-
methoprim-DHFR-NADPH.19 The active site of chick-
en liver DHFR is much wider than that in the E. coli
enzyme by approximately 1.5-2.0 Å. Overall it is also
much more hydrophobic and extensive. The inhibitor
is inserted deeper into the active site of this enzyme and
binds with a butterfly-type conformation. The o-meth-
oxy group binds in what is designated as 2-space as
opposed to 6-space which is crowded by the bulk of Tyr
31. 2-Space is defined as a narrow cleft between the
methylene bridge of the benzyl ring and the NADPH
cofactor; the methoxy substituent is buttressed up
against the C4-C5-C6 edge of the nicotinamide moiety
of the cofactor, hence the weak positive coefficient with
both the MR2 term and the hydrophobic π2 parameter.
5-Space and 6-space are highly constricted by the
presence of Tyr 31, and the positioning of this residue
precludes binding in this general vicinity; substantial
steric hindrance in this area is denoted by the strong,
negative MR5 term. The 3- and 4-methoxy groups are
partially desolvated by an extensive hydrophobic surface
which is comprised of six hydrophobic residues: Phe 34,
Met 52, Ile 60, Leu 67, Val 115, and the side chain of
Thr 56. Val 115 is a strong deterrent to binding in the
3 position; the steep descending slope of the bilinear
equation is testament to its negative impact on binding.
Discu ssion
The pyrimidine ring binds in a manner similar to
DHFR from both bacterial and avian species, with the
exception of a loss of a hydrogen bond between N4 of
the pyrimidine ring and the carbonyl group of Val 115
in the chicken liver complex. It is clear however that
the benzyl moiety lies in different environments in the
ternary complexes; in the L. casei complex, this moiety
projects beyond the binding site and toward the area
occupied by the cofactor, while in the chicken liver
complex, with the butterfly arrangement of I, the
potential for favorable van der Waals contacts is much
greater.
According to eq 9, it is clear that 2-3-binding space
is not equivalent to 6-5-binding space, in terms of
hydrophobicity and steric constraints. The ability of the
2,4-diamino-5-(X-benzyl)pyrimidines to undergo a 180°
rotation about the C7-C1′ bond in order to place the
substituents in a more favorable binding environment

QSAR of 2,4-Diamino-5-(2-X-benzyl)pyrimidines
J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 22 4267
F igu r e 2. Binding of pseudo-TMP to chicken liver DHFR.
is well-documented. The phenomenon of “ring flipping”
has been well-established by 13C NMR studies of the
binding of trimethoprim to L. casei DHFR.7,20,21 Recent
EPR studies on spin-labeled inhibitors in L. casei DHFR
substantiate these results.22 In the case of monosub-
stituents, “ring-flipping” places all the substituents in
the more accessible 6-space. With multisubstituted
analogues, e.g., 2,3-dichloro analogue, the situation is
complex. While the ortho substituent would favor
6-space, the meta substituents would have a tendency
to bind in the more hydrophobic 3-space. The drive to
desolvate the substituents in hydrophobic 3-space over-
rides all other constraints and favors the localization
of the two hydrophobic chloro groups in 2,3-space,
despite the steric hindrance in 2-space. Using this
assumption, the activity of the analogue is well-
predicted and its standard deviation is well within the
average deviation for the whole set. Thus the overall
assignment of hydrophilic meta and all ortho substitu-
ents to 5-6-space and hydrophobic meta substituents
to 3-space results in a better correlation and a more
cohesive and consistent model.
Equations 9 and 18 pertaining to the inhibition of L.
casei and chicken liver DHFR, respectively, are complex
and multifaceted equations which serve to pinpoint
regularities in behavior of the ortho, meta, and para
substituents. These positional isomers vary consider-
ably in their interactions with both the avian and
bacterial enzymes as can be vividly illustrated in the
case of the monomethoxy-substituted I. (See Table 2.)
In the more hydrophobic avian enzyme, as one moves
from the ortho to the para derivatives, the inhibitory
activity drops slightly (∆Ki ) 0.15), while the reverse
(∆Ki ) 1.25) is seen in the case of the more hydrophilic
bacterial enzyme. It is clear from these models that
substitution in the ortho position, while favorable to
binding in the avian enzyme, is not conducive to binding
Ta ble 2. Comparison of Activities of Equilipophilic
2,4-Diamino-5-(X-benzyl)pyrimidines I
log(1/C)
chicken
DHFR
L. casei
DHFR
∆log(1/Ki)
X
SIa
2-OCH3
3-OCH3
4-OCH3
2,3,4-(OCH3)3 (pseudo-TMP)
3,4,5-(OCH3)3 (TMP)
4.54
4.45
4.29
4.79
3.98
5.00
5.93
6.25
6.22
6.88
0.46
1.48
1.96
1.43
2.90
a
SI ) log(1/Ki(L. casei DHFR)) - log(1/Ki(chicken liver DHFR)).
to the bacterial enzyme. The avian enzyme has often
been used as a surrogate for human DHFR since the
two enzymes are highly homologous with no insertions
or deletions. It is thus assumed that they have a similar
backbone structure. Recently it has been shown that
the mouse L1210 enzyme also falls into the same
general class.23
The detrimental behavior of the ortho substituent
toward the L. casei enzyme has also been seen versus
E. coli dihydrofolate reductase. Stuart et al. evaluated
the binding characteristics of pseudo-TMP versus E. coli
and rat liver DHFR.13 Calculated log(1/Ki) values of
TMP versus E. coli and rat liver DHFR were 8.15 and
3.50, respectively, while pseudo-TMP yielded values of
6.03 and 4.20 versus E. coli and rat liver DHFR,
respectively. It is assumed that pseudo-TMP will also
undergo 180° flipping in the case of the E. coli bacterial
enzyme which would place the methoxy groups in the
4, 5, and 6 positions. The steric hindrance encountered

4268 J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 22
Selassie et al.
in 6-space likely hinders maximal binding of this
analogue to E. coli DHFR.
TMP:27 2,3-dihydro-1-(2,4-diamino-5-pyrimidyl)-1H-in-
dene, which would certainly preclude ring flipping.
Versus E. coli and rat liver DHFR, the following log(1/
C) values were obtained respectively: 5.85 and 3.39. The
selectivity index was 2.46 which again substantially
deviates from that of TMP. Its similar activity to TMP
versus the rat enzyme suggests that it can adopt a
similar mode of binding. However, versus the E. coli
enzyme, the constrained analogue does not allow rota-
tion about the C7-C1′ bond as seen in Figure 2, and
thus the analogue is forced to adopt the unfavorable
conformation in 2-space and not 6-space which most of
the flexible analogues favor.
Some of this steric hindrance may be attributed to
the topography of the methoxy-substituted benzene ring.
It is well-established that the 3,5-dimethoxy groups of
TMP lie mainly in the plane of the benzene ring away
from the methoxy group at C4′ of TMP which lies out
of plane.24 In the case of pseudo-TMP, Stuart et al.
indicate that vicinal substituents on either side force
the 2(6)- and 3(5)-methoxy groups out of plane in
opposite directions thus allowing the 4-methoxy group
to attain planarity with the ring but away from the
3-substituent.13 X-ray crystallographic analysis of tet-
ramethoprim (X ) 3,4,5,6-(OCH3)4) and pentamethop-
rim (X ) 2,3,4,5,6-(OCH3)5) indicated that out-of-plane
methoxy groups predominate in these analogues.25 In
tetramethoprim only the 5-methoxy is coplanar with the
benzene ring, while the other methoxy groups are
rotated down, up, and up with respect to the phenyl
plane. In pentamethoprim, the methyl groups of the
five methoxy groups are alternately oriented down, up,
down, up, down. Thus the topography of the methoxy-
substituted analogues is such that TMP has one methyl
group protruding out of the plane, and pseudo-TMP has
two, while tetramethoprim and pentamethoprim have
three and five, respectively. The latter two compounds
were not assayed versus any bacterial DHFR, but
bacteriological screening indicated that TMP was at
least 10 times more active versus E. coli cultures than
tetramethoprim while pentamethoprim was ineffective.
As suggested by the strong dependence on MR terms
in eq 9, polar/dispersive interactions in the meta and
para positions are of critical importance in the overall
response to the bacterial enzyme. Exposure of the polar
oxygen atoms to the hydrophilic milieu in the L. casei
enzyme without shielding by the methyl groups en-
hances overall binding. Also substituents in the ortho
position with enhanced width have a deleterious effect
on binding. Note that TMP has no ortho substituents.
Pseudo-TMP and tetramethoprim both have one each,
while Pentamethoprim has two.
Con clu sion
On the basis of the preceding analysis, we offer the
following generalizations. The type and positioning of
the substituent on the benzyl ring of the inhibitor
nucleus clearly determines the selectivity of the ana-
logue. Within the binding site, the region available to
the ortho and meta substituents in both the bacterial
and avian enzyme is clearly not equivalent. Rotation
or “flipping” about the C1′-C7 bond to maximize van
der Waals interactions at the molecular level is preva-
lent particularly in nonequivalent monosubstituted or
disubstituted analogues. Thus careful analysis via the
QSAR paradigm is useful in revealing anomalies in
binding behavior. A case in point was illustrated by the
2,3-Cl2 and 2,3-dimethoxy analogues where desolvation
of the substituents was of overriding importance in
binding and overruled steric constraints in these par-
ticular cases.
From the drug design perspective, the presence of
ortho substituents does little to enhance selectivity in
these 2,4-diamino-5-(X-benzyl)pyrimidines. In the case
of the ortho-monosubstituted analogues, it is clear that
the lack of large, hydrophilic substituents in the 4,5
position and a moderate hydrophobic moiety in the 3
position severely curtails their antifolate activity. Pre-
liminary work with a carefully designed series of non-
symmetrical trisubstituted analogues shows promising
activity against E. coli DHFR and weak activity versus
avian DHFR.28 The specificity of DHFRs from various
bacterial and mammalian sources has been critically
examined using different tools and approaches for the
last 30 years. Nevertheless the flexibility of the binding
sites in the proteins continues to reveal subtle nuances
in behavior with each new ligand probe. Roberts has
aptly stated that “the conformational flexibility of
proteins and the subtlety of their responses to amino-
acid substitutions and changes in ligand structure
continue to present substantial challenges”.29
Roth et al. also examined the effects of total substitu-
tion on the ring by analyzing the binding characteristics
of two symmetrical I analogues, 2,3,5,6-tetramethyl-4-
methoxy and 2,3,5,6-tetramethyl-4-ethoxy, again versus
E. coli DHFR.26 For the 4-methoxy and 4-ethoxy
methylated analogues, activities of 6.50 and 6.42 were
obtained. Note that 3,5-dimethyl-4-methoxy had a log-
(1/C) value equal to 6.82. Obviously substitution in both
the ortho positions caused a significant (0.32) drop in
activity. The activities of the tetramethyl compounds
were also assessed versus rat liver DHFR. Unfortu-
nately test concentrations higher than the micromolar
level were not utilized; no inhibition was apparent at
this low concentration. However the 3,5-dimethyl-4-
methoxy analogue did yield an activity of approximately
3.60, and it is conceivable that the tetramethyl ana-
logues would lie within 1 log unit since the 3,5-diethyl-
4-methoxy analogue yields a value of 4.24. The selec-
tivity indices, which are the ratios of the antibacterial
activities to antimammalian activities, would be low
(2.30) compared to that of TMP (8.15 - 3.50 ) 4.65).
Exp er im en ta l Section
QSAR An a lysis. The π constants for most of these sub-
stituents are well-documented.30 The partition coefficients of
many of the benzylpyrimidines have been previously mea-
sured. Some partition coefficients are measured, and their π
values are calculated. The multiregression analysis was done
by using the C-QSAR suite of programs (BioByte, Inc.).
En zym e Stu d ies. DHFR from L. casei was purchased from
Biopure Fine Chemicals, Inc. (Boston, MA) as a lyophilized
powder. It was dissolved in 0.05 M Tris buffer, pH 7.20, to
yield approximately 0.5 enzyme unit/mL. Aliquots were stored
at -20 °C and were thawed just before using. The buffer
contained 50 mM mercaptoethanol to prevent oxidation. The
Chan and Roth in an elegant study designed and
evaluated a conformationally restricted analogue of

QSAR of 2,4-Diamino-5-(2-X-benzyl)pyrimidines
J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 22 4269
Sch em e 2. Synthesis of 2,4-Diamino-5-(X-benzyl)pyrimidines I
Ta ble 3. Synthesis of
R-(X-Benzylidene)-â-methoxypropionitriles III
solutions of dihydrofolic acid (FAH2) and NADPH were kept
at 0 °C. The benzylpyrimidines were solubilized with DMSO
such that the final concentration of DMSO did not exceed 0.5%.
DHFR activity with and without added inhibitor was deter-
mined by the standard spectrophotometric assay which moni-
tors the oxidation of NADPH to NADP+ and the concomitant
reduction of FAH2 to FAH4 at 340 nm by using a stopped flow
spectrophotometer. The Ki, log(1/Ki), and 95% confidence
limits were determined by using the jackknife procedure.31
DHFR from chicken liver was a generous gift from B. T.
Kaufman (NIADDK, NIH, Bethesda, MD). Each vial con-
tained 0.19 enzyme unit/mL. Each vial was dissolved in 24
mL of 200 µM NADPH and used as previously described.31
Syn th esis. Melting points were determined with a Buchi
melting point apparatus and are uncorrected. IR and NMR
spectra are in agreement with the structures cited and were
recorded on a Perkin-Elmer 12 or a Perkin-Elmer FTIR 1600
and a GE QE-300 spectrometer. 1H NMR spectra were
obtained in (CD3)2SO and were referenced to TMS. Elemental
analyses (CH) were performed by Galbraith Laboratories, Inc.,
Knoxville, TN, and are within (0.4% of theory unless other-
wise noted. Thin-layer chromatography was performed on
Baker 1B-F silica gel plates.
Syn t h esis of 2,4-Dia m in o-5-(2-X-ben zyl)p yr im id in es.
Most of the benzylpyrimidines in this study were synthesized
by using the general procedure of Stenbuck et al. with minor
variations on this theme.32 (See Scheme 2.) The following
pyrimidines were synthesized by using commercially available
or synthesized 2-X-benzaldehydes: such as X ) 2-F, 2-Cl, 2-Br,
2-CH3, 2-OCH2CH3, 2-OCH2CHdCH2, 2-O(CH2)3CH3, 2-O(CH2)5-
CH3, 2-OCH2C6H5, 2,3-(OCH3)2, and 2,3,4-(OCH3)3. The gen-
eral methodology (method A) is outlined as follows.
Meth od A: Syn th esis of 2,4-Dia m in o-5-(2-br om oben -
zyl)p yr im id in e (71). To a solution of sodium (2 g) in
methanol (50 mL) were added 2-bromobenzaldehyde (9.8 g,
0.066 mol) and â-methoxypropionitrile (5.6 g, 0.066 mol). After
heating at reflux temperature for 7 h, the solution was cooled,
the methanol was removed under reduced pressure, and the
mixture of benzal nitriles was dissolved in ether. The ethereal
solution was washed consecutively by water (2 × 100 mL),
NaHSO3 (2 × 50 mL), and water (2 × 50 mL) and dried
overnight (sodium sulfate). After removal of the drying agent
and the solvent, the brown oil was distilled in vacuo. Weight
of the mixture of benzal nitriles ) 10.4 g (64%); bp ) 160-
176 °C/2.5 mmHg. Table 3 lists the benzal nitriles, III,
obtained by this procedure. They were used in the next step
without further purification.
method
X
yield (%)
bp (°C/mmHg)
A
A
A
A
A
A
A
A
A
A
A
2-F
61
65
63
60
51
36
49
68
70
53
63
123-140/1.4
136-152/1.3
160-176/2.5
129-144/2.2
130-153/2.2
155-169/0.4
142-164/0.2
203-215/0.8
138-145/0.4
100-145/2.0
182-187/0.5
2-Cl
2-Br
2-CH3
2-OCH2CH2
2-O(CH2)3CH3
2-O(CH2)5CH3
2-OCH2C6H5
2-OCH2CHdCH2
2,3-(OCH3)2
2,3,4-(OCH3)3
(2-bromobenzyl)pyrimidine, was recrystallized from absolute
ethanol: mp 250-251 °C (EtOH), 40%; NMR (Me2SO-d6) δ 3.70
(s, 2, CH2), 5.80 (s, 2, NH2), 6.25 (s, 2, NH2), 7.20 (m, 3, Ar),
7.35 (t, 1, Ar), 7.60 (s, 1, pyr C6-H). Anal. (C11H11N4Br).
2,4-Diam in o-5-(2-flu or oben zyl)pyr im idin e (69): mp 181-
182 °C (EtOH-benzene), 3%; NMR (Me2SO-d6) δ 3.62 (s, 2,
CH2), 5.80 (s, 2, NH2), 6.22 (s, 2, NH2), 7.20 (br-m, 4, Ar), 7.40
(d, 1, pyr C6-H). Anal. (C11H11N4F).
2,4-Diam in o-5-(2-ch lor oben zyl)pyr im idin e (70): mp 231-
232 °C (EtOH), 39%; NMR (Me2SO-d6) δ 3.68 (s, 2, CH2), 5.80
(s, 2, NH2), 6.25 (s, 2, NH2), 7.22 (br-m, 4, Ar), 7.45 (d, 1, pyr
C6-H). Anal. (C11H11N4Cl).
2,4-Dia m in o-5-(2-m et h ylb en zyl)p yr im id in e (73): mp
205-207 °C (EtOH-MeOH), 16%; NMR (Me2SO-d6) δ 2.20 (s,
3, CH3), 3.58 (s, 2, CH2), 5.90 (s, 2, NH2), 6.35 (s, 2, NH2), 7.15
(br-m, 4, Ar), 7.00 (m, 1, pyr C6-H). Anal. (C12H14N4).
2,4-Diam in o-5-(2-eth oxyben zyl)pyr im idin e (79): mp 175-
176 °C (EtOH), 4%; NMR (Me2SO-d6) δ 1.35 (t, 3, CH3), 3.55
(s, 2, CH2), 4.05(q, 2, OCH2), 5.80 (s, 2, NH2), 6.20 (s, 2, NH2),
7.00 (br-m, 4, Ar), 7.50 (s, 1, pyr C6-H). Anal. (C13H16N4O).
2,4-Dia m in o-5-(2-p r op en yloxyben zyl)p yr im id in e (80):
mp 175-176 °C (MeOH-benzene), 9%; NMR (Me2SO-d6) δ
3.55 (s, 2, CH2), 4.05 (d, 2, OCH2), 5.35 (d, 1, dCH2), 5.42 (d,
1, dCH2), 5.70 (s, 2, NH2), 6.10 (s, 2, NH2), 6.05 (m, 1, CH),
6.85-7.18 (br-m, 4, Ar), 7.40 (s, 1, pyr C6-H). Anal.
(C14H16N4O).
2,4-Dia m in o-5-(2-b u t oxyb en zyl)p yr im id in e (81): mp
146-147 °C (EtOH-benzene), 5%; NMR (Me2SO-d6) δ 0.95 (t,
3, CH3), 1.45 (m, 2, CH2), 1.70 (m, 2, CH2) 3.55 (s, 2, CH2),
4.00 (t, 2, OCH2), 5.70 (s, 2, NH2), 6.10 (s, 2, NH2), 6.85-7.20
(br-m, 4, Ar), 7.40 (s, 1, pyr C6-H). Anal. (C15H20N4O).
2,4-Dia m in o-5-(2-h exyloxyben zyl)p yr im id in e (82): mp
145-146 °C (EtOH), 2%; NMR (Me2SO-d6) δ 0.88 (t, 3, CH3),
1.30 (m, 4, (CH2)2), 1.45 (m, 2, CH2), 1.70 (m, 2, CH2) 3.55 (s,
2, CH2), 3.99 (t, 2, OCH2), 5.70 (s, 2, NH2), 6.10 (s, 2, NH2),
6.85-7.20 (br-m, 4, Ar), 7.40 (s, 1, pyr C6-H). Anal.
(C17H24N4O).
A solution of guanidine hydrochloride (12 g, 0.126 mol) in
methanol was neutralized by a methanolic solution of sodium
(3 g, 0.130 mol). The mixture of benzal nitriles (10.4 g, 0.04
mol) was dissolved in the guanidine solution, and the resulting
solution was refluxed for 24-48 h. After cooling the solvent
was reduced in volume, and a precipitate was collected. The
precipitate was further purified by passage through an alu-
mina column and elution with an appropriate solvent (2-5%
methanol in chloroform). The resulting solid, 2,4-diamino-5-
2,4-Dia m in o-5-(2-ben zyloxyben zyl)p yr im id in e (83): mp
186-187 °C (MeOH-benzene), 7%; NMR (Me2SO-d6) δ 3.60
(s, 2, CH2), 5.20 (s, 2, OCH2), 5.73 (s, 2, NH2), 6.10 (s, 2, NH2),

4270 J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 22
Selassie et al.
6.90-7.20 (br-m, 4, Ar), 7.35 (s, 5, Ar), 7.40 (s, 1, pyr C6-H).
Anal. (C18H18N4O).
SO-d6) δ 3.64 (s, 2, CH2), 5.80 (s, 2, NH2), 6.30 (s, 2, NH2),
7.00-7.35 (br-m, 4, Ar), 7.90 (d, 1, pyr C6-H). Anal.
(C11H11N4I).
2,4-Dia m in o-5-(2,3-d im eth oxyben zyl)p yr im id in e (84):
mp 194-195 °C (EtOH), 50%; NMR (Me2SO-d6) δ 3.50 (s, 2,
CH2), 3.68 (s, 3, OCH3), 3.75 (s, 3, OCH3), 5.70 (s, 2, NH2),
6.10 (s, 2, NH2), 6.62-6.95 (br-m, 3, Ar), 7.32 (s, 1, pyr C6-
H). Anal. (C13H16N4O2).
2,4-Diamino-5-(2,3,4-trimethoxybenzyl)pyrimidine (P seu-
d o-TMP ) (88): mp 221-223 °C (MeOH), 15%; NMR (Me2SO-
d6) δ 3.40 (s, 2, CH2), 3.70 (s, 9, OCH3), 5.65 (s, 2, NH2), 6.05
(s, 2, NH2), 6.70 (br-m, 2, Ar), 7.30 (s, 1, pyr C6-H). Anal.
(C13H16N4O2).
Meth od B: Syn th esis of 2,4-Dia m in o-5-(2-h yd r oxyben -
zyl)p yr im id in e (77). A solution of 2,4-diamino-5-(2-meth-
oxybenzyl)pyrimidine (11.5 g, 0.05 mol) in 48% HBr (100 mL)
was refluxed under nitrogen for 4 h. After cooling, the solvent
was removed under reduced pressure to yield a red oil. The
oil gradually solidified on standing to yield the hydrobromide
salt of the desired pyrimidine. Recrystallization from ethanol
yielded a pure solid: mp 271-273 °C (MeOH), 61%; NMR (Me2-
SO-d6) δ 3.50 (s, 2, CH2), 6.70-6.85 (m, 2, Ar), 7.15 (m, 2, Ar),
7.30 (s, 1, pyr C6-H) 8.20 (s-br, 2, NH2), 9.50 (s-br, 2, NH2),
12.10 (s-br, 1, NH). Anal. (C11H12N4‚HBr).
Syn th esis of 2,4-Dia m in o-5-(2-ca r boxa m id om eth oxy-
ben zyl)p yr im id in e (76). A suspension of potassium hydrox-
ide (0.3 g, 0.005 mol) and 2,4-diamino-5-(2-hydroxybenzyl)-
pyrimidine (0.7 g, 0.003 mol) was heated at 70-80 °C for 30
min till solution ensued. Then chloroacetamide (0.6 g, 0.006
mol) was added and the solution heated at reflux for 20 h. The
resulting suspension was filtered, and the gray solid was
collected, recrystallized from water, and dried over P2O5: mp
207-209 °C (H2O), 38%; NMR (Me2SO-d6) δ 3.35 (s, 2, NH2),
3.60 (s, 2, CH2), 4.30 (s, 2, OCH2), 5.70 (s, 2, NH2), 6.05 (s, 2,
NH2), 6.70-7.15 (br-m, 4, Ar), 7.40 (s, 1, pyr C6-H). Anal.
(C13H15N5O2).
Syn t h esis of 2,4-Dia m in o-5-(2-(m et h ylsu lfon yloxy)-
ben zyl)p yr im id in e (75). A solution of 2,4-diamino-5-(2-
hydroxybenzyl)pyrimidine (2.2 g, 0.01 mol) and 2 N KOH (15
mL) was cooled over ice. Methanesulfonyl chloride (1.4 g,
0.012 mol) was dissolved in 10 mL of benzene and added
dropwise to the ice-cooled solution of the pyrimidine, with
continuous agitation. After 10 min, a white precipitate was
formed. It was collected, washed with 2 N KOH and water,
and then recrystallized from ethanol: mp 203-204 °C (EtOH),
59%; NMR (Me2SO-d6) δ 3.50 (s, 3, CH3), 3.65 (s, 2, CH2), 5.75
(s, 2, NH2), 6.10 (s, 2, NH2), 7.10-7.30 (br-m, 4, Ar), 7.36 (s, 1,
pyr C6-H). Anal. (C12H14N4O3S).
Meth od C: Syn th esis of 2,4-Dia m in o-5-(2-iod oben zyl)-
p yr im id in e (72). A solution of 2-iodobenzoic acid (31 g, 0.125
mol) in methanol was saturated with HCl gas and refluxed
for 30 h. After removal of the methanol, the oil was dissolved
in ether, washed successively with water (2 × 100 mL), 10%
Na2CO3 (2 × 100 mL), and water (2 × 100 mL), and dried
(MgSO4). After removal of the ether the yellow oil was distilled
to yield methyl 2-iodobenzoate: yield ) 13 g (40%); bp ) 92-
93 °C (0.4 mmHg) (lit. bp 103.5 °C/1 mmHg).33
To a solution of methyl 2-iodobenzoate (13 g, 0.05 mol) in
toluene was added diisobutylaluminum hydride (60 mL of 1.0
M solution in toluene) at -60 °C. After 2 h the mixture was
decomposed with excess saturated NH4Cl solution, poured into
an ice-water mixture, and extracted with ether. The ether
extract was dried over MgSO4 and evaporated to yield 2-io-
dobenzaldehyde: yield ) 4 g (34%); mp ) 38-40 °C (lit. mp )
39 °C).34
To an ethanolic solution of sodium (2 g in 80 mL of ethanol)
was added crude 2-iodobenzaldehyde (4.0 g, 0.017 mol) and
â-methoxypropionitrile (1.6 g, 0.018 mol). The solution was
refluxed for 24 h. Meanwhile a guanidine hydrochloride
solution (5.0 g, 0.05 mol) was neutralized by an ethanolic
solution of sodium (1.5 g in 50 mL of ethanol) and added to
the flask. After continuous heating at reflux for another 48
h, the solution was washed, the ethanol removed, and the
yellow precipitate recrystallized from ethanol to yield the
desired pyrimidine: mp 266-267 °C (EtOH), 9%; NMR (Me2-
Meth od D: Syn th esis of 2,4-Dia m in o-5-(2-m eth oxy-
m eth ylben zyl)p yr im id in e (74). A solution of 2-(R-bromom-
ethyl)benzonitrile (15 g, 0.077 mol) in methanol (100 mL) was
added dropwise to a solution of sodium (1.8 g) in methanol
(100 mL). After refluxing for 5 h, the solution was washed,
the solvent removed, and the oil extracted with ethyl acetate
and washed with water. The ethyl acetate fractions were
combined and dried overnight (MgSO4). The solvent was
removed and the resulting oil distilled to yield 2-(2-methoxym-
ethyl)benzonitrile: yield ) 9.3 g (85%); bp ) 80-85 °C/0.75
mmHg. The nitrile was used without further purification in
the next step.
A mixture of Raney alloy (50; 50; 20 g), nitrile (9.3 g, 0.06
mol), and 50% v/v aqueous formic acid (300 mL) was refluxed
for 36 h. After the mixture was filtered, the filtrate was
diluted with water and extracted with ether. The ether extract
was washed with NaHCO3 (2 × 100 mL) and water (1 × 100
mL) and dried (MgSO4). After removal of the ether a yellow
solid was obtained which gave a positive test with 2,4-
dinitrophenyl hydrazine: yield ) 2.8 g (31%). The benzalde-
hyde was used without further purification.
The crude benzaldehyde (2.8 g, 0.018 mol), â-methoxypro-
pionitrile (0.018 mol), and sodium (1 g) in methanol (150 mL)
were refluxed for 20 h. Then a solution of guanidine hydro-
chloride (3.0 g, 0.03 mol) in sodium methoxide was added to
the benzal nitrile solution, which was further refluxed for 48
h. After cooling the solvent was removed, and the yellow solid
was decolorized with charcoal and recrystallized twice from
ethanol to yield the desired pyrimidine: mp 205-206 °C
(EtOH), 5%; NMR (Me2SO-d6) δ 3.20 (s, 3, CH3), 3.50 (s, 2,
CH2), 4.40 (s, 2, CH2O-), 5.70 (s, 2, NH2), 6.10 (s, 2, NH2),
7.00-7.30 (br-m, 4, Ar), 7.30 (m, 1, pyr C6-H). Anal.
(C13H16N4O).
Meth od E: Syn th esis of 2,4-Dia m in o-5-(2-m eth oxy-
ben zyl)p yr im id in e (78). Derivatives synthesized by this
method included the 2-OCH3, 2,3-Cl2, 2,4-Cl2, and 2,4-(CH3)2
analogues. The general procedure of Poe et al. was used.35 To
a mixture of anisaldehyde (28 g, 0.2 mol), 3-anilinopropionitrile
(38 g, 0.26 mol) and dimethyl sulfoxide (100 mL), which had
been heated to 100 °C, was added a slurry of sodium methoxide
(11 g, 0.20 mol) in dimethyl sulfoxide (60 mL). The temper-
ature was raised to 125-130 °C. for 2 h. After cooling, the
dark brown solution was poured into ice-water (1000 mL) and
left overnight in the freezer. The slurry was filtered, and the
gummy brown oil was triturated with cold ethanol-ether to
yield a pale yellow product: yield ) 38 g (72%); mp ) 117-
119 °C. Generally, the crude intermediates whose yields
hovered between 36% and 50% were used without further
purification in the next step.
Crude 2-methoxy-â-cyano-N-phenylcinnamylamine (26 g,
0.1 mol) was added to an ethanolic solution of guanidine (20 g
of the hydrochloride neutralized by 5 g of sodium in 20 mL of
ethanol). The solution was refluxed for 24 h. The ethanol was
removed under reduced pressure, and the resulting orange oil
was triturated with ethanol until a yellow solid appeared: mp
166-167 °C (EtOH), 50%; NMR (Me2SO-d6) δ 1.35 (t, 3, CH3),
3.55 (s, 2, CH2), 3.80 (s, 3, OCH3), 5.75 (s, 2, NH2), 6.12 (s, 2,
NH2), 7.10 (br-m, 4, Ar), 7.38 (s, 1, pyr C6-H). Anal.
(C12H14N4O).
2,4-Dia m in o-5-(2,3-d ich lor oben zyl)p yr im id in e (85): mp
268-270 °C (95% EtOH), 15%; NMR (Me2SO-d6) δ 3.70 (s, 2,
CH2), 5.80 (s, 2, NH2), 6.20 (s, 2, NH2), 7.00-7.21 (m, 2, Ar),
7.25 (s, 1, pyr C6-H), 7.45 (d, 1, Ar). Anal. (C11H10N4Cl2).
2,4-Dia m in o-5-(2,4-d ich lor oben zyl)p yr im id in e (87): mp
225-227 °C (MeOH-EtOH), 20%; NMR (Me2SO-d6) δ 3.60 (s,
2, CH2), 5.75 (s, 2, NH2), 6.20 (s, 2, NH2), 7.10 (d, 1, Ar), 7.20
(s, 1, pyr C6-H), 7.30-7.55 (d, 2, Ar). Anal. (C11H10N4Cl2).
2,4-Diam in o-5-(2,4-dim eth ylben zyl)pyr im idin e (86): mp
215-216 °C (EtOH), 62%; NMR (Me2SO-d6) δ 2.1 (s, 3, CH3),
2.2 (s, 3, CH3), 3.50 (s, 2, CH2), 5.70 (s, 2, NH2), 6.10 (s, 2,

QSAR of 2,4-Diamino-5-(2-X-benzyl)pyrimidines
J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 22 4271
(10) Stammers, D. K.; Champness, J . N.; Beddell, C. R.; Dann, J . G.;
Eliopoulos, E.; Geddes, A. J .; Ogg, D.; North, A. C. T. The
Structure of Mouse L1210 Dihydrofolate Reductase-Drug Com-
plexes and the Construction of a Model of Human Enzyme.
FEBS Lett. 1987, 218, 178-184.
(11) Volz, K. W.; Matthews, D. A.; Alden, R. A.; Freer, S. T.; Hansch,
C.; Kaufman, B. T.; Kraut, J . Crystal Structure of Avian
Dihydrofolate Reductase Containing Phenyltriazine and NAD-
PH. J . Biol. Chem. 1982, 257, 2528-2536.
(12) Hansch, C.; Leo, A. Steric Effects on Organic Reactions. In
Exploring QSAR: Fundamentals and Applications in Chemistry
and Biology; Heller, S. R., Ed.; American Chemical Society:
Washington, DC, 1995; p 76.
(13) Stuart, A.; Paterson, T.; Roth, B; Aig, E. 2,4-Diamino-5-ben-
zylpyrimidines 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, 607-
673.
(14) Fleischman, S. H.; Brooks, C. L., III. Protein-Drug Interac-
tions: Characterization of Inhibitor Binding in Complexes of
DHFR with Trimethoprim and Related Derivatives. Proteins
1990, 7, 52-61.
(15) Mcdonald, J . J .; Brooks, C. L., III. Theoretical Approach to Drug
Design. 2. Relative Thermodynamics of Inhibitor Binding by
Chicken Dihydrofolate Reductase to Ethyl Derivatives of Tri-
methoprim Substituted at 3′-, 4′-, and 4′-postions. J . Am. Chem.
Soc. 1991, 113, 2295-2301.
Ta ble 4. Synthesis of X-Benzaldehydes Using Method F
X
yield (%) bp, °C (mmHg) lit. bp, °C (mmHg)
2-O(CH2)3CH3
2-O(CH2)5CH3
2-OCH2C6H5
46
75
57
59
90-93 (0.1)
144-146 (0.5)
49-50 (EtOH)
105-108 (1.8)
105 (0.2)
a
b
2-OCH2CHdCH2
136-138 (12)
a
b
Anal. (C13H18O2) C, H. Anal. (C14H12O2) C, H.
NH2), 6.80-6.90 (m, 3, Ar), 7.10 (s, 1, pyr C6-H), 7.45 (d, 1,
Ar). Anal. (C13H16N4).
Syn t h esis of o-Alk oxy-Su b st it u t ed Ben za ld eh yd es.
This general method was utilized for the synthesis of o-alkoxy-
substituted benzaldehydes, which were not commercially
available. (See Table 4.) Potassium hydroxide (11.2 g, 0.20
mol) was dissolved in a mixture of diglyme and ethanol and
refluxed with salicylaldehyde (24.4 g, 0.2 mol) for 30 min. Then
1-bromobutane (27.4 g, 0.2 mol) was added, and the resulting
solution was heated at reflux for 72 h. After cooling the
potassium bromide was filtered off and the solvent removed
under reduced pressure. The solid was washed with water
and extracted with ether. The ethereal solution was washed
successively by KOH (2 × 100 mL) and water (2 × 100 mL)
and dried overnight (MgSO4). After removal of the drying
agent and solvent, the yellow oil was distilled to obtain the
desired benzaldehyde. The following benzaldehydes were
synthesized in this manner: 2-O(CH2)3CH3;36 2-O(CH2)5CH3;37
2-OCH2C6H5 and 2-OCH2CHdCH2.38
(16) Li, R. L.; Hansch, C.; Matthews, D.; Blaney, J . M.; Langridge,
R.; Delcamp, T. J .; Susten, S. S.; Freisheim, J . H. A Comparison
by QSAR, Crystallography and Computer Graphics of the
Inhibition of Various Dihydrofolate Reductases by 5-(X-Benzyl)-
2,4-diamino Pyrimidines. Quant. Struct. Act. Relat. 1982, 1, 1-7.
(17) Pearlman, D. A.; Case, D. A.; Caldwell, J . W.; Ross, W. S.;
Cheatham, T. C.; Debolt, S. E.; Ferguson, D. M.; Siebel, G. L.;
Kollman, P. A. AMBER, A Package of Computer Programs for
Applying Molecular Mechanics, Molecular Dynamics and Free
Energy Calculations to Simulate the Structural and Energetic
Properties of Molecules. Computer Phys. Commun. 1995, 91,
1-41.
Ack n ow led gm en t. This research was funded in
part by NIH Grant A132734-01 (C.S.) and NIH Grant
RR-1081 (T. Ferrin, P.I.). Molecular graphics images
were produced using the Midas Plus suite of programs
at the UCSF Computer Graphics Laboratory.
(18) Huang, C. C.; Couch, A. S.; Patterson, E. F.; Ferrin, T. E.
MidasPlus Release 2.0, 1994.
(19) 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 Structures of E. coli and
Chicken Liver DHFR Containing Bound Trimethoprim. J . Biol.
Chem. 1985, 260, 381-391.
(20) Feeney, J . NMR Studies of Interactions of Ligands with Dihy-
drofolate Reductase. Biochem. Pharmacol. 1990, 40, 141-152.
(21) Searle, M. S.; Forster, M. J .; Birdsall, B.; Roberts, G. C. K.;
Feeney, J .; Cheng, H. T. A.; Kompis, I.; Geddes, A. J . Dynamics
of Trimethoprim Bond to Dihydrofolate Reductase. Proc. Natl.
Acad. Sci. U.S.A. 1988, 85, 3787-3791.
(22) Blakely, R. L.; Piper, J . R.; Mahorag, G.; Appleman, J . R.;
Delcamp, T. J .; Freisheim, J . H.; Kulinski, R. F.; Montgomery,
J . Mobility of the Spin-Labeled Side Chains of Some Novel
Antifolate Inhibitors in their Complexes with Dihydrofolate
Reductase. Eur. J . Biochem. 1991, 196, 271-280.
(23) Stammers, D. K.; Champness, J . N.; Dann, J . G.; Beddell, C. R.
The Three-Dimensional Structure of Mouse L1210 Dihydrofolate
Reductase. In Chemistry and Biology of Pteridines; Blair, J . A.,
Ed.; Walter de Gruyter: New York, 1983; pp 567-57.
(24) Koetzle, T. F.; Williams, G. J . B. The Crystal and Molecular
Structure of the Antifolate Drug Trimethoprim (2,4-diamino-5-
(3,4,5-Trimethoxybenzyl) Pyrimidine). A Neutron Diffraction
Study. J . Am. Chem. Soc. 1976, 98, 2074-2078.
(25) Brossi, A.; Sharma, P. N.; Takahashi, K.; Chiang, J . F.; Karle,
I. L.; Seibert, G. Tetramethoprim and Pentamethoprim: Syn-
thesis Antibacterial Properties and X-ray Structures. Helv.
Chim. Acta 1983, 66, 795-800.
(26) Roth, B.; Rauckman, B. S.; Ferone, R.; Baccanari, D. P.;
Champness, J . N.; Hyde, R. M. 2,4-Diamino-5-benzylpyrimidines
as Antibacterial Agents. 7. Analysis of the Effect of 3,5-dialkyl
Substituent Size And Shape On Binding to Four Different
Dihydrofolate Reductase Enzymes. J . Med. Chem. 1987, 30,
348-356.
Su p p or tin g In for m a tion Ava ila ble: Tables of biological
data and physical/chemical data (5 pages). Ordering informa-
tion is given on any current masthead page.
Refer en ces
(1) Hitchings, A. H. Selective Inhibitors Of Dihydrofolate Reductase
(Nobel Lecture). Angew. Chem., Int. Ed. 1989, 879-885.
(2) Kuyper, L. F. Inhibitors of Dihydrofolate Reductase. In Computer
Aided Drug Design. Methods and Applications; Perun, T. J .,
Propst, C. L., Eds.; Marcel Dekker: New York, 1989; pp 327-
370.
(3) Blaney, J . M.; Hansch, C.; Silipo, C.; Vittoria, A. Structure
Activity Relationships of Dihydrofolate Reductase Inhibitors.
Chem. Rev. 1984, 84, 333-407.
(4) Selassie, C. D.; Klein, T. E. Comparative Quantitative Structure
Activity Relationships of the Inhibition of Dihydrofolate Reduc-
tase. In Comparative Quantitative Structure Activity Relation-
ships; Devillers, J ., Ed.; Taylor and Francis: Washington, DC,
1998; in press.
(5) Selassie, C. D.; Fang, Z. X.; Li, R. L.; Hansch, C.; Debnath, G.;
Klein, T. E.; Langridge, R.; Kaufman, B. T. On the Structure
Selectivity Problem in Drug Design. A Comparative Study of
Benzylpyrimidine Inhibition of Vertebrate and Bacterial Dihy-
drofolate Reductase via Molecular Graphics and Quantitative
Structure Activity Relationships. J . Med. Chem. 1989, 32, 1895-
1905.
(6) Selassie, C. D.; Klein, T. E. Building Bridges: QSAR and
Molecular Graphics. In 3D Drug Design: Theory, Methods and
Applications; Kubinyi, H., Ed.; Escom Science Publishers: Leiden,
1993; pp 256-275.
(27) Chan, J . H.; Roth, B. 2,4-Diamino-5-benzylpyrimidines as
Antibacterial Agents. 14. 2,3-Dihydro-1-(2,4-diamino-5-pyrim-
idyl)-1H-Indenes as Conformationally Restricted Analogues of
Trimethoprim. J . Med. Chem. 1991, 34, 550-555.
(7) Cheung, H. T. A.; Searle, M. S.; Feeney, J .; Birdsall, B.; Roberts,
G. C. K.; Kompis, I.; Hammond, S. J . Trimethoprim Binding to
Lactobacillus casei Dihydrofolate Reductase: A 13C NMR study
using Selectively 13C-enriched Trimethoprim. Biochemistry 1986,
25, 1925-1931.
(8) Kubinyi, H.; Kehrhahn, O. Quantitative Structure Activity
Relationships. VI. Nonlinear Dependence of Biological Activity
on Hydrophobic Character: Calculation Procedures for the
Bilinear Model. Arzneim-Forsch. (Drug Res.) 1978, 28, 598-601.
(9) Selassie, C. D.; Li, R. L.; Poe, M.; Hansch, C. On the Optimization
of Hydrophobic and Hydrophilic Substituent Interactions of 2,4-
diamino-5-(substituted benzyl) Pyrimidines with Dihydrofolate
Reductase. J . Med. Chem. 1991, 34, 46-54.
(28) Selassie, C.; Gan, W. X.; Yi, C.; Liu, M. Unpublished observa-
tions.
(29) Roberts, G. C. K. Understanding the Specificity of the Dihydro-
folate Reductase Binding Site. NATO ASI Ser., Ser. A 1989, 183,
209-220.
(30) Hansch, C.; Leo, A.; Hoekman, D. Hammett Sigmas. In Exploring
QSAR: Hydrophobic, Electronic and Steric Constants; Heller,
S. R., Ed.; American Chemical Society: Washington, DC, 1995;
pp 217-304.

4272 J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 22
Selassie et al.
(31) Dietrich, S. W.; Blaney, J . M.; Reynolds, M. A.; J ow, P. Y. C.;
Hansch, C. Quantitative Structure Selectivity Relationships.
Comparison of the Inhibition of Escherichia coli and Bovine Liver
Dihydrofolate Reductase by 5-(substituted benzyl)-2,4-diamino
Pyrimidines. J . Med. Chem. 1980, 23, 1205-1212.
(32) Stenbuck, P.; Baltzly, R.; Hood, H. M. A New Synthesis of
5-benzyl Pyrimidines. J . Org. Chem. 1963, 28, 1983-1988.
(33) Brown, H. C.; Okamoto, Y.; Ham, G. The Effect of Halogen
Substituents Upon the Rates of Electrophilic Reactions. J . Am.
Chem. Soc. 1957, 79, 1906-1909.
(34) The Systematic Identification of Organic Compounds. A Labora-
tory Manual, 6th ed.; Shriner, R. L., Fuson, R. C., Curtin, D. Y.,
Morrill, T. C., Eds.; J ohn Wiley & Sons: New York, 1980; p 538.
(35) Poe, M.; Ruyle, W. V. Inhibitor of Dihydrofolate Reductase. U.S.
Patent 4,258,405, 1981.
(36) Katz, L.; Karger, L.; Schroeder, W.; Cohen, M. Hydrazine
Derivatives. I. Benzalthio and Bisbenzaldithio-Salicyl Hy-
drazides. J . Org. Chem. 1953, 18, 1380-1402.
(37) Coates, L. V.; Drain, D. J .; McCrae, F. J . The Preparation and
Evaluation of Some Phenolic Ethers as Antifungal Agents. J .
Pharm. Pharmacol. 1959, 11, 240T-249T.
(38) Pansevich-Kolyada, V. I.; Strel’tsov, A. E. Ethers with an Allylic
Double Bond. VII. Allyl Ether of Salicyladehyde in the Grignard
Reaction. Zh. Obshchei. Khim. 1960, 30, 3261-3263.
J M970776J