Full text of DOI:10.1002/cbic.201402192

CHEMBIOCHEM
FULL PAPERS
DOI: 10.1002/cbic.201402192
Repositioning the Substrate Activity Screening (SAS)
Approach as a Fragment-Based Method for Identification
of Weak Binders
Rafaela Gladysz,[a] Matthias Cleenewerck,[a] Jurgen Joossens,[a] Anne-Marie Lambeir,[b]
Koen Augustyns,[a] and Pieter Van der Veken*[a]
Fragment-based drug discovery (FBDD) has evolved into an es-
tablished approach for “hit” identification. Typically, most appli-
cations of FBDD depend on specialised cost- and time-inten-
sive biophysical techniques. The substrate activity screening
(SAS) approach has been proposed as a relatively cheap and
straightforward alternative for identification of fragments for
enzyme inhibitors. We have investigated SAS for the discovery
of inhibitors of oncology target urokinase (uPA). Although our
results support the key hypotheses of SAS, we also encoun-
tered a number of unreported limitations. In response, we pro-
pose an efficient modified methodology: “MSAS” (modified
substrate activity screening). MSAS circumvents the limitations
of SAS and broadens its scope by providing additional frag-
ments and more coherent SAR data. As well as presenting and
validating MSAS, this study expands existing SAR knowledge
for the S1 pocket of uPA and reports new reversible and irre-
versible uPA inhibitor scaffolds.
Introduction
Over the last decade, fragment-based drug discovery (FBDD)
has become an established methodology, delivering several
high-quality leads and clinical candidates and at least one
FDA-approved drug (vemurafenib/PLX4032).[1] FBDD emerged
as an alternative to high-throughput screening (HTS) for “lead”
discovery. FBDD approaches have in common that they con-
struct “lead” molecules from smaller fragments, typically con-
taining fewer than 12 heavy atoms and possessing relatively
low individual affinities. Such fragments in general allow or-
thogonal optimisation to meet predefined criteria for target af-
finity (“ligand-efficiency”) and biopharmaceutical behaviour, as
proposed in, for example, the “rule of three” for fragments.[3]
Another advantage of the approach is that the compound li-
braries used for fragment identification can be multiple orders
of magnitude smaller in size than those required for HTS. This
relates to the fact that drug-like diversity space can be much
more efficiently probed with small fragments than with the
typically larger molecules found in drug-like HTS libraries.[2–5]
There are several recent reviews that extensively document
these concepts with relevant examples from case studies.[6]
Nonetheless, as summarised in a recent opinion article by
Murray et al., FBDD still faces a number of fundamental chal-
lenges.[5] One of these relates to the need for better methodol-
ogy to detect and to study weakly binding fragments. X-ray
crystallography and protein NMR are among the most estab-
lished techniques for this purpose.[7,8] However, their main
drawback is the requirement for substantial amounts of highly
purified target protein and specialised infrastructure that runs
at high financial cost and with medium throughput capacity.
Other biophysical techniques that bypass these limitations
have been applied. These include, among others, surface plas-
mon resonance (SPR), 19F NMR, thermal denaturation, mass
spectrometry, thermal electrophoresis and isothermal titration
calorimetry.[5,9,10]
Specifically for enzyme targets, substrate activity screening
(SAS) was proposed by Ellman and co-workers as an attractive
fragment-based approach to inhibitor discovery.[11–16] Promising
results with different classes of enzyme targets have been re-
ported, mainly by the groups of Ellman and, more recently,
Seebach.[17] Notable examples include identification of non-
peptidic inhibitors for serine and cysteine proteases, receptor
tyrosine kinases and the protein tyrosine phosphatase PtpB of
Mycobacterium tuberculosis.[11–13,17] The SAS approach, demon-
strated for a protease target, consists of three steps. First of all,
a library of small, fragment-sized molecules, each linked to
a scissile fluorogenic amide bond, is screened for substrates of
a target protease (step 1, Scheme 1). Next, the identified sub-
strates are optimised in a separate cycle (step 2) and finally
transformed into inhibitors by replacing the scissile amide
bond with a warhead functionality (step 3).[11] Diversity in the
substrate library comes from drug-like, fragment-sized groups
that function as potential affinity-conferring recognition units
[a] R. Gladysz, M. Cleenewerck, Dr. J. Joossens, Prof. Dr. K. Augustyns,
Prof. Dr. P. Van der Veken
Medicinal Chemistry (UAMC)
Department of Pharmaceutical Sciences, University of Antwerp
Universiteitsplein 1, 2610 Wilrijk (Belgium)
E-mail: pieter.vanderveken@uantwerpen.be
[b] Prof. Dr. A.-M. Lambeir
Laboratory of Medical Biochemistry
Department of Pharmaceutical Sciences, University of Antwerp
Universiteitsplein 1, 2610 Wilrijk (Belgium)
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cbic.201402192.
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2014, 15, 2238 – 2247 2238

CHEMBIOCHEM
FULL PAPERS
www.chembiochem.org
inhibitors therefore continues to
raise our interest. During our
exploration of SAS we encoun-
tered a number of limitations of
the reported approach. Here we
describe a simple and effective
alternative for SAS, which we
have named “MSAS” (modified
substrate activity screening). We
demonstrate that screening the
library for inhibitors of a target
enzyme rather than for its sub-
strates avoids false negatives:
that is, fragments with high po-
Scheme 1. The SAS approach, demonstrated for a protease target.
tential for inhibitor discovery
that are not identified in a SAS
for the target enzyme. These are linked to a functionality that
can be processed by the target enzyme, thereby releasing
a quantifiable reporter molecule. The main rationale of the SAS
methodology is that the cleavage efficiencies (expressed as the
kcat/KM ratio) for the individual library members are positively
correlated with a fragment’s affinity for the enzyme’s transi-
tion-state-stabilising conformation and hence with its potential
for inhibitor design. It is worth mentioning that this principle
had already been recognised decades ago and has been ap-
plied extensively for discovery of substrate-derived enzyme in-
hibitors. SAS, however, does not rely on library molecules that
are direct analogues of a target’s natural substrates. In this
way, it is not biased to deliver inhibitors with an overall bio-
molecule-derived architecture but has the unique potential to
provide fragments with favourable, more drug-like structures
immediately.
assay. We also show that MSAS avoids false positives that can
surface during a regular SAS assay, and runs with better cost
and time efficiency. Furthermore, an FBDD strategy is reported
to transform identified fragments into inhibitors that do not
rely on a warhead functionality for target affinity. Finally, we
also demonstrate with the aid of experimental data that the
classical SAS step in which substrates are translated into inhibi-
tors by addition of a warhead, although intrinsically highly val-
uable, does not per se lead to compounds of practical biophar-
maceutical quality. Finally, the results show that adoption of
the MSAS approach can not only circumvent limitations of the
parent methodology, but can also offer additional potential for
FBDD on enzyme targets.
In practice, it is necessary to optimise fragments after the Results and Discussion
first stage. This is done by creating additional, directed chemi-
Library synthesis
cal diversity around the best substrates identified. Optimised
substrates are ultimately transformed into inhibitors by direct
replacement of the enzyme-processed functionality in a sub-
strate molecule with a mechanism-based warhead or pharma-
cophore.[11,18] Although only superficially examined for this pur-
pose, SAS fragments could also be subjected to a standard
FBDD-optimisation strategy for obtaining small-molecule inhib-
itors that do not draw upon a warhead functionality to gain
target affinity.
The start of our investigations was the synthesis of a SAS li-
brary of 137 fluorogenic N-acyl-7-amino-3-methylcoumarin
substrates (N-acyl AMCs). All the compounds in the library con-
tained fragment-sized N-acyl residues (MW<150). Selection of
around 90 of these residues was done in a non-target-biased
manner, aiming to cover as much of “drug-like” chemical space
as possible: steric, electronic and electrostatic parameters were
taken into account. In addition, several target-biased subsets
were prepared, containing moieties of known uPA inhibitors
and/or fragments that might reasonably be anticipated to bind
to the active centres of trypsin-like enzymes. Inclusion of these
fragments as positive controls was considered most helpful for
investigation of the intrinsic performance of SAS during frag-
ment identification and the internal coherence of results ob-
tained. Although highly interesting, potential issues of this
type have not been investigated earlier. It is also worth men-
tioning that on the basis of the dimensions of fragments and
SAS’s reliance on enzymatic activity, processed substrates can
reasonably be expected to be accommodated in the S1 region
of the enzyme (i.e., the S1 pocket and the parts of the active
centre immediately surrounding it). This consideration was
taken into account during the selection of the positive control
We decided to apply SAS to inhibitor discovery for urokinase
plasminogen activator (uPA), a trypsin-like serine protease that
is overexpressed in metastasising solid tumours.[19,20] The
enzyme is a valuable oncology target, but clinical development
of its inhibitors has been problematic. This is most probably re-
lated to the doubtful biopharmaceutical performance of com-
pounds developed so far and their insufficient selectivity with
respect to other, phylogenetically related trypsin-like proteases.
Nonetheless, the field of urokinase inhibitor discovery still sees
highly interesting developments, such as with recent ap-
proaches based on bicyclic peptide constructs.[21]
Earlier, our group described selective, irreversible inhibitors
of uPA with significant anti-metastatic activity in a rodent
model of breast cancer.[22] Discovering structurally novel uPA
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2014, 15, 2238 – 2247 2239

CHEMBIOCHEM
FULL PAPERS
www.chembiochem.org
set: this set contains mainly basic groups and S1-binding sub-
stituents of known uPA inhibitors.
group and the scissile amide bond in a typical P1(Arg)-contain-
ing peptide substrate of the enzyme. The lipophilic aryl deriva-
tives 6 and 17 are the only non-basic hits in the series, with 6
being processed by uPA with similar efficiency to 2. Notably,
even a small deviation from a hit compound’s structure was
observed to cause a total loss of substrate properties. This is
illustrated by the cleavage efficiency pattern within the series
1–3, 4–6, 9 and 10, and indicates that the robustness of SAS
as a method to identify useful fragments for inhibitor discovery
is not optimal. It is indeed highly conceivable that “unbiased”
SAS libraries will overlook potentially interesting fragments if
either 1) the linker distance between the fragment and the
acyl-AMC functionality or 2) the fragment substitution pattern
does not precisely fit the requirements for stabilisation of the
transition state of substrate conversion. In our opinion, these
findings demonstrate that unmodified application of the SAS
protocol can lead to loss of relevant information and hence
“false negatives”. In addition, predictive application of the ob-
tained processing data by construction of structure–cleavage
efficiency relationships (analogous to structure–activity rela-
tionships in traditional inhibitor discovery) seems compro-
mised by the use of a readout system that is so sensitive to
minute structural changes. Furthermore, we observed that the
SAS protocol requires substantial amounts of target protein
(ꢀ2.5 mg per well), together with long screening times. Both
experimental parameters were used according to Ellman’s
reports and were found to be crucial for detection of slowly
degraded library members. It is also worth highlighting the
critical importance of enzyme purity. In a separate screen of
our library with commercial uPA obtained from human urine,
a series of additional hits characterised by very high turnover
efficiencies was obtained. These compounds, however, were
not cleaved to any extent by the recombinant enzyme
(Scheme 2). During further investigations of this overt discrep-
ancy between the two uPA preparations, we were able to
show that the processing of those compounds could not be
inhibited by addition of a nanomolar uPA inhibitor that we
had reported earlier (UAMC-00122).[22] Application of chromato-
graphic and gel electrophoretic techniques were not helpful
for identifying the catalyst responsible for cleavage in the
human uPA preparation. Nonetheless, these results also indi-
A substantial part of this library could be prepared by a one-
pot protocol, by starting from the individual acyl residues and
7-amino-4-methylcoumarin and using Ghosez’s reagent as the
coupling mediator.[23,24] Of a large series of mild coupling re-
agents that we evaluated (DCC, EDC, TBTU, HATU, TFFH,
PyBrop, Ghosez’s reagent), only the last was found capable of
cleanly and efficiently promoting the reaction with the very
weakly nucleophilic 7-amino-4-methylcoumarin. For several
compounds, additional steps (protection, homologation, func-
tionalisation) were necessary in order to obtain the desired
derivatives. The structures, the synthetic preparation and the
characterisation data of all library members are listed in the
Supporting Information.
SAS experiment for the library of N-acyl AMCs
The prepared library of N-acyl AMCs was first screened for uPA
substrates by a typical SAS protocol. All these experiments
were conducted in duplicate in HEPES buffer at pH 8.2, thus al-
lowing near-maximum enzymatic activity to be combined with
minimal aspecific hydrolysis of N-acyl AMCs. An initial screen-
ing of the library was performed with 200 nm of recombinant
human uPA and the highest substrate concentration allowed
by compound solubility. Although these concentrations varied
for the individual library members, they were generally in the
100–500 mm range. We reasoned that the use of high substrate
concentrations in the exploratory phase of the project would
allow identification of all library members that are processed
by uPA, even those characterised by low kcat values.
Under the initial conditions, eleven N-acyl AMCs from the li-
brary were found to behave as substrates of uPA, albeit with
large differences in cleavage rate. To allow reliable ranking of
cleavage efficiencies, assays for these compounds were repeat-
ed at subsaturating substrate concentrations ([S]<KM). To
avoid the need to determine KM values for all the obtained
hits, we followed the approach proposed by Ellman et al.[11]
Here, only the KM value of the optimal substrate in the series is
determined. Subsequently, all initially obtained hits are investi-
gated again at a concentration below the KM value of the best
substrate, with this serving as a reference relative to which
cleavage efficiencies are reported. We considered the guanidi-
nophenyl-based compound 2 (Table 1), displaying a KM value
of 120 mm, to be the best substrate. Subsequently, all initially
obtained hits were rescreened at 100 mm concentration, and
this reconfirmed guanidinophenyl derivative 2 as the most effi-
ciently cleaved substrate in the series. The cleavage efficiencies
of all eleven hits, relative to compound 2, are summarised in
Table 1 (“substrate screen” columns).
Given uPA’s substrate preferences and the intentional inclu-
sion of a substantial number of basic compounds in the library,
it is not surprising that most other identified hits contain
a basic functionality, most likely accommodated in the acidic
S1 pocket of uPA. Additionally, the distance between this basic
functionality and the acyl-AMC group in each substrate rough-
ly equals the corresponding distance between the guanidine
Scheme 2. Compounds that were processed by a commercial human uPA
preparation obtained from urine, but not by recombinant human uPA.
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2014, 15, 2238 – 2247 2240

CHEMBIOCHEM
FULL PAPERS
www.chembiochem.org
Table 1. Hits obtained after screening of a 137-compound library of N-acyl AMCs).
Cpd Structure
Substrate screen Inhibitor screen Cpd Structure
Substrate screen Inhibitor screen
cleavage
[I] [mm] inhibition
cleavage
[I] [mm] inhibition
R=
efficiency[a] [%]
[%][b]
R=
efficiency[a] [%]
[%][b]
[c]
1
–
100
63.4
2
100
100
50
3
5
–
–
250
100
58
4
6
–
50
25.7
50
33.6
98.7
400
7
9
–
–
50
500
500
25
8
10
12
72.3
65
500
500
500
50
45
39
64.3
30.4
11
34.5
30
13
15
20.5
9
500
500
28
17
14
16
14
–
500
100
20
27.5
17
19
8.2
–
100
100
27
31
18
20
–
–
50
23
100
32.6
21
23
–
–
50
21
22
24
–
250
500
27.3
23
250
20.3
3.2
25
27
–
–
500
100
21.5
24
26
28
–
–
100
100
27
19
[a] Cleavage efficiency is defined as the cleavage rate of a compound relative to the “best” substrate in the library (compound 2). [b] Inhibition is defined
as the % decrease in the processing rate of reference uPA substrate pyro-Glu-Gly-Arg-pNA. [c] “–” indicates that no uPA-mediated cleavage of the com-
pound was observed.
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2014, 15, 2238 – 2247 2241

CHEMBIOCHEM
FULL PAPERS
www.chembiochem.org
cate that the published SAS protocol is susceptible to possible
occurrence of false positives, with the presence of other cata-
lytically active species (e.g., other enzymes occurring as impuri-
ties) being responsible.
more, our alternative method avoids false positives due to
other catalytically active species present in the enzyme prepa-
ration. False positive results caused by compound-induced
aggregation or denaturation of the enzyme were also not ob-
served during the inhibitor screen.
Inspired by these hitherto unaddressed but relevant find-
ings, we devised a fundamentally different experimental setup
for library evaluation. In this approach the inhibitory properties
of the library members are investigated, rather than their sub-
strate properties. A protocol strongly related to the archetypi-
cal assay normally used for enzyme inhibitor evaluation was
elaborated. Here, the library members’ potential to inhibit deg-
radation of a known, peptide-derived chromogenic substrate
of the target is evaluated. We hypothesised that this setup
should be able to uncover all fragments with affinity for uPA,
and not only those characterised by an ideal linker distance or
an optimal substitution pattern. Additionally, selection of the
efficiently processed chromogenic pyro-Glu-Gly-Arg-pNA (KM =
80 mm) for this assay allowed the enzyme concentration to be
lowered tenfold relative to the SAS protocol.[25] The application
of a single, kinetically well-characterised substrate in our opin-
ion also avoids the possibility of “false positive” results and re-
moves the need to verify whether the library members are pro-
cessed by the actual target enzyme or by another catalytically
active species occurring in the enzyme preparation.
These findings confirm that, to identify useful fragments in
a given SAS library, it could be more efficient to evaluate the
inhibitory properties of the library members rather than their
substrate characteristics. More emphasis can in this way be
placed on creating a structurally diverse library because the
need for a number of homologues or close analogues around
each structural feature is reduced. A point-by-point compari-
son of the two screening modes is given in Table 2.
Table 2. Comparison of SAS and inhibitor screening protocols.
Parameter
SAS
Inhibitor screen
enzyme amount per well
screening times
2.5 mg (400 IU)
6 h[a]
0.25 mg (40 IU)
10 min
not observed
not observed
false positives[b]
yes[b]
false negatives
yes[c]
[a] Long screening times can cause errors due to, for example, enzyme
denaturation, autoproteolysis or buffer evaporation.[18] [b] Due to other
catalytically active species present in the enzyme preparation. [c] Result-
ing in incomplete SAR data.
In this experiment, with use of 20 nm uPA, the 137 com-
pounds of the library were screened again at 50–500 mm, with
concentrations depending upon compound solubility as
before. The readout consisted of the evaluation of uPA-mediat-
ed para-nitroaniline release from the chromogenic substrate
pyro-Glu-Gly-Arg-pNA, at 100 mm concentration. The results,
expressed as percentage inhibition of pyro-Glu-Gly-Arg-pNA
cleavage at a given compound concentration, are also sum-
marised in Table 1 (“inhibitor screen” columns). In general, it
deserves mentioning that the affinities displayed by the “hits”
are well within the range that is generally reported for frag-
ments (high micromolar). Furthermore, our alternative ap-
proach identifies all eleven “hits” initially revealed by the tradi-
tional SAS protocol. The relative affinities observed for these
eleven compounds in the inhibition experiment (extrapolated
from their inhibitory potencies) roughly reflect the cleavage ef-
ficiencies of the compounds.
When using a library of AMC amides for fragment identifica-
tion, one might nonetheless speculate on the possibility that
the AMC moiety might interfere during the process of target
binding. Although peptidyl-AMC amides have been used suc-
cessfully for decades in inhibitor discovery, this is not com-
pletely inconceivable in, for example, hypothetical cases in
which the AMC ring is not accommodated in the S1’ region of
the enzyme. Theoretically, such interference could consist
either of 1) a net impeditive effect on fragment binding (e.g.,
by steric hindrance) or, alternatively, 2) a net supportive effect
(e.g., if the AMC ring were to contribute to affinity). We expect
that interference of the first type would surface in the form of
incoherent SAR data, involving “outliers” that unexpectedly do
not show target affinity. The identification of all positive con-
trols in our library and the near complete coverage of com-
pounds that belong to inhibiting structural subgroups of the
library do indicate that, in general, the AMC ring is not signifi-
cantly hampering fragment binding. Similarly, the obtained re-
sults also do not suggest that the AMC ring contributes signifi-
cantly to affinity, because most of the library members evaluat-
ed were devoid of measurable uPA affinity. Taking all these
considerations into account, we expect the AMC portion of the
library members not to interfere significantly with the inhibito-
ry properties of the fragments. Equally illustrative of this is the
fact that the isolated guanidinobenzene fragment 38 (vide
infra) has an affinity broadly comparable to those of AMC-
linked fragments 1–3. So far we have also not observed indica-
tions of interference in a number of other ongoing projects
dealing with inhibitor design for caspases and autophagins
during which the same library was screened for inhibiting frag-
ments.
Most interestingly, though, our modified screening proce-
dure also discloses an additional 17 molecules that inhibit the
release of para-nitroaniline. Inspection of the compounds with
inhibitory properties immediately provides a more coherent
image of structural classes that possess potential for uPA inhib-
itor discovery within the library. As an example, all guanidino-
phenyl (1–3) and guanidinoalkyl (9, 10) homologues present in
the library were identified as inhibitors, whereas SAS had only
selected one of either class. Analogously, all chlorophenyl (4–6)
and closely related lipophilic phenyl derivatives (7, 16–19, 21
and 22) that were present in the library turned up as potential-
ly valuable constituents for new uPA inhibitors. Again, from
the results of the SAS approach, one would conclude that 6
and 17 are two singletons of interest within the library, where-
as they instead belong to a group of closely related structures
that could all be valuable for uPA inhibitor design. Further-
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2014, 15, 2238 – 2247 2242

CHEMBIOCHEM
FULL PAPERS
www.chembiochem.org
It also deserves mentioning that the highest affinities ob-
served within the structural subgroups with inhibitory proper-
ties do not necessarily belong to the compounds that are also
substrates. This is seen, for example, on comparing the slightly
higher inhibitory potency of guanidinophenyl derivative 1 with
that of homologue 2, of which only the latter is processed as
a substrate. We therefore investigated whether apparently
lower affinities of library members displaying substrate proper-
ties might be accounted for by their gradual consumption
during the inhibition experiment. Quantification of cleavage
during the course of an inhibition experiment was determined
by monitoring the release of AMC. The process was, however,
found to be too slow to interfere significantly with determina-
tion of inhibitory potency. This is read out after 10 min, typical-
ly during the linear phase of peptide substrate consumption
(Figure 1). In hypothetical cases in which library members with
exceptionally high kcat/KM values relative to the peptide sub-
strate used might be present, however, such an effect could
not be excluded. In addition, therefore, two control experi-
ments were also carried out for each inhibitory library member
to ascertain whether or not the identified inhibitor series could
include non-competitive, allosteric or irreversible compounds.
Again, the SAS protocol would not allow identification of such
fragments, although they could certainly be of interest to spe-
cific inhibitor discovery programs. For all compounds (1–28),
inhibition was found to decrease with increasing concentration
of the chromogenic substrate, thus indicating competition for
the enzyme’s active site. The percentage inhibition increased
with inhibitor concentration and did not change significantly
with longer inhibitor preincubation times, as would be the
case for slowly and irreversibly binding compounds.
On the basis of all these results, we propose a modified ex-
perimental strategy that combines optimal efficiency and maxi-
mum extraction of useful structural information during screen-
ing of SAS libraries. We have provisionally called this approach
MSAS. Its key steps are represented in Scheme 3. In MSAS, the
library is first screened for inhibitory fragments (Scheme 3,
step 1). This experimental layer will provide SAR data for the
interesting fragment types present within the library. As dem-
onstrated, the hits obtained in the inhibition experiment will
also include library members with substrate properties. Be-
cause step 1 runs with higher time- and cost-efficiency than a
traditional SAS experiment, we propose to perform SAS during
the second phase of MSAS and only for identifying the sub-
strates within the set of hits identified during the inhibitor
screening experiment (Scheme 3, step 2). Furthermore, it is im-
portant to stipulate that MSAS’s experimental setup, like that
of the parent methodology, is not limited to protease inhibitor
discovery. The same strategy can directly be applied to any
other type of enzyme target studied, provided that the mem-
bers of the screened library each contain a suitable enzyme-
processable functionality.
The affinity data obtained after step 1 and the substrate
property data obtained after step 2 can theoretically be vali-
dated in several orthogonal approaches for transforming frag-
ments into inhibitors. To investigate and validate the results of
the MSAS setup further, representative examples of such ap-
proaches are preliminarily explored in the following part. Li-
brary member 2, possessing both significant uPA affinity and
substrate properties, was selected as the common starting
point. The 4-guanidinophenethyl portion of this compound
had already been reported earlier by us as a constituent of
nanomolar and highly selective peptide-derived diaryl phos-
phonate inhibitors of uPA.[22] It was therefore included in the
“biased” portion of the library. Its status as a positive control
element also justified its selection for the experiments dealing
with translating fragments into inhibitors.
For the validation of affinity data produced during step 1,
we present a generally applicable strategy—not previously re-
ported, to the best of our knowledge—for obtaining scaffold-
based inhibitors. With specific regard to protease targets, such
scaffold-based compounds have the potential to circumvent
several of the inherent liabilities of classical, peptide-based
protease inhibitors [mainly related to ADME (absorption, distri-
bution, metabolism and excretion)]. Also, on a much more
general level, many of the recently approved small-molecule
drugs share an overall comparable architecture consisting of
a central scaffold decorated with several substituents that
confer additional affinity for the biomolecules they target. A
generalised routine for obtaining such compounds is pro-
posed. Firstly, a fragment identified during the first step of
MSAS is chemically grafted onto a limited set of different,
drug-like scaffolds (Scheme 4). If a compound with uPA affinity
significantly higher than that of the original fragment is found
within this small set of monosubstituted scaffolds, that com-
pound is selected for further optimisation. During the optimi-
Figure 1. Inhibitory profiles of two selected structurally related MSAS hits.
A) Inhibitory profile of 2, a library member with substrate properties. At
t=10 min, consumption of 2 is minimal, and read-out of inhibitory potency
is not significantly influenced. B) Inhibitory profile of 3, a library member
that is not a substrate. c: uPA+pyro-Glu-Gly-Arg-pNA (control): release of
pNA; c: uPA+pyro-Glu-Gly-Arg-pNA+2/3: release of pNA; c: uPA+
pyro-Glu-Gly-Arg-pNA+2/3: release of AMC.
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2014, 15, 2238 – 2247 2243

CHEMBIOCHEM
FULL PAPERS
www.chembiochem.org
readily be introduced, preferentially in a regioselec-
tive fashion and by combinatorial chemistry tech-
niques.
To elaborate this concept, we chose the imidazo-
pyridine and pyrimidine scaffolds, already present in,
for example, the hypnotic drug zolpidem and the
HIV-RT inhibitor rilpivirine.[26,27] A wealth of efficient
chemical decoration strategies that allow efficient
production of diversely substituted analogues exist
in both cases.[28,29] Firstly, the guanidinophenethyl
fragment was attached through an amine linker
(Scheme 4). The two scaffolded inhibitor fragments
29 (Scheme 5) and 30 were then evaluated; they dis-
played roughly comparable affinities for uPA (9.4Æ
0.8 mm and 20.7Æ1.3 mm, respectively). This corre-
sponds to a relative increase in affinity of about one
order of magnitude relative to library member 2 and
indicates that both the imidazopyridine and pyrimi-
dine scaffolds could be used for construction of scaf-
fold-based uPA inhibitors. Hypothetically, additional
steps necessary for obtaining molecules with further
optimised affinity would consist of introducing one
or several additional substituents.
As pointed out earlier, attention was then devoted
to the original data validation strategy of the SAS
approach. To this end, we grafted the potentially ir-
reversibly binding diphenyl phosphonate warhead
onto the 4-guanidinophenethyl moiety (Scheme 6)
and also, as a control, onto the homologous 4-gua-
nidinophenylmethyl residue of 1 (Table 3). The latter
Scheme 3. Outline of the modified substrate activity screening approach, demonstrated
for a protease target.
sation, one to several additional substituents are introduced
on the selected monosubstituted scaffold to increase target af-
finity further. For maximum efficiency, it is advisable to select
scaffold types onto which several additional substituents can
fragment, although still possessing uPA affinity, was not pro-
cessed as a substrate in the corresponding assay. As suggested
by Ellman et al., this would translate either into an inhibitor
with significantly reduced potency, or at least into a compound
Scheme 4. Introduction of the 4-guanidinophenethyl fragment onto an imidazopyridine and a pyrimidine scaffold to afford 29 and 30, respectively. Scaffold
positions that—from a synthetic point of view—allow easy substituent introduction for further optimisation are marked with asterisks; R1, R2 =additional sub-
stituents.
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2014, 15, 2238 – 2247 2244

CHEMBIOCHEM
FULL PAPERS
www.chembiochem.org
with compromised capability of mechanism-based enzyme
blocking.[12] Compounds 33 and 34 indeed displayed an ap-
proximately 1000-fold difference in potency (Table 3). Further-
more, both compounds were subsequently demonstrated to
be irreversible inhibitors. This typically implies that the inhibi-
tory potencies of 33 and 34 should not be interpreted as
a strict indication of intrinsic affinities but rather as a measure
for their respective second-order rate constants of the irreversi-
ble step in enzyme inactivation. In any case, these results
agree well with SAS’s tenets, and on a broader level, with the
fundamental assumptions of more canonical approaches in
substrate-based drug discovery.
Scheme 5. Synthesis of scaffold-based inhibitor 29 (for synthesis and charac-
terisation see the Supporting Information). Reagents: a) i: MeOH, HClO4
(cat.), RT; ii: TFA/CH2Cl2 1:1, RT.
It is also remarkable that the potency displayed by com-
pound 34 broadly compares with that of diaryl phosphonate
35 (UAMC-00150), reported earlier by us as one of the best
representatives of known uPA inhibitors.[22] This reference in-
hibitor contains an additional methoxycarbamoyl group mim-
icking the P2ÀP1 amide bond of uPA’s peptide substrates. On
the basis of this higher degree of similarity with compound
types that are naturally processed by uPA, one could reasona-
bly assume that both the kinetic and the thermodynamic pa-
rameters of target binding could be favourable in the case of
35, thus resulting in a higher net potency. However, the ab-
sence of the methoxycarbamoyl fragment in 34 does not seem
to interfere significantly with the overall process of target rec-
ognition and irreversible covalent bond formation.
Scheme 6. Synthesis of diaryl phosphonate inhibitor 34 (for synthesis and
characterisation see Supporting Information). Reagents: a) phenol, Et3N, tolu-
ene; b) i: 1-(2-bromoethyl)-4-nitrobenzene, pressure tube, 1208C; ii): zinc
dust, THF, 08C; c) i: N,N’-di-Boc-guanylpyrazole, Et3N, THF, RT; ii: TFA/CH2Cl2
1:1, RT.
Additionally, compounds 36 and 37, containing identical
methoxycarbamoyl substituents, were evaluated as potential
uPA inhibitors. The first, compound 36, lacks a warhead func-
tionality and served mainly to assess the contribution of the
reactive functionality to inhibitory potencies, as discussed earli-
er. Compound 36 has the same IC50 value as phenylguanidine
fragment 38 (IC50 =250 mm); this indicates that the potencies
observed for 33–35 are mainly driven by an efficiently occur-
ring irreversible step after initial binding of the inhibitor to
uPA.[30]
Table 3. Comparison of IC50 values for uPA inhibition of different benzyl-
guanidine-containing compounds derived from 1 and 2.
Cpd
R=
IC50 (uPA) [mm]
1.39Æ0.06
Inhibition type
irreversible[a]
With this information at hand, compound 37 was evaluated
as a second test case for the SAS protocol. This molecule con-
tains a carbonitrile group, a potentially reversible, covalent
warhead type very often used in inhibitors of serine proteases.
Its low-micromolar affinity indicates that the nitrile group in
this molecule might not be suitably oriented to allow covalent
bond formation with uPA in a low-energy inhibitor conforma-
tion. This result, in our opinion, does not raise any critical
doubts as to the validity of the SAS protocol. It nonetheless
warns that for transformation of SAS “hits” into inhibitors, eval-
uation of several warheads might be mandatory in order to
identify a type that performs well with the selected protease
and inhibitor.
33
34
0.0097Æ0.0003
0.0031Æ0.0005
irreversible[a]
irreversible[a]
35[b]
36
37
250
reversible
reversible
Conclusion
In conclusion, we have investigated SAS for the discovery of in-
hibitors of oncology target urokinase (uPA). Although our re-
sults were supportive of the fundamental hypotheses formulat-
ed earlier for SAS, we also encountered a number of hitherto
unreported limitations of the approach. In response, we pro-
5.0Æ0.01
[a] Validated experimentally by dilution assay.[25] [b] Previously reported as
UAMC-00150.[22]
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2014, 15, 2238 – 2247 2245

CHEMBIOCHEM
FULL PAPERS
www.chembiochem.org
out with urokinase chromogenic substrate BIOPHEN CS-61(44)
pose a simple, efficient modified methodology: “MSAS” (modi-
fied substrate activity screening). This methodology not only
circumvents limitations of the parent approach, but also
broadens its scope by providing additional fragments and
more coherent SAR data. As well as introducing MSAS as a gen-
erally applicable method for enzyme inhibitor discovery, this
study has expanded existing SAR knowledge on S1-pocket-
binding fragments of uPA. In addition, hitherto unreported
uPA inhibitor scaffolds are presented and have been used to
obtain new reversible and irreversible compounds.
(pyro-Glu-Gly-Arg-pNA) purchased from Nodia (KM =80 mm).[25]
A
HEPES buffer (Sigma–Aldrich, pH 8.2, 50 mm) was used. All enzy-
matic activity measurements were routinely performed in dupli-
cate. N-Acyl AMC stock solutions and inhibitor stock solutions
(10 mm) were prepared in DMSO and stored at À208C. Enzymatic
assays contained not more than 5% (v/v) of DMSO.
Inhibitor kinetic assays: Enzymatic activity was measured over
5 min at 378C in the presence of urokinase chromogenic substrate
BIOPHEN CS-61(44). Absorbance was monitored at l=405 nm. The
assay mixture contained the N-acyl AMC compound (50–500 mm
depending on the solubility), the non-recombinant urokinase solu-
tion in buffer (ca. 20 nm) and substrate BIOPHEN CS-61(44) (KM =
80 mm, 100 mm) in a final volume of 200 mL. The concentration of
the chromogenic substrate used (100 mm) allowed for a sufficiently
high initial substrate processing rate, while limiting competition
between substrate and inhibitor.
Experimental Section
Reagents were obtained from Sigma–Aldrich or from Acros, Fluoro-
chem or Apollo Scientific and were used without further purifica-
tion. Synthesised compounds were characterised by 1H NMR,
13C NMR and mass spectrometry. 1H NMR and 13C NMR spectra
were recorded with a 400 MHz Bruker Avance DRX 400 spectrome-
ter, and analysed by use of MestReNova analytical chemistry soft-
ware. ES mass spectra were obtained with an Esquire 3000plus ion-
trap mass spectrometer from Bruker Daltonics. Purities were deter-
mined with two diverse HPLC systems based either on mass detec-
tion or on UV detection. A Waters acquity UPLC system coupled to
a Waters TQD ESI mass spectrometer and a Waters TUV detector
was used. Where necessary, flash purification was performed with
a Biotage ISOLERA One flash system equipped with an internal var-
iable dual-wavelength diode array detector (200–400 nm).
Fluorogenic substrate screen against uPA: Screening of the li-
brary of N-acyl AMCs for substrates of uPA was performed over 6 h
at 378C. The excitation wavelength was 383 nm, and the emission
wavelength was 455 nm. Because of false positives appearing in
assays based on a non-recombinant enzyme, the substrate screen-
ing was performed with a recombinant human uPA. Initial screen-
ing of the library was performed at the highest substrate concen-
tration possible (for most of the library members 100–500 mm), uPA
concentration was around 200 nm. Final screening was performed
with subsaturating levels of the substrate.11 Under these conditions
cleavage is assumed to be a first-order process; hence kcat/KM
obs t
values are compliant with the relationship St/S0 =eÀk , where St =
Library of N-acyl aminocoumarins: A library of 137 N-acyl amino-
coumarins was prepared. A large part of the library was synthes-
ised by means of single coupling reactions of the constituting moi-
eties in the presence of a mild acyl chlorinating agent—namely
tetramethyl-a-chloroenamine (“Ghosez’s reagent”)[23,24]—as a cou-
pling mediator. Further details on the library synthesis and the
chemical characterisation of the obtained compounds can be
found in the Supporting Information.
concentration of the substrate remaining at time t, S0 =initial sub-
strate concentration, and kobs =kcat [enzyme]/KM.[33] Final substrate
screen and ranging hits based on the enzymatic cleavage efficiency
were determined at 100 mm substrate and approximately 200 nm
uPA concentration. Relative fluorescence units (RFUs) were mea-
sured for each substrate at regular intervals over a 6 h period of
time with and without enzyme (blank). Blank was subtracted from
the enzymatic activity measurements. The slope of the plotted line
gave the relative kcat/KM value for each substrate.[25] More informa-
tion can be found in the Supporting Information.
Synthesis of scaffolded reversible inhibitors of uPA: The scaffold-
based inhibitors of uPA discussed in this report contain the imida-
zopyridine and pyrimidine scaffold types. The imidazopyridine-scaf-
fold-containing inhibitor 29 was prepared by a general protocol
for the Groebke–Blackburn–Bienaymꢁ reaction for the synthesis of
fused 3-aminoimidazoles.[31,32] The pyrimidine-scaffold-containing
compound 30 was prepared by a standard nucleophilic aromatic
substitution reaction protocol from previously prepared starting
material. More detailed synthetic procedures and chemical charac-
terisation of the structures can be found in the Supporting Infor-
mation.
Determination of IC50 values: Enzymatic activity was measured at
378C with urokinase chromogenic substrate BIOPHEN CS-61(44).
Absorbance was monitored at l=405 nm. Each reaction mixture
had a volume of 200 mL and contained the chromogenic substrate
(250 mm), the non-recombinant enzyme solution (ca. 20 nm) in
buffer (145 mL) and the inhibitor (5 mL). An initial screening at three
inhibitor concentrations (250 mm, 2.5 mm and 25 nm) was per-
formed to estimate the range of the IC50 value. For exact IC50 deter-
mination, at least four inhibitor concentrations above and four
concentrations below the estimated IC50 value were used. IC50
values were determined by fitting the obtained data with a four-
parameter logistics equation with the aid of GraFit7. More informa-
tion can be found in the Supporting Information.
Synthesis of diaryl phosphonate irreversible inhibitors of uPA:
Diaryl phosphonate inhibitors 33 and 34 were prepared by a gener-
al protocol for base-promoted alkylation of H-phosphonates (the
Michaelis–Becker reaction) and by the modified version of the clas-
sical Arbuzov reaction protocol, respectively. More detailed syn-
thetic procedures and the chemical characterisation can be found
in the Supporting Information.
Determination of inhibition type: To follow dissociation of the in-
hibitor·enzyme complex, aliquots of enzyme were incubated at
378C 1) without and 2) with the inhibitor, at a concentration 50
times higher than its IC50. Enzyme was used at 2.5 times higher
concentration than for the IC50 determination. After 15 min, the ali-
quots were diluted 50-fold with the substrate (250 mm) solution in
assay buffer. Dissociation of the enzyme·inhibitor complex was de-
termined spectrophotometrically by monitoring hydrolysis of the
chromogenic substrate over time.[25]
General procedures for biochemical assays: Enzymatic assays
were performed with use of BioTek Microplate Reader (Syner-
gy MX). Data collection and analysis were performed with Gen5 Mi-
croplate Software and Microsoft Excel. Human urokinase-type plas-
minogen activator (uPA) was obtained from Nodia. A fluorogenic
substrate screen against uPA was performed with recombinant
human uPA (R&D Systems). Inhibitor kinetic assays were carried
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2014, 15, 2238 – 2247 2246

CHEMBIOCHEM
FULL PAPERS
www.chembiochem.org
[15] M. J. Leyva, F. Degiacomo, L. S. Kaltenbach, J. Holcomb, N. Zhang, J.
Gafni, H. Park, D. C. Lo, G. S. Salvesen, L. M. Ellerby, J. A. Ellmann, Chem.
Biol. 2010, 17, 1189–1200.
[16] T. D. Baguley, H.-C. Xu, M. Chatterjee, A. C. Nairn, P. J. Lombroso, J. A.
Ellman, J. Med. Chem. 2013, 56, 7636–7650.
[17] J. Chapelat, F. Berst, A. L. Marzinzik, H. Moebitz, P. Drueckes, J. Trappe,
D. Fabbro, D. Seebach, Eur. J. Med. Chem. 2012, 57C, 1–9.
[18] A. W. Patterson, W. J. L. Wood, J. A. Ellman, Nat. Protoc. 2007, 2, 424–
433.
[19] a) T. Steinmetzer, IDrugs 2003, 6, 138–142; b) A. H. Mekkawy, D. L.
Morris, M. H. Pourgholami, Future Oncol. 2009, 5, 1487–1500.
[20] a) K. Dass, A. Ahmad, A. S. Azmi, S. H. Sarkar, F. H. Sarkar, Cancer Treat.
Rev. 2008, 34, 122–136; b) G. Abbenante, D. P. Fairlie, Med. Chem. 2005,
1, 71–104.
[21] a) R. Roodbeen, B. Paaske, L. Jiang, J. K. Jensen, A. Christensen, J. T. Niel-
sen, M. Huang, F. A. A. Mulder, N. Chr. Nielsen, P. A. Andreasen, K. J.
Jensen, ChemBioChem 2013, 14, 2179–2188; b) S. Chen, D. Gfeller, S. A.
Buth, O. Michielin, P. Leiman, P. G. Leiman, C. Heinis, ChemBioChem
2013, 14, 1316–1322.
[22] J. Joossens, O. M. Ali, I. El-Sayed, G. Surpateanu, P. Van der Veken, A.-M.
Lambeir, B. Setyono-Han, J. A. Foekens, A. Schneider, J. Med. Chem.
2007, 50, 6638–6646.
[23] F. Munyemana, A.-M. Frisque-Hesbain, A. Devos, L. Ghosez, Tetrahedron
Lett. 1989, 30, 3077–3080.
[24] C. A. G. N. Montalbetti, V. Falque, Tetrahedron 2005, 61, 10827–10852.
[25] J. Joossens, P. Van der Veken, G. Surpateanu, A.-M. Lambeir, I. El-Sayed,
O. M. Ali, K. Augustyns, A. Haemers, J. Med. Chem. 2006, 49, 5785–
5793.
Acknowledgements
Rafaela Gladysz is grateful to the IWT–Vlaanderen for a scholar-
ship. This work received support from the Fund for Scientific Re-
search–Flanders (FWO–Flanders).
Keywords: drug design · inhibitors · MSAS · SAS · urokinase
[1] G. Bollag, P. Hirth, J. Tsai, J. Zhang, P. N. Ibrahim, H. Cho, W. Spevak, C.
Zhang, Y. Zhang, G. Habets, E. A. Burton, B. Wong, G. Tsang, B. L. West,
B. Powell, R. Shellooe, A. Marimuthu, H. Nguyen, K. Y. J. Zhang, D. R.
Artis, J. Schlessinger, F. Su, B. Higgins, R. Iyer, K. D’Andrea, A. Koehler, M.
Stumm, P. S. Lin, R. J. Lee, J. Grippo, I. Puzanov, K. B. Kim, A. Ribas, G. A.
McArthur, J. A. Sosman, P. B. Chapman, K. T. Flaherty, X. Xu, K. L. Nathan-
son, K. Nolop, Nature 2010, 467, 596–599.
[2] M. Congreve, G. Chessari, D. Tisi, A. J. Woodhead, J. Med. Chem. 2008,
51, 3661–3680.
[3] M. Congreve, R. Carr, C. Murray, H. Jhoti, Drug Discovery Today 2003, 8,
876–877.
[4] a) C. W. Murray, D. C. Rees, Nat. Chem. 2009, 1, 187–192; b) P. J. Hajduk,
J. Greer, Nat. Rev. Drug Discovery 2007, 6, 211–219.
[5] C. W. Murray, M. L. Verdonk, D. C. Rees, Trends Pharmacol. Sci. 2012, 33,
224–232.
[6] D. C. Rees, M. Congreve, C. W. Murray, R. Carr, Nat. Rev. Drug Discovery
2004, 3, 660–672.
[7] C. W. Murray, T. L. Blundell, Curr. Opin. Struct. Biol. 2010, 20, 497–507.
[8] M. J. Hartshorn, C. W. Murray, A. Cleasby, M. Frederickson, I. J. Tickle, H.
Jhoti, J. Med. Chem. 2005, 48, 403–413.
[9] a) W. Huber, F. Mueller, Curr. Pharm. Des. 2006, 12, 3999–4021; b) C.
Lambruschini, M. Veronesi, E. Romeo, G. Garau, T. Bandiera, D. Piomelli,
R. Scarpelli, C. Dalvit, ChemBioChem 2013, 14, 1611–16119; c) C. Dalvit,
Prog. Nucl. Magn. Reson. Spectrosc. 2007, 50, 259–261; d) G. Papeo, P.
Giordano, M.-G. Brasca, F. Buzzo, D. Caronni, F. Ciprandi, N. Mongelli, M.
Veronesi, A. Vulpetti, C. Dalvit, J. Am. Chem. Soc. 2007, 129, 5665–5672.
[10] T. Neumann, H. D. Junker, K. Schmidt, R. Sekul, Curr. Top. Med. Chem.
2007, 7, 1630–1642.
[11] W. J. L. Wood, A. W. Patterson, H. Tsuruoka, R. K. Jain, J. A. Ellman, J. Am.
Chem. Soc. 2005, 127, 15521–15527.
[12] C. M. Salisbury, J. A. Ellman, ChemBioChem 2006, 7, 1034–1037.
[13] M. B. Soellner, K. A. Rawls, C. Grundner, T. Alber, J. A. Ellman, J. Am.
Chem. Soc. 2007, 129, 9613–9615.
[26] D. S. Johnson, J. J. Li, The Art of Drug Synthesis, Wiley, New York, 2007.
[27] M. Sharma, L. D. Saravolatz, J. Antimicrob. Chemother. 2013, 68, 250–
256.
[28] D. J. Brown, R. F. Evans, W. B. Cowden, M. D. Fenn, Chemistry of Hetero-
cyclic Compounds: The Pyrimidines, Wiley, New York, 1994.
[29] E. Ruijter, R. Scheffelaar, R. V. Orru, Angew. Chem. Int. Ed. 2011, 50,
6234–6246; Angew. Chem. 2011, 123, 6358–6371.
[30] J. C. Powers, J. L. Asgian, O. D. Ekici, K. E. James, Chem. Rev. 2002, 102,
4639–4750.
[31] H. A. Bienayme, Angew. Chem. Int. Ed. 1998, 37, 2234–2237; Angew.
Chem. 1998, 110, 2349–2352.
[32] A. El Akkaoui, M.-A. Hiebel, A. Mouaddib, S. Berteina-Raboin, G. Guillau-
met, Tetrahedron 2012, 68, 9131–9138.
[33] F. G. Gervais, J. Biol. Chem. 1998, 273, 17102–17108.
[14] M. Drag, M. Bogyo, J. A. Ellman, G. S. Salvesen, J. Biol. Chem. 2010, 285,
3310–3318.
Received: April 24, 2014
Published online on August 25, 2014
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2014, 15, 2238 – 2247 2247