Library Screening by Means of Mass Spectrometry (MS) Binding Assays-Exemplarily Demonstrated for a Pseudostatic Library Addressing γ-Aminobutyric Acid (GABA) Transporter1 (GAT1)

Article abstract of DOI:10.1002/cmdc.201200201

In the present study, the application of mass spectrometry (MS) binding assays as a tool for library screening is reported. For library generation, dynamic combinatorial chemistry (DCC) was used. These libraries can be screened by means of MS binding assays when appropriate measures are taken to render the libraries pseudostatic. That way, the efficiency of MS binding assays to determine ligand binding in compound screening with the ease of library generation by DCC is combined. The feasibility of this approach is shown for γ-aminobutyric acid (GABA) transporter1 (GAT1) as a target, representing the most important subtype of the GABA transporters. For the screening, hydrazone libraries were employed that were generated in the presence of the target by reacting various sets of aldehydes with a hydrazine derivative that is delineated from piperidine-3-carboxylic acid (nipecotic acid), a common fragment of known GAT1 inhibitors. To ensure that the library generated is pseudostatic, a large excess of the nipecotic acid derivative is employed. As the library is generated in a buffer system suitable for binding and the target is already present, the mixtures can be directly analyzed by MS binding assays-the process of library generation and screening thus becoming simple to perform. The binding affinities of the hits identified by deconvolution were confirmed in conventional competitive MS binding assays performed with single compounds obtained by separate synthesis. In this way, two nipecotic acid derivatives exhibiting a biaryl moiety, 1-{2-[2'-(1,1'-biphenyl-2-ylmethylidene)hydrazine]ethyl}piperidine-3-carboxylic acid and 1-(2-{2'-[1-(2-thiophenylphenyl)methylidene]hydrazine}ethyl)piperidine-3-carboxylic acid, were found to be potent GAT1 ligands exhibiting pK_i values of 6.186±0.028 and 6.229±0.039, respectively. This method enables screening of libraries, whether generated by conventional chemistry or DCC, and is applicable to all kinds of targets including membrane-bound targets such as Gprotein coupled receptors (GPCRs), ion channels and transporters. As such, this strategy displays high potential in the drug discovery process.

Full text of DOI:10.1002/cmdc.201200201

MED
DOI: 10.1002/cmdc.201200201
Library Screening by Means of Mass Spectrometry (MS)
Binding Assays—Exemplarily Demonstrated for
a Pseudostatic Library Addressing g-Aminobutyric Acid
(GABA) Transporter 1 (GAT1)
Miriam Sindelar and Klaus T. Wanner*[a]
Dedicated to Prof. Herbert Mayr with warmest wishes on the occasion of his 65th birthday
In the present study, the application of mass spectrometry
(MS) binding assays as a tool for library screening is reported.
For library generation, dynamic combinatorial chemistry (DCC)
was used. These libraries can be screened by means of MS
binding assays when appropriate measures are taken to render
the libraries pseudostatic. That way, the efficiency of MS bind-
ing assays to determine ligand binding in compound screening
with the ease of library generation by DCC is combined. The
feasibility of this approach is shown for g-aminobutyric acid
(GABA) transporter 1 (GAT1) as a target, representing the most
important subtype of the GABA transporters. For the screen-
ing, hydrazone libraries were employed that were generated in
the presence of the target by reacting various sets of alde-
hydes with a hydrazine derivative that is delineated from pi-
peridine-3-carboxylic acid (nipecotic acid), a common fragment
of known GAT1 inhibitors. To ensure that the library generated
is pseudostatic, a large excess of the nipecotic acid derivative
is employed. As the library is generated in a buffer system suit-
able for binding and the target is already present, the mixtures
can be directly analyzed by MS binding assays—the process of
library generation and screening thus becoming simple to per-
form. The binding affinities of the hits identified by deconvolu-
tion were confirmed in conventional competitive MS binding
assays performed with single compounds obtained by separate
synthesis. In this way, two nipecotic acid derivatives exhibiting
a biaryl moiety, 1-{2-[2’-(1,1’-biphenyl-2-ylmethylidene)hydrazi-
ne]ethyl}piperidine-3-carboxylic acid and 1-(2-{2’-[1-(2-thiophe-
nylphenyl)methylidene]hydrazine}ethyl)piperidine-3-carboxylic
acid, were found to be potent GAT1 ligands exhibiting pKi
values of 6.186Æ0.028 and 6.229Æ0.039, respectively. This
method enables screening of libraries, whether generated by
conventional chemistry or DCC, and is applicable to all kinds of
targets including membrane-bound targets such as G protein
coupled receptors (GPCRs), ion channels and transporters. As
such, this strategy displays high potential in the drug discovery
process.
Introduction
Hit identification is one of the fundamental steps in the early
stages of the drug discovery process. Often large compound
collections that might contain molecules originating from dif-
ferent sources are searched for ligands that bind and modulate
the activity of a biopolymer that has been selected as a thera-
peutic target for the cure of a disease of interest. For function-
ally active targets like enzymes or receptors, luminescence-
and fluorescence-based methods are quite common, which is
mainly due to the high sample throughput that can be ach-
ieved with such techniques. In the last ten years, mass spec-
trometry (MS) has also become an important tool in drug
screening. Different concepts have evolved that utilize MS for
this purpose. In the most common approach, compounds for-
merly bound to the target under equilibrium or near equilibri-
um conditions are liberated and subsequently identified by
MS. These MS-based techniques, which are termed affinity se-
lection mass spectrometry (AS-MS), are quite powerful and
have found widespread application in drug screening; howev-
er, they require monitoring of all constituents of the library by
MS.[1,2]
MS binding assays represent a novel technique to perform
saturation, competition and kinetic experiments. As such, they
are closely related to radioligand binding assays. But with no
labeling of the ligands for quantitation by MS being required,
they avoid all the drawbacks that result from radioactivity asso-
ciated with radiometric assays.[3]
So far, competitive MS binding assays have been widely
used to characterize single compounds in competition studies
by establishing competition curves to create precise affinity
data.[3,4] Being easy and simple to perform, competitive MS
binding assays could also be considered as readouts for drug
screening. For this purpose, competitive MS binding assays
could be reduced to a single-point determination quantifying
the amount of bound reporter ligand competing with com-
[a] M. Sindelar, Prof. Dr. K. T. Wanner
Department fꢀr Pharmazie—Zentrum fꢀr Pharmaforschung
Ludwig-Maximilians-Universitꢁt Mꢀnchen
Butenandtstr. 5–13, 81377 Mꢀnchen (Germany)
E-mail: klaus.wanner@cup.uni-muenchen.de
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cmdc.201200201.
ChemMedChem 0000, 00, 1 – 14
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
1
&
ÞÞ
These are not the final page numbers!

M. Sindelar and K. T. Wanner
MED
pound library members for the binding sites at a given con-
centration. Thus, the analytical demand for this kind of assay is
far less than for the above-mentioned AS-MS methods, which
require the quantification of all library components.[3] Of
course, radioligand binding assays could also be used for drug
screening but are largely avoided for comprehensive drug
screening campaigns because of the radioactively labeled com-
pounds required for this technique. Thus, by avoiding these
drawbacks, competitive MS binding assays should have high
potential for application in drug screening.
tion,[6] and so, the process is generally very easy to perform.
We considered it worthwhile to use DCLs for our project in
which—as already outlined above—demonstrating the useful-
ness of competitive MS binding assays as tools in drug screen-
ing was a major aim. Combining DCC reactions for library gen-
eration and competitive MS binding assays for screening can
be realized in the following manner: The DCL should be gener-
ated in the presence of the target, not to enhance the concen-
trations of best binders by a template effect, but rather to im-
prove the assay performance. In addition, this approach could
possibly help to accelerate the formation of best binders by ki-
netic template effects, thus amplifying their concentrations in
the compound mixture, in case libraries had not reached ther-
modynamic equilibrium because of insufficient reaction time
or inappropriate reaction conditions.
In this study, we demonstrate that competitive MS binding
assays are a useful tool for drug screening. To this end, MS
binding assays are applied to compound libraries generated
by utilizing simple and efficient reactions known from dynamic
combinatorial chemistry (DCC),[5–7] which results in a highly effi-
cient method for library generation and screening, as outlined
below.
For the analysis of DCLs, the libraries should favorably con-
sist of equal amounts of individual constituents. As the compo-
sition of dynamic libraries depends on the thermodynamic sta-
bilities of the individual members, this requires the compo-
nents to be mostly isoenergetic; however, this is rarely ach-
ieved. To overcome this problem, only a one-dimensional li-
brary should be used, in which a set of structurally diverse
building blocks is reacted with a single compound of comple-
mentary chemical reactivity. In order to force the reaction equi-
librium towards the products, the single compound with fixed
structure should be employed in a large excess as compared
with the structurally diverse fragments. In this way, the library
should become pseudostatic, and by starting from a mixture
with equal amounts of the structurally diverse fragments, a li-
brary with all constituents being present in similar amounts
should be obtained (Figure 1).
Dynamic combinatorial libraries (DCLs) are generated by
combining building blocks of suitable reactivity allowing the
formation of all possible ligands that interchange in a thermo-
dynamic equilibrium. Classical, adaptive DCLs are generated in
the presence of the protein target that serves as a template,
shifting the equilibrium towards the best binders. Identification
of amplified compounds is carried out by monitoring the shift
of the composition of the library effected by the target or by
analyzing formed target–ligand complexes by an MS screening
technique, such as electrospray ionization (ESI)-MS/MS,[8] ESI-
Fourier transform ion cyclotron resonance (FTICR)-MS/MS[9] and
ESI-MS with quadrupole time-of-flight (QTOF).[10]
When designing a DCL, the concept of isoenergetic com-
pounds has to be taken into account to ensure the formation
of equal amounts of each library member in an untemplated
DCL.[5] This is of fundamental importance to facilitate the de-
tection of amplification effects and to prevent false negative
results due to low amounts of energetically disfavored but
potent binders. Furthermore, until today, to ensure the detect-
ability of amplification effects, this concept has typically been
applied to targets like enzymes or carbohydrates.[5,11] This is
likely to be because of their good availability in pure form and
their high solubility in aqueous systems, allowing their applica-
tion at relatively high, that is, micromolar, concentra-
tions.
Finally, adaptive DCLs can usually only be applied to targets
available in reasonable amounts, like enzymes, because moni-
toring the library composition is a highly demanding task,
however, the combination of library generation by DCC and
analysis by MS binding assays should be applicable to mem-
brane-bound proteins like G protein coupled receptors
(GPCRs), ligand-gated ion channels and neurotransmitter trans-
porters. Usually, only extremely low concentrations, typically in
the sub-nanomolar range, can be reached for these kinds of
Next to adaptive libraries, there are pre-equilibrat-
ed[6] and iterative[12] forms. Both kinds of DCLs are
generated under thermodynamic conditions but are
analyzed when static, accomplished by changing the
temperature or pH. Of course, these libraries are
amenable to screening by conventional meth-
ods.[13–15]
Clearly, drug screening by DCC represents an intri-
guing concept, however, it is important to note that
the analysis of dynamic libraries for the identification
of the best binders is a challenging and demanding
task.[16] Moreover, the generation of DCLs is com-
monly based on the most fundamental and efficient
chemical reactions of organic chemistry, such as
imine and oxime formation[17] or aldol condensa-
Figure 1. Graphical representation of the designed method comprising generation of
pseudostatic libraries by DCC and analysis by competitive MS binding assays. Initially,
a pseudostatic library is generated by combining a diverse set of building blocks in
equal amounts with a large excess of one compound of complementary chemical reac-
tivity. After equilibration of the mixture, a competitive MS binding assay is performed by
adding a native marker and quantifying the amount of marker bound to the target to
unravel the presence of hits.
&2
&
www.chemmedchem.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 0000, 00, 1 – 14
ÝÝ
These are not the final page numbers!

Library Screening Using MS Binding Assays
MED
targets. To demonstrate that the new method described here
using MS binding assays as a readout can cope well under
these conditions, a membrane-bound protein was employed in
this study. In this context, it should also be added that MS
binding assays can serve not only as readout for hit detection,
but also for hit identification. The latter, however, will be the
subject of further studies.
To demonstrate the feasibility of the concept outlined
above, mGAT1, one of the four transporters known to mediate
transport of GABA across cellular membranes in the brain, was
selected as the target in this study. For this transporter, a com-
petitive MS binding assay, employing the native marker NO711
(6) has previously been established.[18,19]
GAT1 (species independent denomination of mGAT1 accord-
ing to the human genome organization HUGO),[20] representing
the most abundant GABA transporter in the brain, is primarily
located in presynaptic cell membranes of GABAergic neurons
and is thus the most important subtype for the reuptake of
GABA and for the termination of GABAergic neurotransmission.
Decreased GABAergic neurotransmission is assumed to be
a major cause of neuronal diseases like epilepsy, Parkinson’s
disease and sleeping disorders. Accordingly, enhancing GABA
concentration in the synaptic cleft by inhibition of GABA trans-
porters, such as GAT1, to support restoration of the neuro-
transmitter balance in the central nervous system (CNS) repre-
sents a rational approach to the cure of these diseases. In fact,
with the introduction of the GAT1 selective inhibitor tiagabine
(5), this transporter subtype has already been demonstrated to
be a valid drug target in the treatment of epilepsy.[21]
(R,S)-Nipecotic acid ((R,S)-2) and guvacine (3) were first de-
scribed to be potent in vitro inhibitors of the neuronal and
glial GABA transport in the 1970s.[22,23] However, due to their
highly polar nature, these compounds are not able to readily
cross the blood–brain barrier and are not orally active.[24] SK&F-
8997-A (4), tiagabine (5) and NO711 (6) are lipophilic deriva-
tives of amino acids 2 and 3, and represent the first orally
active GAT inhibitors described.[25] These compounds are char-
acterized by the presence of a lipophilic aromatic domain,
which is attached to the amino nitrogen of the cyclic amino
acid with a spacer of a definite length. As a result of this modi-
fication, compounds 4–6 possess a distinctly improved ability
to permeate the blood–brain barrier, they exhibit high subtype
selectivity in favor of GAT1,[26–29] and they also act as anticon-
vulsants in animal models.[30] Of these compounds, tiagabine
(Gabitril; 5), having undergone successful clinical development,
is in use as an add-on medication for the treatment of epilep-
sy.[21,22,31] In a study by Andersen et al. published in 2001,
a number of series of GAT1 inhibitors were reported in which
the lipophilic moiety had been modified.[32] Compounds 7–9,
which are among the most potent derivatives (IC50 =55–
111 nm), represent typical examples that nicely demonstrate
the structural variations tolerated by the target like the inser-
tion of an ether function or the different arrangement of the
aryl rings.
Figure 2. Structures of GABA and GAT-1 inhibitors. The pIC50 values quoted
were calculated from the published IC50 values taken from Reference [30]
(compounds 1–6) and Reference [32] (compounds 7–9).
drophilic and lipophilic halves of the molecules was deemed
to be the most suitable position for the functional group re-
quired for library generation. Hydrazone formation was consid-
ered a suitable reaction. However, under the conditions typical-
ly required for binding assays as such neutral pH, hydrazone li-
braries equilibrate only very slowly and thus do not fully repre-
sent dynamic libraries. That said, the assay conditions em-
ployed for the generation of dynamic and static libraries are
more or less the same. With this in mind, using hydrazone li-
braries to validate the proposed drug screening concept still
seemed justified. Moreover, the use of hydrazone chemistry
could significantly ease the identification of best binders
through their independent resynthesis.
Accordingly, libraries based on hydrazone chemistry and
containing compounds that resemble GAT1 inhibitors 7–9
were chosen as model compounds for this study. In line with
the conditions defined above, reaction of aldehyde libraries 11
with an excess of 10 in the presence of the target mGAT1 was
Inspired by inhibitors 7–9 (Figure 2),[32] we aimed to develop
combinatorial libraries of related structures that, as already
stated, could be generated by DCC. The spacer linking the hy-
ChemMedChem 0000, 00, 1 – 14
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org
&
3
&
ÞÞ
These are not the final page numbers!

M. Sindelar and K. T. Wanner
MED
sodium chloride (1m) for improved binding of the reporter
ligand[19] at varying near-neutral pH values.
To determine the implications of using a phosphate rather
than a TRIS buffer, a series of NO711 saturation experiments at
different pH values close to physiological pH were performed.
The original binding assay carried out using TRIS buffer
(pH 7.1) yielded a Kd value for NO711 of 23.4Æ2.2 nm (n=15,
meanÆSEM).[18,19] The Kd values calculated for NO711 from the
results of the saturation experiments performed with phos-
phate buffer are given in Table 1. The data shows that NO711
binding at pH 7.1 is unaffected by the buffer used (phosphate
Scheme 1. Condensation of nipecotic-acid-derived hydrazine 10 with diverse
aldehydes to give hydrazones with the general structure 12.
suggested for the generation of the desired libraries
(Scheme 1). Subsequent application of MS binding assays for
mGAT1 to the resulting protein–ligand mixtures should identify
active compounds.[18,19]
Table 1. Comparison of Bmax and Kd values in phosphate-buffered solu-
tions at varying pH levels.[a]
pH
6.80
7.10
7.40
7.70
[b]
Bmax
Kd
37.8Æ0.5
20.9Æ3.2
39.1Æ2.8
26.5Æ4.6
33.5Æ0.7
24.7Æ3.5
40.6Æ3.1
42.9Æ7.2
[c]
Results and Discussion
[a] NaH2PO4·H2O (12.5 mm), Na2HPO4·2H2O (12.5 mm), NaCl (1m);
[b] pmolmgÀ1 protein (n=3, meanÆSEM); [c] nm (n=3, meanÆSEM).
Chemistry
Synthesis of N-(2-hydrazinoethyl)nipecotic acid (10) was ac-
complished according to Scheme 2. Nipecotic acid ethyl ester
(13) was transformed into 14 using a literature method.[33] Sub-
buffer: Kd =26.5Æ4.6 nm; TRIS buffer: Kd =23.4Æ2.2 nm). Fur-
thermore, only slight changes in Kd values were observed
when the pH was lowered to 6.8 (20.9Æ3.2) or raised to 7.4
(24.7Æ3.5), however, the Kd value varied more significantly
when the pH of the binding assay was adjusted to 7.70 (42.9Æ
7.2 nm).
To study the influence of the phosphate buffer pH on the
rate of hydrazone formation, a series of experiments at differ-
ent pH levels were performed in which hydrazine 10 was react-
ed with 2-phenylbenzaldehyde (24A). Aldehyde 24A was se-
lected for two reasons: the structure is prototypic for the alde-
hydes intended for use, and the UV absorption of the resulting
hydrazone (24H) differs from that of the free aldehyde to such
an extent that the progress of the reaction can be directly
monitored by following the increase in UV absorption of the
reaction mixture without generating false results caused due
to the decrease of aldehyde UV absorption at the chosen
wavelength. The rate of hydrazone formation gradually in-
creased as the pH of the phosphate buffer was lowered, with
the fastest rate observed at pH 6.8, the lowest pH applied;
however, the rate of hydrazone formation was still reasonably
fast at pH 7.1 (Figure 3a). Therefore, this pH level was selected
for the generation and screening of libraries as it represents
a good compromise between the rate of hydrazone formation
and the stability of the binding assay, the Kd value of NO711
varying only slightly when lowering (pH 6.8) or increasing
(pH 7.4) the pH of the assay conditions.
Scheme 2. Reagents and conditions: a) 1-bromo-2-chloroethane, acetone,
K2CO3, RT, 20 h;[33] b) tert-butyl carbazate, EtOH, reflux, 1 h; c) TFA, CH2Cl2, RT,
30 min; d) 4m HCl, 808C, 1 h; e) ion exchange chromatography.
sequent reaction with N-mono-Boc-protected hydrazine gave
15. Removal of the N-Boc group and hydrolysis of the ester
functionality under acidic conditions provided hydrazine 10.
Assay conditions
As already outlined above, for library screening, a competitive
MS binding assay for mGAT1 based on NO711 (6) as a reporter
ligand should be employed for hit detection.[18,19] To this end,
a suitable buffer system had to be found that does not inter-
fere with library formation, allows rapid library generation, and
is compatible with the binding assay. For the original MS bind-
ing assay, a 2-amino-2-hydroxymethyl-propane-1,3-diol (TRIS;
50 mm) buffer containing sodium chloride (1m) was used. As
the buffer capacity of TRIS in the neutral pH range is rather
low, and TRIS carries a primary amino functionality that could
interfere with the hydrazone formation, this buffer was
thought to be less suitable for the new assay.[18,19] Being nearly
structurally inert and characterized by high buffer capacity, we
decided to study a phosphate buffer (25 mm) containing
For the current study, only small libraries containing four dif-
ferent hydrazones were used, with the concentration of the in-
dividual hydrazones amounting to 10 mm. This concentration
was chosen as the minimum affinity—the highest IC50 value—
a test compound should have to give a positive signal in the
MS binding assay when no other active constituent is present
in the library. In that case, the specific binding of the MS
&4
&
www.chemmedchem.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 0000, 00, 1 – 14
ÝÝ
These are not the final page numbers!

Library Screening Using MS Binding Assays
MED
&
Figure 3. a) Absorbance (A) versus time (t) for the reaction of 24A (10 mm) with hydrazone 10 (100 mm) in phosphate buffer at different pH values: 6.8, ; 7.1,
; 7.4, ; 7.7, ; (n=12). Reactions were monitored by UV: e284nm 10: 180 LmolÀ1 cmÀ1, e284nm 24A: 1700 LmolÀ1 cmÀ1, e284nm 24H: 9720 LmolÀ1 cmÀ1. Absorb-
~
!
^
ance (A) versus time (t) for the reaction of b) 24A, c) 29A, d) 33A, e) 37A and f) 38A (40 mm) with hydrazine 10 (100 mm) in phosphate buffer at pH 7.1
(n=3).
marker NO711 (6) in the MS binding assay serving as a readout
would be decreased to 50% or less. The concentration of nipe-
cotic-acid-derived hydrazine 10 was set to 100 mm, which cor-
responds to a clear excess of the hydrazine over the aldehydes
(ratio=2.5:1), ensuring that in a potential state of equilibrium
the products are favored. To apply hydrazine 10 in a still
higher concentration (i.e., 1 mm) turned out to be less suitable,
as in this case the hydrazine itself decreased marker binding to
62Æ14% (meanÆSD of n=10 experiments with four repli-
cates).
To get a rough estimate of the time necessary to accomplish
an almost complete hydrazone formation under the conditions
defined above for the generation of the library, five electroni-
cally diverse aromatic aldehydes (24A, 29A, 33A, 37A and
38A) were reacted with hydrazine 10 (100 mm) in the absence
of the target. To resemble the conditions of the library forma-
ChemMedChem 0000, 00, 1 – 14
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org
&
5
&
ÞÞ
These are not the final page numbers!

M. Sindelar and K. T. Wanner
MED
tion, the concentrations of individual aldehydes were set to
the total concentration of aldehydes in the assay (40 mm; four
aldehydes each at 10 mm). Like 24A, aldehydes 29A, 33A, 37A
and 38A were selected according to their UV characteristics so
as to allow direct monitoring of the reaction by UV. In each
case, almost complete equilibrium was reached within four
hours (Figure 3b–f). Therefore, an incubation period of four
hours for the generation of the libraries in the presence of the
target, that is, in the final experiments, seemed sufficient.
drazine 10. For a purely additive effect, the decrease in NO711
binding would have to be less than the sum of the decreases
observed for hydrazine 10 and the respective aldehyde library;
however, this is never the case (Figure 5b).
To verify the validity of the result obtained for library 1, this
library was also tested after it had been generated in an inde-
pendent pre-equilibration experiment. For the formation of hy-
drazone library 1, hydrazine 10 was mixed with aldehyde li-
brary 1 at a 100-fold higher concentration for both compo-
nents (1 mm of each aldehyde, 10 mm of hydrazine) in the ab-
1
sence of the target. According to H NMR experiments, hydra-
In situ library generation and screening
zone formation was complete upon mixing (see Supporting
Information). Subsequent dilution with incubation buffer by
a factor of 100 gave the original concentration. This pre-equili-
brated library was then evaluated in the MS binding assay, and
the NO711 specific binding was found to be 31Æ3%. This
result is in good agreement with that obtained for the four-
hour experiments in which library generation takes place in
the presence of the target (26Æ6%).
With regard to the general considerations described above
and the conclusions drawn for the assay conditions, the new
method for the generation and screening of compound libra-
ries was accomplished as follows.
1. Generation of pseudostatic libraries by DCC. For the genera-
tion of libraries, a set of four different aldehydes (each
10 mm) was incubated with hydrazine 10 (100 mm) and
mGAT1 protein (10–20 mg per sample, total volume 250 mL)
in phosphate buffer for 4 h at 378C in a 96-well microtiter
plate.
The same set of experiments with a four-hour incubation
period for library generation and with a pre-equilibrated library
was also carried out for aldehyde library 2. The results for
these two experiments were very similar too; NO711 specific
binding was found to be 29Æ2% after 4 h incubation (library
generation in presence of target) and 25Æ2% when the pre-
equilibrated library was used. As such, it was concluded that
a four-hour incubation period in the presence of the target is
sufficient to reach equilibrium and thus to obtain reliable re-
sults. Accordingly, the remaining hydrazone libraries were only
generated and screened following the original procedure (4 h
incubation in the presence of the target and screening by sub-
sequent MS binding assay). Hydrazone libraries 3 and 5 turned
out to be the most potent, decreasing marker binding to 7.5Æ
3.5% and 13Æ1%, respectively. Though less potent, libraries 4,
8, and 9 also decreased marker binding to less than 50%, sug-
gesting the presence of hydrazones with IC50 values below
10 mm. With NO711 specific binding around 50%, libraries 6
and 7 appear to be the least potent.
2. Hit detection by competitive MS binding assay. For hit detec-
tion subsequent to library generation, NO711 was added to
each sample (final concentration 20 nm) and allowed to
equilibrate for 40 min at 378C. Experiments were terminat-
ed by filtration. Bound MS marker (NO711) was liberated by
denaturing the isolated protein–ligand complexes (by
drying at 508C for 1 h and elution with 300 mL MeOH) and
quantified by LC-ESI-MS/MS. The amount of MS marker
found was used to assess the potency of the library.
In the current study, 36 different aldehydes grouped in nine
libraries each consisting of four members (Figure 4) were ap-
plied to test the feasibility of the concept outlined above. In
each case, the generation and screening of the hydrazone li-
braries (generated by mixing the respective aldehyde library
with hydrazine 10, Figure 5a) was supplemented by a control
experiment in which only the aldehyde library was applied to
reveal the potency exerted by the aldehydes themselves. Simi-
larly, hydrazine 10 was tested at the assay concentration
(100 mm) for its inhibitory potency in the absence of aldehydes.
Figure 5b shows the results of the control and library screen-
ing experiments. In control experiments, neither hydrazine 10
nor any of the aldehydes decreased marker binding to mGAT1
to a significant extent. For example, aldehyde library 1 (16A,
17A, 18A, and 19A) had no considerable effect on NO711
binding, decreasing it slightly to 80Æ3%. For hydrazine 10,
the effect was even less pronounced (92Æ3%). However, hy-
drazone library 1 (generated from aldehyde library 1 and hy-
drazine 10) led to a significant decrease in specific marker
binding to just 26Æ6%, thus indicating that at least one
potent hydrazone inhibitor is present in the mixture. With the
hydrazine alone being almost inactive, the observed decrease
must be due to hydrazone formation and cannot be the result
of an additive effect of inhibition by aldehyde library and hy-
Deconvolution experiments
Of the nine hydrazone libraries, seven gave rise to a decrease
in marker binding of at least 50%, indicating a high number of
hydrazones with IC50 values below 10 mm. For this reason, crite-
ria to detect an active library were tightened. Only compounds
with an IC20 value of less than 10 mm, that is the concentration
that gives rise to only 20% of the maximal activity, which cor-
responds to an IC50 value of 2.5 mm (IC20 =4ꢁIC50), should be
detected with this method. Consequently, the upper limit for
defining a library as active was lowered to 20% remaining
marker binding. This value was only achieved by the two most
potent libraries, libraries 3 and 5.
For the identification of the most active compounds in libra-
ries 3 and 5, the same protocol was followed as for library
screening, except that single compounds were evaluated. Hy-
drazone formation was allowed to take place for four hours as
for library screening (aldehyde: 10 mm; hydrazine 10: 100 mm).
&6
&
www.chemmedchem.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 0000, 00, 1 – 14
ÝÝ
These are not the final page numbers!

Library Screening Using MS Binding Assays
MED
Figure 4. Aldehyde libraries 1–9 employed in drug screening.
The data for the control samples, consisting either only of hy-
drazine derivative 10 or aldehyde, were generated in the same
way. According to the data obtained in the deconvolution ex-
periments, the most potent inhibitors were based on 2-phenyl-
benzaldehyde (24A) (library 3; Figure 6b) and 2-thiophen-2-yl-
benzaldehyde (35A) (library 5; Figure 6d). The corresponding
hydrazones decreased the specific binding of NO711 to ꢀ5%
(24H) and to 8.2Æ0.4% (35H). In addition, library 4 was evalu-
ated as an example of a medium potency library (30Æ3%). Al-
dehyde 31A was found to give rise to the best binding hydra-
zone derivative in this library. However, the inhibitory potency
of 31H was limited, decreasing specific NO711 binding to only
~37Æ1%. To further verify the validity of the new library gen-
eration and screening assay, the individual components of li-
braries 1 and 8 were also analyzed to determine the best bind-
ers. In line with the results obtained for libraries 1 and 8, none
of the library members exhibited significant inhibitory potency
(Figure 6a and e).
Confirmation of deconvolution results
To verify the results of the deconvolution experiments, hit
compounds 24H and 35H, both decreasing marker binding to
less than 20%, and medium potency hydrazones 26H, 31H
ChemMedChem 0000, 00, 1 – 14
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org
&
7
&
ÞÞ
These are not the final page numbers!

M. Sindelar and K. T. Wanner
MED
Figure 5. a) Library formation exemplified by the reaction of aldehyde library 1 with hydrazine 10. b) Specific binding of NO711 in the presence of hydrazine
10, pure aldehyde libraries 1–9 (for library compounds, see Figure 4), and the corresponding hydrazone libraries after an incubation time of 4 h for library
generation and 40 min for marker binding. Data for pre-equilibrated hydrazone libraries 1 and 2 is also shown. Data represents the mean ÆSD of four repli-
cates.
and 45H, each decreasing marker binding to slightly greater
than 20%, were synthesized and characterized in competitive
MS binding experiments to determine the binding affinities
(pKi) for mGAT1. Interestingly, like 24H and 35H, hydrazones
26H, 31H and 45H are derived from ortho-substituted benzal-
dehydes too. Hydrazone 51H was also evaluated as it contains
an ortho-biaryl moiety that, considering the results obtained
for compounds 24H and 35H, could be favorable for binding.
For this compound, a member of the weakly active library 9,
a binding experiment identical to the deconvolution experi-
ment was performed (10 mm 51A and 100 mm hydrazine 10).
However, with a remaining NO711 binding of 65Æ8%, hydra-
zone 51H exhibited only low affinity (data not shown).
Hydrolysis of the hydrazone products is possible in the ab-
sence of excess hydrazine 10, and this could lead to false re-
sults from the binding assays. To exclude this possibility, the
&8
&
www.chemmedchem.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 0000, 00, 1 – 14
ÝÝ
These are not the final page numbers!

Library Screening Using MS Binding Assays
MED
Figure 6. Deconvolution experiments of hydrazone libraries a) 1, b) 3, c) 4, d) 5, and e) 8 including control experiments (aldehyde libraries A, hydrazone libra-
ries H, and hydrazine 10).
stability of three representative but structurally diverse com-
pounds (35H, 36H and 38H) was monitored at pH 7.1 by
1H NMR spectroscopy (see Supporting Information). According
to the results of these experiments, the tested hydrazones are
sufficiently stable to be characterized in conventional competi-
tive MS binding assays, which take two hours to perform and
provide full competition curves and corresponding binding af-
finities. Only hydrazone 36H decomposed slightly within the
two-hour evaluation period. However, with the release of the
corresponding aldehyde being only 5% of the total com-
ChemMedChem 0000, 00, 1 – 14
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org
&
9
&
ÞÞ
These are not the final page numbers!

M. Sindelar and K. T. Wanner
MED
pound, evaluation of the hydrazone should still deliver a relia-
ble Ki value. For the sake of completeness, it should be added
that the equilibrium was reached after eight hours, the ratio of
hydrazone 36H to aldehyde 36A amounting to 84:16. For hy-
drazones 35H and 38H, no decomposition products could be
seen by 1H NMR after two hours. However, the dynamic charac-
ter of hydrazones 35H and 38H became evident, too, upon
extended monitoring of the equilibration. After seven hours
(35H) and one week (38H), equilibrium had been reached
with the hydrazones still clearly predominating, and the ratios
of hydrazone to aldehyde amounting to 96:4 (35H)and 98:2
(38H).
Table 2. Comparison of NO711 binding data determined in deconvolu-
tion experiments (pKi) and functional GAT1 activity (pIC50).
[c]
Compd
R
SB[a] [%]
pKi[b]
pIC50
24H
ꢀ5
6.186Æ0.028
5.308Æ0.096
For the determination of the binding affinities of the above-
mentioned hydrazone derivatives, the MS binding assay for
mGAT1 as described by Zepperitz et al.[18,19] was used, except
that the original TRIS buffer was replaced by the phosphate
buffer employed for screening. Among the compounds tested,
hit compounds 24H and 35H identified in the deconvolution
experiments of libraries 3 and 5 again exhibited the highest
potencies, with pKi values of 6.186Æ0.028 and 6.229Æ0.039
(meanÆSEM, n=3), respectively (Figure 7). For hydrazones
26H, 31H, 45H and 51H, which exhibited specific NO711
35H
26H
31H
45H
8.2Æ0.4
6.229Æ0.039
5.542Æ0.042
5.577Æ0.037
5.445Æ0.075
5.542Æ0.107
5.186Æ0.084
4.895Æ0.152
4.879[d]
27Æ1
37Æ1
24Æ1
51H
65Æ8[e]
4.479Æ0.064
4.022[d]
[a] Specific binding (SB) of NO711 determined in deconvolution experi-
ments under standard assay conditions: aldehyde (10 mm), 10 (100 mm),
4 h equilibration period. Specific binding <5% indicates NO711 binding
was below the lower limit of quantification (50 pm). [b] pKi values were
determined by competitive MS binding assay with NO711 performed in
an mGAT1-expressing HEK293 cell membrane preparation. Data represent
the meanÆSEM of three independent experiments. [c] pIC50 values were
determined by [3H]-GABA uptake assay performed in mGAT1-expressing
HEK293 cells. Data represent the meanÆSEM of three independent ex-
periments. [d] Value is the result of a single experiment. [e] Deconvolution
graph not shown in Figure 6.
~
Figure 7. Representative competition curves for hydrazones 24H ( ), 35H
!
&
*
^
( ), 26H ( ), 31H ( ), 45H ( ) and 51H (ꢁ) obtained from competitive MS
binding assays. Data points represent specific binding of NO711 (meanÆSD
from triplicate values) in the presence of different concentrations of test
compounds. Binding curves were generated by nonlinear regression.
binding values of 27Æ1%, 37Æ1%, 24Æ1%, and 65Æ8%, re-
spectively, in the deconvolution experiments, pKi values of
5.542Æ0.042, 5.577Æ0.037, 5.445Æ0.075, and 4.479Æ0.064
were found (meanÆSEM, nꢁ3). These results not only confirm
the high potency of the best hits 24H and 35H identified in
the deconvolution experiments, but also indicate that the per-
centage values obtained in the deconvolution experiments
provide a reasonable estimate of the pKi values of the respec-
tive compounds, with the rank orders in both series being
comparable.
by employing a [3H]-GABA uptake assay for mGAT1 as de-
scribed in the literature,[34] are summarized in Table 2. As can
be seen from the data, the pIC50 values for [3H]-GABA uptake
by mGAT1 in first approximation run parallel to the pKi values
for binding affinity for this transporter. That the pKi values are
slightly higher than the pIC50 values is a common phenomen-
on for these assays. Furthermore, the potencies of the best hy-
drazone inhibitors 24H and 35H at mGAT1 are close to those
of the model compounds 7–9, which served as prototypes for
the design of the hydrazone libraries. According to these re-
sults, hydrazone libraries are well suited to exploring the struc-
tural motifs required for efficient binding to a target, and hits
In addition, hydrazone derivatives 24H, 35H, 26H, 31H,
45H and 51H were characterized with respect to their func-
tional GAT1 activity. The results of these studies, accomplished
&10
&
www.chemmedchem.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 0000, 00, 1 – 14
ÝÝ
These are not the final page numbers!

Library Screening Using MS Binding Assays
MED
1
formed using Merck silica gel 60 (mesh: 0.040–0.063 mm). H NMR
spectra were recorded at room temperature (or 258C for samples
solved in [D6]DMSO) on a JNMR-GX (JEOL) at 400 or 500 MHz.
1H NMR chemical shifts were internally referenced to tetramethylsi-
lane (TMS) or 1,4-dioxane for samples dissolved in D2O. The spectra
were processed using MestReNova (version 5.1.1–3092 2007; Mes-
trelab Research, Santiago de Compostela, Spain). Broadened sig-
nals are designated by br (sbr, dbr, tbr). Infrared (IR) spectroscopy
was performed with an FTIR Spectrometer 410 (Jasco). Samples
were measured either as KBr pellets or as films on NaCl plates. A
identified as best binders can be considered suitable templates
for the development of stable analogues.
Conclusions
In this study, competitive MS binding assays that have so far
only been used for the characterization of the binding affinity
of single compounds have been demonstrated to be applica-
ble also as a readout in compound screening campaigns. By
employing libraries generated by using dynamic combinatorial
chemistry (DCC), a new kind of library generation and screen-
ing assay has been developed. In this assay, the dynamic libra-
ries are generated in a way that renders them pseudostatic,
with the library composition thus being well defined. Further-
more, the reaction is performed in a medium suitable for the
binding assay and with the target already present. This allows
the resulting mixtures to be directly used for competitive MS
binding assays to determine the activity of the generated libra-
ries. MS binding assays are highly sensitive, and as such, this
method could be applied to targets like membrane-bound
proteins for which typically only low concentrations can be
reached.
Hewlett Packard 5989 A with 59.980 B particle beam LC–MS inter-
face was used for mass spectrometry (ionization: chemical (CH5
or electron impact (70 eV)). High-resolution mass spectrometry
(HRMS) was carried out using an LTQ FT (ThermoFinnigan) or
a JMS GCmate II (Jeol).
+
)
1-(2-Chloroethyl)piperidine-3-carboxylic acid ethyl ester (14)[33] and
2-benzylbenzaldehyde (26A)[35] were prepared as described in the
literature.
1-[2-(2-tert-Butoxycarbonylhydrazino)ethyl]piperidine-3-carbox-
ylic acid ethyl ester (15): A stirred solution of 14 (158 mg,
0.717 mmol, 1.0 equiv) in abs EtOH (3.0 mL) was treated with Et3N
(73.0 mg, 0.721 mmol, 100.0 mL, 1.0 equiv) and tert-butyl carbazate
(304 mg, 2.30 mmol, 3.2 equiv). The mixture was heated at reflux
for at least 1 h (monitored by TLC). After cooling to RT, the solvent
was completely evaporated in vacuo. The crude was purified by CC
(step gradient; n-pentane/EtOAc/Et3N, mixture 1) 1:1:0, mixture
2) 5:5:1) providing 15 as colorless oil (131 mg, 58%): Rf =0.60 (n-
The feasibility of this method was demonstrated for the
membrane-associated protein GAT1, the most abundant g-ami-
nobutyric acid (GABA) transporter in the central nervous
system (CNS), using pseudostatic hydrazone libraries. By com-
bining equal amounts of four different aldehydes and one hy-
drazine derivative (10) in excess in the presence of the target,
pseudostatic hydrazone libraries comprising equal amounts of
the desired hydrazones were generated and subsequently
screened by competitive MS binding assays employing NO711
(6) as a native mGAT1 marker. The binding potencies of the li-
braries were determined from the amount of bound marker
found by LC-ESI-MS/MS as part of the competitive MS binding
assay. In this way, two hit libraries were identified that decrease
marker binding to less than 20%. Subsequent deconvolution
experiments led to the identification of hydrazones 24H and
35H as compounds with the highest target affinity. After indi-
vidual synthesis, full competitive MS binding assays confirmed
the identified hits as potent binders of mGAT1, with pKi values
of 6.186Æ0.028 and 6.229Æ0.039, respectively.
1
pentane/EtOAc/Et3N, 5:5:1); H NMR (500 MHz, CD2Cl2): d=1.23 (t,
J=7.1 Hz, 3H), 1.43 (s, 9H), 1.37–1.50 (m, 1H), 1.50–1.61 (m, 1H),
1.65–1.75 (m, 1H), 1.83–1.93 (m, 1H), 1.95–2.05 (m, 1H), 2.17 (tbr,
J=9.8 Hz, 1H), 2.41–2.50 (m, 2H), 2.53 (tt, J=10.3, 3.8 Hz, 1H),
2.70–2.80 (m, 1H), 2.90 (td, J=6.0, 2.6 Hz, 2H), 2.92–3.00 (m, 1H),
4.10 (q, J=7.1 Hz, 2H), 4.04–4.26 (sbr, 1H), 6.30–6.64 ppm (sbr, 1H);
13C NMR (101 MHz, CD2Cl2): d=14.4, 25.1, 27.4, 28.5, 42.3, 48.6,
~
54.0, 56.0, 57.6, 60.6, 80.1, 157.0, 174.4 ppm; IR (film): n=3319,
2977, 2940, 2811, 1730, 1720, 1454, 1367, 1281, 1252, 1155,
1031 cmÀ1; MS (EI, 70 eV): m/z (%): 315 (65) [M]+, 270 (13), 242
(100), 215 (67), 214 (41), 138 (33); MS (CI, CH5+): m/z (%): 316 (100)
[M+H]+, 260 (77), 216 (26), 170 (77); HRMS-EI (70 eV): m/z [M]+
calcd for C15H29N3O4: 315.2158, found: 315.2162.
1-(2-Hydrazinoethyl)piperidine-3-carboxylic acid (10): A stirred
solution of 15 (353 mg, 1.12 mmol) in CH2Cl2 (2.0 mL) was treated
with trifluoroacetic acid (12.6 g, 110 mmol, 8.50 mL, 100 equiv) at
RT. After 1 h, the solvent was evaporated in vacuo. The resulting
residue was dissolved in 5m aq HCl (10 mmol, 2 mL, 9 equiv), and
the solution was heated at 808C for 2 h. After evaporation to dry-
ness, the resulting HCl salt[36] was purified by ion exchange chro-
matography (Amberlite IRA-120; solvent: 20% aq NH3 solution)
providing 10 as a colorless, amorphous solid (195.4 mg, 93%):
1H NMR (500 MHz, D2O/NaOD): d=1.29 (qd, J=12.9, 4.2 Hz, 1H),
1.50 (qt, J=13.1, 3.8 Hz, 1H), 1.68–1.78 (m, 1H), 1.87–1.95 (m, 1H),
1.99 (dt, J=12.2, 2.8 Hz, 1H), 2.04 (t, J=11.4 Hz, 1H), 2.34 (tt, J=
11.8, 3.6 Hz, 1H), 2.46–2.54 (m, 2H), 2.82–2.93 (m, 3H), 3.01 ppm
(dbr, J=11.3 Hz, 1H); 13C NMR (126 MHz, D2O/NaOD): d=24.6, 28.2,
Having been successfully applied to screening of compound
libraries generated by DCC—a rather challenging task, compet-
itive MS binding assays can be considered as powerful tools
for the identification of active compounds, suitable for most
types of libraries and targets, and therefore, they could find
widespread application in drug discovery.
Experimental Section
Chemistry
~
45.4, 50.6, 53.8, 55.7, 56.7, 183.9 ppm; IR (KBr): n=3700–2300,
2946, 1581, 1451, 1396 cmÀ1; HRMS-ESI (ESI+): m/z [M+CH3CN]+
Solvents used were of analytical grade and freshly distilled before
use except for DMSO. Ethyl nipecotate was purchased from
Sigma–Aldrich and freshly distilled before use. Other purchased re-
agents and reactants were used without further purification. Thin-
layer chromatography (TLC) was carried out on precoated silica gel
F254 glass plates (Merck). Flash chromatography (CC) was per-
calcd for C10H20N4O2: 228.1568, found 228.1705.
Preparation of hydrazones: Experimental protocols and character-
ization data for the hydrazones described here are given in the
Supporting Information.
ChemMedChem 0000, 00, 1 – 14
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org
&
11
&
ÞÞ
These are not the final page numbers!

M. Sindelar and K. T. Wanner
MED
tively. Calibration curves for marker quantitation were generated
using these standards.
MS Binding experiments
mGAT1 Membrane preparation: Membrane preparations of HEK293
cells stably expressing mGAT1[34] were prepared as described previ-
ously and stored at À808C.[19] On the day of the assay, an aliquot
was rapidly thawed and diluted in a 20-fold volume of cold aq
0.9% NaCl (m/v). After centrifugation at 15000 rpm and 48C for
20 min (CP56GII, P70AT, Hitachi Ltd., Tokyo, Japan), the pellet was
resuspended in ice-cold assay buffer (see below) to a protein con-
centration of approximately 0.1 mgmLÀ1 (determined as described
previously).[19]
Deconvolution experiments: The deconvolution experiments were
carried out in the same way as described for library screening,
except that a solution of single aldehyde in 10% DMSO/phosphate
buffer was used (pH 7.1, v/v, 100 mm, 25 mL).
Pre-equilibrated libraries: A solution of four individual aldehydes
(1 mm each) in 10% DMSO/phosphate buffer (pH 7.1, v/v, 10 mL)
and a solution of 10 (10 mm) in phosphate buffer (pH 7.1, 10 mL)
were added to phosphate buffer (pH 7.1, 80 mL). The mixtures were
incubated for 1 h at 378C (shaking water bath) to guarantee full
conversion to the corresponding hydrazones. After addition of
phosphate buffer (pH 7.1, 900 mL), aliquots of the resulting solu-
tions (25 mL) were supplemented with phosphate buffer (pH 7.1,
150 mL), mGAT1 membrane preparation (50 mL), and 200 nm NO711
(25 mL). Subsequently, the samples were processed as described for
library screening.
MS Binding assay: Saturation experiments (at various pH levels)
and competition experiments for isolated hydrazones were per-
formed as described recently,[18,19] but substituting the original in-
cubation buffer for phosphate buffer (12.5 mm Na2HPO4·2H2O,
12.5 mm NaH2PO4·H2O, 1m NaCl adjusted to the respective pH
value with 2m NaOH). Unless otherwise indicated, phosphate
buffer at pH 7.1 was used in binding assays.
Library screening: Library screening experiments were performed
with quadruplicate samples in a total volume of 250 mL in 1.2 mL
polystyrene deep-well plates (Sarstedt, Nꢂmbrecht, Germany). A
solution of four individual aldehydes (100 mm each) in 10% DMSO/
phosphate buffer (v/v, pH 7.1; 25 mL) and a solution of 10 (1 mm)
in phosphate buffer (pH 7.1, 25 mL) were added to additional phos-
phate buffer (pH 7.1, 125 mL). The first incubation period (library
equilibration) was initiated by addition of the mGAT1 membrane
preparation (50 mL) immediately after combining hydrazine and al-
dehydes (delay <1 min). After 4 h at 378C (shaking water bath),
NO711 (25 mL, 200 nm) was added to start the second incubation
period (binding experiment). The binding experiment was stopped
after 40 min at 378C (shaking water bath) by transferring an ali-
quot (200 mL) of each sample onto a 96-well filter plate (Acroprep,
glass fiber, 1.0 mm, 350 mL; Pall Corp, Dreieich, Germany) with a 12-
channel pipette and subsequent vacuum filtration. The filter plate
was washed ice-cold 1m NaCl solution (5ꢁ150 mL per well) using
a 12-channel pipette, dried (60 min, 508C) and cooled to room
temperature. Afterwards, liberation of the marker was achieved by
elution of the filter plate with MeOH (3ꢁ100 mL per well) into
a 1.2 mL polypropylene deep-well plate (Sarstedt, Nꢂmbrecht, Ger-
many). Finally, each sample was supplemented with 1 nm
[2H10]NO711 in MeOH (200 mL) as an internal standard and then
dried for 12–16 h (508C). For quantification by LC-ESI-MS/MS, sam-
ples were reconstituted in 10 mm NH4HCO2 buffer (pH 7.0)/MeOH
(95:5, v/v, 200 mL).
Analysis
LC-ESI-MS/MS: Quantitation by LC-ESI-MS/MS was performed as
described previously using a API 3200 triple-quadrupole mass
spectrometer and by drying and reconstituting the methanolic elu-
ates.[18,19]
Analysis of binding experiments: Marker depletion was negligible
(<10%) in all binding experiments. Equilibrium dissociation con-
stant (Kd) and density of binding sites (Bmax) were calculated from
one-site saturation isotherms of specific binding by means of the
nonlinear curve-fitting program Prism 4.02 (GraphPad Software,
San Diego, CA, USA). Specific binding was defined as the difference
between total and nonspecific binding. Nonspecific binding below
50 pm could not be determined experimentally, but was extrapo-
lated by linear regression for nonspecific NO711 binding concen-
trations ꢁ50 pm. The concentration of a competitor that inhibits
50% of specific binding (IC50) was calculated from competition
curves plotting NO711 specific binding concentrations against the
log of the competitor concentration (eight different concentrations
per competitor) with Prism 4.02 using the equation for one-site
competition and nonlinear curve fitting. Specific binding deter-
mined for control samples in the absence of any competitor was
set to 100%, whereas the bottom level was set to 0%. Ki values
were calculated according to Cheng and Prusoff[37] and are ex-
pressed as pKi values. Unless stated otherwise, all results are ex-
pressed as the meanÆSEM. pKi values were determined in at least
three separate experiments.
Each library screening experiment included samples characterizing
the pure aldehyde libraries (phosphate buffer (pH 7.1, 150 mL), a so-
lution of four individual aldehydes (25 mL; 100 mm each), and 10%
DMSO/phosphate buffer (pH 7.1, v/v), mGAT1 membrane prepara-
tion (50 mL) and 200 nm NO711 (25 mL)) and pure hydrazine 10
(phosphate buffer (pH 7.1, 125 mL), 10% DMSO/phosphate buffer
(pH 7.1, v/v, 25 mL), hydrazine 10 (1 mm) in phosphate buffer
(pH 7.1, 25 mL), mGAT1 membrane preparation (50 mL) and 200 nm
NO711 (25 mL)). Total binding and nonspecific binding of NO711
was determined in analogously constituted samples lacking any in-
hibitor or in the presence of 100 mm GABA, respectively. Addition-
ally, matrix blanks, zero samples and matrix standards were ob-
tained in the same way, performing the binding experiment with-
out NO711 and inhibitors. After filtration and elution, samples were
supplemented with MeOH (200 mL), 1 nm [2H10]NO711 in MeOH
(200 mL) or methanolic calibration standards containing 1 nm
[2H10]NO711 and 50 pm, 100 pm, 500 pm or 1 nm NO711, respec-
UV-monitoring of hydrazone formation: UV monitoring was car-
ried out in a 96-well quartz glass UV plate (Hellma) with a Spetra-
Max M2e (Molecular Devices) plate reader and analyzed with Soft-
Max Pro 5.4 software. The incubation buffer described for the MS
binding assay was used at a pH 7.1 except for comparing various
pH values (pH 6.8, 7.1, 7.4, 7.7). The total volume of the samples
was 300 mL. Each experiment was supplemented with blank sam-
ples that were composed of 270 mL phosphate buffer and 30 mL of
10% DMSO (v/v) in phosphate buffer (pH 7.1).
Comparison of different pH values: The experiment was per-
formed with twelve identically constituted replicates. Phosphate
buffer (240 mL) was supplemented with a solution of 24A (30 mL,
100 mm) in 10% DMSO/phosphate buffer (v/v). To start the reaction,
hydrazine 10 (1 mm, 30 mL) was added. Absorption values at
284 nm were determined immediately after placing the plate into
&12
&
www.chemmedchem.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 0000, 00, 1 – 14
ÝÝ
These are not the final page numbers!

Library Screening Using MS Binding Assays
MED
[17] V. A. Polyakov, M. I. Nelen, N. Nazarpack-Kandlousy, A. D. Ryabov, A. V.
Eliseev, J. Phys. Org. Chem. 1999, 12, 357–363.
the instrument. Further absorption values were recorded at 12 min
intervals for 8 h.
[18] C. Zepperitz, G. Hçfner, K. T. Wanner, ChemMedChem 2006, 1, 208–217.
[19] C. Zepperitz, G. Hçfner, K. T. Wanner, Anal. Bioanal. Chem. 2008, 391,
309–316.
[20] The murine GABA transporter subtypes mGAT1–4 correspond to GAT1,
GAT2, GAT3 and BGT-1, respectively, according to species-independent
nomenclature suggested by HUGO. N. O. Dalby, Eur. J. Pharmacol. 2003,
479, 127–137.
[21] B. S. Meldrum, A. G. Chapman, Epilepsia 1999, 40, S2–S6.
[22] P. Krogsgaard-Larsen, G. A. R. Johnston, J. Neurochem. 1975, 25, 797–
802.
[23] P. Krogsgaard-Larsen, E. Falch, O. M. Larsson, A. Schousboe, Epilepsy Res.
1987, 1, 77–93.
Monitoring of reaction progress: The experiment was performed
with identically constituted triplicates. Phosphate buffer (240 mL)
was supplemented with the desired aldehyde (400 mm, 30 mL) and
10% DMSO/phosphate buffer (v/v). To start the reaction, hydrazine
10 (1 mm, 30 mL) was added. Absorption values at a wavelength of
284 nm for conversion of 24A, 29A, 33A and 37A and 356 nm for
conversion of 38A were determined immediately after placing the
plate in the instrument and then at intervals of 10 min or 20 min
for 11 h.
GABA uptake assays: [3H]GABA uptake assays were performed as
previously described.[34]
[24] H.-H. Frey, C. Popp, W. Loescher, Neuropharmacology 1979, 18, 581–
590.
[25] N. O. Dalby, Neuropharmacology 2000, 39, 2399–2407.
[26] F. E. Ali, W. E. Bondinell, P. A. Dandridge, J. S. Frazee, E. Garvey, G. R.
Girare, C. Kaiser, T. W. Ku, J. L. Lafferty, G. I. Moonsammy, H.-J. Oh, J. A.
Rush, P. E. Setler, O. D. Stringer, J. W. Venslavsky, B. W. Volpe, L. M.
Yunger, C. L. Zirkle, Med. Chem. 1985, 28, 653–660.
[27] C. Braestrup, E. B. Nielsen, U. Sonnewald, L. J. S. Knutsen, K. E. Andersen,
J. A. Jansen, K. Frederiksen, P. H. Andersen, P. A. Mortensen, P. D. Suzdak,
J. Neurochem. 1990, 54, 639–647.
Keywords: binding assays · dynamic combinatorial chemistry ·
hydrazones · mass spectrometry · membrane proteins
[1] A. Annis, C.-C. Chuang, N. Nazef in Mass Spectrometry in Medicinal
Chemistry, (Eds.: K. T. Wanner, G. Hçfner), Wiley-VCH, Weinheim, 2007,
pp. 121–156.
[28] P. D. Suzdak, K. Frederiksen, K. E. Andersen, P. O. Sorensen, L. J. S. Knut-
sen, E. B. Nielsen, Eur. J. Pharmacol. 1992, 224, 189–198.
[29] S. Hog, J. R. Greenwood, K. B. Madsen, O. M. Larsson, B. Frølund, A.
Schousboe, P. Krogsgaard-Larsen, R. P. Clausen, Curr. Topics Med. Chem.
2006, 6, 1861–1882.
[30] L. A. Borden, T. G. M. Dhar, K. E. Smith, R. L. Weinshank, T. A. Branchek, C.
Gluchowski, Eur. J. Pharmacol. Mol. Pharmacol. Sect. 1994, 269, 219–
224.
[2] C. E. Whitehurst, D. A. Annis, Comb. Chem. High Throughput Screening
2008, 11, 427–438.
[3] G. Hçfner, C. Zepperitz, K. T. Wanner in Mass Spectrometry in Medicinal
Chemistry, (Eds.: K. T. Wanner, G. Hçfner), Wiley–VCH, Weinheim, 2007,
pp. 247–283.
[4] G. Hçfner, K. T. Wanner, Angew. Chem. 2003, 115, 5393–5395; Angew.
Chem. Int. Ed. 2003, 42, 5235–5237.
[5] M. Hochguertel, J.-M. Lehn in Fragment-Based Approaches in Drug Dis-
covery, (Eds.: W. Jahnke, D. A. Erlanson), Wiley-VCH, Weinheim, 2006,
pp. 341–364.
[31] D. Schmidt, L. Gram, M. Brodie, G. Krꢄmer, E. Perucca, R. Kꢄlviꢄinen, C. E.
Elger, Epilepsy Res. 2000, 41, 245–251.
[32] K. E. Andersen, J. Lau, B. F. Lundt, H. Petersen, P. O. Huusfeldt, P. D.
Suzdak, D. B. Swedberg, Bioorg. Med. Chem. 2001, 9, 2773–2785.
[33] B. S. Orlek, F. E. Blaney, F. Brown, M. S. G. Clark, M. S. Hadley, J. Hatcher,
G. J. Riley, H. E. Rosenberg, H. J. Wadsworth, P. Wyman, et al., J. Med.
Chem. 1991, 34, 2726–2735.
[34] A. Kragler, G. Hçfner, K. T. Wanner, Eur. J. Med. Chem. 2008, 43, 2404–
2411.
[35] C. B. DeZutter, J. H. Horner, M. Newcomb, J. Phys. Chem. A 2008, 112,
1891–1896.
[6] O. Ramstrçm, J.-M. Lehn, Nat. Rev. Drug Discovery 2002, 1, 26–36.
[7] R. L. E. Furlan, S. Otto, J. M. K. Sanders, Proc. Natl. Acad. Sci. USA 2002,
99, 4801–4804.
[8] S.-A. Poulsen, J. Am. Soc. Mass Spectrom. 2006, 17, 1074–1080.
[9] S.-A. Poulsen, Rapid Commun. Mass Spectrom. 2000, 14, 44–48.
[10] B. M. R. Liꢃnard, J. Med. Chem. 2008, 51, 684–688.
[11] S. Sakai, Y. Shigemasa, T. Sasaki, Tetrahedron Lett. 1997, 38, 8145–8148.
[12] A. Eliseev, M. Nelen, J. Am. Chem. Soc. 1997, 119, 1147–1148.
[13] T. Bunyapaiboonsri, O. Ramstrçm, S. Lohmann, J.-M. Lehn, L. Peng, M.
Goeldner, ChemBioChem 2001, 2, 438–444.
[36] Hydrazine 10 can be stored as the HCl salt at À308C for at least
12 month.
[37] Y.-C. Cheng, W. H. Prusoff, Biochem. Pharmacol. 1973, 22, 3099–3108.
[14] O. Ramstrçm, S. Lohmann, T. Bunyapaiboonsri, J.-M. Lehn, Chem. Eur. J.
2004, 10, 1711–1715.
[15] T. Bunyapaiboonsri, H. Ramstrçm, O. Ramstrçm, J. Haiech, J.-M. Lehn, J.
Med. Chem. 2003, 46, 5803–5811.
Received: April 10, 2012
[16] S. Ladame, Org. Biomol. Chem. 2008, 6, 219–226.
Published online on && &&, 0000
ChemMedChem 0000, 00, 1 – 14
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org
&13
&
ÞÞ
These are not the final page numbers!

FULL PAPERS
M. Sindelar, K. T. Wanner*
Mix and measure: The generation of
compound libraries by dynamic combi-
natorial chemistry in the presence of
a target and subsequent library screen-
ing by competitive mass spectrometry
(MS) binding assays represents a new
and highly efficient approach to drug
discovery. This method, which requires
the compound libraries to be rendered
pseudostatic, has been successfully ap-
plied to mGAT1, the most abundant
GABA transporter in the brain, leading
to potent hits for this target.
&& – &&
Library Screening by Means of Mass
Spectrometry (MS) Binding Assays—
Exemplarily Demonstrated for
a Pseudostatic Library Addressing g-
Aminobutyric Acid (GABA) Transporter
1 (GAT1)
&14
&
www.chemmedchem.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 0000, 00, 1 – 14
ÝÝ
These are not the final page numbers!