Synthesis and evaluation of conformationally restricted inhibitors of aspartate semialdehyde dehydrogenase

Article abstract of DOI:10.1039/c0mb00227e

Inhibitors of the enzyme aspartate semialdehyde dehydrogenase, a key biological target for the generation of a new class of antibiotic compounds, have been developed. To investigate improvements to binding within an inhibitor series, the lowering of the entropic barrier to binding through conformational restriction was investigated. A library of linear and cyclic substrate analogues was generated and computational docking used to aid in structure selection. The cyclic phosphonate inhibitor 18 was thus identified as complimentary to the enzyme active-site. Synthesis and in vitro inhibition assay revealed a K _i of 3.8 mM against natural substrate, where the linear analogue of 18, compound 15, had previously shown no inhibitory activity. Two further inhibitors, phosphate analogue diastereoisomers 17a and 17b, were synthesised and also found to have low millimolar K_i values. As a result of the computational docking investigations, a novel substrate binding interaction was discovered: hydrogen bonding between the substrate (phosphate hydroxy-group as the hydrogen bond donor) and the NADPH cofactor (2′-oxygen as the hydrogen bond acceptor). The Royal Society of Chemistry 2011.

Full text of DOI:10.1039/c0mb00227e

View Article Online / Journal Homepage / Table of Contents for this issue
Dynamic Article Links
Molecular
BioSystems
Cite this: Mol. BioSyst., 2011, 7, 1564–1575
www.rsc.org/molecularbiosystems
PAPER
Synthesis and evaluation of conformationally restricted inhibitors
of aspartate semialdehyde dehydrogenasew
Andrew S. Evitt*a and Russell J. Cox*b
Received 7th October 2010, Accepted 9th February 2011
DOI: 10.1039/c0mb00227e
Inhibitors of the enzyme aspartate semialdehyde dehydrogenase, a key biological target for
the generation of a new class of antibiotic compounds, have been developed. To investigate
improvements to binding within an inhibitor series, the lowering of the entropic barrier to binding
through conformational restriction was investigated. A library of linear and cyclic substrate
analogues was generated and computational docking used to aid in structure selection.
The cyclic phosphonate inhibitor 18 was thus identified as complimentary to the enzyme
active-site. Synthesis and in vitro inhibition assay revealed a Ki of 3.8 mM against natural
substrate, where the linear analogue of 18, compound 15, had previously shown no inhibitory
activity. Two further inhibitors, phosphate analogue diastereoisomers 17a and 17b, were
synthesised and also found to have low millimolar Ki values. As a result of the computational
docking investigations, a novel substrate binding interaction was discovered: hydrogen bonding
between the substrate (phosphate hydroxy-group as the hydrogen bond donor) and the NADPH
cofactor (20-oxygen as the hydrogen bond acceptor).
novel antimicrobial agents. The aspartate pathway is not
Introduction
targeted by the current generation of clinical antibacterial
The aspartate biosynthetic pathway is responsible for the
agents and inhibition of the pathway would be expected
production of four essential amino acids (threonine, isoleucine,
to overcome the resistance that is now undermining
last-line-of-defense medicines.3
methionine and lysine) and important primary metabolites
such as diaminopimelate (DAP) 4 in bacteria (Scheme 1).1,2
Aspartate semialdehyde dehydrogenase (ASADH, EC 1.2.1.11)
The pathway is not found in mammals but is critical to the
is found immediately prior to the first branch-point in the
survival of bacteria and thus presents a potential target for
aspartate pathway. Unlike the enzymes found in the catalytic
steps surrounding ASADH there are no analogous enzymes
that can replace the catalytic function of ASADH, marking
ASADH as a unique weak-point in the pathway.1 The vital
role of ASADH in bacteria has been demonstrated through
the creation of Salmonella enterica mutants deficient in the asd
gene encoding ASADH.4 Mutants lacking asd are auxotrophic
for diaminopimelate 4 and lysine, chemicals which are
required for the construction of the peptidoglycan layer of
the bacterial cell wall.5 Furthermore, genetic deficiency in
ASADH has been used as a strategy for the creation of
successful vaccine strains of Salmonella typhimurium,
Streptococcus mutans and Legionella pneumophila.6–8 By preventing
ASADH function instead through the mechanism of potent
selective small-molecule inhibitors, it is expected to obtain lethality
towards bacteria without side-effects in mammals.
Scheme
1 Key transformations of the aspartate pathway.
AK = aspartokinase.
The mechanism of ASADH is well understood, particularly
through comparisons to extensively studied enzymes such as
serine proteases and the highly analogous 3-phospho-glycer-
aldehyde dehydrogenase.9–13 In the biosynthetic (dephosphoryl-
ating) direction, the thiol of C135 (E. coli numbering) is
deprotonated by H274 to generate an ionic pair (Scheme 2).
a Pharmaceutical Science Division, King’s College London,
Franklin-Wilkins Building, 150 Stamford Street, London, UK,
SE1 9NH
b School of Chemistry, University of Bristol, Cantock’s Close, Bristol,
UK, BS8 1TS
w Electronic supplementary information (ESI) available: See DOI:
10.1039/c0mb00227e
c
1564 Mol. BioSyst., 2011, 7, 1564–1575
This journal is The Royal Society of Chemistry 2011

View Article Online
underlying why the inhibition constants are not appreciably
lower than the KM value of BAP 2 potentially arises from
their structural similarity—inhibitor binding was most likely
mirroring the binding of BAP. This theory is further supported
by the fact that the inhibition was competitive. To further
improve binding, and thus inhibition, we decided to pursue a
strategy of reducing the entropic cost of binding through
conformational restriction. Conformational restriction has
proven to be a useful strategy in certain classes of inhibitor,
showing for example over 400-fold improvement to binding
upon cyclisation of pentapeptide inhibitors of penicillopepsin,
an aspartate protease.17 Cyclisation of inhibitor structures
would also provide a carbon framework upon which to build
further functionality complimentary to the enzyme active-site,
providing the potential for furthering the enthalpic benefit to
binding.
Scheme 2 Simplified mechanism of ASADH.
Computational docking
The C135 thiolate then attacks the b-carbonyl of b-aspartyl-
phosphate (BAP, 2) giving a tetrahedral oxyanion that is likely
stabilised by N135.9 Collapse of the tetrahedral intermediate
expels phosphate to generate the enzyme-bound thiolester inter-
mediate. The thiolester is reduced by hydride transfer from
NADPH, proceeding through a second stabilised tetrahedral
intermediate to release the product aspartate semialdehyde
(ASA 3), inorganic phosphate (Pi) and NADP+ (Scheme 2).
Kinetic studies have shown that the order of binding/release for
BAP/NADPH is random and ASA/Pi/NADP+ proceeds with a
non-obligatory preference for Pi and NADP+ to bind before or
release after ASA.14 NADP(H) and Pi are bound to the enzyme
during the steps in which they are required but it has been noted
that NADPH remains bound throughout the phosphate thiolysis
stage and that Pi remains bound throughout the hydride transfer
stage, steps for which they are respectively not required.
To select inhibitor structures that were complimentary to the
active-site of ASADH, computational simulated docking was
employed.
Selection of ASADH coordinate file
Computational simulated docking was performed using the
FlexX software package (BioSolveIT GmbH)18 which requires
a rigid enzyme file, the coordinates of which are held constant
through the computation. Five ASADH coordinate sets were
selected from the RCSB protein databank in order to study the
suitability of applying a rigid structure simulated docking
method to ASADH (Fig. 2). The sequences were chosen for
The previously designed ASADH inhibitors possessed low
millimolar inhibition constants (Fig. 1).15,16 The reason
Fig. 2 Comparison of ASADH coordinate sets from the RSCB
protein databank. A low overall sequence identity does not affect
the active site sequence and structural identity. Highlighted: the 20OH
groups of the NADPH ribose locate within 1.4 A despite greater
distances in the remainder of the NADPH molecule. Mesh depicts the
binding pocket occupied by substrate. aSequence identity calculated
using ClustalW2. bSMCS: S-methyl cysteine sulfoxide.
Fig. 1 The natural substrate (BAP, with KM value) and previous
inhibitors (with Ki values; all values in mM).15,16 Ki values for both
ASA and inorganic phosphate (Pi) represent competitive inhibitors.
Remaining compounds noted as irreversible inhibitors or inactive as
inhibitors.
c
This journal is The Royal Society of Chemistry 2011
Mol. BioSyst., 2011, 7, 1564–1575 1565

View Article Online
their diversity of source microbe: Gram negative E. coli
(1GL3 and 1T4B), Gram negative H. influenzae (1NWH),
Gram positive S. pneumoniae (2GZ3) and the archaea
M. jannaschii (1YS4). The overall sequence identity of
ASADH varied between 72% (1GL3 against 1NWH) and
12% (1GL3 against 1YS4), correlating with phylogenetic
separation (ClustalW2 alignment of all sequences in ESIw).
Despite overall sequence identities as low as 12%, the sequence
and the structure of residues within the active-site were observed
to be near identical (Fig. 2). The NADPH cofactor overlaid less
reproducibly, although the active hydride and 20OH resided in
similar spatial positions. The coordinate set 1GL3 was selected
for the simulated docking studies as this structure contains the
cofactor NADPH and is from the E. coli isoform of ASADH,
which we use for in vitro inhibition studies.
BAP docking
Initial simulated docking experiments were performed on the
natural substrate BAP 2 in order to examine performance of
the docking software. While the exact binding mode of 2 has
not been observed crystallographically, many key interactions
have been elucidated from site-directed mutagenesis and
intermediate-trapping experiments,9 and the simulated docking
results matched these interactions (Fig. 3a). In particular, the
BAP backbone was stretched between arginine residues R102
and R267, with the phosphate functional group in contact
with R102 and the carboxylate with R267. The protonated
amine located in a pocket of hydrogen bond acceptors
(1.5 A to E241, 2.0 A to Q162 and 2.2 A to C135) and the
substrate carbonyl located within 4 A of the C135 active-site
nucleophile.
The charge states of the enzyme active-site residues were set
automatically by the FlexX algorithm. The protonation states
of the catalytic dyad residues, C135 and H274, were investigated
through manual modification and assessment by simulated
docking (ESIw Fig. 2). It was found that C135 SH torsion angles
and H274 conformation did not affect results significantly, but
that better results were observed when C135–H274 were set as
the charged pair (i.e. C135 anion, H274 cation).
During these simulated docking studies, the protonation
states of the BAP substrate were set as the charged carboxylate
and ammonium groups. Analysis of the pKa of the phosphate
group suggested that the first proton would be absent under
physiological conditions (pKa1 = 1.5–2.0) but the second
protonation state remained in question (pKa2 = 6.5–7.0).19
All three possible protonation states of the phosphate were
therefore examined in separate experiments by simulated
docking. Full protonation led to a loss of interaction with
R102 while full deprotonation maintained good binding with
R102. Single protonation, however, led to electrostatic binding
with R102 and formation of a hydrogen bond between the
BAP phosphate (acting as the hydrogen bond donor) and the
NADPH ribose 20OH (hydrogen bond acceptor, Fig. 3a and b).
The docking results were examined by compiling an overlay
of the top scoring poses. For tightly binding ligands the top
scoring poses should lie in a narrow area of the orientation
space and therefore overlay well (seen, for example, in Fig. 3a
and d) while for a weakly binding structure the top scoring
poses would be spread more widely across the orientation
space (seen, for example, in Fig. 3c).
Fig. 3 The results of computational docking presented as overlays of the top scoring poses; (a) natural substrate BAP 2; (b) BAP, highlighting the
hydrogen-bond between the substrate phosphate and NADP+; (c) poor alignment of putative phosphate inhibitor 17a; (d) good alignment of
putative phosphonate inhibitor 18a.
c
1566 Mol. BioSyst., 2011, 7, 1564–1575
This journal is The Royal Society of Chemistry 2011

View Article Online
Without a hydrogen bond between substrate and NADPH
the alignment, and hence overall binding of substrate, was
adversely affected (ESIw Fig. 3 and 4).
A study of all the available coordinate files in the RSCB
protein databank revealed that no examples where the
NADPH 20OH acts as the hydrogen bond acceptor
have previously been reported. Three examples were found
demonstrating that the NADPH ribose 20OH can act as the
hydrogen bond donor.20–22 Simulated docking was therefore
used to investigate the NADPH 20OH as the hydrogen bond
donor, which was performed by rotating the 20OH torsion so
that the oxygen lone pairs were directed towards the
phosphate binding site. This led to complete loss of the
hydrogen bond, indicating that the 20OH in ASADH-bound
NADPH cannot act as the hydrogen bond donor (data not
shown). As a result of these experiments, the phosphate moiety
was consistently set as singly deprotonated in all following
docking experiments. After highlighting theoretical strongly-
binding structures, determining if single deprotonation under
physiological conditions is likely for that structure would be
performed.
Library docking
A virtual compound library totaling 182 structures was
created, encompassing the natural substrate, previous
inhibitors,15,16 a series of putative linear and cyclic (5-, 6-,
and 7-membered rings) BAP analogues, all bearing functional
groups in a wide variety of locations and stereochemistries
(depicted in Fig. 4 and set out in full in ESIw Fig. 5). The amine
and carboxylic acid groups were set as charged and the
phosphate groups were set as singly deprotonated. This library
was submitted to FlexX and the results were assessed by visual
inspection of binding interactions and then further by overlay
assessment as described for the natural substrate.
Scheme 3 Reagents and conditions: (a) TMS-Cl, DIPEA, DCM,
reflux 2 h then Cbz-Cl, 0 1C to rt, 16 h, quant.; (b) Jones reagent,
acetone, 0 1C, 10 min, 75%; (c) isobutylene, H2SO4 (cat.), CH2Cl2, rt,
16 h, 73%; (d) ethyl diazoacetate, BF3ꢀEt2O, Et2O, rt, 1 h, 79%;
(e) NaCl, DMSO, H2O (cat.), 140 1C, 4 h, 23 in 34% and 24 in 36%;
(f) L-selectride, THF, ꢁ45 1C to rt, 30 min, 25a in 22% and 25b in
40%.
structurally restrained analogue of 15, a compound that had
previously shown no inhibitory activity and therefore provides
a platform on which to examine the effects of cyclisation. In a
similar fashion to the natural substrate, pipecolate 18a was
located between R102 and R267, with the phosphate and
carboxylate groups aligning in an analogous manner to 2.
The protonated ring amine of 18a was well conserved through-
out the overlaid poses but could not get as close to the
E241–Q162–C135 hydrogen bonding pocket as the natural
substrate amine, presumably owing to the conformational
restriction within the parent structure. The protonated
hydroxy-group of the phosphate was again observed to form
a hydrogen-bonding interaction with the NADPH cofactor.
The proposed synthetic route to compound 18a (vide infra)
would also allow for rapid access to the structurally similar
phosphate 17a, and therefore this compound was also
subjected to the simulated docking study (Fig. 3c). 17a did
not perform well in the docking study—the top scoring poses
of 17a did stretch the carboxylate and phosphate between the
two arginine residues of the active site, but were often
misaligned so that the inhibitor was inverted (i.e. the carboxylate
formed interactions with R102 rather than R267). It was also
noted that the amine does not make any consistent contact.
We thus decided to synthesise and examine the phosphate
inhibitor 17a in vitro as a control for the reliability of the
docking software.
The results revealed that several of the 5- and 7-membered
rings could form favourable interactions with the active-site,
but it became apparent that the goodness-of-fit arose from the
flexibility available to these ring sizes. The 5-membered rings
took advantage of choice of puckered ring atom (ESIw Fig. 6)
and the 7-membered rings could sample a wide range of
conformations simply due to their size (ESIw Fig. 7). The
lack of conformational restriction observed in the 5- and
7-membered ring inhibitors led us to believe that binding
would not be improved as there would be little reduction in
entropy. Efforts were therefore focused on the 6-membered
rings, which are restricted largely to one of the two conformers
of the chair form. From the results of the 6-membered ring
structures the functionalised pipecolate derivative 18a
was highlighted as complementary to the ASADH active
site (Fig. 3d). Furthermore, pipecolate 18a represents a
Fig.
4 Template for the creation of cyclic inhibitor library.
All possible stereoisomers were automatically investigated.
c
This journal is The Royal Society of Chemistry 2011
Mol. BioSyst., 2011, 7, 1564–1575 1567

View Article Online
while the trans-phosphate 17a decomposed appreciably over
72 h and thus was assayed immediately after preparation.
For synthesis of phosphonate 18a, the main target high-
lighted in the docking study, a method of constructing a
secondary carbon to phosphorus bond was required. The lack
of a general procedure for bond formation between secondary
carbons and phosphorus, coupled with the great deal of
steric bulk inherent in the cyclohexyl system, indicated a
challenging synthetic conversion. A great deal of effort was
invested into constructing this carbon–phosphorus bond,
including the use of Michaelis–Arbuzov,24,25 Michaelis–Becker,26
Michaelis–Becker with Salvatore modification,27 Perkow,28 and
chlorophosphonium techniques,29 but all failed to produce the
desired material. Thioketone methodology reported by Yoshida
finally allowed access to these compounds (Scheme 5).30,31 The
protected pipecolate-5-ketone 23 was reacted with H2S in a
procedure reported to generate the geminal dithiol.32 It was
found, however, that a mixture comprising of the enethiol 28
and thioketone trimer 29 was obtained, but that this mixture
underwent the desired reaction when refluxed with trimethyl-
phosphite to generate a 2 : 1 mixture of pipecolate derivatives 30a
and 30b. The thioether groups were removed by hydrogenolysis
to give compound 32, but this also led to partial cleavage of the
Cbz group, giving compound 31. Re-attachment of the Cbz
group aided purification by chromatography and increased the
isolated yield of the reaction. Full deprotection then led to the
mixture of trans- and cis-diastereoisomers 18a and 18b. 1H-NMR
confirmed that the 2 : 1 ratio of trans- to cis-isomers that
was set in the trimethylphosphite addition had been conserved
throughout the deprotection, and the mixture of diastereoisomers
was subjected to the inhibition study.
Synthesis
The synthetic procedure of Jung and Avery23 was exploited to
rapidly access ketone 23 and diastereomeric alcohols 25a and
25b (Scheme 3).
Thus, trans-L-hydroxy proline 19 was protected and the hydro-
xyl group oxidised to give ketone 20. Ring expansion with ethyl
diazoacetate gave the regioisomeric pipecolate derivatives 21 and
22 which were not separated. The mixture was decarboxylated
under Krapcho conditions23 and the regioisomeric ketones 23
and 24 then separated by chromatography. We obtained a 15%
overall yield of ketone 23 in comparison to Jung and Avery’s
28%. Jung and Avery then screened a number of reducing agents
showing that the stereoselective synthesis of the cis alcohol 25b
was possible through simply using sodium borohydride. As we
required the trans diastereoisomer 25a we made use of the
reported sterically hindered reducing agent L-selectride, achieving
a ratio of 1 : 2 25a : 25b in 62% yield (Jung and Avery report a
2 : 3 ratio of 25a : 25b in 92% yield).
Reaction of the separate diastereomeric alcohols with
dibenzyl iodophosphonate 26, followed by full deprotection
gave the phosphate inhibitors 17a and 17b as their HCl salts
(Scheme 4). It was noted that the cis-phosphate 17b was stable,
Inhibition assay
The procedure of Cox et al. was used to investigate inhibition
of ASADH in vitro,15 with the modification that bicine buffer
was used in order to minimise imine formation between buffer
and ASA. In brief, the ASADH-catalysed reaction was run in
Scheme 4 Reagents and conditions: (a) pyridine, CH2Cl2, 26 (0.18 M
in CH2Cl2), ꢁ20 1C to rt, 2 h, 60%; (b) TFA : DCM (1 : 1), 2 h, 84%;
(c) H2, Pd/C (10%), EtOH, 20 h, HCl(aq), 49%. 25b was reacted using
the same conditions in (a) 65%; (b) 60% and (c) 76% yield.
Scheme 5 Reagents and conditions: (a) H2S, morpholine (0.1 eq), DMF, 10 1C, 5 h, 28 in 6%, 29 in 49%; (b) P(OMe)3, toluene, reflux, 40 h,
30a : 30b 2 : 1, 67%; (c) Raney nickel, EtOH, 2 h; (d) DIPEA, Cbz-Cl, 0 1C to rt, 20 min, 32a : 32b 2 : 1, 79%; (e) 1 : 1 TFA : CH2Cl2, 2 h, 2 : 1
trans : cis, 96%; (f) Pd/C (10%), EtOH, 20 h, 2 : 1 trans:cis, 71%; (g) HCl (6 N), reflux, 5 h, 18a : 18b 2 : 1, 63%.
c
1568 Mol. BioSyst., 2011, 7, 1564–1575
This journal is The Royal Society of Chemistry 2011

View Article Online
Table
ASADH
1
Constants determined during kinetic investigations of
The results of the inhibition assay did not agree fully
with the predictions of the simulated docking software. The
software had predicted that phosphonate 18a should bind better
than phosphate 17a, however, inhibition was found to be in the
same range. Furthermore, the software had predicted that the
phosphonate should bind in an orientation that competed with
both ASA and Pi—i.e. binding to both regions of the active site.
The in vitro results suggest that the phosphonate and phosphate
inhibitors are only competitive with ASA and not with Pi. This
may indicate that the in vitro binding mode of the inhibitors is
different to that predicted by the software.
Kinetic parameter
ASADH
kcat/sꢁ1
48.5 ꢂ 6.4
KM of ASA/mM
KM of Pi/mM
0.17 ꢂ 0.03
0.032 ꢂ 0.013
kcat/KM of ASA/Mꢁ1 sꢁ1
kcat/KM of Pi/Mꢁ1 sꢁ1
2.36 ꢃ 105 ꢂ 1.07 ꢃ 105
2.65 ꢃ 105 ꢂ 0.77 ꢃ 105
reverse using synthetic ASA 3 as the substrate and NADP+ as
the cofactor. The initial velocity was monitored by recording
the formation of NADPH spectrophotometrically at 340 nm.
Kinetics data for ASA and Pi were collected (Table 1) and
found to be in reasonable agreement with literature values.33
Inhibition data were obtained by varying the concentration of
ASA and of Pi in presence of varying concentrations of
inhibitor. The resulting initial velocity data were fit to the
Hanes–Woolf and Lineweaver–Burk plots (ESIw).34 The mixture
of phosphonates 18a and 18b was found to have Ki values of
2.64 mM vs. ASA and 7.67 mM vs. inorganic phosphate.
Phosphate 17b showed Ki values comparable to the phosphonate
mixture (5.78 mM vs. ASA and 6.49 mM vs. Pi), while the
phosphate 17a was around two-fold poorer as an inhibitor
(13.97 mM vs. ASA and 11.9 mM vs. Pi). All three were found
to inhibit in a competitive or mixed mode against ASA and in an
uncompetitive mode against inorganic phosphate.
The software does, however, predict a previously unknown
hydrogen-bonding interaction between substrate and
NADPH, which could aid in the explanation of three features
of ASADH catalysis. It has been noted that in the catalytic
mechanism, NADPH and Pi perform roles in non-overlapping
elementary steps and there is therefore no requirement for
both to be bound at the same time. Kinetic evidence, however,
shows that Pi remains bound throughout the oxidoreductive
stage and that NADPH remains bound throughout the
phosphate thiolysis stage,14 which could be explained by
favourable binding between NADPH and the phosphate
group. Secondly, a hydrogen bond involving the protonated
OH of the phosphate group would explain the documented
advantage of ASADH inhibitors having a high pKa2.15
Thirdly, the NADPH 20OH has been observed to locate in a
similar spatial position in 1GL3 and 1YS4 when much of the
rest of the NADPH molecule differs. This interaction therefore
requires further real-world investigation.
Discussion
Investigation of the predicted binding of the natural
substrate BAP 2 suggests some interesting aspects of substrate
binding. The g-carbonyl bound relatively far away from the
active site nucleophile (ca. 4 A). The phosphate functional
group is in the phosphate binding region of the active-site,
which forces the carbonyl to adopt a position close to the
phosphate binding site. The phosphate binding site is necessarily
removed from the active site nucleophile to ensure that the acyl
intermediate does not favour the bound phosphate too heavily
over the NADPH hydride. To accompany movement of the
substrate g-carbonyl towards C135, the BAP amine group must
rotate away from C135 to make full use of interaction with the
Q241 residue.
The key finding in this study was that cyclisation of the linear
inhibitor 1515 by a one carbon linker to generate the cyclic
compounds 18a and 18b resulted in a change from no detectable
inhibition to low millimolar inhibition. We propose that this
improvement to inhibition must derive from the reduction of the
entropic barrier to binding arising from the prior locking of the
inhibitor conformation. The observed similarity in inhibition
constants between the cis-phosphate 17b and phosphonate 18
contrasts with the simulated docking results, which suggested that
the extra oxygen of the phosphate should infer worse binding
characteristics. It is likely, therefore, that the simulated docking
did not provide an accurate simulation of real-world binding.
Some information can be ascertained from the binding
modes of the respective inhibitors. Phosphates 17a and 17b
inhibit against ASA by two different modes—compound 17a
shows non-competitive inhibition while compound 17b shows
competitive inhibition. The differences in Ki and in inhibition
mode are most likely related to the different diastereomeric
complimentarity of 17a and 17b with ASADH. Compound 18
was isolated as a mixture of diastereoisomers and the
inhibition mode will therefore reflect this heterogeneity. The
inhibition mode of the mixture 18 against ASA is competitive.
Both diastereoisomers may be competitive inhibitors but it is
more likely that the inhibition matches the diastereomeric
complementarity seen with 17a and 17b—only one diastereo-
isomer is a competitive inhibitor but this dominates over the
effect of the weaker inhibitor. The inhibition mode against
Pi in all cases was uncompetitive, indicating that the inhibitors
do not occupy the same binding site as Pi.
Conclusions
Cyclisation of the linear inhibitor 1515 to the cyclic inhibitor 18
leads to an increase in inhibitory activity which we suspect can
be directly attributed to the decreased entropic cost of binding.
Simulated docking suggests that phosphonate 18a should
bind at both ASA and phosphate binding sites but in vitro
results suggest a different binding mode to that predicted.
A new mode of substrate–ligand binding is highlighted by the
computational docking, whereby the substrate forms a hydrogen
bond with the NADPH cofactor and further confirmation of this
interaction must be performed. An extensive screen of reaction
conditions for the synthesis of secondary carbon to phosphorus
bonds was conducted, encompassing many literature reported
conditions, and it was found that no conditions provided the
desired material except thioketone mediated bond formation.
c
This journal is The Royal Society of Chemistry 2011
Mol. BioSyst., 2011, 7, 1564–1575 1569

View Article Online
samples run as a dilute solution of the stated solvent. All
NMR spectra were referenced to the residual undeuterated
solvent as an internal reference. The following abbreviations
were used to explain the multiplicities: s = singlet, d =
doublet, t = triplet, q = quartet, m = multiplet, br = broad.
High resolution mass spectra (HRMS) were recorded on a
magnetic sector VG Autospec using EI (electron ionisation)
and CI (chemical ionisaton). High resolution mass spectra
were recorded on a Bruker Apex FT–ICR–MS using ESI
(electrospray ionisation). Infrared (IR) spectra were recorded
on a Perkin-Elmer FT-IR Paragon 1000 spectrometer under
neat conditions using a universal attenuated total reflection
(UATR) attachment. Melting points were obtained on an
electrothermal melting point apparatus and are uncorrected.
Optical rotations were measured with a Bellingham+Stanley
ADP220 polarimeter using a 1 dm optical cell. Elemental
analyses were performed in the microanalytical laboratories
of the University of Bristol. GC–MS (gas chromato-
graphy–mass spectrometry) data were obtained using an
Agilent 6890 series gas chromatograph system with an
Agilent 5973 network mass selective detector (EI mode).
A 30 m ꢃ 0.25 mm, 0.25 mm film Gamma Dex column was
used as the stationary phase. Injection performed with injector
at 200 1C, oven temperature started and held at 70 1C for
1 min, ramped at 25 1Cꢀminꢁ1 to 150 1C, ramped at 45 1Cꢀminꢁ1
to 250 1C, held at 250 1C for 3 min, ramped at 45 1Cꢀminꢁ1 to
300 1C and held at 300 1C for 3 min. Total ion current trace
was obtained for positive electron ionisation (EI+). LC–MS
(high performance liquid chromatography–mass spectrometry)
was performed using a Micromass Platform LC (ESI+ mode).
The stationary phase was a 25 cm ꢃ 4.6 mm Phenomenex
Luna 5 m column with 5 mm C18 packing. The solvent system
started at 75% H2O/25% MeCN, ramped to 95% MeCN over
13 min and held for a further 2 min. All cyclic structures
presented are numbered with priority given to the amino
acid moiety: the a-amino carbon is consistently numbered ‘1’.
All data for reproduction of literature syntheses can be found
in the ESIw.
Experimental
Computational
PyMol (DeLano scientific)35. PyMOL Release 0.99 was used
for all enzyme related graphics.
CORINA (Molecular networks)36. CORINA version 3.46
with the Corina.direct version 3.2 graphical user interface was
used to generate three-dimensional structure coordinates of
BAP and all putative inhibitors. When running Corina, all
possible enantiomers and diastereoisomers were generated and
saved in the multimol2 (.mol2) file format.
Simulated docking. The PDB coordinate file 1GL3 was used
in all simulated docking experiments. Subunit A was used in
all cases, and prior to docking the SMCS atom coordinates
were manually removed. Proton coordinates were calculated
either by using the Pymol algorithm or by using the FlexX
software algorithm. Docking was performed using FlexX
version 1.13.5 L implemented in Tripos Sybyl.18 Default
settings were used. The active site was defined as a 10 A sphere
centred around the Cys135 sulfur and the software was
instructed to include the NADPH cofactor in the docking
run. Further simulated docking was performed using the
standalone FlexX version 3.1.3 (64 bit).18 The Corina version
3.46 plugin for FlexX was used to generate ring-flip conformers
for relevant ligands. The active site was set as a 10 A sphere
around the Cys135 sulfur and the software was instructed to
include the NADPH cofactor in the docking run. Cys135
was set as the thiolate, and His274 as the imidazolium. No
pharmacophore was set. The ligand settings were changed so that
the software performed no manipulation of the protons and
charges. The docking and scoring parameters were all left as the
defaults. The results of the simulated docking runs were
analysed in PyMol by constructing overlays of the FlexX pose
coordinates.
Chemistry. All reactions were performed under a nitrogen
atmosphere with dry solvent under anhydrous conditions,
unless otherwise noted. Dry, deoxygenated diethyl ether
(Et2O), tetrahydrofuran (THF), acetonitrile (MeCN), toluene,
dichloromethane (DCM) and hexane were obtained by
passing commercially available pre-dried, oxygen-free
formulations through activated alumina columns. Anhydrous
ethanol was obtained by distillation from Mg/I2. Anhydrous
methanol was purchased commercially (HPLC grade). All
water used was previously deionised. Yields refer to chromato-
graphically and spectroscopically (1H NMR) homogeneous
materials. Reagents were purchased at the highest commercial
quality and used without further purification, unless otherwise
stated. Reactions were monitored by thin-layer chromato-
graphy (TLC) performed on 0.2 mm Merck silica gel glass
plates (60F-254) and visualised using ultraviolet light (254 nm)
and by potassium permanganate and heat as developing
agents. Merck silica gel (60, particle size 0.035–0.070 mm)
was used for flash column chromatography. NMR spectra
were recorded on JEOL delta/GX270 (D270), JEOL delta/
GX400 (D400), JEOL lamba300 (L300), Eclipse+ 400 (E400),
Varian 400 (V400) and Varian 500 (V500) instruments. NMR
spectra were obtained at 298 K unless otherwise stated and
N-Cbz-5-(R)-dibenzylphosphate-(S)-tert-butylpipecolate, 27a.
To an anhydrous 25 ml round-bottom flask, under an inert
atmosphere (N2) and with magnetic stirring, was added the
trans-alcohol 25a (222 mg, 0.66 mmol) and anhydrous DCM
(4.4 ml). The reaction was cooled to ꢁ20 1C (external
temperature) and pyridine (dist. 214 ml, 2.65 mmol) added in
one portion. To this was added dibenzylphosphoryl iodide solu-
tion (0.18 M in DCM, 4.4 ml, 0.79 mmol) dropwise over 15 min.
The solution turned yellow and a precipitate formed over the
period of addition. The mixture was stirred at ꢁ20 1C for 30 min,
then warmed to room temperature and stirred for another 2 h.
The precipitate was filtered off and the filtrate reduced in vacuo.
The resulting oil was extracted between water (6 ml) and
Et2O (25 ml). The layers were separated and the aqueous was
further extracted with Et2O (2 ꢃ 25 ml). The combined organics
c
1570 Mol. BioSyst., 2011, 7, 1564–1575
This journal is The Royal Society of Chemistry 2011

View Article Online
were washed with KHSO4 (0.3 M; 10 ml), NaHCO3 (saturated;
10 ml) and brine (10 ml). The organic layer was dried (MgSO4),
filteredandreducedinvacuo togive the crude product (288 mg) as
a white cloudy oil. Purification was performed by column
0.783 mmol, 65.3%). [a]23 = ꢁ15.0 (c 0.01, DCM); 1 : 1
D
mixture of rotamers; dH (V400, 400 MHz, CDCl3) 7.40–7.28
(15H, m, ArH), 5.25–4.99 (6H, m, 3 ꢃ ArCH2), 4.76 (0.5H, d,
J = 4.5, 0.5 ꢃ H-2eq), 4.63 (0.5H, d, J = 4.7, 0.5 ꢃ H-2eq), 4.37
(0.5H, dd, J = 12.7, 5.3, 0.5 ꢃ H-6eq), 4.29–4.16 (1.5H, m, 0.5 ꢃ
H-6eq and H-5ax), 2.95 (0.5H, dd, J = 12.7, 10.6, 0.5 ꢃ H-6ax),
2.85 (0.5H, dd, J = 11.2, 10.6, 0.5 ꢃ H-6ax), 2.29–2.19 (1H, m,
H-4eq), 2.10–1.99 (1H, m, H-3eq), 1.68–1.57 (1H, m, H-4ax); 1.45
(4.5H, s, 0.5 ꢃ C(CH3)3), 1.42 (4.5H, s, 0.5 ꢃ C(CH3)3),
1.41–1.30 (1H, m, H-3ax), dC (E400, 100 MHz, CDCl3) 169.6
& 169.5 (ester CQO), 155.8 & 155.6 (carbamate CQO), 136.4 &
136.4 (quaternary ArC), 135.8–135.8 (m, 2 ꢃ quaternary
phosphate-ArC), 128.5, 128.4, 128.0, 127.9, 127.8, 127.8, 82.1
& 82.0 (C(CH3)3), 72.5 (d, J = 5.0, 0.5 ꢃ C5), 72.4 (d, J = 5.0,
0.5 ꢃ C5), 69.4–69.2 (m, 2 ꢃ ArCH2P), 67.5 & 67.4 (2 ꢃ
Cbz-ArCH2), 53.8 & 53.5 (C2), 45.9 (d, J = 4.0, 0.5 ꢃ C6),
45.8 (d, J = 5.0, 0.5 ꢃ C6), 28.3 & 28.3 (C3), 27.9 & 27.8
(C(CH3)3), 24.6 & 24.5 (C4); dP (E400, 162 MHz, CDCl3) ꢁ1.1,
ꢁ1.2; nmax (neat) cmꢁ1 3035 (ArC–H), 2976 (C–H), 1731 (ester
CQO), 1706 (carbamate CQO); ESIMS m/z 634 ([M]K+ 19%),
618 ([M]Na+ 100%), 540 (6%), 358 (8%), 206 (6%); HRMS ESI
calcd. for C32H38NO8PNa ([M]Na+) 618.2227, found 618.2257;
anal. calcd. (%) for C32H38NO8P C 64.53, H 6.43, N 2.35,
found C 64.78, H 6.60, N 2.64.
chromatography (10% EtOAc in DCM used as eluent, RF
=
0.13) to give the target phosphate 27a as a clear oil (234 mg, 0.393
mmol, 59.5%). [a]20D = ꢁ20.0 (c 0.07, DCM); 1 : 1 mixture of
rotamers;dH (E400, 400 MHz, CDCl3) 7.37–7.24 (15H, m, ArH),
5.25–4.92 (6H, m, 3 ꢃ ArCH2), 4.90–4.86 (0.5H, m, 0.5 ꢃ H-2eq),
4.74 (0.5H, d, J = 4.6, 0.5 ꢃ H-2eq), 4.66–4.61 (0.5H, m,
0.5 ꢃ H-5eq), 4.56–4.51 (0.5H, m, 0.5 ꢃ H-5eq), 4.34 (0.5H, d,
J = 14.7, 0.5 ꢃ H-6eq), 4.25 (0.5H, d, J = 14.7, 0.5 ꢃ H-6eq), 3.28
(0.5H, d, J = 14.7, 0.5 ꢃ H-6ax), 3.17 (0.5H, d, J = 14.7, 0.5 ꢃ
H-6ax), 2.09–1.83 (3H, m, 2 ꢃ H-4 and H-3eq), 1.52–1.45 (1H, m,
H-3ax); 1.44 (4.5H, s, 0.5 ꢃ C(CH3)3), 1.40 (4.5H, s, 0.5 ꢃ
C(CH3)3), dC (E400, 100 MHz, CDCl3) 170.3 & 170.2 (ester
CQO), 156.4 & 156.0 (carbamate CQO), 136.4 & 136.3
(quaternary Cbz-ArCH), 135.8, (d, J = 6.9, 1 ꢃ quaternary
phosphate-ArC), 135.6, (d, J = 6.9, 1 ꢃ quaternary phosphate-
ArC), 128.5, 128.4, 128.4, 127.9, 127.9, 127.8, 127.8, 127.8, 127.7,
81.9 & 81.8 (C(CH3)3), 70.9 (d, J = 5.4, 0.5 ꢃ C5), 70.8 (d, J =
5.4, 0.5 ꢃ C5), 69.2–69.1 (m, 2 ꢃ ArCH2P), 67.3 & 67.2
(Cbz-ArCH2), 54.3 & 53.9 (C2), 45.9 (d, J = 5.4, 0.5 ꢃ C6),
45.8 (d, J = 5.4, 0.5 ꢃ C6), 27.9 & 27.9 (C(CH3)3), 25.8 (d, J =
3.8, 0.5 ꢃ C4), 25.6 (d, J = 3.8, 0.5 ꢃ C4), 20.3 & 20.1 (C3);
dP (E400, 162 MHz, CDCl3) ꢁ0.88, ꢁ1.01; nmax (neat)cmꢁ1 3034
(ArC–H), 2976 (C–H), 1731 (ester CQO), 1702 (carbamate
CQO); ESIMS m/z 634 ([M]K+ 3%), 618 ([M]Na+ 100%),
596 (16%), 206 (6%); HRMS ESI calcd. for C32H38NO8PNa
([M]Na+) 618.2227, found 618.2230; anal. calcd. (%) for
C32H38NO8P C 64.53, H 6.43, N 2.35, found C 64.23,
H 6.68, N 2.52.
N-Cbz-5-(R)-dibenzylphosphate-(S)-pipecolic acid.
To a 25 ml round-bottom flask with magnetic stirring was
added the protected phosphate 27a (219 mg, 0.368 mmol)
followed by DCM (2 ml) and TFA (2 ml). Stirring was
continued until the reaction was complete by TLC (2 h). The
reaction mixture was evaporated from DCM four times to
remove excess TFA and then subjected to a base–acid cycle:
the resulting oil was dissolved in NaHCO3 (2.5%, 10.8 ml, 3.24
mmol) and extracted with Et2O (2 ꢃ 20 ml). The aqueous was
acidified to pH 1 with HCl (2 N) and extracted with EtOAc
(3 ꢃ 20 ml). The acid extract organics were combined, dried
(MgSO4), filtered and reduced in vacuo to give the target
material as a white foam (166 mg, 0.309 mmol, 84%) which
was used without further purification. dH (V400, 400 MHz,
CDCl3) 8.90–8.09 (3H, br s, exchangeable OH & NH2),
N-Cbz-5-(S)-dibenzylphosphate-(S)-tert-butylpipecolate, 27b.
To an anhydrous 25 ml round-bottom flask, under an
inert atmosphere (N2) and with magnetic stirring, was added
the cis-alcohol 25b (405 mg, 1.20 mmol) and anhydrous DCM
(8 ml). The reaction was cooled to ꢁ20 1C (external temperature)
and pyridine (dist. 387 ml, 4.80 mmol) added in one portion. To
this was added dibenzylphosphoryl iodide solution (0.18 M in
DCM, 8 ml, 1.44 mmol) dropwise over 15 min. The solution
turned yellow and a precipitate formed over the period of
addition. The mixture was stirred at ꢁ20 1C for 30 min, then
warmed to room temperature and stirred for another 2 h. The
precipitate was filtered off and the filtrate reduced in vacuo.
The resulting oil was extracted between water (7 ml) and Et2O
(27 ml). The layers were separated and the aqueous was further
extracted with Et2O (2 ꢃ 27 ml). The combined organics were
washed with KHSO4 (0.3 M; 10 ml), NaHCO3 (saturated; 10 ml)
and brine (10 ml). The organic layer was dried (MgSO4), filtered
and reduced in vacuo to give the crude product (543 mg) as a
white cloudy oil. Purification was performed by column
7.26–7.07 (15H, m, ArH), 5.07–4.75 (6H, m, 3 ꢃ ArCH2
&
H-2eq), 4.58–4.11 (2H, m, H-6eq & H-5ax), 3.17–3.01
(1H, m, 0.5 ꢃ H-5ax), 2.08–1.73 (1H, m, H-3eq & H-4eq
& H-3ax), 1.48–1.34 (1H, m, H-4ax).
N-Cbz-5-(S)-dibenzylphosphate-(S)-pipecolic acid.
To a 25 ml round-bottom flask with magnetic stirring was
added the protected phosphate 27b (644 mg, 1.08 mmol)
followed by DCM (4 ml) and TFA (4 ml). Stirring was
chromatography (15% EtOAc in CHCl3 used as eluent, RF
=
0.30) to give the target phosphate 27b as a clear oil (466 mg,
c
This journal is The Royal Society of Chemistry 2011
Mol. BioSyst., 2011, 7, 1564–1575 1571

View Article Online
continued until the reaction was complete by TLC (2 h). The
reaction mixture was evaporated from DCM four times to
remove excess TFA and then subjected to a base–acid cycle:
the resulting oil was dissolved in NaHCO3 (2.5%, 10.8 ml,
3.24 mmol) and extracted with Et2O (2 ꢃ 20 ml). The aqueous
was acidified to pH 1 with HCl (2 N) and extracted with EtOAc
(3 ꢃ 20 ml). The acid extract organics were combined, dried
(MgSO4), filtered and reduced in vacuo to give the target
material as a white foam (346.9 mg, 0.646 mmol, 60%) which
was used without further purification. dH (V400, 400 MHz,
CDCl3) 9.18–8.95 (3H, br s, exchangeable OH & NH2),
water (25 ml). The resulting solution was acidified with HCl
(2 N, 5 ml) and freeze-dried to provide the target salt 17b
(151 mg, 0.49 mmol, 76%) as a yellow hygroscopic solid.
[a]23D = ꢁ8.40 (c 0.08, H2O); dH (V400, 400 MHz, D2O) 4.47
(1H, dm, J = 8.6, H-5eq), 3.86 (1H, dd, J = 12.5, 3.4, H-2ax),
3.39 (1H, dm, J = 13.7, H-6eq), 3.12 (1H, dm, J = 13.7,
H-6ax), 2.09–1.82 (3H, m, H-4 and H-3eq), 1.75–1.65 (1H, m,
H-3ax); dC (V400, 100 MHz, D2O) 170.5 (carbonyl), 66.7
(d, J = 5.5, C5), 56.1 (C2), 47.7 (d, J = 5.5, C6), 27.0
(d, J = 3.1, C4), 19.8 (C3); dP (V400, 162 MHz, D2O) ꢁ1.4;
nmax (neat) cmꢁ1 3600–2400 (br s, acid O–H and ammonium
N–H), 1728 (acid C|O); ESIMS (negative ion mode) 449
([2M–H]ꢁ 26%), 224 ([M–H]ꢁ 100%); HRMS ESI (negative
ion mode) calcd. for C6H11NO6P ([M–H]ꢁ) 224.0329, found
224.0338.
7.38–7.20 (15H, m, ArH), 5.15–4.71 (6H, m, 3 ꢃ ArCH2
&
H-2eq), 4.43–4.08 (2H, m, H-6eq & H-5ax), 3.03–2.84 (1H, m,
0.5 ꢃ H-5ax), 2.32–2.18 (1H, m, H-3eq), 2.16–2.98 (1H, m,
H-4eq), 1.71–1.56 (3H, m, H-3ax), 1.51–1.35 (1H, m, H-4ax).
Pipecolic acid-5-(R)-phosphate hydrochloric acid salt, 17a.
N-Cbz-5,6-dehydro-5-thiol-tert-butyl-(S)-pipecolate, 28.
To a boiling-tube, under an inert atmosphere (N2) and
with magnetic stirring, was added the 5-keto pipecolate 23
(3.33 g, 10 mmol), DMF (2 ml) and morpholine (86 ml, 1 mmol).
The reaction mixture was equipped with a CO2(s)/acetone
condenser. The reaction vessel was charged with H2S gas, which
was trapped by the condenser and allowed to drip back into the
reaction mixture. The reaction vessel was held in an oil bath that
was maintained at 10 1C. The reaction vessel was recharged with
H2S every 1 h for a total of 5 h. The reaction mixture was then
quenched by pouring onto ice and acidifying with 2 N HCl until
acidity persisted. The acidified aqueous layer was extracted with
EtOAc (3 ꢃ 50 ml), and this organic layer was further washed
with water until the water extracts were of neutral pH. The
organic layer was dried (MgSO4), filtered and reduced in vacuo to
give a yellow oil (3.71 g). Purification was performed by
chromatography (15% EtOAc in hexane, RF of enethiol
28 = 0.25, RF of trimer 29 = 0.14) to give recovered starting
material in 20% yield and the enethiol 28 as a pale yellow solid
(209 mg, 0.60 mmol, 6%). [a]23D = ꢁ45.4 (c 0.09, DCM); mp =
70–74 1C (neat); 1 : 1 mixture of rotamers; dH (L300, 300 MHz,
CDCl3) 7.41–7.32 (5H, m, ArH), 7.22 & 7.11 (1H, s, CH-6), 5.27
& 5.12 (1H, d, J = 12.3, 0.5 ꢃ ArCH2), 5.25 & 5.19 (1H, d, J =
12.1, 0.5 ꢃ ArCH2), 4.82–4.78 & 4.71–4.67 (1H, m, H-2), 2.57 &
2.54 (1H, q, J = 0.9, SH), 2.42–1.87 (4H, m, H-3 & H-4), 1.45
& 1.38 (9H, s, C(CH3)3), dC (V400, 100 MHz, CDCl3) 169.5
& 169.3 (ester C|O), 152.8 & 152.7 (carbamate C|O), 135.8
(quaternary Cbz-ArCH), 128.6, 128.5, 128.3, 128.2, 128.0,
125.9 & 125.8 (C6), 104.5 & 103.9 (C5), 82.0 & 81.9 (C(CH3)3),
68.1 & 67.8 (Cbz-ArCH2), 53.6 & 53.2 (C2), 28.2 & 27.8 (C3),
27.9 & 27.8 (C(CH3)3), 24.3 & 24.1 (C4), nmax (neat) cmꢁ1 3105
alkene C–H, 3037 (Ar C–H), 2979 (C–H), 2945 (C–H), 2889
(C–H), 2843 (C–H), 2544 (S–H), 1736 (ester C|O), 1706 (carbamate
To a 25 ml round-bottom flask, under an inert atmosphere
(N2) and with magnetic stirring, was added the N-Cbz-5-(S)-
dibenzylphosphate-(S)-pipecolic acid (166 mg, 0.309 mmol).
This was dissolved in EtOH (10 ml) and Pd/C (10% 20 mg)
added. The reaction was stirred under a hydrogen atmosphere
for 20 h. The suspension was filtered through celite and
washed with water (25 ml). The resulting solution was acidified
with HCl (2 N, 5 ml) and freeze-dried to provide the target salt
17a (46 mg, 0.15 mmol, 49%) as a yellow hygroscopic solid.
[a]23 = ꢁ10.00 (c 0.008, H2O); dH (V400, 400 MHz, D2O)
D
4.36–4.27 (1H, m, H-5ax), 4.02 (1H, dd, J = 7.3, 4.4, H-2ax),
3.39 (1H, dd, J = 12.7, 3.2, H-6eq), 3.12 (1H, dd, J = 12.7, 7.1,
H-6ax), 2.27–2.16 (1H, m, H-3eq); 1.92–1.80 (2H, m, H-3ax and
H-3eq); 1.77–1.65 (1H, m, H-4eq); dC (V400, 100 MHz, D2O)
170.4 (carbonyl), 67.2 (d, J = 4.7, C5), 54.7 (C2), 45.6
(d, J = 5.5, C6), 26.6 (d, J = 3.9, C4), 21.0 (C3); dP
(V400, 162 MHz, D2O) ꢁ1.2; nmax (neat) cmꢁ1 3600–2400
(br s, acid O–H and ammonium N–H), 1726 (acid C|O);
ESIMS (negative ion mode) m/z 449 ([2M–H]ꢁ 5%), 224
([M–H]ꢁ 100%); HRMS ESI (negative ion mode) calcd. for
C6H11NO6P ([M–H]ꢁ) 224.0329, found 224.0329.
Pipecolic acid-5-(S)-phosphate hydrochloric acid salt, 17b.
To a 25 ml round-bottom flask, under an inert atmosphere
(N2) and with magnetic stirring, was added N-Cbz-5-(R)-
dibenzylphosphate-(S)-pipecolic acid (346.9 mg, 0.646 mmol).
EtOH (15 ml) and Pd/C (10% 20 mg) were added. The
reaction was stirred under a hydrogen atmosphere for 20 h.
The suspension was filtered through celite and washed with
t
C|O); EIMS m/z 349 ([M]+ 13%), 249 ([M–CO2 Bu+H]+ 16%),
204 (30%), 158 (12%), 91 ([Bn]+ 100%); HRMS EI calcd. for
C18H23NO4S ([M]+) 349.1348, found 349.1352. anal. calcd. (%)
for C18H23NO4S3 C 61.87, H 6.63, N 4.01, S 9.18, found
C 61.68, H 6.60, N 4.00, S 9.42.
c
1572 Mol. BioSyst., 2011, 7, 1564–1575
This journal is The Royal Society of Chemistry 2011

View Article Online
N-Cbz-5-thioketone-(S)-tert-butylpipecolate trimer, 29.
3010 (Ar C–H), 2942 (C–H), 1731 (ester C|O), 1702 (carbamate
C|O), 1244 (P|O); S–H absorption absent; CIMS m/z 474
t-
([M]H+ 7%), 418 ([M–tBu+H]H+ 88%), 374 ([M–CO2
t
Bu+H]H+ 53%), 372 ([M–CO2 Bu]+ 55%), 360 (23%), 328
(44%), 236 (21%), 91 ([Bn]+ 100%), (thiol [M]H+ = 460; not
observed); HRMS CI calcd. for C21H33NO7SP ([M]H+)
474.1715, found 474.1715.
N-Cbz-5-dimethylphosphonate-(S)-tert-butylpipecolate, trans
(32a) and cis (32b) isomer mixture.
The trimer 29 was obtained as a highly deliquescent white
foam (1.72 g, 1.64 mmol, 49%). [a]24D = +11.3 (c 0.5, DCM);
dH (L300, 300 MHz, CDCl3) 7.44–7.26 (5H, m, ArH),
5.30–5.06 (2H, m, ArCH2), 4.97–4.57 (1H, m, H-2),
4.34–4.08 (1H, m, H-6eq), 3.43–3.17 (1H, m, H-6ax),
2.55–1.56 (4H, m, H-3 & H-4), 1.51–1.31 (9H, m comprising
at least 10 singlets, C(CH3)3); nmax (neat) cmꢁ1 3067 (ArC–H),
3034 (ArC–H), 2977 (C–H), 2934 (C–H), 1704 (C|O); EIMS
To an anhydrous 50 ml round-bottom flask, under an inert
atmosphere (N2) and with magnetic stirring was added the
2 : 1 mixture of a-thioether phosphonates 30a and 30b
(489 mg, 1.03 mmol) followed by EtOH (anh; 15 ml). Raney
nickel (10 g, 50% slurry in H2O) was washed with EtOH
(anh; 3 ꢃ 5 ml) and added to the reaction mixture. The
reaction mixture was stirred until complete by TLC (2 h)
and filtered through celite. The solvent was removed in vacuo
and re-suspended in DCM (5 ml). The reaction vessel was
cooled in an ice-bath and then DIPEA (400 ml, 2.30 mmol) and
Cbz-Cl (150 ml, 1.05 mmol) added. The reaction mixture was
allowed to warm to room temperature over 20 min and then
extracted between HCl (0.5 N, 2 0 ml) and EtOAc (3 ꢃ 25 ml).
The combined organics were dried (MgSO4), filtered and
reduced in vacuo to give a yellow oil (280 mg). Purification
was performed by column chromatography (20% EtOAc in
hexane, RF = 0.26) gave the target material as a clear oil
(349 mg, 0.82 mmol, 79%). Obtained as a 2 : 1 mixture
of the trans : cis isomers and 1 : 1 mixture of rotamers: dH
(E400, 400 MHz, CDCl3) 7.43–7.41 (5H, m, ArH), 5.27–5.07
(2H, m, ArCH2), 4.88 (0.17H, d, J = 5.9, H-2), 4.81–4.66
(0.67H, m, H-2), 4.75 (0.17H, d, J = 5.9, H-2), 4.38–4.21
(1H, m, H-6eq), 3.80–3.61 (6H, m, P(O)(OCH3)2), 3.59–3.47
(0.33H, m, H-6ax), 3.48–3.39 (0.33H, m, H-6ax), 3.19 (0.17H,
ddd (dd 31P decoupled), J = 12.9, 12.1, 3.9, H-6ax), 3.07
(0.17H, ddd (dd 31P decoupled), J = 12.6, 11.9, 3.9, H-6ax),
3.98–1.69 (5H, m, H-5 & H-3eq & H-3ax & H-4eq & H-4ax),
1.49–1.42 (9H, C(CH3)3), dC (L300, 76 MHz, CDCl3) 170.2 &
169.7 & 169.6 (ester C|O), 155.8 & 155.6 & 155.5 & 155.5
(carbamate C|O), 136.3 & 136.3 & 136.2 (quaternary ArC),
128.3, 128.2, 128.2, 127.8, 127.7, 127.7, 127.6, 81.8 & 81.7 &
81.4 (C(CH3)3), 67.1 & 67.1 & 67.0 (ArCH2), 54.3 & 54.2 &
53.9 (C2), 62.4 & 52.3 & 52.2 & 52.1 (P(O)(OCH3)2), 39.6–39.2
(m, C6), 34.5, 34.3, 32.6, 32.4, 31.2, 31.2, 30.1, 29.3, 28.6, 27.7
& 27.7 (C(CH3)3), 27.5, 26.2, 26.0, 22.8, 22.8, 20.7, 20.6, 20.3,
20.3. dP (L300, 122 MHz, CDCl3) 33.8, 31.3; nmax (neat) cmꢁ1
2954 (C–H), 2869 (C–H), 1731 (ester), 1702 (carbamate);
CIMS m/z 428 ([M]H+ 14%), 372 ([M–tBu + H]H+ 55%),
t
m/z 349 ([thione]+ 15%), 249 ([M–CO2 Bu]+ 29%), 204
(52%), 158 ([M–tBu–Cbz+H]+ 21%), 91 ([Bn]+ 100%); anal.
calcd. (%) for C54H69N3O24P3S3 C 61.87, H 6.63, N 4.01,
S 9.18, found C 62.00, H 6.81, N 4.01, S 9.06.
N-Cbz-5-dimethylphosphonate-5-methylthioether-(S)-tert-
butylpipecolate, trans (30a) and cis (30b) isomer mixture.
To an anhydrous 25 ml round-bottom flask, under an inert
atmosphere (N2) and with magnetic stirring, was added the
trimer 29 (1.0 g, 2.86 mmol). This was followed by anhydrous
toluene (10 ml) and P(OMe)3 (1.23 ml, 10.4 mmol). The
reaction was heated under reflux until no starting material
could be detected by GC–MS (40 h), cooled to room temperature
and solvent reduced by blowing with N2. Purified by column
chromatography (15% EtOAc in hexane, RF of thioether 30a
and 30b = 0.24) gave the target material as a clear oil (899 mg,
1.90 mmol, 67%). Obtained as a 2 : 1 mixture of the trans : cis
isomers: [a]24 = ꢁ22.03 (c 0.3, DCM); dH (E400, 400 MHz,
D
d6-DMSO) 7.42–7.28 (5H, m, ArH), 5.23–5.01 (2H, m,
Cbz-ArCH2), 4.79–4.71 (0.67H, trans-H-2eq), 4.57–4.51 (0.33H,
cis-H-2eq), 4.49–4.30 & 3.24–3.10 (0.67H, m, cis-H-6eq+ax),
4.04–3.96 & 3.49–3.32 (1.33H, m, trans-H-6eq+ax), 3.78–3.57
(6H, m, P(OCH3)2), 2.23 & 2.21 (1H, cis-SCH3), 2.13 & 2.08
(2H, trans-SCH3), 2.25–1.71 (4H, m, H-3 & H-4), 1.41 & 1.37
(6H, C(CH3)3), 1.40 & 1.36 (3H, C(CH3)3); dC (E400, 100 MHz,
d6-DMSO) 170.2 & 170.1 (ester C|O), 155.9 & 155.1 (carbamate
C|O), 137.1 & 137.0 (quaternary Cbz-ArCH), 129.0, 128.9,
128.6, 128.5, 128.4, 128.2, 128.0, 127.9, 82.3 & 79.7 (C(CH3)3),
67.4 & 67.2 & 67.2 & 67.2 (Cbz-ArCH2), 54.8, 54.6, 54.5, 54.5,
54.4, 54.4, 54.2, 54.0, 53.9, 53.9, 45.3 (cis-C6), 45.7 (trans-C6),
28.1, 28.0, 23.9, 23.9, 22.7 & 22.7 (C(CH3)3); dP (L300, 122 MHz,
CDCl3) 28.8 & 28.7 (0.4P), 26.2 & 25.9 (0.6P); nmax (neat) cmꢁ1
t
328 (100%), 326 ([M–CO2 Bu]+ 38%), 282 (67%), 192 (23%),
91 ([Bn]+ 100%); HRMS CI calcd. for C20H31NO7P ([M]H+)
428.1838, found 428.1841; anal. calcd. (%) for C20H30NO7P C
56.20, H 7.07, N 3.28, found C 56.247, H 7.327, N 3.456.
c
This journal is The Royal Society of Chemistry 2011
Mol. BioSyst., 2011, 7, 1564–1575 1573

View Article Online
N-Cbz-5-dimethylphosphonate-(S)-pipecolic acid, cis and
trans isomer mixture.
through celite. The solvent was removed in vacuo to yield the
target material as a clear oil (102 mg, 0.433 mmol, 71%).
Used in the next step without further purification. dH (L300,
300 MHz, CDCl3) 10.22–8.07 (2H, br s, NH & OH; removed
by D2O shake), 3.77 & 3.74 (6H, s, P(OCH3)2), 3.73–2.83 (3H, m,
2 ꢃ H-6 and H-2), 2.48–1.40 (5H, m, 2 ꢃ H-3 and 2 ꢃ H-4);
dP (L300, 122 MHz, CDCl3) 29.9 & 29.8.
5-Phosphonic acid pipecolic acid hydrochloride salt, trans
(18a) and cis (18b) isomer mixture.
To a 25 ml round-bottom flask, under an inert atmosphere
(N2) and with magnetic stirring was added the 2 : 1 mixture of
phosphonates 32a and 32b (398 mg, 0.93 mmol) followed by
DCM (3 ml) and TFA (3 ml). The reaction mixture was stirred
until complete by TLC (2 h) and the solvent removed in vacuo.
The resulting oil was evaporated from DCM (5 ꢃ 10 ml) to
remove residual TFA, giving a brown oil (416 mg). This was
extracted between NaHCO3 (2.5%, 10 ml) and Et2O (3 ꢃ 10 ml).
The aqueous was then acidified using 2 N HCl until acidity
persisted, at which point a white precipitate formed. Extraction
with EtOAc (3 ꢃ 10 ml) removed the precipitate. The combined
EtOAc layers were dried (MgSO4), filtered and reduced in vacuo
to give the deprotected acid as a clear oil (332 mg, 0.89 mmol,
96%). Obtained as a 2 : 1 mixture of the trans : cis isomers and
1 : 1 mixture of rotamers: dH (V400, 400 MHz, CDCl3)
8.95–8.70 (1H, br s, OH), 7.45–7.22 (5H, m, ArH), 5.29–5.07
(2H, m, ArCH2), 5.04–5.77 (1H, m, H-2), 4.38–4.23 (1H, m,
H-6eq), 3.80–3.59 (6H, m, P(O)(OCH3)2), 3.52 (0.33H, dd, J =
14.2, 4.5, H-6ax), 3.43 (0.33H, dd, J = 14.2, 4.5, H-6ax), 3.23
(0.17H, ddd, J = 13.1, 12.6, 4.4, H-6ax), 3.15 (0.17H, ddd, J =
To a 50 ml round-bottom flask with magnetic stirring was
added the cis and trans 5-dimethylphosphonate pipecolic acid
isomeric mixture (138 mg, 0.582 mmol) followed by HCl
(6 N, 10 ml). The reaction was heated to reflux and continued
until complete by LCMS (5 h). The solvent was removed by
blowing with N2 to give the diastereoisomeric mixture 18a
and 18b as a brown hygroscopic solid (HCl salt; 106 mg,
0.366 mmol, 63%). Obtained as a 2 : 1 mixture of the trans : cis
isomers, integration is with respect to each isomer: dH (V400, 400
MHz, D2O) 4.24–4.20 (1H, m, cis H-2), 3.83 (1H, dd, J = 12.47,
trans, cis H-2), 3.51 (1H, dm, J = 1.45, trans H-6eq), 3.39–3.31
(1H, m, cis H-6eq), 3.21 (1H, ddd (dd, P decoupled), J = 12.96,
11.25, 6.11, cis H-6ax), 2.91 (1H, ddd (dd, P decoupled), J =
13.20, 13.20, 4.65, trans H-6ax), 2.31–2.24 (1H, m, trans H-3eq),
2.28–2.15 (1H, m, cis H-3eq), 2.13–1.98 (1H, m, cis H-5),
2.08–1.98 (1H, m, trans H-4eq), 2.06–1.93 (1H, m, trans H-5),
1.92–1.82 (1H, m, cis H-3eq), 1.91–1.80 (1H, m, cis H-3ax),
1.68–1.53 (1H, m, trans H-3ax), 1.62–1.51 (1H, m, trans H-4ax),
1.51–1.39 (1H, m, cis H-4ax); dC (V400, 100 MHz, D2O) 171.3
(cis acid CQO), 170.4 (trans acid CQO), 56.2 (trans C2), 54.1 (cis
R2), 43.9 (trans C6), 41.3 (cis C6), 32.5 (cis C5), 32.0 (trans C5),
25.1 (trans C3), 23.3 (cis C3), 22.6 (trans C4), 19.5 (cis C4);
dP (V400, 162 MHz, CDCl3) 22.5 (cis), 22.0 (trans); nmax (neat)
cmꢁ1 3600–2400 (br s, acid O–H and ammonium N–H), 1730
(acid CQO); ESIMS (negative ion mode) m/z 417 ([2M–H]ꢁ
22%), 208 ([M–H]ꢁ 100%); HRMS ESI (negative ion mode)
calcd. for C6H11NO5P ([M–H]ꢁ) 208.0380, found 208.0388.
13.1, 13.1, 3.9, H-6ax), 2.44–1.52 (5H, m, H-5 & H-3eq & H-3ax
&
H-4eq & H-4ax), dC (L300, 76 MHz, CDCl3) 174.4 & 173.4 &
173.2 (ester CQO), 159.0 & 158.6 & 156.2 & 155.7 (carbamate
CQO), 136.2 & 136.2 (quaternary ArC), 128.6, 128.5, 128.4,
128.0, 128.0, 127.9, 127.8, 127.8, 67.7 & 67.7 (ArCH2), 54.3 &
54.2 & 53.9 (C2), 53.8, 53.5, .53.3, 53.3, 53.2, 53.1, 53.1, 53.0,
40.7, 40.7, 40.4, 40.4, 39.7, 39.4, 34.3, 34.0, 32.8, 32.6, 30.9, 29.5,
26.3, 26.1, 26.0, 22.8, 20.7, 20.6, 20.4, 20.3, 19.8. dP (L300, 122
MHz, CDCl3) 33.9 (trans-P), 22.4 & 32.0 (cis-P); nmax (neat) cmꢁ1
3750–2100 (br s, acid O–H), 2956 (C–H), 1680 (br, ester &
carbamate CQO); CIMS m/z 372 ([M]H+ 8%), 328
([M–CO2+H]H+ 54%), 282 (24%), 192 (30%), 91 ([Bn]+
100%); HRMS CI calcd. for C16H23NO7P ([M]H+) 372.1212,
found 372.1210.
5-Dimethylphosphonate pipecolic acid, cis and trans isomer
mixture.
Enzyme assays. Enzyme assays: stock solutions were made
with deionised water and using ACS grade reagents. ASADH
(homodimer; subunit MW = 39 kDa)12 was expressed and
purified from recombinant E. coli (the same stock and purity
used in previous publications15,16—provided by Dr A. Hadfield,12
University of Bristol). The increase in NADPH concentration
was observed at 340 nm over 100 seconds using a PharmaciaLKB
Ultrospec III spectrophotometer equipped with a water-heated
(37 1C) cuvette holder. The buffer solutions were prewarmed to
37 1C before use by immersion in a water bath. All other stock
solutions were stored on ice. Assays were conducted in a far UV
quartz cuvette that was fully transparent to 340 nm light.
To a 25 ml round-bottom flask, under an inert atmosphere
(N2) and with magnetic stirring, was added the Cbz protected
pipecolic acid 2 : 1 mixture of cis and trans N-Cbz-5-dimethyl-
phosphonate-(S)-pipecolic acid isomers (225 mg, 0.606 mmol)
followed by EtOH (15 ml). 10% Pd/C (50 mg) was added and
the reaction mixture flushed with H2 and the H2 atmosphere
maintained by balloon. The reaction was allowed to stir until
complete by LCMS (20 h) and then worked up by filtration
The assay was performed in the following way (all
concentrations refer to final cuvette concentration): buffer
solution (bicine buffer; 900 ml, 0.2 M, pH 8.6) was introduced
c
1574 Mol. BioSyst., 2011, 7, 1564–1575
This journal is The Royal Society of Chemistry 2011

View Article Online
into a 1000 ml quartz cuvette. The remaining assay components
were then added to make up the final 100 ml. The order of
addition used for KM and Ki determination was ASADH
(0.56 mgꢀmlꢁ1), phosphate (15 mM), inhibitor (0, 0.5, 1, 2 and
5 mM) and NADP+ (150 mM). ASA (50, 100, 200, 1000 and
2500 mM) was added last to start the reaction. The specific
activity recorded for the production of NADPH by ASADH
was 73.74 mmol minꢁ1 mgꢁ1 at 1240 mM ASA, 15 mM Pi and
150 mM NADP+. The cuvette was inverted twice, placed in the
spectrophotometer and the subsequent reaction was monitored
at 340 nm. The spectrophotometer was set up to record the
absorption at 340 nm every 2s. KM, Vmax and Ki, were obtained
from Hanes–Woolf plots (see ESIw).34
16 R. J. Cox, J. S. Gibson and A. T. Hadfield, ChemBioChem, 2005, 6,
2255–2260.
17 A. Khan, J. Parrish, M. Fraser, W. Smith, P. Bartlett and
M. James, Biochemistry, 1998, 37, 16839–16845.
18 FlexX Docking Software, BioSolveIT GmbH, http://www.
biosolveit.de/.
19 L. D. Freedman and G. O. Doak, Chem. Rev., 1957, 57, 479–523.
20 C. Baldock, J. B. Rafferty, S. E. Sedelnikova, P. J. Baker,
A. R. Stuitje, A. R. Slabas, T. R. Hawkes and D. W. Rice, Science,
1996, 274, 2107–2110.
21 C. W. Levy, A. Roujeinikova, S. Sedelnikova, P. J. Baker,
A. R. Stuitje, A. R. Slabas, D. W. Rice and J. B. Rafferty, Nature,
1999, 398, 383–384.
22 D. G. Gourley, A. W. Schuttelkopf, G. A. Leonard, J. Luba,
¨
L. W. Hardy, S. M. Beverley and W. N. Hunter, Nat. Struct. Biol.,
2001, 8, 521–525.
23 J.-C. Jung and M. A. Avery, Tetrahedron: Asymmetry, 2006, 17,
2479–2486.
24 L. D. Quin, A Guide to Organophosphorus Chemistry, Wiley
Interscience, New York, 2000.
Referencesand notes
1 R. E. Viola, Acc. Chem. Res., 2001, 34, 339–349.
2 C. R. Faehnle, J. Le Coq, X. Liu and R. E. Viola, J. Biol. Chem.,
2006, 281, 31031–31040.
3 K. Kumarasamy, et al., Lancet Infect. Dis., 2010, 3099, 1–6.
4 H. Loessner, A. Endmann, M. Rohde, R. Curtiss and S. Weiss,
FEMS Microbiol. Lett., 2006, 265, 81–88.
5 R. J. Cox, Nat. Prod. Rep., 1996, 13, 29–43.
6 G. A. Cardineau and R. Curtiss, J. Biol. Chem., 1987, 262, 3344–3353.
´
7 J. E. Galan, K. Nakayama and R. Curtiss, Gene, 1990, 94, 29–35.
8 O. S. Harb and Y. Abu Kwaik, Infect. Immun., 1998, 66,
1898–1903.
9 S. Black and N. Wright, J. Biol. Chem., 1955, 213, 39–50.
25 W. J. Murphy, Organophosphorus Reagents, Oxford University
Press, Oxford, 2004.
26 R. Rodriguez, R. M. Adlington, S. J. Eade, M. W.
Walter, J. E. Baldwin and J. E. Moses, Tetrahedron, 2007, 63,
4500–4509.
27 R. J. Cohen, D. L. Fox, J. F. Eubank and R. N. Salvatore,
Tetrahedron Lett., 2003, 44, 8617–8621.
28 W. Perkow, Chem. Ber., 1954, 87, 755–758.
29 W. Hoffmann, C. Simmons and J. Glunz, J. Am. Chem. Soc., 1957,
79, 3570–3575.
30 Z. Yoshida, S. Yoneda and T. Kawase, Chem. Lett., 1975, 4,
279–280.
10 J. Biellmann, P. Eid, C. Hirth and H. Jornvall, Eur. J. Biochem.,
1980, 104, 53–58.
31 S. Yoneda, T. Kawase and Z. Yoshida, J. Org. Chem., 1978, 43,
1980–1985.
¨
11 J. Blanco, R. Moore and R. Viola, Proc. Natl. Acad. Sci. U. S. A.,
2003, 100, 12613–12617.
32 R. Mayer, G. Hiller, M. Nitzschke and J. Jentzsch, Angew. Chem.,
Int. Ed. Engl., 1963, 2, 370–373.
12 A. Hadfield, G. Kryger, J. Ouyang, G. A. Petsko, D. Ringe and
R. E. Viola, J. Mol. Biol., 1999, 289, 991–1002.
33 R. A. Moore, W. E. Bocik and R. E. Viola, Protein Expression
Purif., 2002, 25, 189.
13 A. Hadfield, C. Shammas, G. Kryger, D. Ringe, G. A. Petsko,
J. Ouyang and R. E. Viola, Biochemistry, 2001, 40, 14475–14483.
14 W. E. Karsten and R. E. Viola, Biochim. Biophys. Acta, Protein
Struct. Mol. Enzymol., 1991, 1077, 209–219.
´
15 R. J. Cox, J. S. Gibson and M. B. Mayo Martın, ChemBioChem,
2002, 3, 874–886.
34 H. Bisswanger, Enzyme Kinetics; Principles and Methods,
Wiley-VCH, Weinheim, Germany, 2008.
35 The PyMOL Molecular Graphics System, Version 0.99,
DeLano Scientific LLC, http://www.pymol.org/.
36 CORINA, Molecular Networks GmbH, http://www.molecular-net
works.com.
c
This journal is The Royal Society of Chemistry 2011
Mol. BioSyst., 2011, 7, 1564–1575 1575