3-Fluoroaspartate and pyruvoyl-dependant aspartate decarboxylase: Exploiting the unique characteristics of fluorine to probe reactivity and binding

Article abstract of DOI:10.1002/chem.201000622

Fluorine-containing amino acids have been used with great success as mechanism-based inhibitors of pyridoxal phosphate (PLP)-dependent enzymes, and the influence of fluorine on the conformation of molecules has also been extensively studied and practically exploited. In this study, we sought to use these unique characteristics to probe the reactivity and binding of aspartate decarboxylase (ADC) enzymes, which are members of the small class of pyruvoyl-dependant decarboxylases. Since ADC activity has been shown to be essential to the virulence of Mycobacterium tuberculosis, information gained in this manner could be used for the development of inhibitors that selectively target pyruvoyl-dependent enzymes such as ADC, without affecting PLP-dependent enzymes in the host. For this purpose, we synthesized the L-erythro and L-threo isomers of 3-fluoroaspartate and tested their ability to act as substrates and/or inhibitors of the M. tuberculosis and Escherichia coli ADC enzymes. Trapping and MS-based binding analysis was additionally used to confirm that both isomers enter the enzymes' active sites. Our studies show that both isomers undergo single turnover decarboxylation and fluorine elimination reactions to give enamine products that can be trapped within the active site. Interestingly, the enamine/ADC complex that forms from the lerythro (but not the L-threo) isomer is sufficiently stable that it can be observed even without any trapping. This finding suggests that the two 3-fluoroaspartates maintain different conformations within the ADC active site, which leads to the enamine products with configurations of different stabilities. Taken together, our results provide new insights for the development of cofactor-specific inhibitors, and confirm the utility of fluorine as a unique tool for probing reactivity and binding profiles within enzymes.

Full text of DOI:10.1002/chem.201000622

DOI: 10.1002/chem.201000622
3-Fluoroaspartate and Pyruvoyl-Dependant Aspartate Decarboxylase:
Exploiting the Unique Characteristics of Fluorine To Probe
Reactivity and Binding
Jandrꢀ de Villiers, Lizbꢀ Koekemoer, and Erick Strauss*[a]
Abstract: Fluorine-containing amino
acids have been used with great success
as mechanism-based inhibitors of pyri-
doxal phosphate (PLP)-dependent en-
zymes, and the influence of fluorine on
the conformation of molecules has also
been extensively studied and practical-
ly exploited. In this study, we sought to
use these unique characteristics to
probe the reactivity and binding of as-
partate decarboxylase (ADC) enzymes,
which are members of the small class
of pyruvoyl-dependant decarboxylases.
Since ADC activity has been shown to
be essential to the virulence of Myco-
bacterium tuberculosis, information
gained in this manner could be used
for the development of inhibitors that
selectively target pyruvoyl-dependent
enzymes such as ADC, without affect-
ing PLP-dependent enzymes in the
host. For this purpose, we synthesized
the l-erythro and l-threo isomers of 3-
fluoroaspartate and tested their ability
to act as substrates and/or inhibitors of
the M. tuberculosis and Escherichia coli
ADC enzymes. Trapping and MS-based
binding analysis was additionally used
to confirm that both isomers enter the
enzymesꢀ active sites. Our studies show
that both isomers undergo single turn-
over decarboxylation and fluorine
elimination reactions to give enamine
products that can be trapped within the
active site. Interestingly, the enamine/
ADC complex that forms from the l-
erythro (but not the l-threo) isomer is
sufficiently stable that it can be ob-
served even without any trapping. This
finding suggests that the two 3-fluo-
roaspartates maintain different confor-
mations within the ADC active site,
which leads to the enamine products
with configurations of different stabili-
ties. Taken together, our results pro-
vide new insights for the development
of cofactor-specific inhibitors, and con-
firm the utility of fluorine as a unique
tool for probing reactivity and binding
profiles within enzymes.
Keywords: amino acids · conforma-
tion analysis · fluorinated substitu-
ents · inhibitors · lyases
Introduction
chemical knockout of either pantothenate synthetase (PS;
the panC gene product) or aspartate decarboxylase (ADC,
the panD gene product) or both, should yield a similar
result, and provided some justification for targeting these re-
spective enzymes for drug design (Scheme 1). Consequently,
much work has been directed towards the discovery of in-
Coenzyme A (CoA), the cofactor involved in central energy
metabolism and the biosynthesis and degradation of fatty
acids, is essential to the growth of all living organisms.[1]
Moreover, in the case of Mycobacterium tuberculosis, it has
been shown that impairing its ability to biosynthesize pan-
tothenate (vitamin B5, the vitamin precursor of CoA) de
novo by knockout of the panC and panD genes also signifi-
cantly reduces its virulence.[2] This finding suggested that
[a] Dr. J. de Villiers, L. Koekemoer, Prof. E. Strauss
Department of Biochemistry, University of Stellenbosch
Private Bag X1, Matieland 7602 (South Africa)
Fax : (+27)21-808-5863
E-mail: estrauss@sun.ac.za
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/chem.201000622. It contains detailed
descriptions of the materials and methods used, as well as additional
results.
Scheme 1. Biosynthesis of pantothenic acid as a precursor in the biosyn-
thesis of CoA through decarboxylation of l-aspartate catalysed by aspar-
tate decarboxylase (ADC). PS: pantothenate synthetase.
10030
ꢁ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2010, 16, 10030 – 10041

FULL PAPER
hibitors of PS through a number of strategies, including the
synthesis of reaction-intermediate analogues, through com-
pound library screening, or through fragment-based
design.[3] On the other hand, few studies have been directed
towards the design of ADC inhibitors, probably because ex-
tensive-interaction studies of the Escherichia coli ADC en-
zymeꢀs active site have shown that the enzyme only allows
the closest structural analogues of aspartate to bind.[4] This
finding is in agreement with the small size of the active site,
which was defined by X-ray crystallographic studies of the
E. coli, Helicobacter pylori and M. tuberculosis ADC en-
zymes.[5–7]
In this study we set out to investigate the reactivity and
binding of ADC enzymes with two main aims: 1) to design
a potential mechanism-based inhibitor of ADC and 2) to de-
termine whether such a molecule could provide insight into
the differences that exist between the majority of amino
acid decarboxylase enzymes that are dependent on the pyri-
doxal phosphate (PLP) cofactor for activity, and the small
class of pyruvoyl-containing decarboxylases to which ADC
belongs. These latter enzymes do not contain an externally
derived cofactor. Instead, auto-catalytic cleavage of the in-
active pro-enzyme (the p protein) at an internal serine resi-
due results in the formation of a small b chain and a larger
a chain that contains the pyruvoyl group at its N termi-
nus.[5,8,9] The ketone of this pyruvoyl group subsequently
reacts with the amine of the substrate to form an iminium
ion (Schiff base), which can stabilize charge in a manner
similar to that seen for PLP-dependent enzymes
(Scheme 2). However, whereas much is known about how
the conformation of the substrate relative to the cofactor in-
fluences the diversity of reactions catalyzed by PLP-depen-
dent enzymes, our knowledge of similar effects in the case
of pyruvoyl-containing enzymes is limited by the informa-
tion that can be deduced from the available crystal struc-
tures.
We therefore considered the use of a fluorine-containing
substrate analogue of ADC to address both of the aims out-
lined above. In regards to the first aim, this strategy was pur-
sued because fluorinated amino acid analogues have exten-
sively been studied as potential mechanism-based inhibitors
of a number of PLP-dependent enzymes. Among these, the
best characterized is arguably a-(difluoromethyl)ornithine
(DFMO, eflornithine), an inhibitor of ornithine decarboxy-
lase (ODC), which is currently being used clinically in the
treatment of human African trypanosomiasis (sleeping sick-
ness). DFMO traps an ODC active site nucleophile in a co-
valent manner after enzyme-mediated decarboxylation and
elimination of fluoride, thereby inhibiting any further ODC
activity.[10] Based on this information, and on a first assump-
tion that the ADC enzymes do not differ significantly from
their PLP-dependent counterparts, a-(difluoromethyl)aspar-
tate can analogously be considered as a potential inhibitor
of ADC enzymes. However, the results of the binding-inter-
action studies of E. coli ADC have shown that neither a-
methyl- nor b-methylaspartate could be trapped within the
enzyme, indicating that branching at either of these posi-
tions leads to unfavourable interactions within the active
site.[4] As an alternative, we considered using 3-fluoroaspar-
tate as a potential mechanism-based inhibitor because the
same relative position of the fluorine with respect to the a-
carboxylate group is retained and because the small size of
the fluorine atom in the b position should minimally perturb
binding in the active site. This analogue holds a further ad-
vantage in that it can simultaneously be used to address the
second aim, because it has been shown that the introduction
of fluorine into the structure of a small molecule often per-
turbs its conformation.[11] Fluorinated aspartates may there-
fore bind in the active site in a conformation that is different
to that of the natural substrate, which may lead to different
reactivity profiles. Information gained from such differences,
should they be present, may be used as a probe for the
major binding interactions that
exist in the active site of ADC
(and perhaps pyruvoyl-depen-
dent decarboxylases in general).
Conveniently, the conformation
of 3-fluoroaspartates has been
studied extensively, both in so-
lution (using NMR spectroscop-
ic analysis) and in the solid
state (through X-ray diffraction
studies).[12,13] The data from
these studies provided an expe-
dient starting point from which
we could conduct our own com-
parative analysis.
This study is not the first in
which
a b-fluorinated amino
acid analogue has been tested
as a potential inhibitor of a de-
carboxylase enzyme. Vidal-Cros
et al. have similarly tested the
Scheme 2. Mechanism of l-aspartate decarboxylation catalysed by ADC. After formation of an iminium be-
tween the amine of the substrate and the ketone of the N-terminal pyruvoyl group of the activated enzyme,
decarboxylation takes place with subsequent delocalization of the resulting charge into the amide carbonyl.
Reprotonation yields the final product, which is released after hydrolysis of the iminium ion.
Chem. Eur. J. 2010, 16, 10030 – 10041
ꢁ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
10031

E. Strauss et al.
interactions of the two l-isomers of 3-fluoroglutamate with
the PLP-dependent glutamate decarboxylase of E. coli.[14]
Their results, which provided us with a directly analogous
PLP-dependent system with which to compare our pyruvo-
yl-containing system, showed that both isomers acted as sub-
strates and were decarboxylated by the enzyme. However,
whereas the threo isomer yielded the fluorine-containing
product, the erythro isomer additionally lost its fluorine
atom during the course of the reaction and instead gave suc-
cinic semialdehyde as the product, without trapping any
active site nucleophile. This finding, the authors found, indi-
cated that the glutamate–PLP Schiff base maintains a rigid
conformation in the active site, and also indicated that no
nucleophile is suitably located to react with the intermediate
enamine formed after decarboxylation and elimination of
fluorine from the erythro isomer. To determine whether the
same holds true in the case of the pyruvoyl-dependent
ADC, we therefore synthesized both of the l-isomers of 3-
fluoroaspartate and tested their interactions with the E. coli
and M. tuberculosis ADC enzymes.
nation to remove the benzyl protecting groups of 3 because,
in our hands, we found it difficult to isolate the product
after using the hydrogenation conditions used by Charvillon
and Amouroux. Finally, we found that using KNCO instead
of NaN3 as the source of the amine nitrogen in the conver-
sion of 6 into 8 gave higher yields in our hands and allowed
recovery of the final product in one step from 7. The final
synthetic procedure as employed in this study is summarized
in Scheme 3.
Optimization of the epoxide ring-opening reaction, in par-
ticular, resulted in a significant improvement in the yield
compared with the original method as used by Charvillon
and Amouroux. In their methodology, LiBF4 was used as
the Lewis acid catalyst in this reaction, which resulted in the
formation and isolation of 2 in 62% yield after three days.
Because we found the yield for this step to be lower, we
therefore evaluated the effect of various metal and lantha-
nide triflates as catalysts in this reaction. A literature search
revealed that these salts have been used with great success
as Lewis acids in the ring-opening of various epoxides by
amines, which has led to significantly improved yields and
selectivity.[18] We subsequently tested the ring-opening of ep-
oxide 1 with dibenzylamine in the presence of three tri-
Results
flates: YbCATHUNGRTEN(NNUG OTf)2, ZnACHTUNRTGENN(GUN OTf)2 and ScACHTNGUERTN(NUGN OTf)2, and found that
Synthesis of the l-isomers of 3-fluoroaspartate: Whereas a
number of synthetic strategies towards the preparation of
the four stereoisomers of 3-fluoroaspartate have been de-
scribed, most of these suffer from at least one of two draw-
backs: they either make use of
the presence of any of the three additives significantly im-
proved both the yield and time it took for the reaction to
reach completion. Of the three, we found that Yb
ACHTUNGTRENNUNG(OTf)2
gave the most significant improvement, resulting in the for-
fluorodehydroxylation reactions
that depend on highly corrosive
HF/SF4 reagents, or they lead
to the formation of racemic
product mixtures.[13,15,16] We
therefore set out to adapt the
methods that were used by
Charvillon and Amouroux to
synthesize the two d-isomers of
3-fluoroaspartate for synthesis
of the corresponding l-erythro-
and l-threo isomers, that is,
(2R,3R)- and (2R,3S)-3-fluo-
roaspartic acid (4 and 8, respec-
tively).[17] First, we used l-dieth-
yl tartrate (l-DET) instead of
d-DET as a starting material in
the synthesis of both isomers,
so as to effect the required
change in chirality at the a
carbon. Second, we optimized
the ring-opening reaction of ep-
oxide 1 by replacing the origi-
nally used LiBF4 catalyst with
YbACHTUNGTRENNUNG(OTf)2 after finding that the
former required days to give
significant yields of 2. Third, we
used catalytic transfer hydroge- Scheme 3. Synthesis of the two l-3-fluoroaspartic acid isomers.
10032
www.chemeurj.org
ꢁ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2010, 16, 10030 – 10041

3-Fluoroaspartate Probes
FULL PAPER
mation of 2 in an excellent yield of 97% after heating at
reflux for 16 h in THF.
initial decarboxylation may also be followed by elimination
of fluoride (pathway c) to give the enamine product.
We first used 19F NMR spectroscopic analysis to deter-
mine whether an ADC enzyme mediates catalytic defluori-
nation of either of the 3-fluoroaspartate isomers 4 or 8 (by
either pathways b or c). E. coli ADC (EcADC) and M. tu-
berculosis ADC (MtADC) were therefore separately incu-
bated at room temperature in the presence of buffered solu-
tions of either 4 or 8 at pH 8.0, and any changes in the
19F NMR spectrum over a period of 20 min were noted.
However, no change in these spectra could be observed,
suggesting that either loss of fluoride does not take place, or
that the sensitivity of the NMR spectroscopy based assay is
not sufficiently high to detect it (see the Supporting Infor-
mation for details).
We subsequently tested whether the 3-fluoroasparates
were decarboxylated by ADC enzymes by using a CO2-re-
lease assay that couples the formation CO2 to the consump-
tion of NADH (NADH=reduced nicotinamide adenine di-
nucleotide) by using the enzymes phosphoenolpyruvate car-
boxylase (PEPC) and malate dehydrogenase.[19] The de-
crease in NADH concentration can be followed spectropho-
tometrically, and is directly equivalent to the production of
CO2. We chose this assay over previous methods that have
been used to measure ADC activity because these were
nearly all based on derivatization of the substrate and prod-
uct mixtures with fluorescamine followed by HPLC analy-
sis.[8,20] We were concerned that the introduced fluorine (and
the concomitant effect it has on the basicity and nucleophi-
licity of the amine) may influence the extent and rate of de-
rivatization, which would lead to inaccurate activity determi-
nations. In contrast, a CO2-release assay would not suffer
from such a drawback, and should respond similarly to both
the natural and fluorinated aspartates.
Overall, implementation of the modified synthetic proto-
col as outlined in Scheme 3 gave the l-erythro isomer 4 in
an overall yield of 28% (starting from the epoxide 1), which
represents a small improvement compared with the 23%
overall yield for d-erythro-3-fluoroaspartic acid achieved by
Charvillon and Amouroux. This improvement was mainly
due to the optimised epoxide ring-opening reaction. Howev-
er, whereas Charvillon and Amouroux were able to synthe-
sise d-threo-3-fluoroaspartic acid in 26% yield from the
cyclic sulfate 5, we were only able to prepare the corre-
sponding l-threo isomer 8 in 3% overall yield, in spite of
the modifications made to the procedure used to convert 6
into 8. Whether using our protocol or the original, elimina-
tion of fluoride from all the fluorine-containing intermedi-
ates proved to be a significant side reaction that hampered
the isolation of the required products in high yield. Many of
the reports detailing previous attempts to synthesise 3-fluo-
roaspartates and other b-fluorinated amino acids have de-
scribed similar difficulties.[16,17] Nonetheless, the final prod-
ucts proved to be stable in solution: no elimination of fluo-
rine (or formation of fluoride) from either 4 or 8 could be
observed by 19F NMR spectroscopic analysis in buffered sol-
utions at pH 8.0 over periods of up to 20 min (see the Sup-
porting Information).
Catalytic activity assays: We subsequently set out to deter-
mine whether the two prepared 3-fluoroaspartate isomers
can act as substrates for the ADC enzyme. Three possible
reaction pathways were considered in this regard
(Scheme 4). First, decarboxylation can take place (path-
way a) followed by reprotonation, to give a fluorinated b-
alanine as product, similar to the native reaction. Second,
deprotonation at the a carbon by a suitable active-site base
can take place instead of decarboxylation, which would lead
to direct loss of fluoride by elimination (pathway b). Finally,
Whereas the PEPC-based coupled assay has been shown
to give good results with a number of decarboxylase en-
zymes, it is known that C4 dibasic acids, such as l-aspartate,
act as allosteric inhibitors of
PEPC.[21] This assay, therefore,
cannot be used to accurately
determine the kcat values of
ADC enzymes (which use l-as-
partate as substrate). Nonethe-
less, when the activity of
EcADC and MtADC were
measured in the presence of in-
creasing concentrations of l-as-
partate using the PEPC-based
coupled assay, the initial rates
could be fit to the Michaelis–
Menten equation using non-
linear regression analysis. The
resulting kinetic parameters
showed that the KM values de-
termined in this manner agree
well with those reported for the
Scheme 4. Possible reaction pathways for 3-fluoroaspartate in the presence of ADC.
EcADC and MtADC enzymes
Chem. Eur. J. 2010, 16, 10030 – 10041
ꢁ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
10033

E. Strauss et al.
as determined by indirect assay methods, whereas the kcat
values were approximately tenfold lower than the previously
reported values (see the Supporting Information for de-
tails).[8,20] This result showed that although the sensitivity of
the enzyme-coupled assay is lower than could be desired, it
can successfully be used to determine the decarboxylation
activity of the respective ADC enzymes.
We subsequently tested whether the two l-isomers of 3-
fluoroaspartate are substrates for the EcADC and MtADC
enzymes. Initial rates of CO2 release were determined in the
presence of increasing concentrations of either 4 or 8, whilst
keeping the same constant enzyme concentration used to
validate the CO2-release assay as described above. However,
even at the highest concentration of substrate used (2 mm)
no activity could be measured in the presence of either com-
pound, suggesting that neither 4 nor 8 can act as substrates
of ADC enzymes, or that the levels of activity are too low
to be detected by the assay.
with the enzyme than the original starting material. In an at-
tempt to resolve these questions we set out to confirm the
binding of 4 and 8 to EcADC and MtADC. This was ap-
proached by using the same MS-based method used by
Webb et al. to perform the original binding analysis of
EcADC,[4] in which amine-containing compounds that form
a Schiff base (imine) with the pyruvoyl group are trapped
by treatment with NaCNBH3 prior to MS analysis.
Although Webb et al. used MALDI-MS to analyze the
trapped protein complexes, any MS analysis technique that
is suitable for protein analysis could conceivably be used in
a similar manner. We considered using LC–ESIMS to ana-
lyze NaCNBH3-treated ADC complexes towards this end.
To determine whether this technique would be equally suita-
ble, we analyzed reaction mixtures containing MtADC and
l-aspartate, and identical samples subsequently treated with
NaCNBH3, to confirm the formation of the expected cova-
lent adducts. Figure 1 shows the resulting mass spectra of
Inhibitor analysis: Because the activity assays showed that
3-fluoroaspartates are neither catalytically decarboxylated
nor defluorinated by the ADC enzymes, we set out to deter-
mine whether either compound can act as an inhibitor of
the decarboxylation of the native substrate, l-aspartate. We
therefore pre-incubated the EcADC and MtADC enzymes
with 0.5 mm of either 4 or 8, and subsequently added 0.5 mm
l-aspartate to test whether these compounds affected their
decarboxylation activity in any way, as measured by the
PEPC-based coupled assay. In the case of EcADC, no sig-
nificant differences were found in the activity of untreated
enzyme compared with the enzyme treated with either 3-flu-
oroaspartate isomer. MtADC was also unaffected by treat-
ment with the erythro-isomer 4; however, treatment with l-
threo-3-fluoroaspartate (8) did result in an approximate
25% reduction in activity compared with untreated enzyme.
These results suggest that at least one of the 3-fluoroaspar-
tate isomers can inhibit the activity of MtADC, probably by
closely mimicking the bound conformation of the native
substrate.
Figure 1. Binding analysis of MtADC and l-aspartate by LC–ESIMS
analysis of the complexes. A) Mass spectrum of untreated MtADC
(60 mm) as positive control, showing the presence of both the processed
a-chain containing the N-terminal pyruvoyl group (labelled a-Pyr), as
well as the unprocessed a-chain with an N-terminal serine (labelled a-
Ser). B) Mass spectrum of MtADC (60 mm) incubated with l-aspartate
(2 mm). Only the a-Pyr and a-Ser chains are observed. C) Mass spectrum
of MtADC (60 mm) incubated with l-aspartate (2 mm) and subsequently
treated with NaCNBH3 (0.2 mm), showing the unmodified a-Pyr and a-
Ser chains as well as the trapped product (mass increase of +73 com-
pared with unlabelled a-Pyr).
Binding and single-turnover activity analysis by using
ESIMS: The apparent inability of the 3-fluoroaspartates to
act as either substrates or inhibitors (with one exception) of
ADCs led us to consider whether these compounds even
enter the enzymesꢀ active sites, and if they did, whether they
were able to form a Schiff base with the ketone of the pyru-
voyl group. For example, it is conceivable that the known
+
strong intramolecular F···NH3 interaction may prevent one
or both of the 3-fluoroaspartate isomers from taking on a
conformation that will allow it to be bound within the active
site.[11,22] Moreover, the electronic effects of the introduced
fluorine on the basicity and nucleophilicity of the amine
may also negatively affect the formation of the imine. We
also considered the possibility that a fluorinated aspartate
could enter the ADC active site, but that only a single-turn-
over reaction took place. Such a situation would result if a
reaction product formed a more stable iminium complex
the untreated control (A), of MtADC in the presence of l-
aspartate (B), and of MtADC and l-aspartate after treat-
ment with NaCNBH3 (C). Scheme 5 shows the expected
mass increases for such a labelling experiment if either sub-
strate or product is trapped by the NaCNBH3 treatment.
The results show that LC–ESIMS can successfully be used
for analysis of the binding complexes of ADC enzymes. It
also shows that only NaCNBH3-treated MtADC samples
contain a stable small molecule adduct of the native pyruvo-
10034
www.chemeurj.org
ꢁ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2010, 16, 10030 – 10041

3-Fluoroaspartate Probes
FULL PAPER
such product adduct was ob-
served when MtADC was treat-
ed with 8 in the absence of
NaCNBH3 (Figure 3A). None-
theless, similar to the results
obtained with 4, NaCNBH3
treatment of
a
solution of
MtADC and 8 also trapped the
product of decarboxylation and
defluorination, as demonstrated
by the formation of a peak cor-
responding to a protein adduct
with a mass increase of +72.
The results of equivalent ex-
periments using the EcADC
enzyme and the 3-fluoroaspar-
tates 4 and 8 showed similar
binding profiles (see the Sup-
porting Information): the prod-
uct of decarboxylation and de-
Scheme 5. Possible labelling pathways for an ADC enzyme treated with NaCNBH3 in the presence of its natu-
ral substrate, l-aspartate.
fluorination
(pathway c
in
yl-bearing a chain, with a mass increase of +73. This closely
corresponds to the predicted mass increase (+74) for the a
chain covalently labelled with the product of the reaction, as
shown in Scheme 5. The observed small mass discrepancy
was also found in the MALDI-MS method used by Webb
et al. No substrate-containing adducts were observed, indi-
cating that the rate of NaCNBH3-mediated labelling cannot
compete with the rate of decarboxylation of the substrate.
An equivalent experiment using EcADC gave similar results
(see the Supporting Information).
Schemes 4 and 6) is trapped on the enzyme by NaCNBH3
treatment in both cases, but only the erythro-isomer 4 gives
a product that can be observed when no NaCNBH3 is used.
However, in the case of EcADC, LC–ESIMS analysis also
showed the presence of additional mass peaks in some cir-
cumstances, most of which did not correspond with any pre-
dicted protein adduct. Although some of these peaks were
also present in the untreated (negative) controls, we current-
ly do not have an explanation for the fact that such protein
peaks are only observed in some cases.
We subsequently repeated the same experiment using
MtADC and either 4 (2 mm), or 8 (2 mm), and analyzed the
samples both with and without NaCNBH3 treatment.
Scheme 6 shows all the possible labelling pathways for an
ADC enzyme with these isomers, among which are path-
ways leading to trapping of the substrate, trapping of the
product of decarboxylation (pathway a), trapping of the
products of defluorination by elimination of HF (path-
way b), and trapping of the products that result from fluo-
ride elimination after decarboxylation takes place (path-
way c); these routes can be compared with the proposed re-
action pathways shown in Scheme 4.
Taken together, the ESIMS results clearly show that both
4 and 8 enter the active site of both ADC enzymes, where
both undergo decarboxylation followed by defluorination to
give enamine products that can be trapped by treatment
with NaCNBH3. Moreover, in the case of the erythro-isomer
4, the resulting product adduct is sufficiently stable to allow
its detection even without reduction of the Schiff base that
holds it in place in the active site. These findings indicate
that both 3-fluoroaspartates can act as substrates of ADC
enzymes for single turnover, but not catalytic, reactions.
The mass spectra of MtADC after treatment with 4 are
shown in Figure 2. These show that, even in the absence of
NaCNBH3, a protein adduct forms that has a mass increase
of +69 (Figure 2A), which closely corresponds to the ex-
pected mass increase (+70) for an adduct of the pyruvoyl-
bearing a chain and the product of decarboxylation and de-
fluorination (formed through pathway c). This finding shows
that the Schiff base formed between the enzyme and this
product is stable enough to survive and to be detected by
LC–ESIMS analysis, and that it is apparently not displaced
by the excess of 4, which would be present in the reaction
samples. As expected, the same product is also trapped by
NaCNBH3-treatment as demonstrated by the observed pro-
tein adduct with a mass increase of +72 (Figure 2B). No
Discussion
We initiated this investigation into the unique aspects of the
mechanism and active site architecture of the pyruvoyl-de-
pendent ADC enzymes with the design and synthesis of the
two l-isomers of 3-fluoroaspartate (4 and 8) as potential
probes and/or inhibitors. Although we successfully modified,
and in one case improved, a published synthesis of these flu-
orinated amino acids, we were nonetheless surprised at the
extent to which some of the fluorine-containing intermedi-
ates were prone to degradation—most probably through
elimination of fluoride. Nonetheless, we found the final
target molecules to be comparatively stable. This allowed us
to successfully use these molecules in the study of ADC as
Chem. Eur. J. 2010, 16, 10030 – 10041
ꢁ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
10035

E. Strauss et al.
ro and threo isomers of 3-fluo-
roglutamate, which showed that
the enzyme decarboxylates
both isomers in a catalytic fash-
ion.[14] Although the reasons for
this distinction is unclear, in the
absence of co-crystal structures
of these proteins, two possibili-
ties may be suggested: First, the
rate of iminium formation be-
tween an amine and an alde-
hyde (as in the case of PLP-en-
zymes) compared to a ketone
(as for the pyruvoyl-containing
ADC) could be more sensitive
to perturbations of the nucleo-
philicity of the amino group.
Second, the rate of product re-
lease could be at play, because
although ADC forms stable
product complexes, glutamate
decarboxylase shows similar
Vmax values for the two 3-fluo-
roglutamate isomers compared
to its natural substrate.
With the exception of 8,
which demonstrated an ability
to inhibit the decarboxylation
of l-aspartate by MtADC, none
of
the
3-fluoroaspartates
showed any inhibition of either
ADC enzyme under the condi-
tions used (0.5 mm each of sub-
strate and fluorinated ana-
logue). This finding may also be
explained based on the reason-
ing laid out above, since neither
4 nor 8 would be able to com-
pete with l-aspartate for forma-
tion of the requisite Schiff base
if the nucleophilicity of their
Scheme 6. Possible labelling pathways for an ADC enzyme treated with NaCNBH3 in the presence of a 3-fluo-
roaspartate.
originally planned, although it took significant effort to
obtain sufficient amounts of 4 and 8 to do so.
amines is reduced.
Our finding that neither isomer acted as substrates of
ADC that are catalytically turned over is not due to them
being excluded from the enzyme, since the LC–ESIMS bind-
ing experiments clearly demonstrated that both isomers are
able to enter the enzymeꢀs active site. Instead, the intro-
duced electronegative fluorine atom most probably reduces
the nucleophilicity of the amine and, as a result, the rate of
initial attack of these isomers on the ketone of the pyruvoyl
group. Furthermore, after iminium formation has taken
place, decarboxylation and defluorination apparently leads
to the formation of stable products that are not displaced by
the excess of substrate available in solution, suggesting the
occurrence of single turnover reactions. These findings are
in contrast to the results of the study of the reactivity of the
PLP-dependent glutamate decarboxylase towards the eryth-
On the other hand, inhibition of MtADC by 8 may arise
by this isomer closely mimicking the conformation of the
bound natural substrate. This observation is based on a com-
parison between the conformation of an aspartate analogue
(l-isoasparagine) trapped in the active site of Helicobacter
pylori ADC (HpADC; PDB: 1UHE),[6] the solution confor-
mations of 4 and 8 as determined by the NMR spectroscopic
study,[12] and their conformations in the solid state as deter-
mined by X-ray diffraction analysis.[13] These studies show
that in both cases the isomers prefer having the a-proton
and fluoride atom antiperiplanar (Figure 4), in a conforma-
tion that also positions the amine and fluorine atoms in
close proximity and allow a favourable intramolecular
+
F···NH3 interaction. As a result, both 8 and the aspartate
analogue bound in the HpADC active site have their car-
10036
www.chemeurj.org
ꢁ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2010, 16, 10030 – 10041

3-Fluoroaspartate Probes
FULL PAPER
Figure 2. Binding analysis of MtADC and l-erythro-3-fluoroaspartate (4)
by LC–ESIMS analysis of the complexes. A) Mass spectrum of MtADC
(60 mm) incubated with 4 (2 mm), showing the a-Pyr and a-Ser protein
chains as well as a protein adduct with a mass increase (+70) corre-
sponding to an adduct of the a-Pyr protein with the product of a-decar-
boxylation and defluorination (pathway c in Schemes 4 and 6). B) Mass
spectrum of MtADC (60 mm) incubated with 4 (2 mm) and subsequently
treated with NaCNBH3 (0.2 mm). A small peak is observed that corre-
sponds to a protein adduct with a mass increase of +72, the expected
mass of the trapped product.
Figure 3. Binding analysis of MtADC and l-threo-3-fluoroaspartate (8)
by LC–ESIMS analysis of the complexes. A) Mass spectrum of MtADC
(60 mm) incubated with 8 (2 mm), showing only the a-Pyr and a-Ser pro-
tein chains. B) Mass spectrum of MtADC (60 mm) incubated with 8
(2 mm) and subsequently treated with NaCNBH3 (0.2 mm), showing the
a-Pyr/product adduct (expected mass increase of +72).
boxyl groups in an antiperiplanar conformation, as demon-
strated by superimposition of the structure of 8 on that of
the bound analogue in the active site of HpADC (Fig-
ure 5B). The l-threo isomer 8 could therefore compete with
l-aspartate for binding in the MtADC active site, which is
essentially identical to the active site of HpADC.[6,7] The
fact that EcADC is not similarly inhibited may be due to it
having a slightly smaller active site, due to exchange of
Val47 found in MtADC and HpADC proteins for a Trp resi-
due (see Figure 5A for a sequence alignment of the
EcADC, MtADC and HpADC proteins).
The results of the LC–ESIMS binding analysis suggests
that, of the possible reaction pathways presented in
Scheme 4, only decarboxylation followed by fluoride elimi-
nation (pathway c) takes place. The reason why pathway b
(elimination of fluoride by deprotonation at the a carbon) is
not followed may be due to the absence of an active site
base that is correctly positioned to remove the a proton.
This assessment is supported by analysis of the enzymeꢀs
active site, which is highly conserved among all ADCs with
regard to both architecture and identity of the active site
residues (Figure 5).[5] The fact that no products resulting
from decarboxylation/reprotonation (Scheme 4, pathway a)
are observed is most probably due to the conformation of
the iminium enolate promoting fluoride elimination rather
than reprotonation. However, it is also possible that the
products of pathway a are less stable in the active site than
Figure 4. Conformations of l-erythro-3-fluoroaspartate (4) and its l-threo
isomer 8. A Newman projection of the most stable conformation (relative
to other staggered conformations, proportion indicated as percentage) of
each isomer in solution as determined by a previous NMR spectroscopic
study are shown,[12] as well as their structures as determined by X-ray dif-
fraction.[13]
those of pathway c and are therefore less likely to be
trapped and observed. The observation in one EcADC ex-
periment of a peak with a mass closely corresponding to the
expected mass for the reprotonated product adduct (+90)
(see the Supporting Information), lends credence to such an
argument.
Although our results show that fluoride is eliminated
during the course of the reaction, and that an a,b-unsaturat-
ed carboxylic acid (a Michael acceptor) is formed as the
product, no evidence could be found that pointed to an
Chem. Eur. J. 2010, 16, 10030 – 10041
ꢁ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
10037

E. Strauss et al.
boxylase, in which the erythro
isomer of 3-fluoroglutamate
also lost fluorine after decar-
boxylation but similarly did not
covalently trap any enzyme
active site residue.[14] This result
suggests that one of our original
goals—to devise a mechanism-
based inhibitor that would ex-
ploit the mechanistic differen-
ces between pyruvoyl-depen-
dent and PLP-dependent decar-
boxylases—is unlikely to be
achieved by using a b-fluorinat-
ed amino acid as a substrate an-
alogue.
Nonetheless, the most inter-
esting result of the ESIMS
binding analysis study does
point to a different tactic for in-
hibitor development. Specifical-
ly, the LC–ESIMS binding anal-
ysis shows that the l-erythro
isomer 4, but not the l-threo
isomer 8, forms enamine prod-
ucts that are sufficiently stable
to
be
detected
without
NaCNBH3-mediated trapping
(Figure 2A). However, the en-
amine products are expected to
have different configurations,
based on an analysis of the re-
action pathways of 4 and 8.
These show that in the case of 8
the enamine will have a cis con-
formation, but that 4 will lead
to formation of a trans enamine
(Figures 5B and C). The results
of the binding analysis suggests
that the enzyme adduct with
the trans enamine is more
stable than the same adduct
with its cis counterpart, most
probably due to different inter-
actions of these products with
active site residues. For exam-
ple, the carboxyl group of the
cis enamine is likely to clash
sterically with the amide car-
bonyl of Asn72 (indicated by
an arrow in Figure 5). These
findings imply that molecules
that are designed to be able to
Figure 5. Active site and reaction pathway analysis of ADC enzymes with isomers 4 and 8. A) Sequence align-
ment of EcADC, MtADC and HpADC enzymes. Identical residues are highlighted in yellow; the active site
residues are in blue. The arrow shows the position of cleavage leading to formation of the pyruvoyl group.
B) On the left, depiction of the active site of HpADC based on its determined structure (PDB: 1UHE) with
l-isoasparagine bound to the pyruvoyl group (Pyr), and the structure of 8 (in orange) overlaid on it. Residues
forming the floor of the active site are shown as a surface; the residues forming its roof are shown as stick
structures. Active site residues highlighted in blue in A are labelled (numbering as for EcADC and MtADC).
On the right, the reaction pathway analysis of 8 showing formation of a cis-enamine product. C) As for B, but
showing 4 in the active site (in orange) on the left, and predicted formation of its trans-enamine product on
the right. D) Reaction pathway analysis for 4 if it takes on the conformation of the bound substrate by rotation
of its b-carboxyl group.
active site nucleophile being covalently trapped by it. This is
in spite of the predicted close proximity of the lysine residue
(Lys9) to the electrophilic centre. The same result was also
found in the study of the PLP-dependent glutamate decar-
enter the small active sites of ADC enzymes, and which
mimic the configuration of the trans-enamine product, may
be effective inhibitors of these enzymes due to their appar-
ent stability within the active site.
10038
www.chemeurj.org
ꢁ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2010, 16, 10030 – 10041

3-Fluoroaspartate Probes
FULL PAPER
It is important to emphasize that the predicted reaction
pathways are based on an analysis that indicates that the
isomers bind in the active site of ADC in their most stable
conformations as determined by the NMR spectroscopy and
XRD studies (Figure 4). In the case of 8 this conformation
is identical to that of the bound aspartate substrate, which is
positioned in the active site by the iminium bond and ionic
interactions of the a- and b-carboxylates with Lys9 and
Arg54, respectively (Figure 5B). However, this is not the
case with the most stable conformation of 4, which overlaps
in all respects except for its b-carboxylate group (Fig-
tions that the introduced fluorine imposes on their confor-
mation can lead to differentiated reaction profiles. Further-
more, our results suggest that the products that form as a
result of such different reaction profiles also take on differ-
ent configurations, which, in one case, leads to the formation
of a stable protein–product adduct. In combination with the
results from previous binding analysis studies, which
mapped out the active site of EcADC, the findings from our
reaction profile analysis provides further direction for the
development of ADC inhibitors that may distinguish be-
tween pyruvoyl- and PLP-dependent enzymes.
À
ure 5C). Rotation around its C2 C3 bond as shown in Fig-
ure 5D would position this group in a manner that allows
ionic interaction with Arg54, and would lead to formation
of the alternative conformation for this isomer bound in the
ADC active site. Nonetheless, several points indicate that
this rotation does not take place. First, it would result in in-
terruption of the favourable intramolecular F···NH+ interac-
tion, although formation of the ionic interaction between
the b-carboxylate and Arg54 may sufficiently compensate
for this loss in binding energy. Second, decarboxylation of
this conformation would place the fluorine atom anticlinical
relative to the new iminium (Figure 5D). This would pre-
vent fluoride elimination from taking place and would be
expected to lead to trapping of the fluorine-containing prod-
uct, which is not observed. Third, because 4 does form a
stable enamine product by fluoride elimination, the iminium
enolate intermediate must take on one of the conformations
shown boxed in Figure 5D. The first requires rotation of the
b-carboxyl group back to its original position (a futile
energy cycle) and leads to formation of the trans enamine as
predicted originally. The second (if it can be accessed) leads
to formation of a cis enamine that is identical to the predict-
ed product of 8, which would contradict our results that
show that the enamine products that form from the two iso-
mers can clearly be differentiated.
Taken together, our results and this analysis suggest that
the intramolecular F···NH+ interaction strongly influences
the binding of the l-3-fluoroaspartate isomers to the ADC
enzyme, causing them to maintain their most stable solution
conformations even in the active site. As a result, the iso-
mers form enamine products of different configurations, of
which the trans enamine is the more stable. These findings
highlight the extent to which the presence of fluorine in a
molecule can direct binding conformations in enzymes, even
when they are in competition with other potential binding
interactions.
Experimental Section
Activity assay based on fluoride-release: The experiment was based on
literature procedures,[23] which were modified as follows: Reactions were
monitored in NMR tubes containing 60 mm enzyme (EcADC or MtADC)
and 12 mm of either 4 or 8 in 50 mm Tris-HC (pH 7.6) and 5 mm MgCl2 in
D2O (68% v/v). Reactions were run over 20 min, with a spectrum record-
ed every 30 s. The progress of the reaction was assessed by monitoring
the change in intensity of the peak corresponding to the fluorine con-
tained in the substrate over time, as well as by monitoring the formation
of a peak corresponding to inorganic fluoride. As a positive control, com-
pound 8 was also incubated in the absence of enzyme for 20 min, with a
spectrum recorded every 5 min.
Activity assay based on CO2 release: ADC activity was monitored by
using the published coupled enzyme assay that relates CO2 release to the
consumption of NADH.[19] Briefly, each 150 mL reaction mixture con-
tained 2.2 mm PEP, 10 mm MgCl2, 0.32 mm NADH, 2.0 mm DTT, 1.5 U
malate dehydrogenase, 0.5 U PEP carboxylase in 50 mm Tris-HCl
(pH 7.6), and either 4.0 mg EcADC or 3.0 mg MtADC. The concentrations
of the tested substrates (l-aspartate, 4 and 8) were varied between 0 and
2.0 mm. Assays were performed in triplicate in flat-bottomed 96-well
plates at either 258C (for EcADC) or 378C (for MtADC). Reactions
were initiated by the addition of the substrate, and subsequently moni-
tored spectrophotometrically by following the oxidation of NADH to
NAD+ at A340 using an extinction coefficient of 6220mÀ1 cmÀ1 for
,
NADH. The resulting data were fitted to the Michaelis–Menten equation
using non-linear regression (SigmaPlot) to obtain the kinetic parameters
kcat and KM.
ESIMS-based binding analysis: The ESIMS experiments were performed
according to literature procedures,[2,4] with slight modification. A typical
reaction consisted of approximately 60 mm enzyme (EcADC or MtADC)
and 2 mm substrate (l-aspartate, 4 or 8) in 100 mm phosphate buffer
(pH 7.6). Samples were incubated for 2 min before 0.2 mm NaCNBH3
was added to selected samples to trap the substrate and/or product in the
active site. Subsequently, 10 mL of each reaction mixture was used for
LC–ESIMS analysis by using a Waters UPLC 5 mm C18 column (2.1ꢂ
50 mm) on a Waters UPLC with a Waters 996 Photodiode Array Detec-
tor and acetonitrile as mobile phase, followed by MS analysis on a
Waters Quattro Micromass mass spectrometer.
(2R,3R)-Diethyl
oxirane-2,3-dicarboxylate
(1):
l-DET
(4.00 g,
19.2 mmol) was added to a round-bottomed flask and cooled to 08C with
an ice bath. HBr/AcOH (30%, 16 mL) was subsequently added by
means of a dropping funnel over 30 min with stirring. The reaction mix-
ture was stirred for a further 15 min at 08C and then allowed to warm to
RT. The reaction mixture was stirred overnight at RT, after which time it
was poured into ice/water and extracted with Et2O (3ꢂ15 mL). The or-
ganic extracts were washed with H2O (3ꢂ10 mL) and brine (2ꢂ8 mL)
and dried over MgSO4. The drying agent was filtered off and the solvent
was removed in vacuo to give the crude protected bromohydrin, which
was dissolved in a solution of absolute EtOH (30 mL) to which acetyl
chloride (450 mL, 6.30 mmol) was previously carefully added. The reac-
tion mixture was heated to reflux for 7 h, after which it was cooled to RT
Conclusion
In this study, we set out to exploit the unique characteristics
that fluorine imparts to small molecules to probe the reac-
tivity and binding of ADC enzymes. Although our studies
using 3-fluoroaspartates towards this end showed that such
fluorinated substrate analogues cannot act as mechanism-
based inhibitors of pyruvoyl-dependent ADCs, the restric-
Chem. Eur. J. 2010, 16, 10030 – 10041
ꢁ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
10039

E. Strauss et al.
and the solvent was removed in vacuo. The product was purified by flash
chromatography (silica gel; EtOAc/hexane 1:2) to give the bromohydrin
as a colourless oil (4.0 g, 78% yield). 1H NMR (400 MHz, CDCl3, 258C,
TMS): d=1.31 (t, J=7.12 Hz, 6H; 2ꢂCH3), 4.29 (m, 4H; 2ꢂCH2), 4.66
(m, 1H; CH), 4.71 ppm (m, 1H; CH). K2CO3 (14.8 g, 107.1 mmol) was
subsequently added to a round-bottomed flask containing the bromohy-
drin (4.37 g, 16.2 mmol) dissolved in anhydrous acetone (50 mL). The
flask was flushed with N2 (g) and the reaction mixture was stirred at RT
for 2 h. The excess K2CO3 was filtered off and the solvent was removed
in vacuo to give a colourless residue. The residue was dissolved in CHCl3
(50 mL) and treated with active charcoal (14.0 g). The charcoal was fil-
tered off and washed exhaustively with EtOH. The solvent was removed
in vacuo to give the crude product as a colourless oil, which was purified
by flash chromatography (silica gel; EtOAc/hexane 1:3) to give 1 as a
colourless oil (2.45 g, 80% yield; 65% overall from l-DET). 1H NMR
(400 MHz, CDCl3, 258C, TMS): d=1.32 (t, J=7.12 Hz, 6H; 2ꢂCH3),
3.66 (s, 2H; 2ꢂCH), 4.27 ppm (m, 4H; 2ꢂCH2).
flask was subsequently flushed with N2 (g) and the reaction mixture was
stirred for 40 min at RT. The reaction was quenched with a saturated
aqueous solution of CuSO4 (250 mL) and the two layers were separated.
The aqueous layer was extracted with CH2Cl2 (3ꢂ50 mL) and the com-
bined organic layers were dried over Na2SO4. The drying agent was fil-
tered off and the solvent was removed in vacuo to give the intermediate
cyclic sulfite as a light-yellow oil. The oil was dissolved in acetonitrile
(25 mL) and NaIO4 (3.11 g, 14.6 mmol), RuCl3·xH2O (15 mg, 0.057 mmol,
1% cat) and H2O (20 mL) were added. The reaction turned greenish-
brown and was stirred at RT for 40 min. Et2O (250 mL) was added, the
layers were separated and the organic layer was washed with a saturated
aqueous solution of NaHCO3 (2ꢂ50 mL) and dried over Na2SO4, fol-
lowed by filtration through a silica plug. The solvent was removed in
vacuo to give 5 as a colourless oil, which became a white solid (810 mg,
1
52%) upon standing overnight. H NMR (400 MHz, CDCl3, 258C, TMS):
d=1.38 (t, J=7.13 Hz, 6H; CH3), 4.39 (dq, J=5.4, 7.3, 1.8 Hz, 4H; CH2),
5.46 ppm (s, 2H; CH).
ACHTUNGTRENNUNG(2S,3R)-Diethyl 2-(dibenzylamino)-3-hydroxysuccinate (2): YbACHTUGNTREN(NUNG OTf)3
AHCTUNGTERG(NNUN 2S,3S)-Diethyl 2-fluoro-3-hydroxysuccinate (6): Cyclic sulfate 5 (1.18 g,
(93 mg, 0.15 mmol) was dissolved in anhydrous THF (5.5 mL) in a round-
bottomed flask, which was subsequently flushed with N2(g). Epoxide 1
(188 mg, 1.0 mmol) was added and the reaction mixture was stirred for
5 min then dibenzyl amine (384 mL, 2.0 mmol) was added. The solution
was heated to reflux and monitored by TLC until all of the starting mate-
rial had been consumed (ca. 16 h). The reaction was cooled to RT and
quenched with H2O (1.5 mL). The aqueous layer was extracted with
Et2O (3ꢂ2 mL) and the combined organic layers were washed with brine
(1ꢂ2 mL) and dried over K2CO3. The drying agent was filtered off and
the solvent was removed in vacuo to give a colourless oil, which was puri-
fied by flash chromatography (silica gel; MeOH/CH2Cl2, 1%) to give 2
as a colourless oil (375 mg, 97%). 1H NMR (400 MHz, CDCl3, 258C,
TMS): d=1.19 (t, J=7.13 Hz, 3H; CH3), 1.30 (t, J=7.23 Hz, 3H; CH3),
3.80–3.95 (m, 4H; CH2), 3.91 (d, J=4.1 Hz, 1H; CH), 4.16 (q, J=7.1,
6.8 Hz, 2H; CH2), 4.22 (m, 2H; CH2), 4.85 (t, J=3.91 Hz, 1H; CH),
7.20–7.40 ppm (m, 10H; Ar).
4.40 mmol) was dissolved in acetone (22 mL) and tetraethylammonium
fluoride dihydrate (1.22 g, 6.60 mmol) was added. The reaction mixture
was stirred vigorously at RT for 7 h, after which the solvent was removed
in vacuo and the residue was suspended in Et2O (22 mL). H2SO4 (20% v/
v, 22 mL) was added and the reaction mixture was stirred for a further
7 h. The layers were separated and the aqueous layer was extracted with
Et2O (3ꢂ15 mL), which was combined with the organic layer. The com-
bined organic layers were washed with H2O (2ꢂ10 mL) and brine
(10 mL) and dried over MgSO4. The drying agent was filtered off and the
solvent was removed in vacuo to give a brown oil, which was purified by
flash chromatography (silica gel; EtOAc/hexane 1:1) to give 6 as a light-
yellow oil (503 mg, 55%). 1H NMR (300 MHz, CDCl3, 258C, TMS): d=
1.31 (td, J=7.13, 2.34 Hz, 6H; 2ꢂCH3), 4.30 (m, 4H; 2ꢂCH2), 4.69 (dd,
J=21.5, 2.05 Hz, 1H; CH), 5.24 ppm (dd, J=45.30, 2.10 Hz, 1H; CH);
13C NMR (75 MHz, CDCl3, 258C): d=14.1 (2ꢂCH3), 62.5 (d, J=
84.70 Hz; 2ꢂCH2), 71.5 (d, J=21.40 Hz; OH), 89.8 (d, J=193 Hz; CF),
166.1 (C=O), 169.8 ppm (C=O); 19F NMR (300 MHz, CDCl3, 258C): d=
À202.43 ppm (dd, J=23.0, 47.0 Hz, 1F).
ACHTUNGTRENNUNG(2R,3R)-Diethyl 2-(dibenzylamino)-3-fluorosuccinate (3): Compound 2
(805 mg, 2.1 mmol) was dissolved in CH2Cl2 (35 mL) and (diethylamino)-
sulfur trifluoride (DAST; 332 mL, 2.52 mmol) was added dropwise. The
solution was stirred at RT for 30 min then a further portion of DAST
(40 mL, 0.3 equiv) was added and the reaction was stirred for 1 h. K2CO3
(580 mg, 4.2 mmol) was added in one portion and the reaction was stirred
for 5 min and then poured into a saturated aqueous solution of NaHCO3
(35 mL). The aqueous layer was extracted with CH2Cl2 (3ꢂ20 mL) and
the combined organic layers dried over MgSO4. The drying agent was fil-
tered off and the solvent was removed in vacuo to give a yellow oil,
which was purified by flash chromatography (silica gel; MeOH/CH2Cl2,
0.5%) to give 3 as a colourless oil (632 mg, 78%). 1H NMR (400 MHz,
CDCl3, 258C, TMS): d=1.22 (t, J=7.13 Hz, 3H; CH3), 1.34 (t, J=
7.12 Hz, 3H; CH3), 3.81 (dd, J=13.7, 25.7 Hz, 4H; 2ꢂCH2), 4.04 (dd, J=
5.3, 17.5 Hz, 1H; CH), 4.21 (m, 4H; 2ꢂCH2), 5.28 (dd, J=5.3, 42.7 Hz,
1H; CH), 7.29–7.38 ppm (m, 10H; Ar); 19F NMR (400 MHz, CDCl3,
258C): d=À196.20 ppm (dd, J=22.5, 47.7 Hz, 1F).
(2S,3R)-Diethyl 2-fluoro-3-isocyanatosuccinate (7): Compound
6
(130 mg, 0.625 mmol) was dissolved in anhydrous CH2Cl2 (6.5 mL) and
cooled to À788C. The reaction mixture was stirred for 10 min at À788C,
then triflic anhydride (126 mL, 0.750 mmol) was added dropwise. The re-
action was stirred for a further 10 min at À788C, then diisopropylethyla-
mine (128 mL, 0.750 mmol) was added. The reaction was allowed to warm
to RT, after which the solvent was removed in vacuo (without heating) to
give a colourless oil, which was dissolved in anhydrous DMF (6.5 mL)
and cooled to À108C using an ice/salt bath. KNCO (507 mg, 6.25 mmol)
was added to the reaction, which was then stirred for 3–4 h at À108C.
EtOAc (10 mL) and H2O (8 mL) were added and the layers were sepa-
rated. The aqueous layer was washed with EtOAc (10 mL) and the com-
bined organic layers were washed with H2O (4ꢂ5 mL) and brine
(10 mL), and dried over MgSO4. The drying agent was removed by filtra-
tion and the solvent was removed in vacuo to give a dark-yellow oil. The
product was purified by flash chromatography (silica gel; EtOAc/hexane
1:2) to give 7 as a colourless oil (30 mg, 21%). 1H NMR (400 MHz,
CDCl3, 258C, TMS): d=1.31 (m, 6H; 2ꢂCH3), 4.28 (m, 4H; 2ꢂCH2),
4.70 (dddd, J=15.8, 5.66, 2.15 Hz, 1H; CF), 5.23 ppm (dd, J=45.3,
2.15 Hz, 1H; CNCO); 13C NMR (75 MHz, CDCl3, 258C): d=14.0 (d, J=
4.40 Hz; 2ꢂCH3), 62.5 (d, J=65.20 Hz; 2ꢂCH2), 71.5 (d, J=21.60 Hz;
NCO), 89.8 (d, J=193.5 Hz; CF), 169.7 ppm (C=O); 19F NMR (300 MHz,
CDCl3, 258C): d=À196.00 ppm (dd, J = 23.7, 43.7 Hz).
ACHTUNGTRENNUNG(2R,3R)-3-Fluoroaspartic acid (4): Compound 3 (347 mg, 0.895 mmol)
was dissolved in HCOOH/MeOH (4.4%, 20 mL) and stirred at RT. Pd
black (200 mg) was added and the reaction was stirred until all starting
material was consumed (based on TLC analysis; MeOH/CH2Cl2, 1%).
The Pd black was removed by filtration through a cotton wool plug and
washed with MeOH (3ꢂ10 mL). The solvent was removed in vacuo to
give a colourless residue, which was dissolved in 5m HCl (20 mL) and
heated to reflux for 24 h. The reaction mixture was cooled to RT and the
solvent was removed by lyophilisation until a constant mass was achieved
to give 4 as a pure white solid (50 mg, 37%). 1H NMR (400 MHz, D2O,
258C): d=4.3 (dd, J=2.54, 26.6 Hz, 1H; CH), 5.16 ppm (dd, J=2.54,
49.2 Hz, 1H; CH); 19F NMR (400 MHz, CDCl3, 258C): d=À198.50 ppm
(dd, J=22.9. 49.0 Hz, 1F).
AHCTUNGERTG(NNUN 2R,3S)-3-Fluoroaspartic acid (8): Compound 7 (30 mg, 0.128 mmol) was
dissolved in 5m HCl (4 mL) and the reaction was heated to reflux for
24 h. The reaction was cooled down to RT and the solvent was removed
by lyophilisation until a constant mass was achieved, to give 8 as a pure
1
white solid (4.5 mg, 23%). H NMR (400 MHz, D2O, 25 8C): d=4.83 (dd,
ACHTUNGTRENNUNG(2R,3R)-Diethyl-2,3-dihydroxy cyclic sulfate (5): l-DET (1.0 mL,
J=2.1, 21.1 Hz, 1H; CH), 5.40 ppm (dd, J=2.1, 45.0 Hz, 1H; CHF);
13C NMR (100 MHz, D2O, 25 8C): d=71.1 (d, J=21.6 Hz, CN), 90.3 (d,
J=186.9 Hz, CF), 171.2 (d, J=24.2 Hz, COOH), 173.4 ppm (d, J=
5.84 mmol) was dissolved in CH2Cl2 (25 mL) and pyridine (705 mL,
8.76 mmol) was added. The reaction mixture was stirred for 5 min at RT
then SOCl2 (508 mL, 7.0 mmol) was carefully added dropwise at RT. The
10040
www.chemeurj.org
ꢁ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2010, 16, 10030 – 10041

3-Fluoroaspartate Probes
FULL PAPER
10.1 Hz, COOH); 19F NMR (400 MHz, D2O, 25 8C): d=À197.88 ppm (dd,
J=22.3, 47.3 Hz, 1F).
Slawin, T. Lebl, F. Chorki, J. P. Issberner, T. van Mourik, J. M. Hey-
gate, J. J. Lambert, L. A. Etherington, K. T. Sillar, D. OꢀHagan,
ChemBioChem 2007, 8, 2265–2274.
[12] S. Hamman, M. C. Salon, C. Beguin, Org. Magn. Reson. 1982, 20,
78–82.
Acknowledgements
[13] a) A. M. Stern, B. M. Foxman, A. H. Tashjian, R. H. Abeles, J. Med.
Chem. 1982, 25, 544–550; b) M. J. Wanner, J. J. M. Hageman, G. J.
Koomen, U. K. Pandit, J. Med. Chem. 1980, 23, 85–87; c) K. Gou-
bitz, H. Schenk, C. H. Stam, Cryst. Struct. Commun. 1981, 10, 1029–
1030.
This work was supported by grants from the South African Medical Re-
search Council (MRC) and the International Centre for Genetic Engi-
neering and Biotechnology (ICGEB) to E.S.
[14] A. Vidal-Cros, M. Gaudry, A. Marquet, Biochem. J. 1985, 229, 675–
678.
[15] a) J. Kollonitsch, S. Marburg, L. M. Perkins, J. Org. Chem. 1979, 44,
771–777; b) J. Kollonitsch, S. Marburg, L. Perkins, J. Org. Chem.
1975, 40, 3808–3809; c) K. Matsumoto, Y. Ozaki, T. Iwasaki, H. Ho-
rikawa, M. Miyoshi, Experientia 1979, 35, 850–851.
[16] M. Hudlicky, J. Fluorine Chem. 1988, 40, 99–108.
[17] F. B. Charvillon, R. Amouroux, Tetrahedron Lett. 1996, 37, 5103–
5106.
[1] a) T. P. Begley, C. Kinsland, E. Strauss, Vitam. Horm. 2001, 61, 157–
171; b) E. Strauss in Comprehensive Natural Products II: Chemistry
and Biology (Eds.: L. Mander, H.-W. Liu), Elsevier, Oxford, 2010,
Vol. 7, pp. 351–410.
[2] K. V. Sambandamurthy, X. Wang, B. Chen, G. R. Russell, S. Derrick,
M. F. Collins, L. S. Morris, R. W. Jacobs, Jr., Nat. Med. 2002, 8,
1171–1174.
[18] a) S. Kobayashi, Eur. J. Org. Chem. 1999, 15–27; b) M. Chini, P.
Crotti, L. Favero, F. Macchia, M. Pineschi, Tetrahedron Lett. 1994,
35, 433–436; c) A. McCluskey, S. K. Leitch, J. Garner, C. E. Caden,
T. A. Hill, L. R. Odell, S. G. Stewart, Tetrahedron Lett. 2005, 46,
8229–8232.
[3] A. W. Hung, H. L. Silvestre, S. Wen, A. Ciulli, T. L. Blundell, C.
Abell, Angew. Chem. 2009, 121, 8604–8608; Angew. Chem. Int. Ed.
2009, 48, 8452–8456.
[4] M. E. Webb, E. Stephens, A. G. Smith, C. Abell, Chem. Commun.
2003, 2416–2417.
[19] a) A. Osterman, N. V. Grishin, L. N. Kinch, M. A. Phillips, Biochem-
istry 1994, 33, 13662–13667; b) A. L. Osterman, H. B. Brooks, J.
Rizo, M. A. Phillips, Biochemistry 1997, 36, 4558–4567.
[20] S. Chopra, H. Pai, A. Ranganathan, Protein Expression Purif. 2002,
25, 533–540.
[5] M. E. Webb, A. G. Smith, C. Abell, Nat. Prod. Rep. 2004, 21, 695–
721.
[6] B. I. Lee, S. W. Suh, J. Mol. Biol. 2004, 340, 1–7.
[7] G. Gopalan, S. Chopra, A. Ranganathan, K. Swaminathan, Proteins
Struct. Funct. Bioinf. 2006, 65, 796–802.
[21] Y. Kai, H. Matsumura, T. Inoue, K. Terada, Y. Nagara, T. Yoshina-
ga, A. Kihara, K. Tsumura, K. Izui, Proc. Natl. Acad. Sci. USA
1999, 96, 823–828.
[22] a) M. Morgenthaler, E. Schweizer, A. Hoffmann-Roder, F. Benini,
R. E. Martin, G. Jaeschke, B. Wagner, H. Fischer, S. Bendels, D.
Zimmerli, J. Schneider, F. Diederich, M. Kansy, K. Muller, Chem-
MedChem 2007, 2, 1100–1115; b) A. Sun, D. C. Lankin, K. Hardcas-
tle, J. P. Snyder, Chem. Eur. J. 2005, 11, 1579–1591.
[23] M. H. Gelb, Y. Lin, M. A. Pickard, Y. Song, J. C. Vederas, J. Am.
Chem. Soc. 1990, 112, 4932–4942.
[8] M. K. Ramjee, U. Genschel, C. Abell, A. G. Smith, Biochem. J.
1997, 323, 661–669.
[9] A. Albert, V. Dhanaraj, U. Genschel, G. L. Khan, M. K. Ramjee, R.
Pulido, B. L. Sibanda, F. von Delft, M. Witty, T. L. Blundell, A. G.
Smith, C. Abell, Nat. Struct. Biol. 1998, 5, 289–293.
[10] a) R. Poulin, L. Lu, B. Ackermann, P. Bey, A. E. Pegg, J. Biol.
Chem. 1992, 267, 150–158; b) N. Qu, N. A. Ignatenko, P. Yamauchi,
D. E. Stringer, C. Levenson, P. Shannon, S. Perrin, E. W. Gerner,
Biochem. J. 2003, 375, 465–470; c) B. W. Metcalf, P. Bey, C. Danzin,
M. J. Jung, P. Casara, J. P. Vevert, J. Am. Chem. Soc. 1978, 100,
2551–2553.
[11] a) N. E. J. Gooseman, D. OꢀHagan, M. J. G. Peach, A. M. Z. Slawin,
D. J. Tozer, R. J. Young, Angew. Chem. 2007, 119, 6008–6012;
Angew. Chem. Int. Ed. 2007, 46, 5904–5908; b) G. Deniau, A. M. Z.
Received: March 10, 2010
Revised: May 20, 2010
Published online: July 19, 2010
Chem. Eur. J. 2010, 16, 10030 – 10041
ꢁ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
10041