The use of biochemical and biophysical tools for triage of high-throughput screening hits - aa case study with escherichia coli phosphopantetheine adenylyltransferase

Article abstract of DOI:10.1111/j.1747-0285.2010.00957.x

High-throughput screening is utilized by pharmaceutical researchers and, increasingly, academic investigators to identify agents that act upon enzymes, receptors, and cellular processes. Screening hits include molecules that specifically bind the target and a greater number of non-specific compounds. It is necessary to 'triage' these hits to identify the subset worthy of further exploration. As part of our antibacterial drug discovery effort, we applied a suite of biochemical and biophysical tools to accelerate the triage process. We describe application of these tools to a series of 9-oxo-4,9-dihydropyrazolo[5, 1-b]quinazoline-2-carboxylic acids (PQ) hits from a screen of Escherichia coli phosphopantetheine adenylyltransferase (PPAT). Initial confirmation of specific binding to phosphopantetheine adenylyltransferase was obtained using biochemical and biophysical tools, including a novel orthogonal assay, isothermal titration calorimetry, and saturation transfer difference NMR. To identify the phosphopantetheine adenylyltransferase sub-site bound by these inhibitors, two techniques were utilized: steady-state enzyme kinetics and a novel ¹⁹F NMR method in which fluorine-containing fragments that bind the ATP and/or phosphopantetheine sites serve as competitive reporter probes. These data are consistent with PQs binding the ATP sub-site. In addition to identification of a series of PPAT inhibitors, the described hit triage process is broadly applicable to other enzyme targets in which milligram quantities of purified target protein are available.

Full text of DOI:10.1111/j.1747-0285.2010.00957.x

ª 2010 John Wiley & Sons A/S
doi: 10.1111/j.1747-0285.2010.00957.x
Chem Biol Drug Des 2010; 75: 444–454
Research Article
The Use of Biochemical and Biophysical Tools for
Triage of High-Throughput Screening Hits – A
Case Study with Escherichia coli
Phosphopantetheine Adenylyltransferase
J. Richard Miller1,2,*, Venkataraman
Key words: 19F NMR, antibacterials, biophysics, calorimetry, phos-
phopantetheine, screening, triage
Thanabal1,3, Michael M. Melnick1,4
,
Manjinder Lall1,3, Charles Donovan1,
Ronald W. Sarver1,5, Doh-Yeel Lee1,
Jeff Ohren1,3 and Don Emerson1,6
Received 28 October 2009, revised 19 January 2010 and accepted for
publication 28 January 2010
1Pfizer, Inc. Ann Arbor, MI 48105, USA
2Merck Research Laboratories Boston, MA 02115, USA
3Pfizer, Inc. Groton, CT 06340, USA
Increasing bacterial resistance to antibiotics has driven efforts to
find new agents that circumvent existing resistance mechanisms
(1,2). Most efforts have focused on one or more of the following
three approaches: (i) significant but incremental improvements of
existing antibiotic classes; (ii) identification of novel chemical matter
affecting an established antibacterial drug target (e.g. DNA topo-
isomerase, translation, etc.); or (iii) identification of inhibitors of
novel drug targets. The sequencing of multiple bacterial genomes
uncovered a wealth of potential targets that led many pharmaceuti-
cal and biotech companies to pursue this third option.
4Kalexsyn, Inc. Kalamazoo, MI 49008, USA
5Neogen, Inc. Lansing, MI 48912, USA
6Resonance Technologies, LLC Saline, MI 48176, USA
*Corresponding author: James R. Miller, James.Miller2@merck.com
High-throughput screening is utilized by pharma-
ceutical researchers and, increasingly, academic
investigators to identify agents that act upon
enzymes, receptors, and cellular processes.
Screening hits include molecules that specifically
bind the target and a greater number of non-spe-
cific compounds. It is necessary to ’triage’ these
hits to identify the subset worthy of further explo-
ration. As part of our antibacterial drug discovery
effort, we applied a suite of biochemical and bio-
physical tools to accelerate the triage process. We
describe application of these tools to a series of
9-oxo-4,9-dihydropyrazolo[5,1-b]quinazoline-2-car-
boxylic acids (PQ) hits from a screen of Escherichia
A commonly used method to identify small-molecule inhibitors of
enzyme drug targets is in vitro high-throughput screening (HTS), in
which collections of 105–106 individual compounds are tested for
their ability to affect the activity of the target enzyme. Recently,
public initiatives such as the Molecular Libraries Screening Centers
Network (part of the NIH Roadmap for medical research) (3) and
the efforts of institutions such as the McMaster University Screen-
ing lab (4), the Scripps Research Institute (5), St. Jude Children's
Research Hospital (6), and many others have made HTS readily
accessible to investigators outside of pharmaceutical companies.
coli
(PPAT). Initial confirmation of specific binding to
phosphopantetheine adenylyltransferase was
obtained using biochemical and biophysical tools,
including novel orthogonal assay, isothermal
phosphopantetheine
adenylyltransferase
In a typical HTS campaign against an enzyme target, an in vitro
assay is developed with the purified protein(s), the assay is opti-
mized and miniaturized, and is then used to assess the inhibitory
effect of the molecules comprising a screening library. Depending
on the target and assay, the output of such a HTS screen is typi-
cally a few hundred to few thousand compounds that reproducibly
inhibit the assay signal. This list of 'hits' provides the substrate for
further SAR (structure activity relationships) work. Unfortunately,
with even the best-designed screening assay, the activity of many
hits is not the result of a specific interaction with a desirable bind-
ing site on the target enzyme. Instead, the activity is attributable to
non-specific effects such as the small-molecule hit interfering with
the assay itself (e.g. quenching of the output signal, sequestering
of the enzyme's substrate, insolubility of inhibitor in assay buffer,
etc.). Alternately, some small molecules are promiscuous and
a
titration calorimetry, and saturation transfer dif-
ference NMR. To identify the phosphopantetheine
adenylyltransferase sub-site bound by these inhibi-
tors, two techniques were utilized: steady-state
enzyme kinetics and a novel 19F NMR method in
which fluorine-containing fragments that bind the
ATP and ⁄ or phosphopantetheine sites serve as
competitive reporter probes. These data are con-
sistent with PQs binding the ATP sub-site. In addi-
tion to identification of a series of PPAT inhibitors,
the described hit triage process is broadly applica-
ble to other enzyme targets in which milligram
quantities of purified target protein are available.
444

Use of Biochemical and Biophysical Tools for Triage
non-specifically inhibit the target through the formation of macro-
molecular aggregates (7). These aggregates may promote protein
unfolding and ⁄ or sequester the protein molecules and thereby inac-
tivate enzymes (8). In light of these and other undesirable mecha-
nisms of inhibition, the first significant challenge of a drug
discovery program emerging from HTS is to triage the list of hits to
identify the set of molecules that specifically bind the target and
are worthy of additional consideration.
To demonstrate their practical application, we describe use of these
tools within the context of an actual HTS triage effort around an
essential enzyme target within the Coenzyme A (CoA) biosynthetic
pathway, phosphopantetheine adenylyltransferase (PPAT). PPAT cata-
lyzes the penultimate step of the CoA biosynthetic pathway; trans-
fer of an AMP moiety from ATP to phosphopantetheine (PhP)
yielding dephosphocoenzyme A (dPCoA, Figure 1) and inorganic
pyrophosphate. Numerous crystal structures of PPAT with and with-
out substrate ligands are available (9,10) and the kinetic mechanism
of the E. coli enzyme has been investigated (11,12). Several other
positive features led to the selection of PPAT as a target for high-
throughput screening. These included a structurally attractive active
site, a high degree of PPAT conservation across various bacterial
species, distinct differences in active site residues between the
human and bacterial enzymes, ready availability of large quantities
of purified enzyme, and a straightforward enzyme-coupled kinetic
assay (12).
Over the past 15 years, Pfizer, and several of its legacy organiza-
tions (Warner-Lambert and Pharmacia Corporation), have conducted
more than 60 HTS campaigns against antibacterial targets. During
this time, a number of biochemical and biophysical assays were
developed and ⁄ or implemented that greatly accelerated HTS triage.
In light of the appealing attributes of PPAT as an antibacterial drug
discovery target, we performed a HTS of approximately one million
compounds to identify PPAT inhibitors. Among the inhibitors identi-
fied, the most promising was a class of ATP-competitive 9-oxo-4,
9-dihydropyrazolo[5,1-b]quinazoline-2-carboxylic acids (PQs). In this
manuscript, we describe their discovery, validation, and character-
ization using a combination of enzymological and biophysical triage
tools (Figure 2). In addition, we describe a novel adaptation of frag-
ment screening and 19F-NMR to determine the PQ binding sub-site.
Although applied to PPAT in this manuscript, the general triage
strategy and the tools we utilize are broadly applicable to other
enzyme-based drug discovery efforts.
Materials and Methods
Biochemical reagents
With the exception of phosphopantetheine and 7-methyl-8-thiogu-
anosine (MESG), all biochemical reagents were from Sigma–Aldrich
and were the highest purity available. Phosphopantetheine and
MESG were sourced as described previously (12).
Figure 1: Coenzyme A (CoA) biosynthetic pathway. aPPC-S and
PPC-DC are phosphopantothenoylcysteine synthetase and phospho-
pantothenoylcysteine decarboxylase, respectively. Both activities
reside on a single polypeptide in Escherichia coli.
Production of Escherichia coli PPAT
PPAT from E. coli was prepared as described previously (12).
Confirm molecular
interaction
Establish hit
tractability
Define binding site
& inhibition modality
Validate inhibition
Generate hit list
Figure 2: Sequential stages of an integrated high-throughput screening (HTS) hit triage utilizing biochemical and biophysical techniques.
Chem Biol Drug Des 2010; 75: 444–454 445

Miller et al.
Production of Enterococcus faecalis PPAT
IC50 determinations
The PPAT from Enterococcus faecalis strain V583 (ATCC# 700802D)
was amplified from genomic DNA and cloned into the pET15b deriv-
ative pPW2 (Affinium Pharmaceuticals, Toronto, Ontario, Canada)
and included both an N-terminal hexahistidine tag and a thrombin
cleavage site. Expression and purification of this protein was identi-
cal to that previously described for E. coli PPAT (12). Assays of
E. faecalis PPAT were performed as described for E. coli PPAT except
that the ATP concentration was doubled to 40 lM because of the
twofold higher Km(ATP) for the E. faecalis PPAT (data not shown).
In both assays, inhibitor potency was assessed by duplicate
10-point titrations of inhibitor from 100 to 1 lM. Inhibition data
were fit to the standard IC50 equation:
vi
vo
1
¼
ð1Þ
n
1 þ ð½Iꢀ=IC50Þ
where vi is the reaction velocity at a given concentration of inhibi-
tor [I], vo is the uninhibited velocity, and n is the Hill slope. IC50
assay replicates showed differences of £15%.
High-throughput screening assay (forward-
direction assay)
Isothermal titration calorimetry
The thermodynamics of PPAT ligand binding was determined using
a VP-isothermal titration calorimetry (ITC) (Microcal, Northampton,
MA, USA) as previously described (12) except that DMSO was
added to the cell to give an equivalent concentration to that in the
syringe when necessary. Ligands to be titrated were solvated at
30 mM in DMSO then diluted to 200–1000 lM in buffer (50-mM
Tris, pH 8.0, 0.5-mM TCEP). For ITC studies conducted at pH 5–7, a
50-mM acetate buffer was utilized in place of Tris.
PPAT screening assays were performed in clear 384-well plates
(Corning #3702) containing in each well either 1-lL DMSO (zero per-
cent inhibition control), 1 lL of variable concentrations of test com-
pounds solvated in DMSO, or 1 lL of a 100-lM solution of a
dipeptide inhibitor of PPAT [Compound 8 from (13)] in DMSO, which
serves as the fully inhibited control. To the entire plate, 20 lL of
assay solution 1 (Table S1) was added using a multidrop dispenser
(Titretek, Huntsville, AL, USA) and the plate was incubated for 5 min
at room temperature. The plate was delivered to a second multidrop
dispenser where 30 lL of solution 2 (Table S1) was added. The
plate was then taken to a Spectramax plate reader (Molecular
devices, Sunnyvale, CA, USA) and the change in absorbance at
360 nm was measured for 5 min. The rate of absorbance change
was corrected for the background rate observed in the fully inhibited
control. The corrected rate was then expressed as a percentage of
the corrected rate observed in the uninhibited (DMSO only) control.
Steady-state kinetic inhibition assays
Assays to determine inhibition modality were performed using the
same assay format as the forward-direction HTS assay described pre-
viously. Methodology was as described previously for product inhibi-
tion experiments (12) except that PQ inhibitors were utilized instead
of PPAT reaction products. Steady-state kinetic and inhibition data
were globally fit using the Enzyme Kinetics Module of Sigma Plot
(Jandel Scientific, San Rafael, CA, USA), which allows comparison of
fits to various inhibition schemes including competitive, non-competi-
tive, uncompetitive, etc. The best model was selected based on the
AIC.c [corrected Akaike information criterion, (14)], which allows direct
comparison of models with different numbers of fitted parameters.
Reverse (ATP-generating) PPAT assay using
firefly luciferase
The reverse assay was performed in polypropylene 96-well plates
(Corning #3365) containing in each well either 1-lL DMSO (zero
percent inhibition control), 1 lL of variable concentrations of test
compounds solvated in DMSO, or 1 lL of a 100-lM solution of a
dipeptide inhibitor of PPAT (Compound 8 from (13) in DMSO,
which serves as the fully inhibited control. To the entire plate,
20 lL of assay solution 1 (Table S2) was added using a multidrop
dispenser (Titretek) and the plate incubated for 5 min at room
temperature. The plate was delivered to a second multidrop dis-
penser where 30 lL of solution 2 (Table S2) was added. The plate
was then taken to a Beckman Multimek fitted with a 96-tip head.
Aliquots (5 lL) of the reaction mix were removed at four equally
spaced time intervals and transferred into a white 384-well plate
(Corning #3705) containing 80 lL of solution 3 (Table S2) in each
well. The plate was transferred to a Victor2 reader (Perkin Elmer,
Wallac Turku, Finland) and the luminescence was immediately
measured. The rate of change in luminescence was calculated
from the time-point data. The rate was then expressed as a per-
centage of the rate observed in the uninhibited (DMSO only) con-
trol. Following reading of the plate, ATP was added to each well
to give a final concentration of 20 lM and the luminescence was
measured again. Wells that did not show a dramatic increase in
luminescence were noted as interfering with the assay and were
not used to fit inhibition data.
NMR methods
471 MHz 19F(1H-decoupled) NMR spectra were collected at 25 ꢀC on
a Bruker DRX 500 MHz spectrometer equipped with a conventional
5-mm SEF probe optimized for 19F detection. The spectra were
collected using the CPMGC (PMG, Carr–Purcell–Meiboom–Gill) spin
echo pulse sequence (15,16)with a t value of 20 ms and a total spin
echo time of 320 ms. Data sets were the average of 128 scans.
Synthetic methods and spectral data
Synthetic procedures to produce analogues not previously described
in the literature (17,18) are provided in the supplementary material.
Results
Identification ⁄ characterization of PQs by
high-throughput screening
A previously described enzyme-coupled spectrophotometric phos-
phate detection assay (12) (Figure 3A) was adapted to a 384-well
446
Chem Biol Drug Des 2010; 75: 444–454

Use of Biochemical and Biophysical Tools for Triage
A
Figure 3: Enzyme-coupled pho-
sphopantetheine adenylyltransfer-
ase (PPAT) assays with different
methods of product detection. (A)
Production of pyrophosphate in the
forward-direction PPAT reaction is
coupled to cleavage of a chromo-
genic nucleoside. (B) Production of
ATP in the reverse-direction PPAT
reaction is coupled to the genera-
tion of light by firefly luciferase.
B
high-throughput format to permit large-scale screening for PPAT
inhibitors. The assay conditions were biased to identify ATP-compet-
itive PPAT inhibitors and ⁄ or PhP-competitive inhibitors and are
detailed in Table S1. The concentration of ATP used in the screen
was 30 lM, well below the Km(ATP) of 220 lM (12). The concentra-
tion of PhP in the screen was held constant at 20 lM, fivefold
higher than the Km of PhP (12). Although this concentration of PhP
was high enough to exclude very weak PhP-competitive inhibitors
as hits, this was the minimum concentration of PhP needed to
maintain a linear rate of product formation across the desired read-
time of 5–10 min (data not shown).
erably less sensitive than the HTS assay in Figure 3A (12). Further-
more, NADH formation is detected at a wavelength 20 nm lower
than the primary assay and can be more susceptible to interference
by inhibitor absorbance.
To avoid these issues and to provide another means to validate the
PQ inhibitors, we developed a novel PPAT assay utilizing firefly lucifer-
ase to quantify the ATP generated by the PPAT reverse reaction (i.e.
using dPCoA and inorganic pyrophosphate as substrates, Figure 3B).
In the presence of ATP, firefly luciferase exhibits 'flash' kinetics with
luminescence peaking then rapidly declining (19). However, when the
luciferase inhibitor CoA is present, 'glow' kinetics are observed,
where the peak luminescence is reduced but luminescence longevity
is dramatically increased (19,20). This property is often employed in
commercially available ATP-detection kits that advertise a 'stable'
luminescence signal. These assays are configured such that the lumi-
nescence produced is proportional to the absolute ATP concentration
of a quenched reaction mixture. This results in a discontinuous assay
and therefore kinetic analysis of the PPAT reaction requires measure-
ment of ATP concentrations at multiple time-points.
Approximately, 1 million compounds were screened individually
against PPAT. Several classes of inhibitors were identified, with the
pyrazoloquinolones (PQs) shown in Table 1 being the most interesting.
Members of this class of PQ inhibitors have previously been investi-
gated for their antiallergic activity in animal models (17,18). The PQ
series exhibited a range of E. coli PPAT IC50 values (Tables 1 and 2)
suggesting the series possesses exploitable SAR. A plate-based neph-
elometer system (BMG LABTECH, Offenburg, Germany) was employed
alongside these IC50 assays to ensure that IC50s were well separated
from the solubility limits of the compounds (data not shown).
We utilized laboratory automation to capture four equally spaced
time-points during the PPAT reverse reaction, enabling us to deter-
mine ATP concentrations and calculate the rate of ATP production.
However, as the ATP concentrations of these samples needed to be
analyzed after the samples were collected, a method to quench the
PPAT reaction was required. Initially, we hypothesized that we could
use CoA to both inhibit PPAT (12) and produce luciferase 'glow'
kinetics. Inclusion of CoA at concentrations that fully inhibited the
PPAT reverse reaction resulted in an unstable luminescence signal
(data not shown) and was cost prohibitive for large-scale screening.
We instead utilized high concentrations of PhP, which is a product
inhibitor (12) of the PPAT reverse reaction, as a quenching agent.
The final concentration of PhP utilized (200 lM) does not inhibit or
activate luciferase (21) but does quench PPAT reactions sufficiently
well that the luciferase signal (i.e. the ATP concentration) is stable
for the 5–10 min required to read the luminescence of the assay
PQs inhibit PPAT in a reverse (ATP-generating)
PPAT assay
Hits identified by HTS can exhibit inhibition because of non-specific
interference with the assay modality as described previously. To
build confidence that the PQs are legitimate PPAT inhibitors, an
alternative plate-based assay format was developed. E. coli PPAT
has an equilibrium constant of approximately 1 (12), allowing use
of the reverse reaction (ATP-generating) as a physiologically rele-
vant alternative assay. In this assay, ATP is produced from dPCoA
and inorganic pyrophosphate. We (12) and others (11) have coupled
the PPAT reverse reaction to NADH production using hexokinase
and glucose-6-phosphate dehydrogenase. PPAT turnover in this
assay results in an increase in absorbance at 340 nm, but is consid-
Chem Biol Drug Des 2010; 75: 444–454
447

Miller et al.
Table 1: In vitro inhibition of Escherichia coli and Enterococcus faecalis phosphopantetheine adenylyltransferase (PPAT) by 4- and 7-substi-
tuted pyrazoloquinolones
O
1
8
X
N
N
7
6
2
CO2H
4
3
N
5
H
Forward direction assay
Reverse direction assay
Compd
X
IC50 (lM) E. coli PPATa
IC50 (lM) E. faecalis PPATa
IC50 (lM) E. coli PPAT
6b
H
CH3
Cl
F
Br
I
SCH3
SOCH3
SO2CH3
SCH(CH3)2
N(CH3)2
OCH3
253
244
17
103
6.2
5.1
13
166
92
14
>167
NTe
>75
>167
209
>137
>65
20
>230
NT
NT
NT
22
7b
8b,d
9b
10c
11c
12b,d
13b
14b
15b
16b
17b
18b
11
8.8
NT
NT
5.0
7.1
NT
NT
NT
249
157
>167
NT
49
>150
160
O
N
Cl
N
N
CO2H
aData with a greater than (>) sign indicate that PPAT inhibition was less than 50% at the highest tested or highest soluble concentration.
bCompound was present in the Pfizer compound library at the time of high-throughput screening (HTS).
cCompound was newly synthesized to expand series structure activity relationships (SAR).
dCompound was resynthesized to confirm activity.
eNT denotes compound was not tested.
plate. Also it is important to note that pyrophosphate, a PPAT
reverse reaction substrate, is an inhibitor of luciferase luminescence
production at concentrations utilized in the assay (19). As pyrophos-
phate was in large excess of the amount of ATP formed, the
change in pyrophosphate concentration during the assay was negli-
gible and thus the inhibitory effect on luciferase was uniform
across all samples. Finally, a small amount of CoA was included to
ensure 'glow' kinetics was maintained.
between the two assays. However, because of the differences in
the ratios of substrate concentrations to the substrate Km values,
there is no expectation that the absolute IC50 values be identical
between the two assays.
Selectivity and specificity of the PQ inhibitors
Given our interest in discovering selective broad-spectrum antibacte-
rial agents, it was important to determine the potency of the PQ
inhibitors against a PPAT from a divergent yet medically relevant
pathogen. As the primary screen was performed against the PPAT
from the Gram-negative E. coli, we selected the E. faecalis PPAT as
a reasonable representative of Gram-positive PPAT enzymes. The
E. faecalis enzyme displays only 44% protein sequence identity to
the E. coli PPAT. Interestingly, with the exception of 13 and 16, all
compounds were inactive versus E. faecalis PPAT (Tables 1 and 2).
Selectivity against the human PPAT is also critically important, and
none of the PQs tested inhibited the human PPAT at concentrations
up to 100 lM (data not shown).
As some small-molecule inhibitors quench and ⁄ or inhibit luciferase
luminescence and would therefore appear to inhibit the PPAT reac-
tion, we also screened for this interference by adding a fixed
amount of ATP (20 lM final concentration) to every well of the
assay plate after reading the luciferase luminescence resulting from
the PPAT reverse reaction. The amount of ATP added is in excess to
that generated in the PPAT reaction, but well below that which
would saturate the luciferase enzyme (data not shown). Compound-
specific quenching was readily apparent as these wells did not
show a dramatic increase in luminescence upon ATP addition.
The PQs inhibited PPAT in this reverse reaction assay (Tables 1 and
2) and did not inhibit luciferase (data not shown), indicating that
their inhibition was not because of interference with the primary
assay. For most analogues tested in the reverse PPAT assay (exclud-
ing compound 16), the rank order potencies corresponded well
Newly resynthesized PQs show PPAT inhibitory
activity
Although considerable effort is made to ensure the accurate iden-
tity and structural integrity of compounds in screening files, it is
448
Chem Biol Drug Des 2010; 75: 444–454

Use of Biochemical and Biophysical Tools for Triage
Table 2: In Vitro inhibition of Escherichia coli and Enterococcus faecalis phosphopantetheine adenylyltransferase (PPAT) by 3- and 6-substi-
tuted pyrazoloquinolones
O
1
8
N
N
7
2
CO2H
X 6
4
3
N
5
H
Forward direction assay
Reverse direction assay
Compound
X
IC50 (lM) E. coli PPATa
IC50 (lM) E. faecalis PPATa
IC50 (lM) E. coli PPAT
6b
H
253
15
54
121
124
49
>167
>167
>167
100
>167
212
>230
NTd
200
NT
>125
93
19c
20c
21b
22b
23b
24b
CH3
CF3
NH2
Br
Cl
O
106
>167
NT
N
N
CO2H
Cl
N
H
aData with a greater than (>) sign indicate that phosphopantetheine adenylyltransferase (PPAT) inhibition was less than 50% at the highest tested or highest
soluble concentration.
bCompound was present in the Pfizer compound library at the time of high-throughput screening (HTS).
cCompound was newly synthesized to expand series structure activity relationships (SAR).
dNT denotes compound was not tested.
not uncommon for structures to be incorrectly assigned or for sam-
ples to have partially degraded – particularly if they are many years
old (e.g. the PQ analogues in our file were approximately 20 years
old). Additionally, contaminants from the synthetic route (i.e. purifi-
cation resins, heavy metals, etc.) can be present. In each of these
cases, the contaminants could be the actual inhibitory agent. To
rule this out, it is critical to resynthesize representative or close-in
analogues using modern and preferably orthogonal synthetic
schemes that minimize or eliminate these potential contaminants.
SAR of PQ inhibitors
Although a number of PQ analogues were present in the Pfizer corpo-
rate collection, we wished to further explore the SAR of this series
and additional analogues were synthesized using the route described
in Figure 4. Table 1 summarizes the SAR for analogues that were var-
ied in the 7- and 4-positions. When X is hydrogen, the IC50 against
E. coli PPAT was found to be 253 lM. Interestingly, the activity
improved by a factor of 50-fold when either the proton, Me or F
groups, which were essentially equipotent, were replaced with either
bromine or iodine, with the 7-iodo displaying an IC50 against E. coli
PPAT of 5 lM. This suggests that substituents in the 7-position may
be interacting with a carbonyl oxygen in the PQ binding pocket
through an induced dipole effect (22,23). Replacement of the proton
with a SMe group also resulted in an approximately 20-fold increase
in activity over the unsubstituted analogue and this was maintained
when the methyl was replaced with an isopropyl group. However,
replacement of the SMe group with OMe or oxidation of SMe pro-
duced an approximately 10-fold loss in activity. This finding also lends
credence to the induced dipole interaction hypothesis. Replacing the
The chemistry reported by Sircar (17,18) and shown in Figure 4 was
utilized to prepare additional analogues. The route used substituted
anthranilic acids 1 as the starting material. Conversion of 1 to isatoic
anhydride 2 was readily accomplished with triphosgene in THF. Reac-
tion of 2 with excess hydrazine hydrate produced hydrazide 3 that
could be reacted with 4 followed by base hydrolysis to yield the final
PQs 5 in acceptable yield. Several PQ analogues were resynthesized
as described in Figure 4 and were observed to inhibit PPAT in both
the forward and reverse-direction assays (Tables 1 and 2).
Figure 4: Reagents and conditions for synthesis of pyrazoloquinolones (PQs): (a) triphosgene, tetrahydrofuran at room temperature; (b)
excess hydrazine hydrate; (c) H2O, reflux, followed by Na2CO3.
Chem Biol Drug Des 2010; 75: 444–454
449

Miller et al.
proton with a dimethylamine in the 7-position was moderately toler-
ated. Replacement of the proton at the 4-position of the 7-chloro ana-
logue resulted in a nearly 10-fold decrease in potency, indicating the
requirement for a hydrogen bond donor at this position.
Table 2 summarizes the SAR for analogues that were varied in the 6-
and 3-positions. In general, adding moderate bulk in the 6-position,
e.g. Me or CF3 resulted in improved potency over the unsubstituted
PQs. The 6-position seems to reside in a somewhat smaller and more
hydrophobic pocket as opposed to the 7-position. For example, the Br
and amine substituents were essentially equipotent and possessed
IC50 values that were about half that of the smaller Cl or CF3 groups.
Unfortunately, despite exhaustive efforts, we were not able to deter-
mine a co-crystal structure or even generate a plausible docking
model for the PQs that were able to explain all the SAR. It is likely
that the PQs bind toward the extremity of the ATP binding site as
tested compounds were competitive with ATP but not with AMP (data
not shown), but a conclusive evaluation of the PQ binding site in PPAT
will have to wait for additional structural studies.
PPAT binds PQs
When newly resynthesized hits show inhibition in the primary and
orthogonal assay(s), there is increased confidence in their validity.
However, these data do not demonstrate specific interaction with
the target protein. To provide evidence that PQs directly interact
with PPAT, we utilized ITC to examine the interaction of compound
12. In the same buffer and pH conditions used in the enzyme
assays, 12 bound the apo-enzyme with a dissociation constant (Kd)
of 24.4 € 0.7 lM (Figure 5), similar to the IC50 in the enzyme assays
(Table 1). Binding to PPAT was enthalpically driven
()5.2 € 0.1 kcal ⁄ mol) and exhibited 1:1 binding stoichiometry, simi-
lar to ATP, PhP, and dPCoA (12).
Figure 5: Binding of 12 to Escherichia coli phosphopantetheine
adenylyltransferase (PPAT) as monitored by isothermal titration calo-
rimetry (ITC). The top panel shows the change in calorimeter power
required to maintain constant temperature after each injection of
12. The bottom panel shows the integrated heats of binding cor-
rected for heats of dilution, and the solid line is the best fit of the
data to a single-site binding model.
exhibited sufficient chemical attractiveness ⁄ ligand efficiency to pur-
sue with additional medicinal chemistry resources (data not shown).
However, several hits contained trifluoromethyl substituents and
their binding to PPAT can be readily monitored by 19F NMR (see
below).
PQs are ATP competitive inhibitors of PPAT
Previously discussed evidence argues that PQs are specific inhibi-
tors of PPAT but do not provide information about the modality of
inhibition. Therefore, PPAT inhibition by PQs was examined by
steady-state kinetics using the HTS-format assay (Figure 3A). When
increasing concentrations of PQ inhibitor 11 were tested against
PPAT at variable concentrations of one substrate and fixed, saturat-
ing concentrations of the other substrate, 11 was competitive with
ATP (Ki of 5.1 € 0.6 lM) as shown in Figure 6A. At saturating con-
centrations of ATP, 11 had no significant effect on the rate of PPAT
turnover across a wide range of phosphopantetheine concentrations
(Figure 6B). The Ki value of 11 determined from this study was
comparable to other assessments of binding potency (Table 1).
Previous work by other authors has shown that weakly binding fluo-
rine-containing molecules can serve as 'spy' probes that can be
competed off by a more tightly bound ligand (24) and that fluorine-
containing fragment libraries have potential for fragment-based drug
discovery (25).The use of 19F NMR to monitor ligand–protein inter-
actions provides simplified NMR spectra without interference from
the proton signals (24,26) We sought to take advantage of the
favourable NMR properties of fluorine to validate the binding of
the PQ inhibitors to PPAT.
The fluorine-containing probes 25 and 26 (Figure 7) identified in
the STD-NMR fragment screen interact with PPAT as gauged by 19F
NMR (Figure 8). Both compounds have PPAT IC50s greater than
100 lM in all assays (data not shown). Despite their weak affinity,
a CPMG spin echo based pulse sequence showed that they are dis-
placed upon addition of ATP and ⁄ or phosphopantetheine, respec-
tively (Figure 8).
Use of trifluoromethyl-containing fragments to
probe inhibitor binding sites
As part of our efforts to identify novel PPAT inhibitors, we also per-
formed a saturation transfer difference nuclear magnetic resonance
spectroscopy (STD-NMR) screen of a small library of low molecular
weight fragments (general properties of the fragment library are
discussed in the supplementary material). This screen identified a
number of molecules that interacted with PPAT. None of these
We have utilized these weakly binding CF3-containing fragments as
reporter probes to classify compound 10 as an ATP site binder. Fig-
ure 8A shows the 19F (1H decoupled) spectra of 25 and 26. The
450
Chem Biol Drug Des 2010; 75: 444–454

Use of Biochemical and Biophysical Tools for Triage
A
Figure 7: Active site probe fragments for use in phosphopant-
etheine adenylyltransferase (PPAT) 19F NMR competition experi-
ments.
is possible that 25 binds PPAT in a position that overlaps both the
ATP and the PhP pocket. Figure 9 demonstrates using these tools to
classify the PQ hits. 19F spectra of both ATP and PhP reporter frag-
ments alone and in the presence of PPAT are shown in Figures 9A
and 9B respectively. Addition of 10 to a solution containing PPAT,
25, and 26 selectively displaces only the ATP probe 25 as evi-
denced by the appearance of its 19F signal at )76.56 ppm (Fig-
ure 9C). As observed above, the peak height of 25 does not fully
return upon addition of 10, providing additional evidence for a sec-
ond non-active site interaction of 25 with PPAT.
B
These data (in conjunction with the steady-state inhibition kinetics
described previously) strongly suggest that the PQs interact with
the ATP pocket of PPAT. However, it is possible that the PQs may
act in an allosteric fashion; binding outside the active site and
remotely affecting the interaction of ATP and ⁄ or fragment 25.
PQ binding affinity is pH dependent and
implications for Structure Based Drug Design
(SBDD)
Figure 6: Steady-state kinetics of phosphopantetheine adenylyl-
transferase (PPAT) inhibition by 11. The assay shown in Figure 3A
was used to determine the rate of PPAT turnover at variable con-
centrations of substrate and inhibitor. (A) The concentration of ATP
was varied at a constant saturating concentration of PhP (200 lM).
The rates of PPAT turnover were then measured at the following
concentrations of compound 11: 0 lM (filled circles), 3 lM (unfilled
circles), 6 lM (filled triangles), 12 lM (unfilled triangles). (B) PPAT
rates with variable concentrations of PhP and saturating concentra-
tions of ATP (2 mM). Concentrations of 11 are as shown in panel
A.
Despite considerable effort, we were unable to obtain co-structures
of PQ inhibitors with PPAT. The conditions under which PPAT readily
crystallizes are acidic (pH <6), whereas our assay conditions are at
pH 8.0. To determine whether this significant difference in pH could
explain our inability to obtain co-structures, we used ITC to mea-
sure the binding affinity of a representative PQ as a function of pH
(Table 3). Although changes in affinity could be as a result of
changes in the ionization state of the inhibitor or the protein, PQ
12 exhibited no detectable binding at pH 6.0. Similarly, the potent
PPAT inhibitor CoA exhibited a 100-fold weaker binding potency at
pH 6.0 versus that at pH 8.0 as measured by ITC. Although signifi-
cantly weaker at the lower pH, CoA binds with sufficient affinity to
obtain a PPAT:CoA co-crystal structure (10).
two signals at )76.56 and )76.08 correspond to the CF3 groups of
fragments 25 and 26, respectively. Upon addition of PPAT, these
19F signals disappear because of the faster relaxation of the CF3
group that occurs upon binding to PPAT (Figure 8B). Addition of ATP
to a mixture of PPAT and compounds 25 and 26 displaces only
compound 25 as evidenced by the selective appearance of the 19F
signal at )76.56 ppm (Figure 8C). Note that the 19F peak of com-
pound 25 does not return to its original intensity. This may indicate
that 25 weakly binds PPAT at an additional site that is not affected
by addition of substrate. Addition of PhP to the mixture of PPAT,
25, and 26 displaces both fragments as evidenced by the appear-
ance of both the 19F signals (Figure 8D). The ATP sub-site specific
signal at )76.56 again exhibits only partial intensity recovery indi-
cating that 25 is not fully competed out by PhP, whereas 26 is. It
Conclusions and Future Directions
The data and examples we provide demonstrate tools and tech-
niques that can be used to increase confidence in the validity of
hits from HTS campaigns. The essential characteristics of a tracta-
ble screening hit are: (i) does not interfere with the primary assay;
(ii) structurally intact and retains activity upon resynthesis; (iii)
close-in analogues possess a range of potency indicating definable
SAR; (iv) specifically interacts with the target enzyme at a discrete
site. Our general approach to establishing these criteria is depicted
in Figure 2 and the quantitative outcome of applying these
Chem Biol Drug Des 2010; 75: 444–454
451

Miller et al.
Figure 8: 19F NMR spectra of fluorine-containing fragments with and without addition of phosphopantetheine adenylyltransferase (PPAT)
and ⁄ or PPAT substrates. (A) 471 MHz 19F(1H-decoupled) NMR spectrum of 100 lM each of 25 and 26 in 10-mM HEPES-d18, pH 8.0, 25 ꢀC.
(B) Addition of 4-lM (final concentration) PPAT causes loss of the indicated peaks because of the fast relaxation that results from PPAT bind-
ing. (C) Addition of 400-lM ATP to the sample in (B) selectively displaces the fragment 25 as evidenced by the appearance of the signal at
)76.56 ppm. (D) Addition of 200-lM PhP to the sample in (B) displacing both 25 and 26 as evidenced by the appearance of signals at
)76.08 and )76.56 ppm. All the spectra were collected using the CPMG spin echo sequence with a total spin of 320 ms. 19F chemical shifts
were referenced with respect to trifluoroethanol.
Figure 9: 19F NMR spectra of fluorine-containing fragments with and without addition of phosphopantetheine adenylyltransferase (PPAT)
and ⁄ or pyrazoloquinolone (PQ) inhibitors. (A) 471 MHz 19F(1H-decoupled) NMR spectrum of 100 lM each of 25 and 26 in 10-mM Hepes-d18,
pH 8.0, 25 ꢀC. (B) Addition of 4-lM PPAT causes the peaks designated by arrows to disappear, due to fast relaxation as a result of binding to
PPAT. (C) Addition of 100 lM of 10 to the sample in (B). Compound 10 selectively competes with the ATP probe fragment 25 as evidenced
by the appearance of the signal at )76.56 ppm.
452
Chem Biol Drug Des 2010; 75: 444–454

Use of Biochemical and Biophysical Tools for Triage
Table 3: Dependence of CoA and pyrazoloquinolone (PQ) binding
affinity on pH as determined by isothermal titration calorimetry
to reinforce the importance of collecting data for several members
of a high-priority compound series rather than just the most potent
analogue. Although we have presented these studies with different
analogues, all of the featured analogues exhibited favourable bind-
ing data in each test (data not shown).
Kd (lM)
pH
Coenzyme A (CoA)
Compound 12
8.0
7.0
6.0
5.0
0.09
4.6
10
24
22
>200
>200
Often, practical considerations such as limited availability of ana-
logues, poor inhibitor solubility, or incompatibility of a target with
one or more tools ⁄ assays preclude confirming all validation criteria
for each series of hits. When a full suite of assay data is not avail-
able for every compound of interest, the data in hand can be used
to 'bin' hit series (or singletons) into high-, medium-, or low-priority
groups for follow-up. Other project-specific considerations, such as
selectivity and attractive physicochemical properties are also impor-
tant and should receive due consideration.
72
approaches to the PPAT screening hits is shown in Figure 10.
Although elimination of the majority of hits occurred in the early
stages of triage (primary and orthogonal ⁄ secondary assays), the use
of biophysical approaches such as STD-NMR and isothermal titra-
tion calorimetry eliminated all but a few unique compounds (often
referred to as singletons) or series of related compounds that
required more intensive enzymological characterization and synthetic
follow-up.
The PQ series of PPAT inhibitors meet the four points mentioned
previously and are therefore suitable for further exploration by syn-
thetic medicinal chemistry. They possess several attractive physico-
chemical and pharmacokinetic properties (data not shown) such as
a high aqueous solubility, low LogD (-2), stability in the presence of
human liver microsomes, no significant cytochrome P-450 inhibition,
and do not inhibit the PPAT reaction of the bi-functional human CoA
synthase. This selectivity against the human ortholog is not surpris-
ing given the low conservation between the two genes (data not
shown). The PQs were not observed to inhibit other ATP-binding
proteins such as protein kinases based on their failure to 'hit' these
targets in numerous previous HTS campaigns (data not shown).
Despite these positive features, the compounds exhibit a dramati-
cally pH-dependent binding profile and lack co-crystal structure(s) to
guide medicinal chemistry efforts. They also do not possess anti-
bacterial activity, likely because of their modest potencies against
PPAT or their inability to penetrate the bacterial cell wall. This sort
of 'mixed' profile for a compound series (i.e. some supporting data
and some inconclusive data) is common for hits emerging from HTS
and decisions must be made about the risk of proceeding with
these assets and liabilities in place.
We have used a variety of tools including a novel application of 19F
NMR and fragment screening to establish the previously described
four validation criteria for the PQ series of PPAT inhibitors. We
chose to present kinetic, thermodynamic, and NMR binding data for
three different PQs rather than data for the most potent compound
742 294 compounds screened
1533 confirmed hits
with additional sample (>5 mg)
available
147 chemically ‘attractive’ hits1
with IC50 <200 µM in
coupled PNP/MESG assay (Figure 3A)
99 Hits1 with IC50 <200 µM in
luciferase assay (Figure 3B)
References
47 Hits1 with correct mass ion by LC/MS and consistent NMR spectrum
1. Payne D.J., Gwynn M.N., Holmes D.J., Pompliano D.L. (2007)
Drugs for bad bugs: confronting the challenges of antibacterial
discovery. Nat Rev Drug Discov;6:29–40.
2. Silver L.L. (2005) A retrospective on the failures and successes
of antibacterial drug discovery. IDrugs;8:651–655.
3. Austin C.P., Brady L.S., Insel T.R., Collins F.S. (2004) NIH Molecu-
lar Libraries Initiative. Science;306:1138–1139.
4. Brown E.D. (2003) Screening in academe: a perspective on
implementation of university-based small molecule screening. J
Biomol Screen;8:377–379.
28 Hits1 confirmed to bind by NMR and/or ITC
- 13 hits with no SAR depth
- 9 hits reacted with PhP
- 3 synthetically intractable or did
not confirm upto resynthesis
3 distinct series remained
- 1 ATP competitive (PQs)
- 1 irreversible PhP site binder
- 1 with mixed inhibition against
ATP and PhP
5. Sheppard T.L. (2007) Scripps Florida. Nat Chem Biol;3:451.
6. Chen T., Guy R.K. (2008) High-Throughput Screening and Drug
Discovery at St. Jude Children's Research Hospital. Cell
Notes;22:3–4.
1
In the noted cases, the term hits counts clusters of structurally related
compounds (such as the pyrazoloquinolones) as a single hit
Figure 10: Quantitative detail of the phosphopantetheine ade-
nylyltransferase (PPAT) high-throughput screening (HTS) triage pro-
cess.
7. McGovern S.L., Caselli E., Grigorieff N., Shoichet B.K. (2002) A
common mechanism underlying promiscuous inhibitors from vir-
tual and high-throughput screening. J Med Chem;45:1712–1722.
Chem Biol Drug Des 2010; 75: 444–454
453

Miller et al.
8. Coan K.E., Maltby D.A., Burlingame A.L., Shoichet B.K. (2009)
Promiscuous aggregate-based inhibitors promote enzyme unfold-
ing. J Med Chem;52:2067–2075.
21. Ford S.R., Buck L.M., Leach F.R. (1995) Does the sulfhydryl or
the adenine moiety of CoA enhance firefly luciferase activity?
Biochim Biophys Acta;1252:180–184.
9. Izard T. (2002) The crystal structures of phosphopantetheine ade-
nylyltransferase with bound substrates reveal the enzyme's cata-
lytic mechanism. J Mol Biol;315:487–495.
22. Voth A.R., Khuu P., Oishi K., Ho P.S. (2009) Halogen bonds as
orthogonal molecular interactions to hydrogen bonds. Nat
Chem;1:74–79.
10. Izard T. (2003) A novel adenylate binding site confers phospho-
pantetheine adenylyltransferase interactions with coenzyme A. J
Bacteriol;185:4074–4080.
23. Lu Y., Shi T., Wang Y., Yang H., Yan X., Luo X. et al. (2009) Hal-
ogen Bondingꢀ-A Novel Interaction for Rational Drug Design?
J Med Chem;52:2854–2862.
11. Geerlof A., Lewendon A., Shaw W.V. (1999) Purification and
characterization of phosphopantetheine adenylyltransferase from
Escherichia coli. J Biol Chem;274:27105–27111.
12. Miller J.R., Ohren J., Sarver R.W., Mueller W.T., de Dreu P.,
Case H. and Venkataraman Thanabal (2007) Phosphopantetheine
adenylyltransferase from Escherichia coli: investigation of the
kinetic mechanism and role in regulation of coenzyme A biosyn-
thesis. J Bacteriol;189:8196–8205.
24. Dalvit C., Fagerness P.E., Hadden D.T.A., Sarver R.W., Stockman
B.J. (2003) Fluorine-NMR experiments for high-throughput
screening: theoretical aspects, practical considerations, and
range of applicability. J Am Chem Soc;125:7696–7703.
25. Vulpetti A., Hommel U., Landrum G., Lewis R., Dalvit C. (2009)
Design and NMR-Based Screening of LEF, a Library of Chemical
Fragments with Different Local Environment of Fluorine. J Am
Chem Soc;131:12949–12959.
13. Zhao L., Allanson N.M., Thomson S.P., Maclean J.K., Barker J.J.,
Primrose W.U. et al. (2003) Inhibitors of phosphopantetheine
adenylyltransferase. Eur J Med Chem;38:345–349.
26. Dalvit C. (2007) Ligand- and substrate-based 19F NMR screen-
ing: principles and applications to drug discovery. Prog Nucl
Magn Reson Spectrosc;51:243–271.
14. Akaike H. (1976) An information criterion (AIC). Math Sci;14:5–9.
15. Carr H.Y., Purcell E.M. (1954) Effects of Diffusion on Free Pre-
cession in Nuclear Magnetic Resonance Experiments. Phys
Rev;94:630–638.
16. Meiboom S., Gill D. (1958) Modified spin-echo method for mea-
suring spin-relaxation rates. Rev Sci Intrum;29:688–691.
17. Sircar J.C., Capiris T., Kesten S.J., Herzig D.J. (1981) Pyrazol-
o[5,1-b]quinazolin-9-ones: a new series of antiallergic agents. J
Med Chem;24:735–742.
Supporting Information
Additional supporting information may be found in the online ver-
sion of this article.
Data S1. Synthesis and analytical assessment of compounds.
Table S1. Composition of solutions used in the forward direction
PPAT HTS assay.
18. Sircar J.C., Capiris T., Kesten S.J. (1981) Reactions of 2H-3,1-
Benzoxazine-2,4-(1H)dione (Isatoic Anhydride) with Anions of 1,4-
Dihydro-5H-Pyrazol-5-ones: Synthesis of Pyrazolo [5,1-b]quinazo-
lin-9-ones. J Heterocycl Chem;18:117–121.
Table S2. Composition of solutions used in the reverse-direction
PPAT assay.
19. Ford S.R., Chenault K.H., Bunton L.S., Hampton G.J., McCarthy
J., Hall M.S., Pangburn S.J., Buck L.M., Leach F.R. (1996) Use of
firefly luciferase for ATP measurement: other nucleotides
enhance turnover. J Biolumin Chemilumin;11:149–167.
20. Airth R.L., Rhodes W.C., Mc Elroy W.D. (1958) The function of
coenzyme A in luminescence. Biochim Biophys Acta;27:519–532.
Please note: Wiley-Blackwell is not responsible for the content or
functionality of any supporting materials supplied by the authors.
Any queries (other than missing material) should be directed to the
corresponding author for the article.
454
Chem Biol Drug Des 2010; 75: 444–454