Enantiomer and conformer recognition of (+) and (-)-disparlure and their analogs by the pheromone binding proteins of the gypsy moth, Lymantria dispar

Article abstract of DOI:10.1016/j.bmc.2013.01.043

Adult female gypsy moths produce a sex pheromone (+)-(7R,8S)-2-methyl-7,8- epoxyoctadecane, (+)-disparlure, to attract male gypsy moths. To better understand the recognition of (+)-disparlure by the male's olfactory system, we synthesized racemic and enantiopure oxa and thia analogs of (+)-disparlure (ee > 98%). Ab initio calculations of the conformeric landscapes around the dihedral angles C_5-6-7-8 and C_7-8-9-10 of (+)-disparlure and corresponding dihedral angles of analogs revealed that introduction of the heteroatom changes the conformeric landscape around these important epitopes. The energy difference between HOMO and LUMO decreased after oxygen or sulfur was introduced into the backbone. Consistent with this, an enhancement of binding affinity between sulfur analogs and the pheromone-binding proteins (PBPs) was observed in vitro. Docking of the pheromone and analogs onto models of the two known PBPs of the gypsy moth revealed that the internal binding pocket of PBP1 showed higher selectivity than that of PBP2, consistent with in vitro binding assays. Further energy analysis revealed that enantiomers adopted different conformations with different energies when docked in the internal binding pocket of PBPs, resulting in enantiomer discrimination of PBPs towards disparlure and its analogs.

Full text of DOI:10.1016/j.bmc.2013.01.043

Bioorganic & Medicinal Chemistry 21 (2013) 1811–1822
Contents lists available at SciVerse ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Enantiomer and conformer recognition of (+) and (ꢀ)-disparlure
and their analogs by the pheromone binding proteins of
the gypsy moth, Lymantria dispar
⇑
Yang Yu, Erika Plettner
Department of Chemistry, Simon Fraser University, 8888 University Drive, Burnaby, British Columbia, Canada
a r t i c l e i n f o
a b s t r a c t
Article history:
Adult female gypsy moths produce a sex pheromone (+)-(7R,8S)-2-methyl-7,8-epoxyoctadecane, (+)-dis-
parlure, to attract male gypsy moths. To better understand the recognition of (+)-disparlure by the male’s
olfactory system, we synthesized racemic and enantiopure oxa and thia analogs of (+)-disparlure
(ee > 98%). Ab initio calculations of the conformeric landscapes around the dihedral angles C5–6–7–8 and
C7–8–9–10 of (+)-disparlure and corresponding dihedral angles of analogs revealed that introduction of
the heteroatom changes the conformeric landscape around these important epitopes. The energy differ-
ence between HOMO and LUMO decreased after oxygen or sulfur was introduced into the backbone. Con-
sistent with this, an enhancement of binding affinity between sulfur analogs and the pheromone-binding
proteins (PBPs) was observed in vitro. Docking of the pheromone and analogs onto models of the two
known PBPs of the gypsy moth revealed that the internal binding pocket of PBP1 showed higher selectiv-
ity than that of PBP2, consistent with in vitro binding assays. Further energy analysis revealed that enan-
tiomers adopted different conformations with different energies when docked in the internal binding
pocket of PBPs, resulting in enantiomer discrimination of PBPs towards disparlure and its analogs.
Ó 2013 Elsevier Ltd. All rights reserved.
Received 30 November 2012
Revised 11 January 2013
Accepted 18 January 2013
Available online 31 January 2013
Keywords:
Thioether
Ether
Epoxide
Enantioselectivity
HOMO
LUMO
1. Introduction
previous studies have shown that PBP1 binds (ꢀ)-1 more strongly
than (+)-1, whereas PBP2 was the opposite.7,8 The extent to which
Gypsy moth, Lymantria dispar, larvae cause serious defoliation
of forests in Asia, Europe, and North America. The major compo-
nent of the sex pheromone produced by the adult female moths
is (+)-disparlure, (+)-(7R,8S)-2-methyl-7,8-epoxyoctadecane (+)-1
(Fig. 1).1,2 The nun moth, a related species, produces not only (+)-
1 to attract male nun moths but also (ꢀ)-disparlure, (ꢀ)-1. This
enantiomer is neither attractive nor repellent to the male nun
moths but prevents the upwind flight behavior of male gypsy
moths when presented with (+)-1.3,4 The number of male gypsy
moths lured into traps decreases significantly, even when only
1% (ꢀ)-disparlure is mixed with (+)-disparlure (98% ee).5 This
remarkable behavioral enatioselectivity is due to the high selectiv-
ity of the moth’s olfactory system.
The antennae of moths are covered with sensory hairs (sensilla),
hollow structures that contain 2–3 dendrites of sensory neurons.
The neurons in sensilla are surrounded by sensillar lymph, an elec-
trolyte solution that is rich in pheromone-binding proteins (PBPs).
Two PBPs, LdisPBP1 and LdisPBP2 (PBP1 and PBP2 in the following
discussion), have been identified from gypsy moth antennae.6 Our
PBPs contribute to the selectivity and sensitivity of sensory hairs is
not known, but studies with the silk moth have revealed that the
odorant receptors (ORs), expressed on the neuronal dendrites in
the sensory hairs, are a major determinant of the response profile
of that neuron.9 Furthermore, studies with the fruit fly, Drosophila
melanogaster, have revealed that the PBP and the OR are both re-
quired for hairs to respond.10,11 Therefore, there are two major
determinants that are likely involved in pheromone molecular rec-
ognition in insects, including stereoisomer recognition in the gypsy
moth: (1) the interaction between the odorant and the PBPs and
(2) the interaction between the PBP-odorant complex and proteins
in the dendritic membrane.10
PBPs bind pheromones selectively, for transport of the phero-
mone from the cuticle of the hair to the neuron and for ultimate
removal of pheromone from the lymph.12 Several studies have sug-
gested that PBPs might have neuron-activating and non-activating
(scavenging) conformations, and that these depend on the li-
gand(s) bound.10,13 Our kinetic binding studies with PBPs suggest
that PBPs bind pheromone in two stages: (1) non-selective external
binding, after interaction between the PBP and the pheromone, and
(2) slow and selective internal binding.13,14 A third step appears to
be a very slow aggregation of PBP-pheromone complexes into
⇑
Corresponding author. Tel.: +1 778 782 3586; fax: +1 778 782 3765.
E-mail address: plettner@sfu.ca (E. Plettner).
0968-0896/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.bmc.2013.01.043

1812
Y. Yu, E. Plettner / Bioorg. Med. Chem. 21 (2013) 1811–1822
factors, such as entropic contributions or dynamic trajectories of
the pheromones binding to ORs or PBPs, also play a role in
recognition.
7
8
1
2
5
3
18
O
O
9
A strategy for preparation of disparlure analogues has been the
replacement of the epoxide oxygen with various other functional-
ities, such as aziridine (NH), CH2 or CCl2.23 Some of these com-
pounds elicited EAG responses and/or bound to the PBPs,12 but
none were active in the field.23 Inspired by the observation that li-
pid insect pheromones appeared to be recognized in a bent, cisoid
conformation,22,24 a series of conformationally constrained analogs
of disparlure have been synthesized.20 Although these compounds
elicited weak to no antennal responses themselves, subsets of
these compounds showed promising agonistic effects in the EAG.
In our previous initial study, disparlure analogs with oxygen
replacements within the hydrocarbon chains, namely (ꢀ) and
(+)-5-oxa analogs (2), caused strong EAG responses and had a bind-
ing affinity with PBP1 similar to (ꢀ)-1.8 We were curious whether
the replacement of the 10-carbon with an oxygen atom (3) and the
5- or 10-carbon atom with a sulfur atom (4 and 5, Fig. 1) would af-
fect the enantioselectivity of the two PBPs (Fig. 1).
The objective of this work was to prepare (+) and (ꢀ) analogs
that differ only subtly from (+)-1 or (ꢀ)-1, but that have altered
conformational regimes. We chose heteroatom replacements at
the 5- and 10-positions, because these are expected to introduce
new dipole and van der Waals interactions within the molecules
(see below), as well as with binding partners. In the current study,
we synthesized enantiopure (+)-1, (ꢀ)-1 and their respective 5-
and 10-oxa and thia analogs to probe the enantioselectivity and
conformation selectivity of the olfactory system of the gypsy moth.
We outline here how the oxa and thia analogs alter conformational
regimes and PBP-binding properties. Specifically, we delineate a
three-step process for the study of these analogs. (1) We present
ab initio calculations of the conformational landscapes around
the 5–6–7–8 and 7–8–9–10 bonds of the pheromone and its analog
enantiomers. (2) We have performed in vitro binding studies and
in silico docking studies with the pheromone and its analog enan-
tiomers and interpret the results from the in vitro studies in terms
of the major docking study results. Finally, we demonstrate how
the altered in vitro selectivity of the PBPs can be explained by
the calculated binding interactions of PBP–ligand complexes.
(+)-1
(-)-2
(+)-3
(-)-1
(+)-2
(-)-3
O
O
O
O
O
O
O
O
S
O
S
O
O
(-)-4
(+)-4
O
S
S
(+)-5
(-)-5
Figure 1. Natural pheromone (+)-1 and target analogs.
dimers and multimers that may be inactive.13 Therefore, the dy-
namic trajectories of the ligands, from the free form to the various
PBP-bound forms likely contribute to the dynamic aspects of the
olfactory responses: the latency period, activation and deactivation
rates. These aspects have been shown to determine the coordi-
nated activation of multiple neurons and the overall strength of
the olfactory responses.15
Enantioselectivity plays an important role in insect pheromone
chemical ecology16 but the mechanisms of olfactory enantioselec-
tivity in insects are still poorly understood. It has been hypothe-
sized that pheromones have active conformations, and that these
are at or near energy minima.17,18 Previous work has shown that
a pheromone with a Z alkene can adopt particular chiral conforma-
tions that are recognized by the insects.19 We hypothesize that (+)-
1 will also adopt a particular conformational regime that reflects
the asymmetry of the molecule and that is the active form. Previ-
ous work with disparlure analogs revealed that a cisoid conforma-
tion around the epoxide moiety might be the recognized one and
also that the shorter, methyl branched, chain is distinguished from
the n-decyl chain, in the active space.20
Previous structure–activity studies with moths that have unsat-
urated, non-chiral acetate pheromones have parted from the
hypothesis that the pheromone is recognized at or near its lowest
energy conformation.21,22 These studies have been quite successful
at explaining the relationship between the electroantennogram
(EAG) activities and the differences between global energy minima
of pheromone analogs and their conformations when forced into
the space occupied by the global minimum of the pheromone.
The larger the energy investment an analog had to make, to fit into
the space delineated by the pheromone’s global minimum, the
lower the activity was. Our own similar study, with conformation-
ally constrained odorants, has shown that the active space of the
gypsy moth pheromone is likely delineated by a cisoid, extended
conformation of (+)-1.20 This active space is most likely reflective
of the combined binding selectivities of the OR and the PBP. Several
analogs of pheromones do not fit into the active space of the pher-
omone, suggesting that the energy minima of the pheromone are
not the only structures recognized by the system and that other
2. Results and discussion
2.1. Synthesis of oxa and thia analogs
The key step of synthesis of oxa and thia analogs was the enzy-
matic resolution of intermediates 6 and 7. We have successfully
separated the racemate of 2,3-epoxide alcohol 6 using an MCTA-I
chiral column, however this method is time consuming and low
in loading ratio.8 Lipase catalyzed kinetic resolution of 2,3-epoxide
alcohols in the synthesis of enantiopure disparlure has been re-
ported before,25–28 but the reported ee values were not high en-
ough for the purpose of our study. Therefore, five lipases, Amano
lipase AK (AK), Amano lipase AY-30 (AY), Amano lipase PS (PS),
Aspergillus niger lipase (AN), and type II porcine pancreatic lipase
(PP) were screened for kinetic resolution of ( )-6 (Supplementary
Tables S1–S3 and Scheme S3). Lipases AK, AY, PS and AN com-
pletely acylated the substrate without any enantioselectivity,
while lipase PP catalyzed 55% conversion with a 96% ee of sub-
strate in two hours with an E value of 32 when diethyl ether was
used as the solvent and acetic anhydride as the acyl donor. Time
course experiments suggested that the one-hour reaction can give
the optimal yields for both enantiomers (Supplementary Fig. S1). A
second resolution increased the ee of (+)-6 to higher than 98%. The
substrate (ꢀ)-6 (80% ee, after first resolution) was resolved again to

Y. Yu, E. Plettner / Bioorg. Med. Chem. 21 (2013) 1811–1822
1813
OH
OTs
Et3N / DMAP / CH2Cl2
TsCl, 98%
O
O
6
8
S
HS
NaH / THF
72%
O
4
Scheme 1. Synthesis of 5-thia analog 4.
Br
O
O
OH
O
NaH / DMF / THF
72 %
7
3
TsCl
95 %
Et3N / DMAP / CH2Cl2
HS
O
O
S
OTs
NaH / THF
70 %
9
5
Scheme 2. Synthesis of 10-oxa and -thia analogs 3 and 5.
Table 1
Dissociation (Kd) binding assays using PBP1 and PBP2 with disparlure and its oxa and thia analogs
PBPs
1
2
3
4
5
PBP 1
PBP 1
PBP 1
PBP 2
PBP 2
PBP 2
(+)
(ꢀ)
( )
(+)
(ꢀ)
( )
13.3 1.9 (5)
7.3 1.1 (5)
15.9 1.3 (5)
12.8 3.0 (5)
29.9 5.8 (5)
11.4 2.1 (5)
25.2 2.1 (4)
17.6 2.0 (5)
43.8 5.3 (5)
46.6 5.7 (4)
35.1 1.5 (4)
10.3 3.7 (5)
48.3 9.1 (5)
11.6 1.3 (5)
26.3 5.7 (5)
22.9 2.8 (5)
18.3 3.7 (5)
21.6 3.3 (5)
8.5 1.2 (5)
7.7 1.4 (5)
41.9 10.4 (5)
17.7 3.8 (6)
5.7 0.8 (4)
21.4 4.0 (8)
22.7 4.2 (8)
23.6 9.1 (5)
8.6 1.1 (8)
15.5 2.5 (7)
61.9 24.6 (5)
46.2 12.1 (5)
Assays were carried out in N replicates, and values represent the average S.E. (N). (Kd,
l
M, pH 7.0 0.1). The protein concentration was 4
lM. The protein and ligand were
incubated in a 1:1 ratio.
give a final ee of 99%. Racemate ( )-7 was also resolved using lipase
PP in two resolutions to achieve (ꢀ)-7 of 98% ee and three resolu-
tions to obtain (+)-7 of 99% ee. Lipase PP can be re-used without
significant loss of enantioselectivity (Supplementary Fig. S2). How-
ever, the enzyme was used repeatedly only for the first round of ki-
netic resolution to make sure the final ees exceeded 98%.
the 10-methylene group with oxygen or sulfur did not reverse
the optical rotations of analogs 3 and 5 relative to 1.
2.2. Binding with PBPs
Binding assays of these analogs (Table 1) to two known gypsy
moth pheromone-binding proteins, PBP1 and PBP2, were carried
out at pH 7.0 in insect Ringer, using reported procedures.7,8 The
binding affinities of PBP1 for (ꢀ)-2 and (+)-2 were about half of
the affinity of PBP1 with (ꢀ)-1 and (+)-1, respectively. The binding
affinity of PBP1 for (ꢀ)-2 was significantly higher than that of PBP1
to (+)-2 (t-test, P = 0.04). The binding affinities of PBP1 for (+)-3 and
(ꢀ)-3 were about a quarter and a half of the affinities when PBP1
bound with (+)-1 and (ꢀ)-1, respectively. Furthermore, PBP1
bound (ꢀ)-3 more strongly than (+)-3 (P = 0.015). It has been re-
ported that PBP1 had higher binding affinity to (ꢀ)-1 than (+)-1.7
Our current results also showed that PBP1 had enantioselelective
preference towards not only (ꢀ)-1 (P = 0.03, t-test with (+)-1 in
current study) but also (ꢀ)-2 and (ꢀ)-3. However, (ꢀ)-2 had the
opposite absolute configuration than (ꢀ)-1. PBP1 bound (+)-4
and (ꢀ)-4 with similar affinity as (ꢀ)-1 but had weaker affinity
The syntheses of racemic and enantiopure oxa analog 3 and thia
analogs 4 and 5 are shown in Schemes 1 and 2. Tosylation of alco-
hol 6 gave compound 8 in which the OTs leaving group was easily
displaced by the nucleophile 3-methylbutane-1-thiol. Thus, thia
analog 4 was formed smoothly in one hour at room temperature
(Scheme 1). Notably, (ꢀ)-4 has the same absolute configuration
as (+)-1. The replacement of the 5 methylene group with sulfur re-
versed the optical rotation in 5-thia analog (ꢀ)-4 relative to the
enantiomer of (+)-1, which was noticed in our previous synthesis
of (ꢀ)-2.8 Williamson’s etherification of epoxy alcohol 7 with 1-
bromooctane provided the racemic 10-oxa analog 3 in good yield
(Scheme 2). The addition of DMF to the reaction mixture in THF in-
creased the yield compared with our previous synthesis.8 Follow-
ing the same methods as outlined for 4, 10-thia analog 5 was
synthesized via tosylate 9 (Scheme 2). However, replacement of

1814
Y. Yu, E. Plettner / Bioorg. Med. Chem. 21 (2013) 1811–1822
towards (+)-5 and (ꢀ)-5. Interestingly, PBP1 bound both (+)-4 and
(ꢀ)-4 or (+)-5 and (ꢀ)-5 equally (P = 0.68 and 0.79, respectively). In
contrast, PBP2 bound (ꢀ)-4 more strongly than (+)-4 (P = 0.02) and
(+)-5 more strongly than (ꢀ)-5 (P = 0.02). PBP2 did not show
enantioselectivity between (+)-2 and (ꢀ)-2 or (+)-3 and (ꢀ)-3
(P = 0.14 and 0.34, respectively), but, consistent with previous re-
ports, PBP2 showed higher binding affinity for (+)-1 than (ꢀ)-1
(P = 0.04). Therefore, PBP1 distinguished enantiomers of oxa ana-
logs better than those of thia analogs and PBP2 distinguished enan-
tiomers of thia analogs better than those of oxa analogs. It seems
PBPs are the first layer of the olfactory system to discriminate
not only enantiomers of the natural pheromone but also a variety
of heteroatom analogs.
The subtle change in the structure of disparlure with one het-
eroatom changed binding affinities and reversed the recognition
of absolute configuration around the epoxide moiety relative to
the original hydrocarbon pheromone. However, the enantioselec-
tivity of PBPs towards oxa and thia analogs with the same absolute
configuration remained the same. Racemates showed three differ-
ent effects for different combinations of PBP and ligand. First, PBP1
and racemic 1, 2, and 4, as well as PBP2 and racemic 4 and 5
showed negative blend effects, that is, much weaker binding of
the racemate than of the individual enantiomers. Second, PBP2
with 2 showed a positive blend effect, that is, stronger binding of
the racemate than of the individual enantiomers. Third, PBP1 with
3 and 5, as well as PBP2 with 3 showed no blend effect. Lastly, PBP2
had a positive blend effect with (ꢀ)-1 but not with (+)-1. Such
blend effects have been noticed in other studies7,12,29,30 and they
have been attributed to the formation of dimers and multi-
mers7,12,29 or to the multi-step binding process that has been de-
tected for both binding proteins with disparlure13 or
fluorescent ligand.14
a
2.3. Ab initio calculations of (+)-disparlure and its analogs
The energy penalty hypothesis, successfully applied in activity
studies of pheromone analogs, states that the greater the energy
cost for the analogs to adopt the bioactive conformation, the lower
the activity in antennal recordings.18,31–33 We propose that this
also applies to binding affinity assays. Therefore, computational
calculations and simulation experiments were carried out. The ab
initio calculation of 1 and its analogs were made feasible with
the resources of High Performance Computing (HPC) on Westgrid,
a subsidiary of Compute Canada. Both enantiomers of disparlure
gave almost identical potential energy surfaces (PES) through the
scanning of dihedral angles C5–6–7–8 (DHA1) and C7–8–9–10 (DHA2)
in 36 degree increments (Fig. 2). The slight differences in the sur-
faces are due to the nature of quantum mechanical calculations.
The global minimum conformer of (+)-1 in this scan was at node
[1,1], DHA1 = 144, DHA2 = 94, one maximum conformer was at
Figure 2. The potential energy surfaces (PESs) of (+)-1 and (ꢀ)-1. Each PES was calculated using density functional theory (DFT, B3LYP wave function, basis set 6-31G) by
variation of dihedral angles C5–6–7–8 (DHA1) and C7–8–9–10 (DHA2) in 36 degree increments. Top: the potential energy surfaces (PESs) of (+)-1 (left) and (ꢀ)-1 (right). The green
node in each plot represents the global minimum conformer of the PES. The structures of the global minimum conformers of (+)-1 (bottom left) and (ꢀ)-1 (bottom right) are
shown.

Y. Yu, E. Plettner / Bioorg. Med. Chem. 21 (2013) 1811–1822
1815
Figure 3. The potential energy surfaces (PESs) of (ꢀ)-2, (+)-3, (ꢀ)-4 and (+)-5. Each PES was calculated using density functional theory (DFT, B3LYP wave function, basis set 6-
31G) by variation of dihedral angles O5–C6–7–8 (DHA1) and C7–8–9–10 (DHA2), C5–6–7–8 (DHA1) and C7–8–9O10 (DHA2), S5–C6–7–8 (DHA1) and C7–8–9–10 (DHA2), and C5–6–7–8
(DHA1) and C7–8–9S10 (DHA2) of (ꢀ)-2, (+)-3, (ꢀ)-4 and (+)-5, respectively. The dihedral angle increment was 36 degrees. The PESs are (ꢀ)-2 (top left), (+)-3 (top right), (ꢀ)-4
(bottom left) and (+)-5 (bottom right). The green node in each plot represents the global minimum conformer.
node [6,3], DHA1 = ꢀ36, DHA2 = 22, and a second small barrier
enantiomers of 1 in the conformational space around the dihedral
angles sampled (Fig. 3). The major differences were: (1) the occur-
rence of two maxima of (ꢀ)-2 at node [3,3], DHA1 = 92, DHA2 = 21,
for the first maximum and node [6,3], DHA1 = ꢀ16, DHA2 = 21 for
the second maximum, and (2) there was a third barrier, approxi-
mately half as high as the global maximum, at node [3,9],
DHA1 = 92, DHA2 = ꢀ195. The PES pattern of (+)-3 was about same
as (ꢀ)-2 with two maxima at node [3,3], DHA1 = 92, DHA2 = 21, for
the first maximum, and node [6,3], DHA1 = ꢀ16, DHA2 = 21 for the
second maximum. The PES of (ꢀ)-4 followed the same landscape as
(+)-1, with one maximum at node [6,3], DHA1 = ꢀ32, DHA2 = 20,
and node [8,8], DHA1 = ꢀ104, DHA2 = ꢀ160 for the global mini-
mum. A noticeable change was a steep decline from the energy
maximum to the global minimum as dihedral angles rotated in
the PES of (ꢀ)-4. The PES of (+)-5 was similar to that of (+)-1 with
1
(approximately
/ as high as the global maximum) occurred at
3
node [9,3], DHA1 = ꢀ144, DHA2 = 22.
Any small differences in energy or dihedral angles between
enantiomers reflect the errors in the calculation, and not real dif-
ferences between enantiomers. The differences between the analog
enantiomers and the pheromone, however, are larger than these
calculation errors. These differences in the energy landscape are
also apparent in the relative positions of the various nodes, with
regard to energy (Supplementary Table S4) and dihedral angle vec-
tor (Supplementary Table S5). The PESs of both enantiomers are al-
most identical as shown in the examples of (+)-1 and (ꢀ)-1 above
(Fig. 2). Therefore, only PESs of (ꢀ)-2, (+)-3, (ꢀ)-4 and (+)-5 are
shown in Figure 3, as these four enantiomers have the same abso-
lute configuration as (+)-1. The enantiomers of 2 differed from the

1816
Y. Yu, E. Plettner / Bioorg. Med. Chem. 21 (2013) 1811–1822
Figure 4. The highest occupied molecular orbital (HOMO) of (+)-1, (ꢀ)-2, (+)-3, (ꢀ)-4 and (+)-5. Blue arrows are dipole moment vectors. For surface display, red represents
negative isodensity values, green represents positive isodensity values. Contours of the HOMO are shown.
Figure 5. The lowest unoccupied molecular orbital (LUMO) of (+)-1, (ꢀ)-2, (+)-3, (ꢀ)-4 and (+)-5. Blue arrows are dipole moment vectors. For surface display, red represents
negative isodensity values, green represents positive isodensity values. Contours of the LUMO are shown.

Y. Yu, E. Plettner / Bioorg. Med. Chem. 21 (2013) 1811–1822
1817
Table 2
PBP1
PBP2
Frontier-orbital energy (eV) of (+)-disparlure and its analogs
HOMO
LUMO
D|H–L|
α7
II
α2
α1
α7
(+)-1
(ꢀ)-2
(+)-3
(ꢀ)-4
(+)-5
ꢀ0.25046
ꢀ0.25176
ꢀ0.25033
ꢀ0.22250
ꢀ0.22423
0.08967
0.06091
0.06511
0.03039
0.02883
0.34013
0.31267
0.31544
0.25289
0.25306
II
α1
III
I
α2
α3
α5
A
V
I
III
α4
α4
α6
α5
α3
IV
one global maximum at node [6,3], DHA1 = ꢀ39, DHA2 = 19, and a
global minimum at node [11,11], DHA1 = ꢀ219, DHA2 = ꢀ269. The
energies of local minima of (ꢀ)-2, (ꢀ)-4 and (+)-5 were much high-
er than that of the global minimum (average differences were 3.9,
2.5 and 3.4 kJ/mol, respectively), whereas in (+)-1 and (+)-3 the
energies of the local minima were quite close to the global mini-
mum of the PES (average differences were 0.81 and 1.1 kJ/mol,
respectively) (Supplementary Table S4).
α6
IV
α7
II
• 7
α2
α1
II
α1
α5
α2
III
α3
B
V
VII
α3
I
The contribution of the oxygen and sulfur to the HOMO and
LUMO, both features a binding partner would interact with (for
example, in the acceptance of H-bonds, sulfur–sulfur or van der
Waals interactions), are shown in Figures 4 and 5 as well as Table 2.
Comparing the HOMO of (+)-1 to those of analogs, two major dif-
ferences become apparent. First, the electron density along the al-
kyl chains is more evenly distributed between the two pendant
chains in the HOMO of (+)-1 than in all analogs. Second, in (+)-1
the epoxide lone pairs are more prominent in the electron density
than in analogs. In (ꢀ)-2 and (+)-3, the oxygen in the chain draws a
large portion of electron density to the shorter and longer chains,
respectively, reducing the importance of the epoxide lone pairs
in the interaction with PBPs. The electron density was completely
drawn to the sulfur atom in the cases of (ꢀ)-4 and (+)-5. The LUMO
of (+)-1 was localized on the longer chain close to the epoxide oxy-
gen and extended to the middle of the long chain. In contrast, the
LUMO of (ꢀ)-2 spread out from oxygen to the end of the shorter
chain. Analog (+)-3 had LUMO close to the oxygen on the longer
chain. The distributions of the LUMOs of (ꢀ)-4 and (+)-5 were sim-
ilar to those in (ꢀ)-2 and (+)-3, respectively, but more centrally lo-
cated around the sulfur atom. Of all the HOMO and LUMO pairs of
oxa and thia analogs, those of (ꢀ)-2 resembled those of (+)-1 the
best.
The application of frontier orbitals to understand the biological
activity has been reported previously.34–36 A correlation has been
observed between LUMO energy and hypolipidemic activity of
phthalimide and related analogs.34 We therefore wanted to com-
pare differences between (+)-1 and its analogs in terms of HOMO
and LUMO energies with binding affinity. As we expected, the
introduction of heavier heteroatoms, with higher electronegativity
than the CH2 carbon, decreased the LUMO energy, to 0.03 and
0.06 eV lower than that of (+)-1 for oxa and thia analogs, respec-
tively. However, the HOMO energy of oxa analogs did not increase
and that of thia analogs increased by 0.03 eV. It has been recog-
nized that the lower the LUMO the easier for a molecule to accept
electrons, which will increase the binding affinity of the molecule
by increasing the electrostatic interaction between it and nucleo-
philic residues of pheromone binding proteins. In contrast, higher
HOMO energy will make the analogs donate electron density more
easily to electrophilic residues of pheromone binding proteins,
which will increase the binding affinity as well. As a result, the
smaller the energy difference between HOMO and LUMO the stron-
ger the H-bonding and dipole/dipole contributions to binding
α5
I
α4
α4
α6
IV
VI
α6
IV
Figure 6. Binding sites on PBP1 and PBP2. The models A (PBP1 and PBP2) were
obtained from Swiss Model online homology modeling program with the 4 or 6 C-
terminal amino acids deleted, respectively. The models B (PBP1 and PBP2) were
obtained from an MOE homology modeling program with the full length of amino
acids. Seven
a helices are numbered from a1 to a7. Binding sites are shown as
dummy atoms and numbered from I to VII when appropriate. (A) PBP1, Green, inner
binding site (I); yellow, C terminal site (II); pink, helix 2 shoulder site (III); red,
channel site (IV); blue, helix six shoulder site (V). (A) PBP2, green, inner binding site
(I); yellow, C-terminal site (II); pink, loop 2 shoulder site (III); red, channel site (IV).
(B) PBP1, green, inner binding site (I); yellow, C terminal site (II); pink, left helix 1
shoulder site (III); red, channel site (IV); violet, external channel site (VI). (B) PBP2,
green, inner binding site 1 (I); yellow, C-terminal site (II); red, channel site (IV);
pink, loop 2 shoulder site (V); dark green, inner site 2 (VII).
However, with the highest value of
DEHOMO-LUMO, (+)-1 bound to
both PBPs most strongly, indicating that electrostatic binding en-
ergy, from H-bonding or dipole/dipole interactions is not the most
important contribution to (+)-1-PBP binding. An extended confor-
mation of a long-chain molecule, such as 1, can have large van
der Waals interactions which may compensate for the weak di-
pole/dipole and H-bonding interactions with binding proteins.
2.4. In silico simulation of binding between PBPs and (+)-1 and
analogs
To understand the observed enantioselective binding affinity,
homology modeling of PBPs and subsequent docking simulations
using Swiss Model37,38 and MOE (Montréal, Canada) were per-
formed. The highest scored conformer from docking simulations
would represent the strongest binding conformer in the binding
assays. According to the energy penalty hypothesis, the total en-
ergy difference between the highest scored conformer and its cor-
responding global minimum conformer is the energy cost for the
analog to pass from the free form in solution to the bound form
in the protein binding pocket. The energy cost of each enantiomer
in a pair should be different if an enantioselectivity in the binding
assays is observed.
We built the protein homology models of PBP1 and PBP2 be-
cause their crystal structures are not yet available. The four and
six C-terminal amino acids (of PBP1 and PBP2, respectively) in
the homology model (Fig. 6 A) produced by Swiss Model were de-
leted to be compatible with the template protein crystal (PDB
code: 2WCJ.A.). The full length sequences of PBP1 and PBP2 were
might be. For example, although the
DEHOMO-LUMO values of thia
analogs were only 0.06 eV smaller than those of oxa analogs, the
binding affinities of thia analogs in general were two times higher
than those of oxa analogs. Strong S-S interactions may be a sub-
stantial contribution to the high binding affinities of thia analogs.

1818
Y. Yu, E. Plettner / Bioorg. Med. Chem. 21 (2013) 1811–1822
submitted for homology modeling using MOE (Fig. 6 B). Both the
short and full length PBP1 models obtained from Swiss Model
and MOE modeling were similar and all gave reasonable Rama-
chandran plots (Supplementary Fig. S3). The short length PBP2
model basically adopted the template protein structure but was
slightly loose compared with the PBP2 model obtained from the
full length sequence. The binding sites of all four PBP models were
searched using MOE with default settings. The single internal dock-
ing site of PBP1 had 164 contact atoms in the short length model
(site I, Fig. 6, A) and 148 contact atoms in the full length model (site
I, Fig. 6, B). In both models this docking site started with MET5-
MET8THR9, and it ended with PHE119GLU122 (full length model)
or ALA116PHE119 (short length model). There were three small
docking sites around the C-terminus of short length PBP1, one di-
rectly under the C-terminus (site II) and two on each shoulder site
of the C-terminus (sites III and V, respectively, Fig. 6, A). The full
length PBP1 model had one binding site directly under the C-termi-
nus and one shoulder site below the C-terminus (site II and III,
respectively, Fig. 6, B). In the short length PBP2, a single large inter-
nal binding site having 216 contact atoms (site I, Fig. 6, A) in the
binding pocket together with several small surface docking sites
were found. The large single docking site was divided into two
smaller sites with 102 and 129 contact atoms in the PBP2 full
length model (sites I and IV, respectively, Fig. 6, B). Site I (102 con-
tact atoms) is deep inside the protein structure, and site IV (129
contact atoms) is a channel connecting the outside to the inside
pockets.
binding sites with compacted conformations (Supplementary
Figs. S4 and S5 for short length model and Figs. S7–S16 for full
length model). Lys94, a key residue in the interaction between li-
gands and PBP1, donated a hydrogen atom forming a H-bond with
(ꢀ)-1,39 (+)-2, (ꢀ)-2, (+)-4, (ꢀ)-4, and (+)-5, as well as forming two
H-bonds with (+)-3 and (ꢀ)-3. No H-bonds were formed between
Lys94 and (+)-1 and (ꢀ)-5, when these ligands docked in PBP1,
indicating Lys94 plays an important role in the mechanism of the
enantioselectivity between these enantiomers (Figs 7 and 8). The
conformations of disparlure and analogs docked in the internal site
of PBP2 were stretched poses (Supplementary Figs. S4 and S6 for
short length model; Figs. S17–S26 for full length model). When
PBP2 was docked with (+)-1, (+)-2 and (ꢀ)-2, an H-bond was
formed between Gln111 and the epoxide oxygen of (+)-1 and O5
of (+)-2, but no H-bond was formed between PBP2 and (ꢀ)-2 (Figs
7 and 8). PBP2 used the amino acid Asn115 to form H-bonds with
epoxide oxygen atoms of (ꢀ)-1, (ꢀ)-3, (+)-5 and (ꢀ)-5 as well as
O10 of (+)-3. In addition to an Asn115-S10 interaction between
PBP2 and (+)-4, an H-
p interaction was also presented between
Phe12 and H11 of (+)-4. Three H-acceptor interactions were de-
tected between Asn115-Oepoxide, Asn115-S5 and Gln111-S5 of (ꢀ)-
4 when it was docked on PBP2. Thus, two amino acid residues
Gln111 and Asn115 are key recognition centers for the orientation
of ligands docked in PBP2, which leads to the possible enantiose-
lectivity of PBP2.
Differences in total energy (DE) between global minima obtained
from PES and the docked poses obtained from short length and full
length models were calculated and the results are presented in
Tables 3, S6 and S7. To understand the enantioselectivity, we calcu-
Docking simulations were carried out with all the binding sites
identified by the MOE program using an induced fit protocol. Both
enantiomers of disparlure and analogs docked in PBP1 internal
lated the difference between the
DE values for enantiomeric pairs
Figure 7. Cut-away views of 3D models of the interactions between 1 and PBPs. Ligands are green skeletons and amino acid residues (Lys94 in PBP1 as well as Gln111 and
Asn115 in PBP2) that had H-bonding interaction with ligands are highlighted in blue. Hydrogen atoms were hidden for clarity. (A) PBP1 with (+)-1, (B) PBP1 with (ꢀ)-1, (C)
PBP2 with (+)-1, (D) PBP2 with (ꢀ)-1.

Y. Yu, E. Plettner / Bioorg. Med. Chem. 21 (2013) 1811–1822
1819
Figure 8. Cut-away views of 3D models of the interactions between 2 and PBPs. Ligands are green skeletons and amino acid residues (Lys94 in PBP1 and Gln111 in PBP2) that
had H-bonding interaction with ligands are highlighted in blue. Hydrogen atoms were hidden for clarity. (A) PBP1 with (+)-2, (B) PBP1 with (ꢀ)-2, (C) PBP2 with (+)-2, and (D)
PBP2 with (ꢀ)-2.
Table 3
The differences in total energy (
(ꢀ)-4, (+)-5 and (ꢀ)-5 docked in the inner site of models of PBP1 and PBP2
D
E, kJ/mol) between the best bound conformers and their respective global minima, of compounds (+)-1, (ꢀ)-1, (+)-2, (ꢀ)-2, (+)-3, (ꢀ)-3, (+)-4,
SWISS-model
MOE model
Average of both models
range
D
E
DD
E
D
E
DD
E
DDE
PBPs
1
2
1
2
1
2
1
2
1
2
(+)-1
(ꢀ)-1
(ꢀ)-2
(+)-2
(+)-3
(ꢀ)-3
(ꢀ)-4
(+)-4
(+)-5
(ꢀ)-5
242
212
140
188
227
202
224
213
232
237
89.7
67.4
88.0
90.6
115
101
170
125
162
185
30
22
ꢀ3
14
45
132
126
160
143
153
149
159
167
155
156
173
115
110
123
137
141
135
157
135
139
6
58
18 12
40 18
ꢀ48
25
17
3
ꢀ13
ꢀ4
ꢀ16 22
ꢀ8
5
9
14 11
5
11
ꢀ8
ꢀ1
ꢀ21
ꢀ4
2
9
2
12 23
ꢀ13 10
ꢀ5
ꢀ23
ꢀ3
(
DDE) and took the average of both models. Even with slight differ-
was 40 kJ/mol. For oxa analogs, the large DDE values between
enantiomers bound to PBP1 and the small DDE values between
enantiomers bound to PBP2 correlate well with the binding assay
results, in that PBP1 discriminated between enantiomers of ana-
logs 2 and 3 but PBP2 did not.
ences between two models, the DDE values were large. The values
from docking of PBP2 were less reliable than the PBP1 values due
to larger differences between the two models. However, the general
trend of the enantioselectivity in terms of DDE fits with binding as-
say results.
The
DE values of the minus series of disparlure and oxa analogs,
For the inner binding site of PBP1, the
PBP1 was 18 kJ/mol lower than that of (+)-1 (Table 3). For PBP2,
the average of DDE of enantiomer pairs between (ꢀ)-1 and (+)-1
D
E of (ꢀ)-1 docked to
bound to PBP1 were lower than those of the plus series enantio-
mers, indicating that less energy might be needed for minus enan-
tiomers to bind to the internal site of PBP1. It is interesting to note

1820
Y. Yu, E. Plettner / Bioorg. Med. Chem. 21 (2013) 1811–1822
that (ꢀ)-2, which has the absolute configuration of (+)-1, bound
more strongly to PBP1 than (+)-2, which mimics (ꢀ)-1. Thus, the
heteroatom replacement caused a reversal of enantioselectivity
in the PBP1 and 2-analog combination. The replacement of the
methylene group at the C-5 position with an oxygen atom inverted
the electronic density distribution of (ꢀ)-2 as seen in its LUMO
(Fig. 5), which may have contributed to the inversion of its interac-
tion with PBP1.
Poses from PBP2 docked with (+)-1 and (ꢀ)-1 also gave large
DDE values that were consistent with the enantioselective bind-
ing preference of PBP2 to (+)-1. The docking poses of oxa analogs
with PBP2 had small DDEs, suggesting that they would bind to
PBP2 with similar affinity according to the energy penalty
hypothesis. Consistently, the binding assay results with PBP2
showed that PBP2 did not discriminate between the oxa analog
enantiomers.
The DDEs calculated from surface binding sites (Supplementary
Table S6 and S7) were not very consistent with the enantioselectiv-
ity shown in binding assays. This is not surprising because binding
assays measured the internal binding interaction at equilibrium. In
our previous kinetic studies, the second step of binding (entering
the internal binding pocket from an external site) was determined
to be the rate limiting step.13 The DDEs between (ꢀ)-4 and (+)-4 as
well as (ꢀ)-5 and (+)-5 when docked into the internal binding site
of PBP2 were much larger than the DDEs when they docked into
PBP1. Higher energy difference means higher selectivity according
the energy penalty hypothesis. Therefore, PBP2 will bind sulfur
analogs more selectively than PBP1, which was observed in the
binding assay experiments. Based on the above analysis, the
enantioselectivity of PBP1 and PBP2 happened at the internal bind-
ing sites. The energy penalty of each enantiomer from its global
minimum to the docked poses correlated well with the enantiose-
lectivity of pheromone binding proteins towards natural and syn-
thetic analogs.
One may ask whether the way that ligands enter or exit the
inner binding site contributes to the enantiodiscrimination of
PBPs on enantiomeric pairs. In our previous study of PBP2 bind-
ing with ligands, the C-terminus was proposed to be involved in
a two-step binding mechanism.14 In the current study, a binding
site under the C terminus (site II) was identified in PBP2. The
DDEs of the poses of (+)-1 and (ꢀ)-1 docked in between site II
and site I were 22 and ꢀ23 kJ/mol, respectively, indicating the
C-terminus had a definite effect when ligand was trying to enter
the internal binding site, and exit it. The poses in the PBP1 inter-
(+)-1 from the internal binding site, (+)-1 exited from the shoulder
of PBP1, indicating that the C-terminus of PBP1 was not involved in
ligand binding, as detected previously with PBP1 and N-phenyl-1-
naphthylamine (NPN).14
Finally, the energy descriptors of the docked protein/ligand
complex, ligands and interaction energy were calculated from
MOE (Supplementary Table S8 and S9). The potential energy cal-
culated from MOE was the sum of bond stretch potential energy,
angle bend potential energy, bond stretch-bend cross-term po-
tential energy, out-of-plane potential energy, torsion (proper
and improper) potential energy, the van der Waals component
of the potential energy and the electronic component of the po-
tential energy. We can see that the interaction energy came from
only van der Waals and electronic components of potential en-
ergy for both PBP1 and PBP2 docked with ligands. The sulfur
analogs had higher electronic components than oxa analogs
which in turn were higher than both enantiomers of disparlure
when docked in PBP1. The van der Waals component of the po-
tential energy was about the same for almost all analogs and
disparlure when docked in PBP1. Oxa analog (+)-2 was an excep-
tion with much lower van der Waals and electronic components.
When disparlure and its analogs docked into the internal binding
pocket of PBP2 the electronic component of potential energy was
smaller than when docked with PBP1, whereas the van der
Waals energy component of potential energy was higher than
when docked in PBP1. The stretched poses of ligands in the large
PBP2 internal binding pocket maximized the van der Waals
interactions between PBP2 and ligands. To further understand
the correlation between docking results and binding affinity as-
says, the total potential energy of each protein/ligand complex
was plotted against Kd (Fig. 9). Two types of potential energy/
binding affinity relationships were noticed with good linear fit-
ting. The top line represented a group of protein/ligand com-
plexes with higher potential energy and stronger binding of
ligands, compared with the group of the lower linear fitting. In
both groups, the binding affinity decreased as the potential en-
ergy of the complex increased, consistent with the energy pen-
alty hypothesis.
The potential energy of the complex reflects the enthalpy
investment necessary for binding in the above discussion. Enthalpy
alone explains most of our results, however, entropy is important
in the fine tuning of ligand-protein binding affinities.40,41 We
390
nal binding site had significantly higher
DEs, about 91 and 64 kJ/
y = 1.2x + 352
R² = 0.86
mol, than the poses in the entry site II when docking with (+)-1
and (ꢀ)-1, respectively, indicating that the C-terminus is not a
rate limiting barrier for the ligand to enter the internal binding
380
site of PBP1. This result fits well with
a previous report
that the binding affinity of PBP1 is not affected by truncation
of the C-terminus, as detected with both GC and fluorescence
assays.14
y = 0.86x + 344
370
R² = 0.87
HESB
The DDEs at the channel site (IV) of PBP2 were very similar for
enantiomers of disparlure (1) and 5-oxa analog (2), indicating no
enantiomer discrimination happened there (Table S7). On the other
hand, the DDEs between enantiomers of thia analogs had compa-
rably large values (16 and 13 kJ/mol for 4 and 5, respectively) on
the channel site when docked with PBP2. It seems that the channel
site might be contributing to the enantioselectivity of PBP2 with
thia analogs as well as the internal binding site, if thia analogs
do enter the PBP2 internal binding site through the channel site.
In a dynamic trajectory simulation of (+)-1 in binding site I of
PBP2, ligand (+)-1 exited from the bottom of PBP2 if the C and N
termini are considered as the top of the protein. It is also a reason-
able alternative pathway for ligands to enter or exit the internal
binding site through channel site IV. In the trajectory of PBP1/
MESB
360
LEMB
OL
350
0
10
20
30
40
50
60
Kd (µM)
Figure 9. Correlations between potential energy of protein/ligand complexes and
dissociation constants (Kd). Diamonds: highest energy complexes with the stron-
gest binding affinities (HESB); dots: medium energy complexes with strong binding
affinities (MESB); squares: low energy complexes with medium binding affinities
(LEMB); triangle: a possible outlier (OL).

Y. Yu, E. Plettner / Bioorg. Med. Chem. 21 (2013) 1811–1822
1821
attribute the difference between the upper and lower linear groups
to entropy. The group of PBP/ligand combinations with the smaller
slope in the Epot versus Kd plot (0.86, Fig. 9) had less favorable en-
tropy contributions than the group with the larger slope (1.2,
Fig. 9). There are three additional complexes in Figure 9: PBP1/
(+)-3 with a very low potential energy but the weakest binding
affinity, complexes PBP2/(ꢀ)-4 and PBP2/(ꢀ)-5 with the highest
potential energy but the highest binding affinity. When the poses
of (+)-3 and (ꢀ)-3 were compared it is clear that in the PBP1/(+)-
3 complex, (+)-3 fit deeply inside the binding pocket without sol-
vent interaction and with a very tightly bent (S-shaped) conforma-
tion. In contrast, (ꢀ)-3 bound to PBP1 in a more extended
conformation, with some solvent contact. With fewer degrees of
freedom the entropy of (+)-3 must be disfavored compared with
that of (ꢀ)-3 when bound to PBP1. Therefore, PBP1 had a very
low binding affinity for (+)-3, even though the complex had very
low potential energy. On the other hand, the PBP2/(ꢀ)-4 and
PBP2/(ꢀ)-5 complexes had the highest potential energies but with
lowest Kd values. Ligands (ꢀ)-4 and (ꢀ)-5, complexed with PBP2,
had stretched conformations and were located in open binding
pockets, resulting in more degrees of freedom, which should con-
tribute favorably to the binding entropy. As a result, both com-
pounds bound to PBP2 more strongly than their enantiomers.
1.63–1.46 (m, 6H), 1.46–1.22 (m, 14H), 1.18 (m, 2H), 0.87 (t,
J = 6.8 Hz, 3H), 0.86 (d, J = 6.4 Hz, 6H). 13C NMR (100 MHz, CDCl3)
d 71.6, 68.8, 56.2, 55.3, 38.5, 31.8, 29.7, 29.4, 29.2, 28.1, 27.9,
27.2, 26.9, 26.1, 22.64, 22.60, 22.58, 14.1. HRMS (ESI) m/z
[M+Na]+: Calcd for C18H36O2Na, 307.2608. Found: 307.2614. GC:
18.87 min. GC/MS (EI): m/z 285 (M++1, 7), 155 (28), 137 (42), 69
(100). (ꢀ)-7 (50 mg, 0.29 mmol) was treated similarly to give (+)-
3
(62.6 mg, 76%),
½
a 2D0
+1.5 (c 1.2, CHCl3), (+)-7 (110 mg,
ꢁ
0.64 mmol) to (ꢀ)-3 (92.4 mg, 51%), ½a D20
ꢀ1.5 (c 1.1, CHCl3).
ꢁ
4.1.2. ( )-2-Decyl-3-(isopentylthiomethyl)oxirane (( )-(4)), (+)-
(2R,3S)-2-decyl-3-(isopentylthiomethyl)oxirane ((+)-(4)), and
(ꢀ)-(2S,3R)-2-decyl-3-(isopentylthiomethyl)-oxirane ((ꢀ)-(4))
A suspension of NaH (60% in mineral oil, 159 mg, 3.98 mmol) in
THF (2.0 mL) was added to a solution of 3-methylbutane-1-thiol
(298 lL, 2.39 mmol) in THF (2.0 mL) at 0 °C and stirred for 5 min.
A solution of ( )-8 (587 mg, 1.59 mmol) in THF (4.0 mL) was added
dropwise via cannulation. The reaction mixture was warmed to rt
and stirred for 1 h. After being cooled to 0 °C the reaction mixture
was quenched with ice cold brine and extracted with diethyl ether.
The organic layer was dried with MgSO4 and concentrated. The res-
idue was purified by column chromatography (ethyl acetate/hex-
ane, 4:96) to give a colorless liquid ( )-4 (343 mg, 72%). 1H NMR
(400 MHz, CDCl3) d 3.14 (ddd, J = 8.4, 5.6, 4.0 Hz, 1H), 2.96 (m,
1H), 2.70 (dd, J = 14.0, 6.4 Hz, 1H), 2.64–2.56 (m, 3H), 1.68 (m,
1H), 1.53–1.46 (m, 6H), 1.37–1.20 (m, 14H), 0.90 (d, J = 6.8 Hz,
6H), 0.88 (t, J = 7.2 Hz, 3H). 13C NMR (100 MHz, CDCl3) d 57.4,
56.6, 38.7, 31.9, 30.5, 30.2, 29.6, 29.53, 29.50, 29.48, 29.3, 27.8,
3. Conclusions
In conclusion, highly enantiopure oxa and thia analogs as well
as disparlure were synthesized chemoenzymatically. In silico sim-
ulation and ab initio calculation of interactions of PBPs/ligands
indicated that the enantioselectivity of PBP1 and PBP2 happened
at the internal binding sites. The enantiomers of the 5-oxa analog
bound with opposite enantioselectivity to PBP1 than the phero-
mone or the other analogs. We found that this was due to a change
in the distribution of the frontier orbitals, away from the epoxide
and onto the shorter chain. The observed enantioselective binding
of PBPs with both natural and synthetic ligands was due to the con-
formation adopted by each enantiomer to fit the orientation of the
internal binding pocket. Several amino acid residues, such as Lys94
in PBP1 as well as Gln111 and Asn115 in PBP2, likely play a key
role in controlling the binding orientation and, therefore, the
enantioselectivity of the two PBPs towards pheromone and hetero-
atom analogs.
27.4, 26.5, 22.7, 22.3, 22.2, 14.1. Anal. Calcd for
C18H36OS
(300.25): C, 71.93; H, 12.07. Found: C, 72.08; H, 11.91. HRMS
(ESI) m/z [M+Na]+: Calcd for C18H36OSNa, 323.2379. Found:
323.2385. GC: decomp. (ꢀ)-(8) (587 mg, 1.59 mmol) was treated
similarly to give (+)-4 (266 mg, 56%), ½a D20
ꢁ
+28.6 (c 1.23, CHCl3),
(+)-(8) (585 mg, 1.59 mmol) to (ꢀ)-4 (254 mg, 53%), ½a D20
ꢀ28.9
ꢁ
(c 1.05, CHCl3).
4.1.3. cis-2,3-Epoxy-8-methyl-1-(octylthio)nonane (( )-5),
(2R,3S)-2,3-epoxy-8-methyl-1-(octylthio)nonane ((+)-5), and
(2S,3R)-2,3-epoxy-8-methyl-1-(octylthio)nonane ((ꢀ)-5)
( )-9 (150 mg, 0.46 mmol) was treated similarly as above (in
Section 4.1.2) to give a colorless liquid ( )-5 (96 mg, 70%). 1H
NMR (400 MHz, CDCl3) d 3.13 (ddd, J = 6.0, 6.0, 4.0 Hz, 1H), 2.97
(m, 1H), 2.69 (dd, J = 14.0, 6.0 Hz, 1H), 2.60 (m, 3H), 1.64–1.44
(m, 7H), 1.44–1.30 (m, 4H), 1.30–1.22 (m, 8H), 1.20 (m, 2H), 0.88
(t, J = 6.4 Hz, 3H), 0.87 (d, J = 6.8 Hz, 6H). 13C NMR (100 MHz,
CDCl3) d 57.4, 56.6, 38.8, 32.5, 31.8, 30.2, 29.8, 29.2, 28.8, 27.88,
27.87, 27.3, 26.8, 22.64, 22.60, 22.59, 14.1 HRMS (ESI) m/z
[M+Na]+: Calcd for C18H36OSNa, 323.2379. Found: 323.2388. GC:
decomp. (ꢀ)-9 (102 mg, 0.31 mmol) was treated similarly to give
4. Experimental section
4.1. Chemistry
General information can be found in Supplementary data.
(+)-5 (84.6 mg, 90%), ½a D20
ꢁ
+7.5 (c 1.3, CHCl3), (+)-9 (186 mg,
4.1.1. cis-2,3-Epoxy-8-methyl-1-octyloxynonane (3), (2S,3R)-2,3-
epoxy-8-methyl-1-octyloxynonane ((+)-3), (2R,3S)-2,3-epoxy-8-
methyl-1-octyloxynonane ((ꢀ)-3)
0.57 mmol) to (ꢀ)-5 (125 mg, 74%), ½a D20
ꢀ7.5 (c 1.3, CHCl3).
ꢁ
NaH (60% in mineral oil, 23 mg, 0.58 mmol) was washed with
hexane then added as a suspension in THF (0.5 mL) into a solution
of 7 (50 mg, 0.29 mmol) in DMF (1 mL) at 0 °C. After 5 min 1-bro-
mooctane (85 mg, 0.44 mmol) was added dropwise. The reaction
mixture was stirred for 5 h and another portion of NaH (60% in
mineral oil, 23 mg, 0.58 mmol) and 1-bromooctane (85 mg,
0.44 mmol) were added. The reaction mixture was stirred over-
night, quenched with water then extracted with diethyl ether.
The organic layer was washed with brine, dried with MgSO4, and
concentrated to give a yellowish liquid which was purified by col-
umn chromatography (hexane/ethyl acetate, 95:5) to give a color-
less liquid 3 (60.0 mg, 72%). 1H NMR (400 MHz, CDCl3) d 3.63 (dd,
J = 11.2, 4.4 Hz, 1H), 3.48 (m, 3H), 3.13 (m, 1H), 2.97 (m, 1H),
4.2. Ab initio calculations
The Potential Energy Surfaces (PES) of (+)-1 and its analogs
were calculated using Density functional theory (DFT) B3LYP wave
function with basis set 6-31G. The calculation was performed using
Gaussian 09W version software implemented on a high perfor-
mance computer (HPC) from grex.westgrid.ca, Compute Canada.
The optimization and single-point calculations of structures were
run on either an HPC or a local PC using Gaussian 09W version soft-
ware. The B3LYP wave function and a 6-31G basis set were em-
ployed. The global minimum conformations of disparlure and its
analogs were obtained from PES scans and their corresponding to-
tal energies were readily available from Gaussian software.

1822
Y. Yu, E. Plettner / Bioorg. Med. Chem. 21 (2013) 1811–1822
4.3. In silico simulation of ligand-PBP interactions
References and notes
1. Bierl, B. A.; Beroza, M.; Collier, C. W. Science 1970, 170, 87.
2. Bierl, B. A.; Beroza, M.; Collier, C. W. J. Econ. Entomol. 1972, 65, 659.
3. Grant, G. G.; Langevin, D.; Liska, J.; Kapitola, P.; Chong, J. M. Naturwissenschaften
1996, 83, 328.
4. Gries, G.; Gries, R.; Khaskin, G.; Slessor, K. N.; Grant, G. G.; Liska, J.; Kapitola, P.
Naturwissenschaften 1996, 83, 382.
5. Miller, J. R.; Mori, K.; Roelofs, W. L. J. Insect Physiol. 1977, 23, 1447.
6. Vogt, R. G.; Kohne, A. C.; Dubnau, J. T.; Prestwich, G. D. J. Neurosci. 1989, 9, 3332.
7. Plettner, E.; Lazar, J.; Prestwich, E. G.; Prestwich, G. D. Biochemistry 2000, 39,
8953.
8. Inkster, J. A. H.; Ling, I.; Honson, N. S.; Jacquet, L.; Gries, R.; Plettner, E.
Tetrahedron: Asymmetry 2005, 16, 3773.
9. Nakagawa, T.; Sakurai, T.; Nishioka, T.; Touhara, K. Science 2005, 307, 1638.
10. Laughlin, J. D.; Ha, T. S.; Jones, D. N. M.; Smith, D. P. Cell 2008, 133, 1255.
11. Ha, T. S.; Smith, D. P. J. Neurosci. 2006, 26, 8727.
12. Honson, N.; Johnson, M. A.; Oliver, J. E.; Prestwich, G. D.; Plettner, E. Chem.
Senses 2003, 28, 479.
The models of PBP1 and PBP2 were made based on the crystal
structure of GOBP2 (PDB ID:2WCJA) using homology modelling on-
line software SWISS-MODEL Workspace and Molecular Operating
Environment (MOE), Chemical Computing Group’s (MOE 2011)
software, Montréal, Canada. Computer-assisted docking and trajec-
tory dynamic simulation experiments were carried out with the
MMFF94X force field using MOE with default parameters, unless
otherwise stated. The PDB files of the highest scored conformers
obtained from the docking simulation using MOE were opened in
Gaussian View 5.0 and subsequently submitted to Gaussian 09W
software for their total energy calculations as was described in
Section 4.2.
13. Gong, Y.; Pace, T. C. S.; Castillo, C.; Bohne, C.; O’Neill, M. A.; Plettner, E. Chem.
Biol. 2009, 16, 162.
4.4. Pheromone binding assays
14. Gong, Y. M.; Tang, H.; Bohne, C.; Plettner, E. Biochemistry 2010, 49, 793.
15. Martin, J. P.; Beyerlein, A.; Dacks, A. M.; Reisenman, C. E.; Riffell, J. A.; Lei, H.;
Hildebrand, J. G. Prog. Neurobiol. 2011, 95, 427.
Binding assays: The binding assay used insect Ringer’s solution
(For 100 ml solution: NaH2PO4ꢂH2O, 69 mg; Na2SO4, 137 mg;
KHCO3, 45 mg; MgCl2ꢂ6H2O, 369 mg; CaCl2, 42 mg; KCl, 46 mg.
The above solution was titrated with a solution of NaHPO4ꢂ7H2O,
16. Mori, K. Bioorg. Med. Chem. 2007, 15, 7505.
17. Norinder, U.; Gustavsson, A.-L.; Liljefors, T. J. Chem. Ecol. 1997, 23, 2917.
18. Gustavsson, A. L.; Larsson, M. C.; Hansson, B. S.; Liljefors, T. Bioorg. Med. Chem.
1997, 5, 2173.
19. Chapman, O. L.; Mattes, K. C.; Sheridan, R. S.; Klun, J. A. J. Am. Chem. Soc. 1978,
100, 4878.
20. Chen, H.; Gong, Y. M.; Gries, R. M.; Plettner, E. Bioorg. Med. Chem. 2010, 18,
2920.
21. Bengtsson, M.; Liljefors, T.; Hansson, B. S. Bioorg. Chem. 1987, 15, 409.
22. Bestmann, H. J.; Roth, K.-D.; Rehefeld, C.; Leinemann, B.; Kern, F.; Vostrowsky,
O. Bioorg. Med. Chem. 1996, 4, 473.
23. Dickens, J. C.; Oliver, J. E.; Mastro, V. C. J. Chem. Ecol. 1997, 23, 2197.
24. Liljefors, T.; Bengtsson, M.; Hansson, B. S. J. Chem. Ecol. 1987, 13, 2023.
25. Bianchi, D.; Cabri, W.; Cesti, P.; Francalanci, F.; Rama, F. Tetrahedron Lett. 1988,
29, 2455.
26. Fukusaki, E.; Senda, S.; Nakazono, Y.; Yuasa, H.; Omata, T. J. Ferment. Bioeng.
1992, 73, 280.
27. Fukusaki, E.; Satoda, S. J. Mol. Catal. B: Enzym. 1997, 2, 257.
28. Fukusaki, E. I.; Satoda, S.; Senda, S.; Omata, T. J. Biosci. Bioeng. 1999, 87, 103.
29. Gong, Y. M.; Plettner, E. Chem. Senses 2011, 36, 291.
30. Paduraru, P. A.; Popoff, R. T. W.; Nair, R.; Gries, R.; Gries, G.; Plettner, E. J. Comb.
Chem. 2008, 10, 123.
31. Bengtsson, M.; Liljefors, T.; Hansson, B. S. Bioorg. Chem. 1987, 15, 409.
32. Jönsson, S.; Hansson, B. S.; Liljefors, T. Bioorg. Med. Chem. 1996, 4, 499.
33. Hayase, S.; Renou, M.; Itoh, T. Eur. J. Org. Chem. 2005, 13, 2777.
34. Ramos, M. N.; Neto, B. D. J. Comput. Chem. 1990, 11, 569.
35. Debnath, A. K.; Decompadre, R. L. L.; Debnath, G.; Shusterman, A. J.; Hansch, C.
J. Med. Chem. 1991, 34, 786.
134 mg/100 ml, to pH 7.0). PBP (4
sect Ringer’s solution (total volume 400
hour at 4 °C. The aliquot (200 L) was passed through a small
(30 mg, 200 L void volume) column of Bio-Gel P-2 (Bio-Rad, Her-
cules, CA, 2 kDa exclusion limit) in a 200 L pipet tip with a cotton
plug. Each tip was flushed once with an additional 200 L of IRS
buffer. The entire filtrate and the remaining protein ligand complex
in IRS were extracted with ethyl acetate/hexane (1:1, 200
lM) and ligand (4
lM) in an in-
l
L) were incubated for one
l
l
l
l
lL ꢃ 2)
separately. The volumes of the extracts were measured. The con-
centrations of ligands in the entire filtrate and the remaining com-
plex were calculated based on the peak areas in GC chromatograms
compared against the calibration curves. The dissociation constant
ð½PBPꢁiꢀ½LꢁbÞꢂð½Lꢁtꢀ½Lꢁb
ÞWhere
½Lꢁb
½PBPꢁꢂ½Lꢁ
was calculated as follows:Kd
¼
¼
½PBPꢂLꢁ
[PBP]i: initial concentration of PBP, [L]b: concentration of bound li-
gand, [L]t: concentration of total ligand.
Acknowledgments
We would like to thank Mr. Chunhai Mu, for the discussion of ab
initio calculation of analogs. This project was funded by Natural
Sciences and Engineering Research Council of Canada (NSERC) stra-
tegic Grants STGP35104-5-2007 and STGP396484-10 to E.P.
36. Shimazaki, K.; Mori, M.; Okada, K.; Chuman, T.; Goto, H.; Osawa, E.; Sakakibara,
K.; Hirota, M. J. Chem. Soc., Perkin Trans. 2 1992, 5, 811.
37. Guex, N.; Peitsch, M. C. Electrophoresis 1997, 18, 2714.
38. Arnold, K.; Bordoli, L.; Kopp, J.; Schwede, T. Bioinformatics 2006, 22, 195.
39. Yu, Y.; Ma, F.; Cao, Y.; Zhang, J.; Zhang, Y.; Duan, S.; Wei, Y.; Zhu, S.; Chen, N. Int.
J. Biol. Sci. 2012, 8, 12.
40. Diehl, C.; Engstrom, O.; Delaine, T.; Hakansson, M.; Genheden, S.; Modig, K.;
Leffler, H.; Ryde, U.; Nilsson, U. J.; Akke, M. J. Am. Chem. Soc. 2010, 132, 14577.
41. Sharp, K. Protein Sci. 2001, 10, 661.
Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.bmc.2013.01.043.
These data include MOL files and InChiKeys of the most important
compounds described in this article.