Conformational and configurational behaviour of κ-agonistic 3,7-diazabicyclo[3.3.1]nonan-9-ones - Synthesis, nuclear magnetic resonance studies and semiempirical PM3 calculations

Article abstract of DOI:10.1039/a806641h

2,4-Diaryl substituted 3,7-diazabicyclo[3.3.1]nonan-9-one 1,5-diesters were found to have a high affinity for κ-opioid receptors. To develop highly potent analgesics, the purpose of this study was the synthesis and the structural characterisation of the novel 2,4-bis(4-nitrophenyl), 2,4-bis(3-nitrophenyl), 2,4-bis(4-quinolyl), 2,4-bis(2-quinolyl), 2,4-bis(1-naphthyl) and 2,4-bis(2-naphthyl) substituted 3,7-diazabicyclo[3.3.1]nonan-9-one 1,5-diesters by means of NMR spectroscopy and molecular modelling. It could be proved that several derivatives undergo trans-cis-isomerisation of the aromatic rings linked to the rigid skeleton whereas others show rotational isomerisation. Semiempirical quantum-chemical PM3 calculations were performed to analyse the thermodynamic stability of the isomers as well as the mechanism of the trans-cis-or cis-trans-conversion.

Full text of DOI:10.1039/a806641h

View Article Online / Journal Homepage / Table of Contents for this issue
Conformational and configurational behaviour of ꢀ-agonistic
3,7-diazabicyclo[3.3.1]nonan-9-ones—synthesis, nuclear magnetic
resonance studies and semiempirical PM3 calculations
Tom Siener,a Ulrike Holzgrabe,*a Susanne Drosihnb and Wolfgang Brandtb
a Institut für Pharmazie und Lebensmittelchemie, Universität Würzburg, Am Hubland,
D-97074 Würzburg, Germany
b Fachbereich Biochemie/Biotechnologie, Universität Halle, Kurt-Mothes-Str. 3, 06120 Halle,
Germany
Received (in Cambridge, UK) 24th August 1998, Accepted 5th July 1999
2,4-Diaryl substituted 3,7-diazabicyclo[3.3.1]nonan-9-one 1,5-diesters were found to have a high affinity for
κ-opioid receptors. To develop highly potent analgesics, the purpose of this study was the synthesis and the
structural characterisation of the novel 2,4-bis(4-nitrophenyl), 2,4-bis(3-nitrophenyl), 2,4-bis(4-quinolyl), 2,4-bis(2-
quinolyl), 2,4-bis(1-naphthyl) and 2,4-bis(2-naphthyl) substituted 3,7-diazabicyclo[3.3.1]nonan-9-one 1,5-diesters
by means of NMR spectroscopy and molecular modelling. It could be proved that several derivatives undergo trans–
cis-isomerisation of the aromatic rings linked to the rigid skeleton whereas others show rotational isomerisation.
Semiempirical quantum-chemical PM3 calculations were performed to analyse the thermodynamic stability of the
isomers as well as the mechanism of the trans–cis- or cis–trans-conversion.
Introduction
2,4-Diaryl substituted 3,7-diazabicyclo[3.3.1]nonan-9-one 1,5-
diesters were found to have a high affinity for κ-opioid receptors
with agonistic properties and, thus, are attracting interest as
highly potent analgesic drugs.1 Even though the bicyclo-
nonanones are almost rigid features, several conformational
and configurational isomers can be found, concerning the con-
formation of the bicyclic system as well as the substituents on
the skeleton: theoretically, the bicyclic system is able to adopt a
chair-chair (cc), a chair-boat (cb), a boat-chair (bc) and a boat-
boat (bb) conformation.2,3 In the case of the aforementioned
highly substituted 3,7-diazabicycles and 3-oxa-7-azabicycles, a
bb conformation was found to be energetically unfavourable.4,5
Thus, this conformation was not taken into account. A cb con-
formation (c in the higher substituted piperidone) can be
observed either in molecules with bulky substituents at N74,6
or in double protonated diazabicycles.5 In this series of
compounds, a bc conformation (b in the higher substituted
piperidone) is impossible unless an epimerisation of the aryl
substituents has taken place in an initial c conformation.7 In
addition to the various ring conformations, restricted rotations
of the aryl substituents at C2 and C4 were observed. Especially
in the case of N3-methylation, the rotational barrier was high
enough to prevent the rings from rotating.8 In very few
cases, instead of the symmetrical cis-configuration of the aryl
substituents, a trans-configuration was observed.9
Since the diazabicyclononanones should specifically bind to a
receptor protein, the spatial arrangement of the molecule is
pivotal for a high biological effect. As part of a larger project
dealing with the development of highly potent analgesics,1 the
purpose of this study was to elucidate the structure of isomers
isolated, the reaction pathway of the isomerism, and to find out
whether the diazabicycles are able to isomerise under physio-
logical conditions.
Scheme 1 2,4-Diaryl substituted dimethyl 3,7-dimethyl-9-oxo-3,7-
diazabicyclo[3.3.1]nonane-1,5-dicarboxylates.
of NMR spectroscopy in solution. For the sake of comparison,
the 2,4-diphenyl substituted compound9 1 was taken into
account. In order to explain the experimental findings and to
clarify whether thermodynamic stability differences or kinetic
processes govern the stereochemistry, force field calculations as
well as semiempirical quantum-chemical PM3 calculations
were performed. Moreover, the mechanism of cis–trans-
isomerisation was theoretically studied.
As probes, 2,4-bis(4-nitrophenyl), 2,4-bis(3-nitrophenyl), 2,4-
bis(4-quinolyl), 2,4-bis(2-quinolyl), 2,4-bis(1-naphthyl) and 2,4-
bis(2-naphthyl) substituted dimethyl 3,7-dimethyl-9-oxo-3,7-
diazabicyclo[3.3.1]nonane-1,5-dicarboxylates 2–7 (Scheme 1)
were synthesized and their stereochemistry elucidated by means
Results
NMR investigations
All compounds could be achieved by a double Mannich
reaction (cf. ref. 9). In the first step, 1 mole of dimethyl 3-oxo-
J. Chem. Soc., Perkin Trans. 2, 1999, 1827–1834
1827

View Article Online
Table 1 1H NMR data for the compounds 2–7
Compound
H2/4
H6/8
-N4CH3
-N7CH3
-OCH3
3.49/3.81
2 (trans)
4.92/5.03
2.71–2.75/3.14–3.19
m
1.85
s
2.33
s
s
s
m
s
s
(cis)
4.68
s
4.95/5.05
2.69/3.04
d, 12.6/d, 12.6
2.62–3.09
1.85
s
1.88
s
2.37
s
2.36/2.43/2.51
s
3.72
s
3.51/3.90
3 (trans)
s
s
m
s
s
s
s
(cis)
4.69
s
2.65–2.78/2.99–3.13
1.88
s
2.36/2.43/2.51
3.72/3.79
m
m
s
s
s
s
s
4
5
6
7
4.86/5.28/5.47
2.63–2.71/3.45–3.50
1.77/1.81
2.47/2.54
3.42/3.66
s
s
s
m
m
s
s
s
s
s
s
5.04/5.33
2.80/3.12/3.17/3.75
d, 12.0/d, 12.0/d, 10.8/d, 10.8
2.61–2.68/3.46–3.68
2.12
s
1.80/1.85
2.29
s
2.49/2.55
3.49/3.99
s
s
s
s
4.86/5.29/5.47
3.44
s
3.76
s
s
s
s
m
m
s
s
s
s
4.75
s
2.68–2.72/3.29–3.33
1.92/1.97
s
2.47/2.48
m
m
s
s
Table 2 13C NMR data for the compounds 2–7
Compound
C1/5
C2/4
C6/8
C9
C᎐O
᎐
-OCH3
-NCH3
2 (trans)
61.52/63.60
62.73
63.64/66.21
62.91
62.44/64.14
62.25/62.64
62.83/63.28/64.94
68.44/74.14
71.83
68.46/74.24
71.56/72.03
65.72/66.25/76.47
70.52/73.17
66.41/66.90/77.68
60.44/66.36
59.72
60.56/61.82
59.76
60.53/61.09/61.40
61.97/66.22
60.95/61.40
202.03
202.34
201.99
202.37
202.17/202.53
200.86
167.68/168.13
167.37
167.67/168.11
167.47
167.03/167.34
169.46
167.85/168.13
52.42/52.96
52.73
52.44/53.02
52.67
52.38
52.30
40.53/44.13
43.31/44.56
40.69/43.32/44.03
43.32/44.15/44.34
43.18/43.90/44.59
40.75/44.30
(cis)
3 (trans)
(cis)
4
5
6
203.76/204.28
52.07
42.66/42.83/
43.82/44.59
7
63.95
72.81
60.10
204.12
168.21
52.30
43.10/43.34/44.42
crystals were obtained, rhombic crystals characterised by a
melting point of 215 ЊC and an Rf value of 0.6 on TLC, and
amorphous crystals characterized by a melting point of 180 ЊC
and a lower Rf value of 0.51. The amorphous powder could be
converted into rhombic crystals by recrystallisation from
methanol. The latter crystals gave a simple NMR spectrum
which is characteristic of a symmetrically substituted diaza-
bicycle with a cis-configuration of the phenyl rings. Logically,
the second set of signals belongs to a diazabicycle with the
phenyl rings trans to each other. Temperature-dependent NMR
measurement of an isomeric mixture of 2 gave further evidence
of cis–trans-isomerism. In DMSO-d6 an increase in the signals
of the cis-isomer was observed in connection with a simul-
taneous decrease in the other set of signals. The reaction
was found to be irreversible. This observation indicates that
the cis-isomer is thermodynamically more stable than the
trans-isomer.
It is worth mentioning that in the case of the cis-isomer the
ortho- and meta- hydrogens of either side of the phenyl ring
gave two different AB systems: one system is located at 7.34 and
8.15 with a coupling constant of 8.7 Hz. NOE difference
experiments showed these hydrogens to be positioned close to
H2/4 of the bicyclic system. The signals of the second AB sys-
tem are close together at δ = 8.35 and 8.40 ppm, and can be
assigned to the hydrogens which point to the “bottom” (endo) of
the bicyclic skeleton. The results indicate that the phenyl ring
takes a fixed, almost perpendicular position to the piperidone
ring, which is in accordance with findings obtained e.g. for
the 2,4-bis(3-chlorophenyl)-3,7-dimethyl-3,7-diazabicyclonon-
anone.8 Since no coalescence of the aryl signals was observed
up to 90 ЊC, it was difficult to determine the rotational barrier
by recording temperature-dependent NMR spectra.
Fig. 1 1H NMR spectrum (300 MHz) of compound 2.
glutarate, 1 mole of methylamine and 2 moles of the corre-
spondingly substituted benzaldehyde were refluxed in ethanol
to give the piperidone diester. Subsequent conversion of the
thus obtained piperidones with formaldehyde and methylamine
resulted in the 3,7-diazabicycles. NMR data of all compounds
are displayed in Table 1 and Table 2.
From the simplicity of the NMR spectrum in the case of the
2,4-diphenyl substituted dimethyl 3,7-dimethyl-9-oxo-3,7-diaza-
bicyclo[3.3.1]nonane-1,5-dicarboxylate 1, a symmetrical com-
pound can be deduced.9 In contrast, the NMR spectrum of the
4-nitrophenyl substituted compound 2, obtained from the first
fraction of crystals of the reaction solution, is characterised by
a double set of signals (Fig. 1). Interestingly, for the hydrogens
at positions 2 and 4, three singlets at δ = 4.68, 4.92, and 5.03
ppm with an integration ratio of 1:2:2 were measured. The
question arose whether either of the double set of signals
belongs to rotational isomers as described for the 2,4-bis(3-
chlorophenyl)-3,7-dimethyl-3,7-diazabicyclononanone8 or to
cis–trans-isomers as observed in the case of the 2,4-bis(2-
pyridyl)-3,7-dimethyl-3,7-diazabicycle.9 The following findings
indicate that the two sets of NMR signals are caused by two
configurational isomers rather than conformational isomerism.
By means of fractional crystallisation, two different sorts of
However, in order to determine the activation barrier of
the trans–cis-isomerisation, compound 2 was dissolved in
methanol-d4 and the solution was refluxed in the NMR tube.
1828
J. Chem. Soc., Perkin Trans. 2, 1999, 1827–1834

View Article Online
Fig. 3 Definition of hydrogens in the 4-quinoline residue.
was mirrored in the corresponding signals for the aromatic
protons.
Again the question arose whether these different sets of sig-
nals belonged to configurational cis–trans-isomers or rotational
isomers: according to the simple set of signals, one isomer
seems to be symmetrical, whereas the double set of signals may
belong to an asymmetrical isomer. Interestingly, in the spectra
of the 4-quinolyl and 1-naphthyl derivatives 4 and 6, a doublet
of doublets could be found at δ = 10.2 and δ = 10.4 ppm,
respectively. This rather strong downfield shift of a single
aromatic proton might be caused by the ring current effect of
either aromatic ring, which only seems to be possible where the
aryl rings are in a cis-configuration. To prove this hypothesis,
H,H-COSY and NOE experiments were performed to elucidate
the spatial neighbourhood of this proton. First, the aromatic
protons were assigned to the signals by the H,H-COSY experi-
ment. From the results it was further hypothesized that both
isomers are characterised by a cis-configuration of the aryl
residues: according to the simple set of signals, one isomer has
to be symmetrical (protons Ha, Hb, etc.) with both bicyclic aryl
rings pointing to the outside (exo) of the skeleton. In the isomer
with the double set of signals, one aryl ring seems to take an
orientation similar to the symmetrical compound (protons HaЈ,
HbЈ, etc.), and the other aryl ring is rotated by about 180Њ (pro-
tons HaЉ, HbЉ, etc., protons are defined in Fig. 3). The different
orientation of the aryl rings creates the asymmetry; temper-
ature dependent NMR measurements have shown that these
apparent conformers do not interconvert at high temperatures.
Thus, the isomerism can be denoted as an atropisomerism
about a sp2–sp3 single bond.13
Fig. 2 1H NMR spectrum (300 MHz) of compound 4.
NMR spectra were recorded every 16 min (after cooling to
ambient temperature) and the ratio of the trans- and cis-
isomers was determined from the integrals of the signals for
H2/4. The ratio was plotted against the time. The rate constant
k of the reaction was calculated from eqn. (1).10 After 110 min
ln 2 ln 2
k(64.7 ЊC) =
=
= 3.3 × 10Ϫ4 sϪ1
(1)
t
2100s
½
almost all trans-isomer was converted into the cis-isomer. The
half-life of the reaction was found to be 2100 s, which resulted
in a reaction constant k(64.7 ЊC) = 3.3 × 10Ϫ4 sϪ1
.
In order to calculate the activation barrier, the boiling point
of methanol (64.7 ЊC) and the t1/2 was subjected to the
logarithmic form (3) of the EYRING equation (2), where ∆G‡
2T
NA × h 2T
∆G‡
k =
e
(2)
is the free activation enthalpy, R is the universal gas constant
(8.31 J KϪ1), T is the absolute temperature (K), NA is
Avogadro’s number (6.022 × 1023 molϪ1) and h is the Planck
constant (6.6256 × 10Ϫ34 J s). From eqn. (3) the free activation
The NOE experiments support the hypothesis of sym-
metrical and asymmetrical cis-isomers (Fig. 4). On saturating
the signal at 5.47 ppm which was assigned to H2/4 in the
putative symmetrical compound, one positive NOE effect with
the aromatic proton Hc was found. This is only possible in the
above mentioned orientation of the 4-quinolyl ring of the
symmetrical compound. Furthermore, a negative NOE effect
could be detected, which can be explained as a three-spin-
system between H2/4, the aromatic proton Hc and the aromatic
proton Hd. Saturating the signals at 4.86 or 5.28 ppm which
were supposed to belong to the protons H2/4 in the asymmetric
compound led to a strong NOE effect at 5.28 or 4.86 ppm,
respectively. This clearly demonstrates the spatial proximity of
H2 and H4, realized in the cis-configuration. The signal at 5.28
ppm (H2) further showed an NOE effect to the aromatic proton
HcЈ and negative effects to HdЈ (compared with the symmetrical
compound) and HbЉ. The signal at 4.86 ppm (H4) showed an
NOE effect to HbЉ and a negative effect to HaЉ. Taking the con-
necting pathways in the H,H-COSY into account, the proton
showing the strong downfield shift could be assigned to the
proton HcЉ. It showed strong NOE effects to the protons HdЉ
and HbЈ (not displayed). This strong downfield shift of HcЉ was
likely to be caused by the spatial proximity of the phenyl rings
in a way that the hydrogens of either ring feel the deshielding
ring current effect of the other.
k
∆G‡ = 19.14 T
ͫ10.32 Ϫ logͩ ͪͬ
(3)
t
enthalpy was calculated to be 25 kcal molϪ1, indicating a rather
high free activation barrier.
In the case of the 3-nitrophenyl substituted compound 3, a
similar cis–trans-isomerism was observed. From recrystallis-
ation of a mixture of cis–trans-isomers in methanol, the pure
cis-isomer was obtained. In contrast, in the case of 2-quinolyl
substituted compound 5, recrystallisation gave only the trans-
isomer. Due to the asymmetry of the molecule, the NMR spec-
trum shows a “double” set of signals in comparison to the
cis-isomers. The trans-configuration could be verified by the
measurement of NOE experiments. Saturation of the signals of
H2 and H4, respectively, showed only a very weak NOE effect
for either hydrogen, whereas the saturation of the signal at
δ = 5.04 ppm resulted in a strong positive NOE with the signal
at δ (H6/8) = 3.75 ppm and a negative NOE with the signal at δ
(H6/8) = 3.17 ppm; the latter might be due to a three-spin effect
(cf. ref. 11) and could only be observed where the signal belongs
to a hydrogen (H2 or H4) in an axial position (cf. ref. 12).
Logically, the saturation of the signal at δ = 5.33 ppm, which
has to be assigned to the hydrogen H2/4 in the equatorial
position, does not show any NOE effect with signals belonging
to H6/8.
PM3 calculations of the diazabicycle 4 in the cc conform-
ation and cis-configuration concerning the quinoline rings
revealed two stable cis-geometries that differ in their heats of
formation only by about 5 kcal molϪ1, one is the symmetrical
The spectra of the 4-quinolyl 4 (Fig. 2) and 1-naphthyl
derivative 6, both having the aromatic system linked through
the α-position to the diazabicycle, showed three signals at
4.86, 5.28, 5.47 ppm (ratio 1:1:2) and 4.86, 5.29, 5.47 ppm
(ratio 1:1:3) for the protons H2/4, respectively. This ratio
arrangement of the aromatic rings (∆H = Ϫ67.9 kcal molϪ1
)
(Fig. 5), and the other one is an asymmetric arrangement with
one quinolyl ring rotated by about 180Њ (Fig. 6) (∆H = Ϫ63.2
J. Chem. Soc., Perkin Trans. 2, 1999, 1827–1834
1829

View Article Online
Fig. 6 Structure of the energetically stable asymmetric cis-configur-
ation of 4, indicating the shortest distance between the quinoline
substituents.
gens observed in the NMR spectra. In turn, if the distance
between these protons is larger—as is the case with compound
1—no such downfield shift can be observed.
Taking the results up to this point together, it can be stated
that the quinolyl substituted compounds suffer from stereo-
chemical overcrowding and, thus, form stable rotational iso-
mers. Due to the high similarity in spatial properties between
the naphthyl compound 6 (repulsions between HcЈ atoms and
the N3 methyl group) and the quinolyl analogue 4, stable
rotational isomers can also be expected.
Theoretical calculations
In order to gain more insight into the stereochemical behav-
iour of the diazabicyclononanones, all conformations of the
skeleton studied herein, chair-chair (cc), chair-boat(N7) (cb),
boat(N3)-chair (bc) and boat-boat (bb) were subjected to force
field and PM3 calculations. As expected from previously per-
formed calculations for the analogous oxaazabicycles,4 in each
case the bc and bb conformations appeared to be energetically
unfavourable. Comparing the energies of all the compounds
and conformations, the cc conformations are slightly more
stable than the cb conformations of the bicyclononanones;
therefore, only the cc will be considered in the following discus-
sions. Furthermore, an equatorial and axial orientation of the
N3-methyl group to the piperidine ring were taken into con-
sideration. In all cases the equatorial conformation appeared to
be more stable than the axial orientation of the methyl group.
The energies of the more stable conformations are listed in
Table 3.
Fig. 4 NOE difference spectra of compound 4.
Thermodynamics of the cis–trans-equilibrium The conform-
ations, i.e. the dihedral angles describing the orientation of the
aromatic rings, obtained for the most stable conformations are
very similar to the already published structures of 1.5 Compar-
ing all compounds with regard to the trans–cis differences of
the heats of formation, the aforementioned experimental find-
ings can be partially understood: for 1 the ∆trans–cis∆H value of
1.8 kcal molϪ1 indicates a preference for a cis-conformation.
The enthalpy differences between trans- and cis-configurations
for 2 and 3 were found to be smaller (2 ∆trans–cis∆H = 1.0 kcal
molϪ1, 3 ∆trans–cis∆H = 1.5 kcal molϪ1; 1.0 kcal molϪ1 in the case
of the asymmetric forms). These differences in the heats of
formation of the two alternative configurations cannot com-
pletely explain the experimentally measured trans–cis-ratios of
2 and 3 in comparison with 1; however, they favour a cis-
configuration which was obtained after recrystallisation.
Fig. 5 Structure of the energetically stable symmetric cis-configur-
ation of 4, indicating the shortest distance between the quinoline
substituents.
kcal molϪ1), both suggested by the NMR spectra. We analysed
the possibility of interconverting the symmetric or the asym-
metric arrangement into either form by rotation of the
quinoline rings in steps of 5Њ. In each case, a strong spatial
hindrance at least between the Hc proton of the quinoline rings
and the protons of the N3 linked methyl group appeared. This
finding nicely explains the stability of the isomers that was
found experimentally. In Figs. 5 and 6 the short distances
between the protons of both rings are marked. In the case of
the symmetrical cis-conformation, such a short distance (2.6 Å)
is observed between both HbЈ atoms. In the asymmetric cis-
conformation an even shorter distance (2.4 Å) between HbЈ and
HcЉ appeared. This explains the downfield shift of these hydro-
For compound 4, two structures with cis-configurations
characterised by a symmetrical and an asymmetrical position
1830
J. Chem. Soc., Perkin Trans. 2, 1999, 1827–1834

View Article Online
Table 3 Heats of formation (in kcal molϪ1) of low energy conform-
ations of bicyclononanone derivatives
Thermodynamics
Compound
∆Hcis
∆Htrans
∆trans–cis∆H
1
only cis
sym
2
؊120.4
cis/trans
؊136.0
cis/trans
؊136.7
Ϫ136.6
only cis
؊67.9
Ϫ63.2
trans
؊70.2
Ϫ67.7
only cis
؊81.6
Ϫ76.8
Ϫ118.6
cis after recrystallisation
Ϫ135.0
1.8
sym
3
sym
asym
4
sym
asym
5
sym
asym
6
1.0
Ϫ135.2
Ϫ135.6
1.5
1.0
Ϫ62.7
Ϫ67.8
5.2
Ϫ4.6
Ϫ68.9
Ϫ69.3
1.3
Ϫ1.6
Fig. 7 Diagram representing the dependence of the length of the
C1–C2 bond on the heat of formation (PM3 calculations) of 2.
sym
asym
Ϫ81.5
Ϫ74.8
0.1
2.0
hydrogen/deuterium exchange should be observed in deuterated
solvent. However, an NMR study of the isomerisation (see
above) performed in methanol-d4 did not show a H/D exchange.
Therefore, only a retro-Mannich reaction needs to be discussed
in order to explain the observed isomerism.
As the initial step of the retro-Mannich reaction, a proton-
ation of the keto carbonyl group is a prerequisite to stabilise the
resulting ring-opened product by formation of an enol (see
Scheme 2). To gain an insight into the probability for such a
of the 4-quinolyl ring were found in the experiments. For the
sake of comparison, both conformations and their correspond-
ing trans-configurations have been calculated. In the symmetric
conformation, the cis-configuration appeared to be highly
favoured over the trans (4 ∆trans–cis∆H = 5.2 kcal molϪ1), whereas
in the case of the “asymmetric” conformation (the aryl rings
are rotated by 180Њ against each other), the opposite is true
(∆trans–cis∆H = Ϫ4.6 kcal molϪ1). However, the symmetric cis-
and the “asymmetric” trans-isomer are energetically equivalent.
Nevertheless, experimentally only the cis-configuration of 4
was observed. The repulsion of the HcЈ hydrogen atom having a
rather close contact of 2.5 Å to both nitrogen atoms of the
bicycle skeleton is likely to be the reason for the thermo-
dynamically unfavourable stability of the asymmetric cis-
conformation. The trans-configuration lacks such a close
spatial interaction. For the asymmetric compound, the clarifi-
cation of this contradiction will be given in the kinetics part.
Similar contradictory results were obtained in the case of
the 2-quinolyl substituted compound 5, which could only be
isolated in the trans-configuration. Both configurations were
analysed. In agreement with the experiment, the heats of for-
mation for the “asymmetric” compound predicted trans to be
slightly more stable than cis (∆trans–cis∆H = Ϫ1.6 kcal molϪ1).
However, in the symmetrical conformation the cis-configur-
ation is favoured (∆trans–cis∆H = 1.3 kcal molϪ1) and is energetic-
ally comparable to both trans-conformations.
Since the thermodynamic examinations could not give satis-
fying explanations of all experimentally observed ratios of
isomers, a subsequent kinetic study was performed.
Scheme 2 Cleavage of the C1–C2 or C4–C5 bond and theoretical
mesomeric stabilisation of the resulting anion.
Kinetics of the trans–cis-isomerisation mechanism. From the
synthesis a mixture of trans/cis isomers of 2 was obtained.
Heating a solution of the isomers up to 40 ЊC always results in a
complete conversion of the trans-isomer into the cis-isomer of 2
which is in accordance with the NMR measurements men-
tioned above. It is not possible to isolate the pure trans-isomer.
The bicyclononanones are relatively rigid compounds. Two
main mechanisms seem to be possible in principle. One is char-
acterised by opening and closing of a covalent bond between
C1/C2 and C4/5, respectively, corresponding to a retro-
Mannich reaction. In order to switch from one to the other
configuration, in the ring-opened form a rotation of the
H–C2(C4) bond around the N(3)–C2(C4) bond has to occur
and seems to be possible. At first glance, a second mechan-
ism might be possible: it is characterised by a temporary
deprotonation of C2 or C4, leading to a planar sp2 carbon
atom. Reprotonation from the backside would lead to the
isomerism. If the latter mechanism were to take place, a
carbonyl protonation, the heats of formation of the keto-
protonated structures were compared with the corresponding
N3-protonated compounds (Table 4). As is expected from
chemical understanding, in each case the N3-protonated com-
pounds are more stable than those with a hydroxy group. Inter-
estingly, the differences of the heats of formation are in the
range (about 16 to 33 kcal molϪ1) of the experimentally
observed activation barrier for a trans–cis isomerisation.
Starting from the carbonyl protonated forms, reaction
coordinate calculations to lengthen the C1–C2 bond in steps of
0.1 Å to a distance of 3.7 Å were performed. As an example, the
resulting energy pathway is schematically shown for 2 in Fig. 7.
After overcoming a transition state at 1.8 Å with an energy
effort of 6.1 kcal molϪ1 the bond cleaves with energy gain. A
comparison of the heats of formation of the optimised ring-
opened forms of all compounds with the ones of the ring-
closed form is listed in Table 5. All possible conformations have
been calculated including two asymmetric forms for the cis- and
J. Chem. Soc., Perkin Trans. 2, 1999, 1827–1834
1831

View Article Online
Table 4 Comparison of the proton affinities of the N3-nitrogen atom and the carbonyl oxygen atom
cis
trans
∆Hf
N3-prot
∆Hf
enol
∆Hf
N3-prot
∆Hf
enol
∆∆Hf,enol-N3-prot
∆∆Hf,enol-N3-prot
1
2
3
sym
sym
25.8
25.4
53.8
28.0
22.6
28.1
27.1
51.7
23.6
20.4
48.0
47.5
sym
asym
21.4
22.1
45.6
45.6
24.2
23.5
23.9
24.8
45.7
43.2
30.8
18.4
4
5
6
sym
asym
82.5
88.4
108.8
108.5
26.3
25.7
92.6
84.4
113.4
106.6
20.8
22.2
sym
asym
72.6
73.7
101.6
103.2
29.0
29.5
74.7
72.9
92.9
98.3
18.2
25.4
sym
asym
61.8
67.9
90.4
95.8
28.6
27.9
64.0
72.2
93.9
88.1
29.9
15.9
Table 5 Enol-forms of several bicyclononanone derivatives corresponding to the Mannich or retro-Mannich reaction, respectively
cis
trans
∆Hf
open
∆Hf
closed
∆∆Hf,
closed–open
∆Hf
open
∆Hf
closed
∆∆Hf,
closed–open
∆∆∆Hf,
cis–trans
1
2
3
25.8
53.8
28.1
51.7
sym
sym
31.7
34.3
22.1
13.7
32.8
34.2
18.9
13.3
3.2
0.4
48.0
47.5
sym
asym
29.6
30.5
45.6
45.6
16.0
15.1
29.8
30.2
45.7
43.2
15.9
13.0
0.1
2.1
4
5
6
sym
asym
90.5
91.4
108.8
108.5
18.3
17.1
89.3
92.5
113.4
106.6
24.1
14.1
Ϫ5.8
3.0
sym
asym
81.2
80.2
101.6
103.2
20.4
23.0
81.3
81.9
92.9
98.3
11.6
16.4
8.8
6.6
sym
asym
69.2
71.3
90.4
95.8
21.2
24.5
71.2
70.3
93.9
88.1
22.7
17.8
Ϫ1.5
6.7
Fig. 8 Diagram representing the dependence of the rotation of the
dihedral angle C4–N3–C2–H (see Fig. 9) on the heat of formation
(PM3 calculations) of 2.
Fig. 9 The structure of the open enol-form of compound 2 in a quasi
cis-configuration. The arrow indicates the rotation of the bond to
switch from a cis- to a trans-configuration.
four for the trans-configuration. In Table 5, however, only those
structures are considered which appeared more stable in
the neutral compounds and which may be relevant for the
mechanistic analysis.
In addition, the process of isomerisation per se, that is the
rotation of the C2(C4)–H bond, needs to be discussed. For this
purpose the dihedral angle C4–N3–C2–H was rotated in steps
of 10Њ (see Fig. 8). Two barriers were observed, each character-
ised by an almost perpendicular orientation of the C–H bond
to the plane of C4–N3–C2 whereby the rotation passing a
repulsion with the N3-methyl group is slightly lower (∆Hf =
34.1 kcal molϪ1) than in the direction of the former C1–C2
bond (∆Hf = 40.7 kcal molϪ1).
Taking the energy efforts and gains of each step together the
retro-Mannich reaction is probably the mechanism of the
trans–cis or cis–trans isomerisation. Even though the entropy
contributions or the solvent stabilisation influences were not
considered (polar solute supports polar transition states), the
barrier for the isomerisation is in the range of the experi-
1832
J. Chem. Soc., Perkin Trans. 2, 1999, 1827–1834

View Article Online
difficult to explain the additional formation of the asymmetric
cis-form. Comparing the heats of formation of the ring-opened
enol-forms, the asymmetric cis-structure is 1.1 kcal molϪ1
favoured over the corresponding trans-configuration, which
would explain the experimental result. However, in contradic-
tion to this, the difference of the heats of formation between the
closed and open compound is 3 kcal molϪ1 higher for the cis-
than for the trans-configuration. Maybe these somewhat con-
tradictory results illustrate the limitations of semiempirical
PM3 calculations performed in the gas phase. Nevertheless, the
thermodynamic preference of the open asymmetric cis- over
the trans-form of 4 seems to be responsible for the formation of
the corresponding neutral bicyclononanone with an asym-
metric cis-configuration.
Overall, a retro-Mannich reaction is very probably the mech-
anism responsible for trans–cis- and cis–trans-isomerisations.
By means of semiempirical PM3 calculations we could
show that thermodynamic as well as kinetic aspects play an
important role in the formation of stable bicyclononanones.
In addition, it can be concluded that all isomers for each
compound are stable enough for pharmacological studies. No
isomerisation should be expected. However, the complex
stereochemical behaviour has to be taken into account when
analysing structure–activity relationships of these compounds
as κ-selective opioids. Preliminary pharmacological screening
has shown the 2-quinolyl derivative to have high affinity for the
κ-receptors.
Fig. 10 Attractive electrostatic interactions between the hydroxy
group and the nitrogen atom of the 2-quinolyl ring of the symmetric
trans-form of 5.
mentally determined value. Two contradictions remain to be
answered: firstly, even though in the case of the 4-quinolyl
compound (4) the symmetric cis-form and “asymmetric” trans-
configuration were found to be energetically equivalent, only
the cis-conformation was observed. Secondly, what, in turn, is
the reason for the stabilisation of the trans-configuration of the
2-quinolyl analogue (5), exclusively isolated from the synthesis.
Compound 5 will be discussed first. Analogously to the ener-
getically more stable cis-configurations of the neutral bicyclic
system, the ring-opened enol forms are also slightly more stable
than the structures related to the trans-configuration. In the
following discussion, only the differences in the heats of form-
ation of the optimised closed and open enol forms were
considered as the barrier, keeping in mind that an additional
barrier of about 6 kcal molϪ1 (Scheme 2) has to be added.
Interestingly, the activation barrier which has to be overcome in
order to form the closed bicycle is, in the case of the cis-
configuration, considerably disfavoured in comparison to the
formation of a trans-configuration for 5: by far the lowest activ-
ation barrier to form the symmetric trans-configuration is
found to be only 11.6 kcal molϪ1 in comparison to 20.4 or 23.0
kcal molϪ1 for the cis-configuration, respectively. This effect is
caused by a stabilisation of the enol trans structure by attractive
electrostatic interactions between the hydrogen atom of the
hydroxy group and the nitrogen atom of the 2-quinolyl ring
due to the relatively short distance of 2.9 Å (see Fig. 10). No
such stabilisation could be found for the symmetrical cis-
form. In addition, if the differences of heats of formation
between the open and closed enolic cis- and trans-configur-
ations (∆∆∆Hf (cis–trans)) were compared with all other com-
pounds (Table 5), such a large difference cannot be observed.
Thus, for kinetic reasons, the conclusion can be drawn that
the symmetric trans-configuration has to be formed in the case
of 5.
Experimental
Chemicals and materials
Melting points were determined with a Dr Tottoli melting point
apparatus (Büchi, Switzerland) and were not corrected. IR
spectra, recorded as KBr discs, were obtained using a Perkin-
Elmer 298 spectrometer. Analysis indicated by the symbols of
the elements was within 0.4% of the theoretical value. TLC
was carried out using silica gel 60 F254 (Merck No. 5554).
As eluent a mixture of cyclohexane–ethyl acetate–methanol
(10:4:1) was used. Dry solvents were used throughout.
General procedure for the synthesis of the 2,6-diaryl substituted
piperidones
Dimethyl 3-oxoglutarate (0.02 mol) dissolved in 10 ml meth-
anol is added dropwise to a solution of 0.04 mol arylaldehyde
and 0.02 mol aqueous methylamine (40%) in 20 ml methanol
at 0 ЊC. After 12 hours at 5 ЊC the product is obtained. It is
recrystallised from ethanol.
General procedure for the synthesis of dimethyl 2,4-diphenyl-3,7-
dimethyl-9-oxo-3,7-diazabicyclo[3.3.1]nonane-1,5-dicarboxylate
1, dimethyl 2,4-bis(4-nitrophenyl)-3,7-dimethyl-9-oxo-3,7-diaza-
bicyclo[3.3.1]nonane-1,5-dicarboxylate 2, dimethyl 2,4-bis(3-
nitrophenyl)-3,7-dimethyl-9-oxo-3,7-diazabicyclo[3.3.1]nonane-
1,5-dicarboxylate 3 and dimethyl 2,4-bis(2-quinolyl)-3,7-di-
methyl-9-oxo-3,7-diazabicyclo[3.3.1]nonane-1,5-dicarboxylate 5
Piperidone (0.0025 mol) is dissolved in 20 ml methanol under
reflux. Aqueous formaldehyde (0.6 ml, 35%) and 0.4 ml aque-
ous methylamine (40%) are added and the solution is refluxed
for 10 minutes. After 3 hours at room temperature the solvent is
evaporated. The remaining oil is dissolved in ethanol and ether.
The product crystallises after 1–2 days. It is recrystallised from
ethanol. Analytical data of compounds 2–7 are described in
Table 6.
However, if the proton is removed from the hydroxy group in
the trans-form and the stable bicyclononanone with the 9-keto
group is rebuilt, the above-mentioned attractive interaction will
change to repulsive interactions which are due to the strong
negative partial charge of the carbonyl oxygen atom and the
negative charge of the nitrogen of the 2-quinolyl ring. Thus, the
trans-form becomes thermodynamically less stable than the cis-
configuration.
Taking the thermodynamic and kinetic discussions together,
it can be concluded that the preferred formation of the trans
configuration of 5 is kinetically controlled. For the 4-quinolyl
substituted compound 4, a mixture of symmetrical and asym-
metrical cis-configurations was found experimentally. It is
Preparation of the dimethyl 2,4-bis(4-quinolyl)-3,7-dimethyl-9-
oxo-3,7-diazabicyclo[3.3.1]nonane-1,5-dicarboxylate 4
Piperidone (0.0025 mol) is dissolved in 20 ml methanol
under reflux. Aqueous formaldehyde (1.2 ml, 35%) and 0.8 ml
J. Chem. Soc., Perkin Trans. 2, 1999, 1827–1834
1833

View Article Online
Table 6 Analytical data for the compounds 2–7
Found (%) (Required)
Compound
(Formula)
Yielda
(%)
Mp/ЊC
C
H
N
2 (C25H26N4O9)
3 (C25H26N4O9)
4 (C31H30N4O5)
5 (C31H30N4O5)
6 (C33H32N2O5)
7 (C33H32N2O5)
90
67
23
98
22
26
180 (trans), 215 (cis)
215 (decomp.)
202 (decomp.)
145 (decomp.)
197 (decomp.)
171–173
56.80 (57.02)
56.90 (57.02)
68.99 (69.12)
68.87 (69.12)
73.68 (73.85)
73.63 (73.85)
5.14 (4.98)
4.99 (4.98)
5.75 (5.62)
5.42 (5.62)
6.10 (6.01)
6.08 (6.01)
10.47 (10.65)
10.57 (10.65)
10.24 (10.41)
10.32 (10.41)
5.22 (5.22)
5.22 (5.22)
a Yields were not optimised.
aqueous methylamine (40%) are added and the solution is
refluxed for 20 minutes. After 3 hours at room temperature the
solvent is evaporated. The remaining oil is dissolved in ethanol.
The solvent is allowed to evaporate from the open beaker at
room temperature. It is recrystallised from ethanol in the same
manner (for analytical data see Table 6).
Electrostatic interactions were taken into consideration by
using a relative permittivity of ε = 4r with a distance dependent
function which, in our experience, gives relevant conformations
in agreement with solution conformations obtained by NMR
spectroscopy. The semiempirical calculations were carried out
using the PM318 hamiltonian as implemented within the
MOPAC 7.0 program in SYBYL. For comparison we also used
AM1 sometimes. The differences in heats of formations which
are of importance to the conclusions to be drawn in the paper
are in the same range as obtained with PM3. However, the
rather high deviations from experimental values of the heats of
formations of some nitro substituted benzene rings (nitro-
benzene, 3-nitrotoluene, 4-nitrotoluene) in the case of AM1
(9.78 kcal molϪ1 to 13.37 kcal molϪ1) in comparison to only
0.86 kcal molϪ1 to Ϫ2.83 kcal molϪ1 using PM3 led to the pref-
erence of PM3. The most stable conformations obtained from
the force field calculations were optimised with PM3 using
the keywords PRECISE, SCFCRT = 1.D-12, EF, NOINTER,
GEO-OK and GNORM = 0.1.
Preparation of the dimethyl 2,4-bis(1-naphthyl)-3,7-dimethyl-9-
oxo-3,7-diazabicyclo[3.3.1]nonane-1,5-dicarboxylate 6
Piperidone (0.0025 mol) is dissolved in 40 ml methanol under
reflux. Aqueous formaldehyde (1.2 ml, 35%) and 0.8 ml aque-
ous methylamine (40%) are added after 5, 20, 40 and 60 min-
utes. After 90 minutes the reaction is interrupted. The product
crystallises at room temperature. It is recrystallised from
ethanol (for analytical data see Table 6).
Preparation of the dimethyl 2,4-bis(2-naphthyl)-3,7-dimethyl-9-
oxo-3,7-diazabicyclo[3.3.1]nonane-1,5-dicarboxylate 7
Piperidone (0.0025 mol) is dissolved in 40 ml methanol under
reflux. Aqueous formaldehyde (1.2 ml, 35%) and 0.8 ml aque-
ous methylamine (40%) are added and the solution is refluxed
for 30 minutes. The product crystallises after 3–4 days at room
temperature. It is recrystallised from ethanol (for analytical
data see Table 6).
Acknowledgements
Thanks are due to the DFG for financial support and Mrs
Ilona Knoblauch for carefully measuring the NMR spectra.
References
NMR measurements
1 B. Kögel, T. Christoph, E. Friderichs, H.-H. Hennies, T. Mathiesen,
J. Schneider and U. Holzgrabe, CNS Drug Rev., 1998, 4, 54.
2 R. Jeyaraman and S. Avila, Chem. Rev., 1981, 81, 149.
3 N. S. Zefirov and V. A. Palyulin, Top. Stereochem., 1991, 20, 171.
4 U. Holzgrabe-Ashauer and T. Busch, Z. Naturforsch., Teil B, 1988,
43, 873.
5 W. Brandt, S. Drosihn, M. Haurand, U. Holzgrabe and C.
Nachtsheim, Arch. Pharm. Pharm. Med. Chem., 1996, 329, 311.
6 R. Haller and H. Unholzer, Arch. Pharm. (Weinheim), 1971, 304,
654.
7 H. Küppers, Hesse, U. Holzgrabe and R. Haller, Z. Naturforsch.,
Teil B, 1987, 42, 221.
8 U. Holzgrabe and E. Erciyas, Arch. Pharm. (Weinheim), 1992, 325,
657.
9 A. Samhammer, U. Holzgrabe and R. Haller, Arch. Pharm.
(Weinheim), 1989, 322, 551.
10 H. Günther, NMR-Spektroskopie, VCH, Weinheim, 1983, p. 228.
11 A. E. Derome, Modern NMR Techniques for Chemistry Research,
Pergamon Press, Oxford, 1987, p. 109.
12 H. Günther, NMR Spektroskopie, Thieme, Stuttgart, 1983, p. 75.
13 E. L. Eliel and S. H. Wilen, Stereochemistry of Organic Compounds,
Wiley, New York, 1994, p. 1150.
14 Tripos Associates Inc., 1699 S. Hanley Road, Suite 3.
15 M. Clark, R. D. Cramer III and N. Van Opdenbosch, J. Comput.
Chem., 1989, 10, 982.
All 1H and 13C NMR experiments spectra were performed on a
Varian XL 300 FT NMR spectrometer operating at 299.956
MHz (1H) and 75 MHz (13C) with a sample temperature of
30 ЊC. In the case of the 1H NMR spectra, a varying number of
scans (depending on the experiment) with a frequency range of
2200 Hz were collected into 65 000 data points, giving a digital
resolution of 0.33 Hz pointϪ1. An appropriate Gaussian func-
tion was applied before Fourier transformation to enhance the
spectra resolution. Abbreviations for data quoted are: s,
singlet; d, doublet; t, triplet; q, quartet; m, multiplet. 1H and 13
C
NMR assignments given for each compound were confirmed by
randomly running HETCOR and H,H-COSY experiments.
Theoretical methods
All molecules were constructed with the molecular modelling
program SYBYL 6.414 on Silicon Graphics Workstations. Each
flexible molecule was subjected to an extensive conformational
search by systematically rotating each rotatable bond by 30Њ
increments. An energy cut-off of 7 kcal molϪ1 above the energy
minimum was applied. All conformations obtained from the
search were grouped into families based on similarities of their
dihedral angles ( 30Њ). Using the TRIPOS force field15 and the
Powell minimizer contained in SYBYL/MAXIMIN, the lowest
energy member of each conformational family was then exten-
sively minimized. The Gasteiger method (PEOE) was used to
calculate the partial charge distribution of the molecules.16,17
16 J. Gasteiger and M. Marsili, Tetrahedron, 1980, 36, 3219.
17 J. Gasteiger and H. Saller, Angew. Chem., 1985, 97, 699.
18 J. J. P. Stewart, J. Comput. Chem., 1989, 10, 209.
Paper 8/06641H
1834
J. Chem. Soc., Perkin Trans. 2, 1999, 1827–1834