Rationally designing safer anilines: The challenging case of 4-aminobiphenyls

Article abstract of DOI:10.1021/jm3001295

We describe how we have been able to design 4-aminobiphenyls that are nonmutagenic (inactive in the Ames test). No such 4-aminobiphenyls were known to us, but insights provided by quantum mechanical calculations have permitted us to design and synthesize some examples. Importantly, the quantum mechanical calculations could be combined with predictions of other properties of the compounds that contained the 4-aminobiphenyls so that these remained druglike. Having found compounds that are not active, the calculations can provide insight into which factors (electronic and conformational in this case) are important. The calculations provided SAR-like information that was able guide the design of further examples of 4-aminobiphenyls that are not active in the Ames test.

Full text of DOI:10.1021/jm3001295

Article
pubs.acs.org/jmc
Rationally Designing Safer Anilines: The Challenging Case of
4-Aminobiphenyls
Alan M. Birch, Sam Groombridge, Robert Law, Andrew G. Leach,* Christine D. Mee,
and Carolin Schramm
AstraZeneca Pharmaceuticals, Mereside, Alderley Park, Macclesfield, SK10 4TG, U.K.
S
* Supporting Information
ABSTRACT: We describe how we have been able to design 4-
aminobiphenyls that are nonmutagenic (inactive in the Ames test). No
such 4-aminobiphenyls were known to us, but insights provided by
quantum mechanical calculations have permitted us to design and
synthesize some examples. Importantly, the quantum mechanical
calculations could be combined with predictions of other properties of
the compounds that contained the 4-aminobiphenyls so that these
remained druglike. Having found compounds that are not active, the
calculations can provide insight into which factors (electronic and
conformational in this case) are important. The calculations provided
SAR-like information that was able guide the design of further examples of 4-aminobiphenyls that are not active in the Ames test.
situation where an Ames positive amino aromatic is embedded
in a more complex structure, concern persists that the free
amine could be liberated through common metabolic processes
such as amidolysis and N-dealkylation. It is therefore preferable
to ensure that only safe aromatic amines are embedded in
potential drugs.
INTRODUCTION
■
During the course of a recent drug discovery program aimed at
finding potential treatments for diabetes, a chemical series
containing an embedded 4-aminobiphenyl was found to have a
number of attractive properties. However, 4-aminobiphenyl
itself is a known human carcinogen and many related
compounds are found to be active in a bacterial mutagenicity
(Ames) test. A compound containing an Ames active aromatic
amine was not considered acceptable for this indication, and so
4-aminobiphenyls were sought that were not active. In recent
studies it had been found that the electrophilicity (computed
with quantum mechanics) of metabolites derived from aromatic
amines correlates with the likelihood of the aromatic amines
being active in the Ames test.1 The insight provided by these
calculations has permitted us to design, make, and assess the
first 4-aminobiphenyls that we are aware of that are not active
in the Ames test. More importantly, this has been done in a
rational fashion and has incorporated a view of the predicted
properties of the compound that would contain the 4-
aminobiphenyl. Hence, the physical and pharmaceutical
properties of target molecules were not sacrificed in order to
remove potential mutagenic metabolites. We believe that this
provides a new paradigm for designing safer compounds while
remaining within druglike chemical space.
The Ames test is part of the standard test battery of genetic
toxicity tests in the International Conference on Harmonisation
set of technical requirements for registration of pharmaceuticals
for human use and, therefore, is required by regulators
worldwide to support clinical trials and registration.3 In this
test,4−6 compounds are applied to strains of S. typhimurium or
E. coli that have been mutated so that they are unable to grow
in the absence of a particular exogenous amino acid. In the two-
strain test reported here, the TA98 and TA100 strains of S.
typhimurium require histidine to grow. Compounds that cause
particular mutations of these bacteria are able to cause reversion
to the state that is able to produce histidine. Hence, exposure of
these bacteria to mutagens is manifest as a colony that is able to
grow in a histidine free environment. In particular, TA98
detects frameshift mutations while TA100 detects base-pair
substitution mutations.
Many compounds, including most aromatic amines, are not
inherently mutagenic, but metabolism generates species that are
inherently mutagenic. This is achieved in the Ames test by
incorporating S9 fraction from the livers of rats in which
metabolic activity has been induced; in this study, this
employed Aroclor 1254.6,7 Aromatic amines are generally
thought to yield mutagens after N-hydroxylation often with
subsequent O-conjugation to create an electrophilic species
BACKGROUND
■
Many marketed and candidate drug molecules contain amino-
substituted aromatic substructures, occasionally with a free
amino group but most commonly where the amine has been
incorporated into amide or sulfonamide groups or into a
heterocyclic ring. Amino aromatics may give rise to safety
concerns, as many have been shown to be mutagenic in the
Ames test and therefore may be carcinogenic in animals.2 In the
Received: January 30, 2012
Published: April 4, 2012
© 2012 American Chemical Society
3923
dx.doi.org/10.1021/jm3001295 | J. Med. Chem. 2012, 55, 3923−3933

Journal of Medicinal Chemistry
Article
(such as an ester, sulfate, or phosphate) that is able to react
with DNA (Scheme 1).8−15
employing examples known from the literature or by exhaustive
profiling of putative aromatic amine metabolites.19 As a
consequence, the properties of the resulting compounds in
which the aromatic amines are embedded have often been
dictated by the tolerated aromatic amines. It would be
preferential to be able to design compounds whose risk of
embedding Ames positive aromatic amines has already been
mitigated. In this way, a more holistic view of the properties of
any designed compound can be taken. Researchers at Bristol-
Myers Squibb have disclosed how predictions from a QSAR
based model could be used to guide design, although it is
difficult to have such an approach suggest what compound to
make next (the inverse QSAR problem).20
Scheme 1. Cascade of Reactions That Is Postulated to Link
Aromatic Amines with DNA Damage and Mutation (X =
COR, SO2R, PO(R)2, etc.)
4-Amino-1,1′-biphenyl compounds give rise to particular
concern, as the parent molecule 1a (Figure 1) is an established
In 2009, Leach et al. described how the probability of an
aromatic amine being active in the Ames test could be
correlated with the dissociation energy for the process shown in
Scheme 2.1 Similar correlations have been found by
Scheme 2. Dissociation Process Studied by Quantum
Mechanical Calculations To Model Mutagenicity in the
Ames Test
Figure 1. Some 4-aminobiphenyls that have been described as causing
mutagenesis and related genetic toxicity and an MCH-1 compound
that was abandoned.
others.13,14,18,21,22 It is possible that Scheme 2 represents the
relevant rate-limiting step in the biochemical process that leads
to mutation in the assay conditions. It is equally possible that
this reaction energy simply provides a general view of the
electrophilicity of the assortment of metabolites that might be
formed from any given aromatic amine and that might react
through a range of mechanisms. In particular, the acetate ester
had been selected for computational convenience as a surrogate
for O-conjugates that are good leaving groups such as esters in
general, sulfates, and phosphates. Other calculations performed
by Leach et al. sought to model the liability toward cytochrome
P450 mediated oxidation and the reactivity of the nitroso
compound that results from further oxidation of the hydroxyl-
amine. Neither of these was found to be as reliable a guide to
activity in the Ames test as the reaction in Scheme 2. From the
perspective of compound design it is important to note that this
property has been found to correlate with activity in the Ames
test more significantly than any other individual descriptor and
that it is chemically understandable such that compounds can
be proposed in a rational way to modulate it. In this report we
disclose how this approach was applied to design nonmutagenic
(Ames negative) aminobiphenyls.
human carcinogen2 and 4-amino-2′,3-dimethyl-1,1′-biphenyl 1b
has been used extensively to induce a variety of cancers in
rodent models.16 Workers at Schering-Plough reported a series
of biarylurea MCH-1 antagonists in which one promising
compound 2 contained embedded the 4-amino-3′-cyano
derivative 1c.17 The free aniline 1c was found to be highly
mutagenic in the Ames test, so compound 2 was abandoned
even though in vivo drug metabolism studies showed no
evidence for liberation of 1c. It was decided that any risk of
aminobiphenyl exposure, regardless of extent, was unacceptable
in the context of antiobesity therapy. A number of other
aminobiphenyls were described in the Organon data set
disclosed by Bentzien et al. and illustrated in Figure 2.18 All
COMPOUND SYNTHESIS
■
Figure 2. Aminobiphenyls reported by Bentzien et al. to be active in
Compounds that were evaluated in the Ames test (Table 2)
were prepared by Suzuki−Miyaura coupling (Schemes 3−5)
using microwave irradiation. The boronic acids described by the
generic structure in Scheme 3 were in several cases unavailable
or gave poor results in the protocol described. In these cases
they were substituted by commercial trifluoroborate salts
(Scheme 5) or boron pinacol esters synthesized from the
corresponding aryl halides (Scheme 4). These reagents have
been shown to slowly release the free boronic acid into the
reaction mixture, thereby minimizing its degradation.23 Both
aniline and nitrobenzene starting materials were utilized with,
in the latter case, hydrogenation of the crude nitrobiphenyl,
yielding the aminobiphenyl product.
an Ames test.18
of these compounds were described as positive in the Ames
test, suggesting that within this particular chemical class it is
particularly challenging to avoid mutagenicity; indeed no
aminobiphenyls were known to us that were not positive in
an Ames test.
In previous diabetes programs at AstraZeneca in which
compounds with embedded aromatic amines have caused
concern, the approach taken to remove the risk of mutagenicity
has been to exchange the aromatic amine for others that are
known to not cause a response in the Ames test, either
3924
dx.doi.org/10.1021/jm3001295 | J. Med. Chem. 2012, 55, 3923−3933

Journal of Medicinal Chemistry
Article
1
Scheme 3. Synthesis of Compounds Using Methods A and
B
P(active) =
a
(0.056Ethermal−8.012)
(1)
1 + e
In eq 1, P(active) is the proportion of compounds that are
active and is equated here to the probability of a new
compound being active and Ethermal is the energy change
including thermal corrections (for 298 K) computed for the
reaction given in Scheme 2 by B3LYP/6-31G*.27−30 The
quantity P(active) has two roles: when used retrospectively to
describe compounds whose activity is known, it indicates the
proportion of compounds with equivalent Ethermal values that
would be active and hence can indicate whether the
compound’s activity is surprising or not; when used
prospectively, it is equated to the probability that a compound
that has not yet been made and tested will be found to be active
in the Ames test.
The key observation is that aromatic amines with a higher
value of Ethermal will have lower probability of being active. At no
point does that probability drop to 0% or reach 100%. Indeed,
as the overlapping distributions in the bar chart in Figure 3a
show, there are active and inactive aromatic amines represented
across most of the Ethermal range. In addition to any inaccuracy
in the calculations or issues with how relevant gas phase
calculations might be, there remain many other factors that
determine whether an aromatic amine is active in the Ames test.
Aromatic amines can be inactive in the test even when the
computed electrophilicty is high for several reasons including
the following: they may not be soluble; they may not be
metabolized or may be metabolized in a fashion that does not
generate mutagens; the activated metabolite may not be stable
or permeable enough when activated to reach the DNA.
Similarly, aromatic amines that are computed to not be
electrophilic enough to be active may nonetheless be active in
the Ames test by using a different mechanism such as reacting
via a different metabolite or by intercalating with DNA. Such
factors confound, in part at least, analysis of how well the
relevant electrophilic reactivity is computed. Despite these
complexities, it is worth recalling the notion enunciated by Box
that “all models are wrong, but some are useful”.31 It is hoped
to highlight the utility of this model in the remainder of the
text.
a
Reagents and conditions: Method A: PdCl2(dppf)−CH2Cl2, K3PO4,
DME, EtOH, H2O. Method B: premixed PdCl2(dtbpf)/K3PO4,
MeCN, H2O.
RESULTS AND DISCUSSION
■
During the course of a drug discovery project aimed at the
production of an antidiabetic compound, optimization was
performed on a series that contained an embedded 4-
aminobiphenyl. Naturally, this prompted concerns about the
potential mutagenicity of such compounds should they be
liberated. As genotoxic compounds are subject to very tight
control, arguments about the possibility or otherwise of these
embedded anilines being released are unlikely to be relevant in
a program aimed at chronic medication for diabetics.3,17,19 The
approach taken was therefore to identify 4-aminobiphenyls that
are Ames negative.
However, the potency and pharmaceutical properties of the
elaborated compounds containing those anilines were also of
concern. During earlier phases of the project, a significant
number of substitution patterns had been proposed as
alternatives to the simple biphenyl group and all of these had
been tracked in an in-house database as part of our Web-based
collaborative design tools.24 The various structures that had
been proposed (for reasons unrelated to mutagenicity) which
ought to include many with relatively attractive properties were
extracted from this database and transformed into a virtual
database of 327 aromatic amines.
As described by Leach et al., the higher the computed energy
change is for the reaction in Scheme 2, the lower is the
proportion of aromatic amines that are active in the Ames test.1
As shown in Figure 3a for 308 of the 312 aromatic amines
described by Leach et al., the dissociation energy changes for
the reaction in Scheme 2 tend to be lower for compounds that
are found to be active in an Ames test (shown as red bars)
compared to those found to be inactive (shown as green
bars).25 In Figure 3b these individual bars are plotted as ratios
of active to inactive compounds in bins that are 10 kcal/mol
wide (the point at X kcal/mol is the ratio of active to inactive
for all aromatic amines with a computed energy change in the
range X − 5 kcal/mol to X + 5 kcal/mol). The statistical
software package JMP is able to fit a continuous function
linking the energy change to this ratio, and this is given as eq 1
and plotted in Figure 3b.26
A set of 142 4-aminobiaryls that were found in the design
database were subject to QM calculations. The details are
provided in the section Computational Methods. The range of
probabilities of being active that were obtained highlighted the
challenge that would be faced (Figure 4). Whereas in general it
is observed that approximately 35% of all aromatic amines that
have been tested are active in the AstraZeneca Ames test, the
mean probability of being active in the Ames test computed for
these anilines is 52%.
a
Scheme 4. Synthesis of Compounds Using Method C
a
Reagents and conditions: (i) PdCl2(dppf)−CH2Cl2, CH3COOK, bis(pinacolato)diboron, dioxane; (ii) PdCl2(dppf)−CH2Cl2, K3PO4, DME,
EtOH, H2O.
3925
dx.doi.org/10.1021/jm3001295 | J. Med. Chem. 2012, 55, 3923−3933

Journal of Medicinal Chemistry
Article
a
Scheme 5. Synthesis of Compounds Using Method D
a
Reagents and conditions: (i) PdCl2(dppf)−CH2Cl2, K3PO4, DME, EtOH, H2O; (ii) 10% Pd on C, H2, EtOAc or EtOH.
Figure 5. Computed properties (molecular weight at the top and
clogP at the bottom) for a range of compounds with embedded
anilines are plotted against the P(active) for the embedded aniline.
Figure 3. (a) Distribution of dissociation energies including thermal
corrections for the process in Scheme 2 for 308 compounds found to
be active (red bars) or inactive (green bars) in the AstraZeneca Ames
test. (b) Proportion of the same 308 compounds in bins 10 kcal/mol
wide that are active in the Ames test at different values of Ethermal are
shown as red points, and the curve fitted through the unbinned data
set is shown in blue.
clogP less than 3, thought to be reasonable upper limits for
these properties for this particular project. Selecting those
anilines with a probability of less than 50% of being active in
the Ames test and satisfying these two property filters suggest
that 26 of the 142 potential compounds are most likely to be of
interest.
The initial output as summarized in Figures 4 and 5 provided
more than just suggestions of useful compounds. It suggested
some SAR that was used to develop other candidates for
synthesis. For instance, the point corresponding to compound
3 is circled in the plots in Figure 5. This aniline (Figure 6) has
an attractively low probability of being active in the Ames test,
but the elaborated compound resulting from it was computed
to have a lipophilicity that is higher than desired. Therefore, the
chlorine atom that was not believed to be critical for primary
potency was deleted; compound 3 was not selected for
Figure 4. Distribution of probability of being active, P(active), found
for a set of aminobiphenyls of interest to an antidiabetes drug
discovery project.
In terms of compound design, the power of being able to
compute this probability is that it can now be considered
alongside other properties of the final elaborated molecule. In
Figure 5, the molecular weight and clogP of each compound
with an embedded aniline are plotted against the probability of
being active in the Ames test for the aniline. In blue are the
compounds that have a molecular weight less than 500 and
Figure 6. An aminobiphenyl (3) highlighted by a circle in Figure 5.
Compound 4 was selected as being a less lipophilic example that
retains a low probability of being active in the Ames test.
3926
dx.doi.org/10.1021/jm3001295 | J. Med. Chem. 2012, 55, 3923−3933

Journal of Medicinal Chemistry
Article
synthesis but was replaced by compound 4 which was
computed to have a probability of being active of 32%.
Examining the data set in more detail highlighted some
interesting SAR among the various difluoro substituted anilines
that were included (Figure 7). While the physical properties of
to be active in the Ames test), and the SAR outlined above
suggested strongly that 11 would have a decreased chance of
being active in the Ames test. Two further anilines that were
included in compounds of interest to the project, 12 and 13,
were also modified in this way to give 14 and 15. The
compounds containing 12 and 13 were accessed by a route that
did not go through the aromatic amine, and so neither of them
were made or tested in the Ames test. The selected compounds,
11, 14, and 15, were computed to have probabilities of being
active of 41%, 44%, and 30%, respectively (lower than their
unmodified relatives 10, 12, and 13).
Having selected compounds 4, 11, 14, and 15 for synthesis,
these were prepared and tested in a two-strain Ames test.
Pleasingly, three of these four anilines were found to be inactive
in the Ames test. Compound 14 was the only one that was
active. Considering the precedent for Ames activity in
aminobiphenyl compounds, we considered this a significant
achievement.
Further exploration of the SAR for these compounds probed
the effect of a range of small substituents (F, Cl, Me, CN,
CONHMe, NHAc, OMe, and SO2Me) at all of the positions
around the biphenyl core either alone or in combination with
the meta fluorine highlighted already. In Table 1, the value of
the change in Ethermal caused by addition of these substituents at
each position is given compared to the parent 4-aminobiphenyl,
which has Ethermal of 138.8 kcal/mol. These changes are color
coded, with those that increase this quantity shaded green and
those that decrease this property shaded red. The change
caused by the combination of each group with the meta fluorine
is also provided and shows that this group generally increases
Figure 7. Examples of difluoro substituted 4-aminobiphenyls that were
included in the initial enumeration and that suggested some key SARs.
Beneath each structure is their computed probability of being active.
this set of compounds are very similar, the probability of being
active in the Ames test varies significantly and shows a
systematic advantage of having fluorine meta to the aniline
group. Comparing 8 with 5−7 shows that a fluorine at the meta
position on the same ring as the aniline group is more beneficial
than a fluorine at any of the positions on the distal ring. Among
the positions on that distal ring, the meta position also is the
best place to put fluorine as shown by 7 compared to 5 and 6.
Meanwhile, having a second fluorine in a meta position on the
same ring as the aniline, as in 9, enhances the effect even
further.
In light of this information, it was decided to modify the
aniline 10, which was at that point the one embedded in one of
the preferred compounds for primary potency, by the addition
of a fluorine to give aniline 11 (Figure 8). Aniline 10 had been
computed to have an Ethermal of 143.6 kcal/mol corresponding
to 49% probability of being active (and was subsequently found
E
thermal by about 4.9 kcal/mol no matter which other groups it is
combined with. An inspection of the curve in Figure 3b shows
that, particularly in the region of the inflection point at 143
kcal/mol, a change of this size could cause a radical decrease in
the likelihood of a compound being active in the Ames test.
The various contributions given in Table 1 generally
correlate well with the σ values provided by Hansch et al.,
and plots of these correlations are provided in the Supporting
Information. It is of note that the correlation is better with σp
than with σm for groups at R2 but not for R3′, which suggests
that R3′ behaves more like a “normal” meta position than R2
where substituents will also have an effect on the conformation
of the biphenyl bond. Studies in the wider set of anilines have
found that σ values have a value within specific series but do
not provide the generality that is achieved with the quantum
mechanical calculations. One significant outlier in the
correlations with σp is CONHMe at the R2′ position. The
geometry optimization of the nitrenium species for this yielded
the spirocyclic cation 16. It remains untested whether this effect
is of experimental relevance.
To help to ground this SAR in experiment, it was decided to
make and test the set of compounds in which both F and Cl are
introduced at each position around the biphenyl framework. As
both substituents are computed to be beneficial at all points, it
was hoped that a number of simple 4-aminobiphenyls might be
Figure 8. Anilines 10, 12, and 13 were of particular interest to a drug
discovery project and were modified in light of computed SAR to 11,
14, and 15 which are less likely to be active in the Ames test.
3927
dx.doi.org/10.1021/jm3001295 | J. Med. Chem. 2012, 55, 3923−3933

Journal of Medicinal Chemistry
Article
Table 1. Changes in Ethermal in kcal/mol Compared to Parent 4-Aminobiphenyl Caused by Addition of (a) Substituents at
Various Positions (Columns 2−6) and (b) the Same Comparisons for the Dual Effect of Adding the Substituent along with a
a
Fluorine at the Meta Position Shown
a
Cells shaded green are ones where Ethermal increases, and those shaded red are ones where Ethermal decreases.
Figure 9. 4-Aminobiphenyl compounds in which F and Cl are systematically placed at each position that were made and tested. Beneath each
structure is their computed probability of being active.
identified that are not active in the Ames test. These
compounds and their computed probabilities of being active
in the Ames test are shown in Figure 9. Most have a computed
probability of about 50% of being active in the Ames test. The
most likely to be active are the fluoro substituted examples,
particularly 17 and 24. The least likely to be active is the chloro
substituted 20. When these compounds were tested exper-
imentally, it was surprising that all of them except for 20 were
active in the Ames test. This suggests that the computational
method does a good job of ranking potential candidates but
also hints that the biphenyl architecture of these compounds
may make them even more likely to be active than an average
aromatic amine with equivalent Ethermal
.
This SAR suggested that all of the singly substituted
aminobiphenyls that had been of interest to the drug discovery
project are Ames positive. However, the beneficial effect of the
meta-fluoro had been established experimentally by comparing
11 and 10. A range of other anilines were therefore selected for
synthesis, all including this feature and other substitution
patterns computed to be beneficial. This set included
compound 8 shown in Figure 7. These were selected both
opportunistically based on availability of reagents and on the
grounds of SAR relating to the primary target. The structures of
this set of compounds are shown in Figure 10, along with their
computed probability of being active, all of which are low for
Figure 10. A set of 4-aminobiphenyls that were selected for synthesis.
Beneath each structure is their computed probability of being active.
this class. These were also all made and tested along with
compound 6, which provides a comparator for compound 26 to
probe the effect of the meta-fluoro group. All of the compounds
26−29 were found not to be active in the Ames test, while by
contrast compound 6, which is the same as 26 but lacking the
meta-fluoro, was found to be active in the Ames test, further
emphasizing the beneficial effect of this fluorine substituent.
During the course of these investigations, 20 anilines had
been made, purity checked, and tested in a two-strain Ames
3928
dx.doi.org/10.1021/jm3001295 | J. Med. Chem. 2012, 55, 3923−3933

Journal of Medicinal Chemistry
Article
Table 2. Summary Results of the Two-Strain Ames Test for the Aminobiphenyl Compounds Designed, Made, and Tested As
Described
a
The maximum increase over control is given along with the corresponding compound quantity in μg/plate.
test, the results of which are summarized in Table 2.32 The
details of each experiment are provided as Supporting
Information. These data reveal that in all cases, anilines in
this set are inactive in the absence of the S9 fraction, indicating
that all of them require metabolic activation to exert their
mutagenic effect. All examples, apart from compound 14, have
the same result (either active or inactive) in both the TA98 and
TA100 strains. Compound 14 shows a weak effect in the TA98
strain and is negative in the TA100 strain. In general this set of
anilines have a larger numerical effect in the TA98 strain than in
TA100.
One question that arises concerns the origin of both the
elevated risk with aminobiphenyls and the reduced risk caused
by the meta-fluorine atom that are reflected in changes in
3929
dx.doi.org/10.1021/jm3001295 | J. Med. Chem. 2012, 55, 3923−3933

Journal of Medicinal Chemistry
Article
Figure 11. Structures optimized with B3LYP/6-31G* for the ArNHOAc and ArNH+ species involved in Scheme 2. 30a,b correspond to the
derivatives of 4-aminobiphenyl. 31a,b correspond to 2-fluoro-4-aminobiphenyl, and 32a,b correspond to 2-chloro-4-aminobiphenyl. Below the
structures are shown the profiles for rotation about the biphenyl bond in each structure with blue curves for ArNHOAc and red curves for the
nitrenium ion. The x-axis is the inter-ring dihedral, and the y-axis is energy in kcal/mol relative to the lowest energy point.
Ethermal. There are a number of factors that may be important.
The σp of phenyl is −0.01 and the σm of fluoro is 0.34,
suggesting that phenyl at the para position is weakly electron
donating which would stabilize a nitrenium or partially positive
charged equivalent while fluorine at the meta position is
electron withdrawing and would have the opposite effect.33 It is
worth noting that the effect of fluorine at the para position
would be much diminished (it has a σp of 0.06), suggesting that
the innate electronegativity of fluorine, which will tend to
withdraw electrons in the plane of the aromatic ring, is
counteracted by back-donation into the π-system at the para
(and presumably ortho) position.33
One property of the aminobiphenyls that is slightly different
from the general set of simple aromatic amines is that they have
a higher lipophilicity. All of the 4-aminobiphenyls listed in
Table 2 have clogP between 3 and 4 except for 14. One effect
of this is that these anilines and many of their metabolites will
have a combination of molecular weight and lipophilicity that
permits them to readily permeate the range of membranes that
would otherwise prevent them from getting from the site at
which metabolism takes place to the site at which the damage
to DNA takes place.34,34,35 The higher lipophilicity of these
compounds will also tend to make them more readily
metabolized. A final property that is slightly different for this
set compared to aromatic amines in general is that substitution
around the biphenyl bond can have a conformational effect that
might change a substituent’s ability to modulate the electro-
philicity. Quantum mechanical calculations can provide insight
into such phenomena. Two sets of calculations have been
performed, one in which the various biphenyl structures are
constrained to be flat and the second in which the energy
profile for rotation about the biphenyl bond is examined.36
When both the ArNHOAc species and the nitrenium ions
are constrained to have a dihedral angle of 0° between the two
phenyl rings, the dissociation energy corresponding to Scheme
2 goes down by 1.7 kcal/mol (from 142.7 to 141.0 kcal/mol)
for the species with no substituents, 30a dissociating to 30b,
suggesting that the flat conformation benefits more from the
electron donating effect of the phenyl ring than the twisted one.
The structures of the lowest energy conformations of the two
key species 30a and 30b are shown in Figure 11 and reveal that
there tends to be a flattening of the dihedral in the nitrenium
ion. The ability to delocalize the cation more effectively
throughout the framework in a flatter structure is in direct
opposition to the steric clashing around the bond linking the
two phenyl rings.
For the corresponding species with a meta-fluoro substituent
(31a and 31b), the dissociation energy goes down by only 1.4
kcal/mol (from 147.6 to 146.2 kcal/mol) when the inter-ring
dihedral is constrained to be 0°, suggesting that the fluorine
actually reduces the effect of the twist that is present in even the
unsubstituted case. This effect also manifests itself in the
geometry of the nitrenium species: for the unsubstituted
example, 30b, the bond linking the two phenyl rings has a
length of 1.441 Å but is 0.003 Å shorter when a fluorine is
added to give 31b, although this shorter distance is offset
slightly by an increased dihedral angle between the rings. By
contrast, with a meta-chloro substituent, the dissociation energy
goes down by 2.7 kcal/mol (from 150.4 to 147.7 kcal/mol)
when all species are constrained to be flat, suggesting that the
extra twist induced by the chlorine and manifest in significantly
larger dihedral angles in 32a and 32b makes a significant
contribution to the decreased probability of being active for this
particular substitution pattern in the aminobiphenyls.
Changes in lipophilicity and the degree of twisting in the
aniline could also contribute to changes in the liability of the
compounds toward metabolism, and so it is interesting to note
that neither property is able to provide any further distinction
among active and inactive compounds. The degree of twisting
in either the reactant ArNHOAc or the product nitrenium
correlates reasonably well with Ethermal, but neither is able to
rationalize the differences between active and inactive
compounds in the Ames test as effectively as Ethermal. In
particular, compounds 10 and 14 are both computed to have
3930
dx.doi.org/10.1021/jm3001295 | J. Med. Chem. 2012, 55, 3923−3933

Journal of Medicinal Chemistry
Article
NMR spectra were recorded on a Bruker Ultrashield 400 Plus at 400
MHz in the indicated deuterated solvent. Chemical shifts are reported
in ppm relative to tetramethylsilane (TMS) (0.00 ppm) or solvent
peaks as the internal reference, and coupling constant (J) values are
reported in hertz (Hz). Merck precoated thin layer chromatography
(TLC) plates were used for TLC analysis. For workup, solutions were
dried over anhydrous magnesium sulfate and the solvent was removed
by rotary evaporation under reduced pressure. Microwave reactions
were carried out in a Biotage initiator microwave reactor. General
methods by which all compounds were prepared are described for
exemplar compounds. Full experimental details for all other
compounds are provided in the Supporting Information.
significant twisting in both species but are active in the Ames
test. It is also of interest that adding a fluorine meta to an
aromatic amine is not sufficient to make it inactive either in 4-
aminobiphenyls (for example, compound 19 is active in the
Ames test) or more generally.
Within the compound set described here, all 4-amino-
biphenyls with Ethermal above 145 kcal/mol are inactive in the
Ames test while those with Ethermal lower than this cutoff are
active with one exception. This is compound 14, which has a
molecular weight of 329. It was noted by McCarren et al. that
the specificity of this kind of prediction is diminished for
aromatic amines that have a molecular weight greater than
250;22 compound 14 is the only one in this set with molecular
weight in this range. Although Ethermal discriminates very well
between active and inactive aminobiphenyls, this does not
preclude other electrophilic metabolites such as the arylnitroso
deriving from further oxidation from being relevant. Calcu-
lations in which methylamine was used as a model nucleophile
adding to ArNO to give either ArN(OH)NHMe or ArN
NMe were performed to examine if they were also capable of
achieving this discrimination and were found not to be able to
do so (computed energy differences are in the Supporting
Information).
General Method A: 2,2′-Difluorobiphenyl-4-amine (8). A
mixture of 2-fluorophenylboronic acid (2.06 g, 14.74 mmol), 4-bromo-
3-fluoroaniline (2.00 g, 10.53 mmol), PdCl2(dppf)−CH2Cl2 adduct
(0.430 g, 0.53 mmol), and tripotassium phosphate (2.68 g, 12.63
mmol) in DME (10 mL), ethanol (5 mL), and water (2.5 mL) was
heated to 110 °C for 60 min by microwave irradiation. The suspension
was cooled, filtered through Celite, and evaporated to dryness. The
residue was diluted with 2-methyltetrahydrofuran (60 mL), washed
sequentially with water (30 mL) and saturated brine (30 mL), dried
over MgSO4, and concentrated to afford crude product. This was
purified by chromatography (EtOAc/isohexane = 5−20%) and then
by preparative basic HPLC, eluting with 5−95% water−acetonitrile
(+0.5% NH3) to provide the title compound (0.178 g, 8%) as a white
1
solid. H NMR (CDCl3) 3.83 (s, 2H), 6.50 (ddd, J = 2.3, 10.0, 14.0
Hz, 2H), 7.09−7.2 (m, 3H), 7.26−7.38 (m, 2H). MS m/z: (M + H)+
= 206.08. HPLC purity = 100%.
CONCLUSIONS
■
A computed probability of being active in the Ames test can
help design safer aromatic amines, even in challenging chemical
series such as 4-aminobiphenyls. These calculations can be
combined with property calculations for the compounds that
contain the embedded aromatic amines to make an improved
design process. These calculations can be used to obtain SAR
that can suggest safer aromatic amines to make. In particular,
adding fluorine or chlorine meta to the aromatic amine has
been found to reduce the risk of activity in the Ames test. In the
aminobiphenyls, reduced Ames risk as reflected in higher
Ethermal values can be attributed to the blend of electron
donating and withdrawing properties of substituents and their
conformational effect. In particular, the meta-fluorine sub-
stituent flanking a para-phenyl group has an electron
withdrawing effect that is reduced by conformational effects
whereas for an equivalent chlorine substituent the two effects
complement one another. By providing an idea of the
likelihood of aromatic amines being active in the Ames test,
this approach allows a potential liability to be avoided in a
rational way. As a consequence, this property can be considered
alongside other computed properties when designing molecules
to enable an improved holistic approach to drug design.
General Method B: 4′-Fluorobiphenyl-4-amine (24). A
mixture of 4-iodoaniline (0.750 g, 3.42 mmol), 4-fluorophenylboronic
acid (0.623 g, 4.45 mmol), and premixed 1,1-bis(di-tert-
butylphosphino)ferrocenepalladium dichloride/tripotassium phos-
phate (5 mol % catalyst and 200 mol % base) (1.57 g) in acetonitrile
(12 mL) and water (4 mL) was heated to 100 °C for 3 h by microwave
irradiation. The suspension was cooled, filtered through Celite, and
evaporated to dryness. The residue was diluted with EtOAc (100 mL)
and washed sequentially with water (100 mL) and saturated brine
(100 mL), dried over MgSO4, and concentrated to afford crude
product. The crude product was purified by chromatography (EtOAc/
isohexane = 10−70%) and then by preparative basic HPLC, eluting
with 5−95% water−acetonitrile (+0.5% NH3) to provide the title
compound (0.435 g, 67%) as a white solid. 1H NMR (DMSO-d6) 5.18
(s, 2H), 6.58−6.66 (m, 2H), 7.12−7.21 (m, 2H), 7.25−7.34 (m, 2H),
7.48−7.58 (m, 2H). MS m/z: (MH)+ = 188.28. HPLC purity = 100%.
General Method C(i): 3-Fluoro-4-(4,4,5,5-tetramethyl-1,3,2-
dioxaborolan-2-yl)aniline. A mixture of 4-bromo-3-fluoroaniline
(5.00 g, 26.31 mmol), potassium acetate (7.75 g, 78.94 mmol),
PdCl2(dppf)−CH2Cl2 adduct (1.50 g, 1.84 mmol), and bis-
(pinacolato)diboron (8.02 g, 31.58 mmol) in dioxane (100 mL),
ethanol (5 mL), and water (2.5 mL) was heated at 80 °C for 24 h.
This was then evaporated to dryness, dissolved in EtOAC (400 mL)
and water (400 mL), and filtered through Celite. The organic layer was
separated, dried over MgSO4, and concentrated to afford crude
product which was purified by chromatography (CH2Cl2) to provide
the title compound (2.80 g, 45%) as a yellow solid. 1H NMR (CDCl3)
1.33 (s, 12H), 3.93 (s, 2H), 6.30 (dd, J = 2.1, 11.3 Hz, 1H), 6.41 (dd, J
= 2.1, 8.1 Hz, 1H), 7.51 (dd, J = 7.0, 8.0 Hz, 1H). MS m/z: (M + H)+
= 238.45.
EXPERIMENTAL METHODS
■
All solvents and chemicals used were reagent grade. Anhydrous
solvents were purchased from Sigma Aldrich. Flash column
chromatography was carried out using Redisep or Crawford prepacked
silica cartridges (4−330 g) and elution was with an Isco Companion
system. Following isolation, compounds were purified to >95% purity
(UV and NMR) by silica gel chromatography and reverse phase
preparative high performance liquid chromatography (HPLC)
purification carried out using a Waters XBridge Prep C18 OBD
column, 5 μm silica, 50 mm diameter, 150 mm length.
Method C(ii): 4′-Amino-4-chloro-2′-fluorobiphenyl-2-car-
bonitrile (28). General method A was followed with 3-fluoro-4-
(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)aniline (0.750 g, 3.16
mmol) in place of a boronic acid and 2-bromo-5-chlorobenzonitrile
(0.685 g, 3.16 mmol). The crude product was purified by
chromatography (EtOAc/isohexane = 5−20%) to provide the title
compound (0.118 g, 15%) as a white solid. 1H NMR (CDCl3) 3.95 (s,
2H), 6.51 (ddd, J = 2.2, 10.1, 14.2 Hz, 2H), 7.17 (t, J = 8.4, 8.4 Hz,
1H), 7.41 (dd, J = 1.3, 8.5 Hz, 1H), 7.57 (dd, J = 2.3, 8.5 Hz, 1H), 7.70
(d, J = 2.2 Hz, 1H). MS m/z: (M + H)+ = 247.21. HPLC purity =
100%.
Mass spectrometry data were recorded by liquid chromatography−
mass spectrometry (LCMS) on a Waters 2790 separations module
with a Phenomenex 5 μm C18 50 mm × 2 mm column, a Waters 996
photodiode array detector, and Waters Micromass ZQ mass
spectrometer, with detection by UV at 254 nm. Purity was >95% for
all test compounds as determined by this HPLC/MS method. 1H
3931
dx.doi.org/10.1021/jm3001295 | J. Med. Chem. 2012, 55, 3923−3933

Journal of Medicinal Chemistry
Article
General Method D(i): 2,2′,6′-Trifluorobiphenyl-4-amine (26).
1-Bromo-2-fluoro-4-nitrobenzene (1.00 g, 4.55 mmol), potassium
(2,6-difluorophenyl)trifluoroborate (1.50 g, 6.82 mmol),
PdCl2(dppf)−CH2Cl2 adduct (0.186 g, 0.23 mmol), and tripotassium
phosphate (1.16 g, 5.45 mmol) were suspended in DME (16 mL),
ethanol (8 mL), and water (4 mL) and degassed under vacuum, and
the atmosphere was replaced with nitrogen. The mixture was heated to
110 °C for 2.5 h in the microwave reactor and cooled to room
temperature. The reaction mixture was evaporated to dryness,
redissolved in EtOAc (100 mL), and washed with water (100 mL)
and saturated brine (100 mL). The organic layer was filtered through
Celite, dried over MgSO4, filtered, and evaporated. The crude product
was purified by chromatography (0−15% EtOAc in isohexane) to
afford 2,2′,6′-trifluoro-4-nitrobiphenyl (0.600 g, 52.1%) as a beige solid.
General Method D(ii). Palladium (10%) on carbon (37.8 mg, 0.04
mmol) was added to 2,2′,6′-trifluoro-4-nitrobiphenyl (0.600 g, 2.37
mmol) in ethanol (100 mL) under nitrogen. The reaction flask was
evacuated and purged 3 times, and the suspension was stirred under an
atmosphere of hydrogen at room temperature for 20 h and then
filtered through Celite and evaporated. It was then purified by
chromatography (5−30% EtOAc in isohexane) and then HPLC,
eluting with 5−95% water−acetonitrile (+0.5% NH3) to afford 2,2′,6′-
ACKNOWLEDGMENTS
■
The authors are grateful to Mike O’Donovan for careful reading
of and improvements to the manuscript.
ABBREVIATIONS USED
■
SAR, structure−activity relationship; DME, 1,2-dimethoxy-
ethane; DMSO, dimethylsulfoxide; dppf, 1,1′-bis-
(diphenylphosphino)ferrocene; dtbpf, 1,1′-bis(di-tert-
butylphosphino)ferrocene; EtOAc, ethyl acetate; DCM, di-
chloromethane; rt, room temperature
REFERENCES
■
(1) Leach, A. G.; Cann, R.; Tomasi, S. Reaction energies computed
with density functional theory correspond with a whole organism
effect; modelling the Ames test for mutagenicity. Chem. Commun.
(Cambridge, U. K.) 2009, 1094−1096.
(2) Some Aromatic Amines, Organic Dyes, and Related Exposures; IARC
Monographs on the Evaluation of Carcinogenic Risks to Humans, Vol.
99; International Agency for Research on Cancer: Lyon, France, 2010;
pp 1−658.
1
trifluorobiphenyl-4-amine (361 mg, 68.2%) as a white solid. H NMR
(3) ICH Expert Working Group. Guidance on Genotoxicity Testing
and Data Interpretation for Pharmaceuticals Intended for Human Use.
International Conference on Harmonisation of Technical Require-
ments for Registration of Pharmaceuticals for Human Use, ICH
S2(R1); ICH: Geneva, 2011.
(4) Ames, B. N.; Whitfield, H. J., Jr. Frameshift mutagenesis in
Salmonella. Cold Spring Harbor Symp. Quant. Biol. 1966, 31, 221−225.
(5) Ames, B. N.; Lee, F. D.; Durston, W. E. Improved bacterial test
system for the detection and classification of mutagens and
carcinogens. Proc. Nat. Acad. Sci. U.S.A. 1973, 70, 782−786.
(6) Mortelmans, K.; Zeiger, E. The Ames Salmonella/microsome
mutagenicity assay. Mutat. Res. 2000, 455, 29−60.
(7) Easterbrook, J.; Fackett, D.; Li, A. P. A comparison of Aroclor
1254-induced and uninduced rat liver microsomes to human liver
microsomes in phenytoin O-deethylation, coumarin 7-hydroxylation,
tolbutamide 4-hydroxylation, S-mephenytoin 4′-hydroxylation, chlor-
oxazone 6-hydroxylation and testosterone 6β-hydroxylation. Chem.-
Biol. Interact 2001, 134, 243−249.
(8) Beland, F. A.; Kadlubar, F. F. Formation and persistence of
arylamine DNA adducts in vivo. Environ. Health Perspect. 1985, 62,
19−30.
(9) Kadlubar, F. F.; Beland, F. A. Chemical properties of ultimate
carcinogenic metabolites of arylamines and arylamides. ACS Symp. Ser.
1985, 283, 341−370.
(10) Wild, D. A novel pathway to the ultimate mutagens of aromatic
amino and nitro compounds. Environ. Health Perspect. 1990, 88, 27−
31.
(11) Turesky, R. J.; Guengerich, F. P.; Guillouzo, A.; Langouet, S.
Metabolism of heterocyclic aromatic amines by human hepatocytes
and cytochrome P4501A2. Mutat. Res., Fundam. Mol. Mech. Mutagen.
2002, 506−507, 187−195.
(12) Novak, M.; Rajagopal, S. Correlations of nitrenium ion
selectivities with quantitative mutagenicity and carcinogenicity of the
corresponding amines. Chem. Res. Toxicol. 2002, 15, 1495−1503.
(13) Borosky, G. L. Ultimate carcinogenic metabolites from aromatic
and heterocyclic aromatic amines: a computational study in relation to
their mutagenic potency. Chem. Res. Toxicol. 2007, 20, 171−180.
(14) Borosky, G. L. Carcinogenic carbocyclic and heterocyclic
aromatic amines: a DFT study concerning their mutagenic potency. J.
Mol. Graphics Modell. 2008, 27, 459−465.
(DMSO-d6) 5.64 (s, 2H), 6.45 (ddd, J = 2.2, 10.5, 14.8 Hz, 2H), 7.02
(t, J = 8.4, 8.4 Hz, 1H), 7.15 (t, J = 8.1, 8.1 Hz, 2H), 7.42 (tt, J = 6.6,
6.6, 8.2, 8.2 Hz, 1H). MS m/z (M − H)− = 222.36. HPLC purity =
100%.
Ames Test. Standard plate incorporation assays were performed
using Salmonella typhimurium strains TA98 and TA100 according to
published methods.6,37 Tests were performed with amounts up to a
maximum of 5000 μg/plate in the absence and presence of S9 from
the livers of rats pretreated with Aroclor 1254 (purchased from Moltox
Inc., Boone, NC, U.S.). All test compounds were dissolved in
dimethylsulfoxide (DMSO). In all tests, there were three plates for the
solvent control, for each level of test compound, and for positive
control groups. The S9-independent positive controls were the
following: for TA100, sodium azide, 0.5 μg/plate; for TA98, 2-
nitrofluorene. The positive control in tests with S9 was 2-amino-
anthracene at 2 μg/plate for both strains. All the compounds were
tested at AstraZeneca, Alderley Park, U.K.
Computational Methods. All calculations were performed in
Gaussian 09 and employed the B3LYP functional and 6-31G* basis
set.27−30 Starting geometries were generated from SMILES strings by
CORINA.38 Two conformations about the Ar−N bond were
generated for each species. All optimized geometries were subjected
to frequency calculations to verify them as minima. Full coordinates
and energies for each species are provided in the Supporting
Information.
ASSOCIATED CONTENT
■
S
* Supporting Information
Full experimental details for Ames tests summarized in Table 2,
dependence upon Hammett parameters of Ethermal, experimental
details for other compounds, geometries and energies from QM
calculations, and dihedral angles and nitroso energy changes.
This material is available free of charge via the Internet at
http://pubs.acs.org.
AUTHOR INFORMATION
■
Corresponding Author
(15) Himmelstein, M. W.; Boogaard, P. J.; Cadet, J.; Farmer, P. B.;
Kim, J. H.; Martin, E. A.; Persaud, R.; Shuker, D. E. G. Creating
context for the use of DNA adduct data in cancer risk assessment: II.
Overview of methods of identification and quantitation of DNA
damage. Crit. Rev. Toxicol. 2009, 39, 679−694.
*Phone: +44 1625 231853. E-mail: andrew.leach@astrazeneca.
com.
Notes
The authors declare no competing financial interest.
3932
dx.doi.org/10.1021/jm3001295 | J. Med. Chem. 2012, 55, 3923−3933

Journal of Medicinal Chemistry
Article
(16) Siraki, A. G.; Chan, T. S.; Galati, G.; Teng, S.; O’Brien, P. J. N-
Oxidation of aromatic amines by intracellular oxidases. Drug Metab.
Rev. 2002, 34, 549−564.
(17) McBriar, M. D.; Guzik, H.; Shapiro, S.; Paruchova, J.; Xu, R.;
Palani, A.; Clader, J. W.; Cox, K.; Greenlee, W. J.; Hawes, B. E.;
Kowalski, T. J.; O’Neill, K.; Spar, B. D.; Weig, B.; Weston, D. J.; Farley,
C.; Cook, J. Discovery of orally efficacious melanin-concentrating
hormone receptor-1 antagonists as antiobesity agents. Synthesis, SAR,
and biological evaluation of bicyclo[3.1.0]hexyl ureas. J. Med. Chem.
2006, 49, 2294−2310.
(18) Bentzien, J.; Hickey, E. R.; Kemper, R. A.; Brewer, M. L.;
Dyekjaer, J. D.; East, S. P.; Whittaker, M. An in silico method for
predicting ames activities of primary aromatic amines by calculating
the stabilities of nitrenium ions. J. Chem. Inf. Model. 2010, 50, 274−
297.
(19) Block, M. H.; Boyer, S.; Brailsford, W.; Brittain, D. R.; Carroll,
D.; Chapman, S.; Clarke, D. S.; Donald, C. S.; Foote, K. M.; Godfrey,
L.; Ladner, A.; Marsham, P. R.; Masters, D. J.; Mee, C. D.; O’Donovan,
M. R.; Pease, J. E.; Pickup, A. G.; Rayner, J. W.; Roberts, A.; Schofield,
P.; Suleman, A.; Turnbull, A. V. Discovery and optimization of a series
of carbazole ureas as NPY5 antagonists for the treatment of obesity. J.
Med. Chem. 2002, 45, 3509−3523.
Foresman, J. B.; Ortiz, J. V.; Cioslowski, J.; Fox, D. J. Gaussian 09,
revision A.1; Gaussian, Inc.: Wallingford, CT, 2009.
(31) Box, G. E. P. Empirical Model-Building and Response Surfaces;
Wiley: New York, 1987.
(32) O’Donovan, M. R.; Mee, C. D.; Fenner, S.; Teasdale, A.;
Phillips, D. H. Boronic acids: a novel class of bacterial mutagen. Mutat.
Res., Genet. Toxicol. Environ. Mutagen. 2011, 724, 1−6.
(33) Hansch, C.; Leo, A.; Unger, S. H.; Kim, K. H.; Nikaitani, D.;
Lien, E. J. Aromatic substituent constants for structure−activity
correlations. J. Med. Chem. 1973, 16, 1207−1216.
(34) Waring, M. J. Defining optimum lipophilicity and molecular
weight ranges for drug candidates: molecular weight dependent lower
log D limits based on permeability. Bioorg. Med. Chem. Lett. 2009, 19,
2844−2851.
(35) Camenisch, G.; Alsenz, J.; van de Waterbeemd, H.; Folkers, G.
Estimation of permeability by passive diffusion through Caco-2 cell
monolayers using the drugs' lipophilicity and molecular weight. Eur. J.
Pharm. Sci 1998, 6, 313−319.
(36) Calculations in which constraints are imposed cannot
incorporate vibrational corrections, as the simple harmonic potential
approximation breaks down. So the energies reported in this section
are uncorrected electronic energies and should not be compared
directly with energy changes reported elsewhere in this paper.
(37) Maron, D. M.; Ames, B. N. Revised methods for the Salmonella
mutagenicity test. Mutat. Res., Environ. Mutagen. Relat. Subj. 1983, 113,
173−215.
(20) Hamann, L. G.; Manfredi, M. C.; Sun, C.; Krystek, S. R.; Huang,
Y.; Bi, Y.; Augeri, D. J.; Wang, T.; Zou, Y.; Betebenner, D. A.; Fura, A.;
Seethala, R.; Golla, R.; Kuhns, J. E.; Lupisella, J. A.; Darienzo, C. J.;
Custer, L. L.; Price, J. L.; Johnson, J. M.; Biller, S. A.; Zahler, R.;
Ostrowski, J. Tandem optimization of target activity and elimination of
mutagenic potential in a potent series of N-aryl bicyclic hydantoin-
based selective androgen receptor modulators. Bioorg. Med. Chem. Lett.
2007, 17, 1860−1864.
(38) CORINA, Molecular Networks; GmbH Computer Chemie
Langemarckplatz 1: Erlangen, Germany.
(21) Borosky, G. L. Quantum chemical studies on ultimate
carcinogenic metabolites from polycyclic aromatic hydrocarbons.
Curr. Med. Chem. 2008, 15, 2901−2920.
(22) McCarren, P.; Bebernitz, G. R.; Gedeck, P.; Glowienke, S.;
Grondine, M. S.; Kirman, L. C.; Klickstein, J.; Schuster, H. F.;
Whitehead, L. Avoidance of the Ames test liability for aryl-amines via
computation. Bioorg. Med. Chem. 2011, 19, 3173−3182.
(23) Lennox, A. J. J.; Lloyd-Jones, G. C. The slow-release strategy in
Suzuki−Miyaura coupling. Isr. J. Chem. 2010, 50, 664−674.
(24) Plowright, A. T.; Johnstone, C.; Kihlberg, J.; Pettersson, J.;
Robb, G.; Thompson, R. A. Hypothesis driven drug design: improving
quality and effectiveness of the design-make-test-analyse cycle. Drug
Discovery Today 2012, 17, 56−62.
(25) Four of the 312 anilines originally studied had been tested more
than once and found to have different classifications as active or
inactive. Rather than perform a detailed analysis of each case, these
four examples have been excluded from this calibration curve.
(26) JMP, version 8; SAS Institute Inc.: Cary, NC.
(27) Becke, A. D. Density-functional thermochemistry. III. The role
of exact exchange. J. Chem. Phys. 1993, 98, 5648−5652.
(28) Lee, C.; Yang, W.; Parr, R. G. Development of the Colle−
Salvetti correlation-energy formula into a functional of the electron
density. Phys. Rev. B: Condens. Matter 1988, 37, 785−789.
(29) Hariharan, P. C.; Pople, J. A. Influence of polarization functions
on MO hydrogenation energies. Theor. Chim. Acta 1973, 28, 213−222.
(30) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.;
Robb, M. A.; Cheeseman, J. R.; Scalmani, G.; Barone, V.; Mennucci,
B.; Petersson, G. A.; Nakatsuji, H.; Caricato, M.; Li, X.; Hratchian, H.
P.; Izmaylov, A. F.; Bloino, J.; Zheng, G.; Sonnenberg, J. L.; Hada, M.;
Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima,
T.; Honda, Y.; Kitao, O.; Nakai, H.; Vreven, T.; Montgomery, J. A., Jr.;
Peralta, J. E.; Ogliaro, F.; Bearpark, M.; Heyd, J. J.; Brothers, E.; Kudin,
K. N.; Staroverov, V. N.; Kobayashi, R.; Normand, J.; Raghavachari, K.;
Rendell, A.; Burant, J. C.; Iyengar, S. S.; Tomasi, J.; Cossi, M.; Rega,
N.; Millam, N. J.; Klene, M.; Knox, J. E.; Cross, J. B.; Bakken, V.;
Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.;
Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Martin, R. L.;
Morokuma, K.; Zakrzewski, V. G.; Voth, G. A.; Salvador, P.;
̈
Dannenberg, J. J.; Dapprich, S.; Daniels, A. D.; Farkas, O.;
3933
dx.doi.org/10.1021/jm3001295 | J. Med. Chem. 2012, 55, 3923−3933