Chemical proteomics-based drug design: Target and antitarget fishing with a catechol-rhodanine privileged scaffold for NAD(P)(H) binding proteins

Article abstract of DOI:10.1021/jm8002284

Drugs typically exert their desired and undesired biological effects by virtue of binding interactions with protein target(s) and antitarget(s), respectively. Strategies are therefore needed to efficiently manipulate and monitor cross-target binding profiles (e.g., imatinib and isoniazid) as an integrated part of the drug design process. Herein we present such a strategy, which reverses the target → lead rational drug design paradigm. Enabling this approach is a catechol-rhodanine privileged scaffold for dehydrogenases, which is easily tuned for affinity and specificity toward desired targets. This scaffold crosses bacterial (E. coli) cell walls, and proteome-wide studies demonstrate it does indeed bind to and identify NAD(P)(H)-binding proteins that are potential drug targets in Mycobacterium tuberculosis and antitargets (or targets) in human liver. This approach to drug discovery addresses key difficulties earlier in the process by only pursuing targets for which a chemical lead and optimization strategy are available, to permit rapid tuning of target/antitarget binding profiles.

Full text of DOI:10.1021/jm8002284

J. Med. Chem. 2008, 51, 4571–4580
4571
Chemical Proteomics-Based Drug Design: Target and Antitarget Fishing with a
Catechol-Rhodanine Privileged Scaffold for NAD(P)(H) Binding Proteins
Xia Ge,† Bassam Wakim,‡ and Daniel S. Sem*,†
Department of Chemistry, Chemical Proteomics Facility at Marquette, P.O. Box 1881, Marquette UniVersity, Milwaukee, Wisconsin 53201, and
Medical College of Wisconsin, Milwaukee, Wisconsin 53226
ReceiVed March 3, 2008
Drugs typically exert their desired and undesired biological effects by virtue of binding interactions with
protein target(s) and antitarget(s), respectively. Strategies are therefore needed to efficiently manipulate and
monitor cross-target binding profiles (e.g., imatinib and isoniazid) as an integrated part of the drug design
process. Herein we present such a strategy, which reverses the target f lead rational drug design paradigm.
Enabling this approach is a catechol-rhodanine priVileged scaffold for dehydrogenases, which is easily
tuned for affinity and specificity toward desired targets. This scaffold crosses bacterial (E. coli) cell walls,
and proteome-wide studies demonstrate it does indeed bind to and identify NAD(P)(H)-binding proteins
that are potential drug targets in Mycobacterium tuberculosis and antitargets (or targets) in human liver.
This approach to drug discovery addresses key difficulties earlier in the process by only pursuing targets for
which a chemical lead and optimization strategy are available, to permit rapid tuning of target/antitarget
binding profiles.
The drug discovery process is costly and often inefficient.1
Genomics and proteomics advances have presented the promise
of improving efficiency, but this has largely translated into the
identification of new drug targets, not new drugs. What is needed
is a better coupling of the chemistry of drug design to advances
in genomics and proteomics. To partially address this, chemical
genetic approaches have been developed,2,3 where enzyme
inhibitors are used to knock out protein function. One advantage
of chemical genetics over traditional genetics is that besides
providing phenotypic data in the context of a whole organism,
it yields an inhibitor for subsequent optimization in the drug
discovery process. Still, this process is problematic in two ways:
(a) one cannot be certain that the inhibitor binds only to the
intended target, and (b) it is highly inefficient because new
inhibitors must be designed for each new target of interest. The
first question is relevant because binding to other proteins
(antitargets) can lead to toxic side effects. Further complicating
matters, in other cases such as imatinib4–6 (4-[(4-methyl-1-
piperazinyl)methyl]-N-[4-methyl-3-[[4-(3-pyridinyl)-2-pyrimidi-
nyl]amino]phenyl]benzamide methanesulfonate) and isoniazid,7
off-target binding is actually thought to contribute to drug
efficacy, thereby calling into question the one-target/one-drug
dogma that serves as the foundation for rational drug design.
The second question is relevant because the process of designing
potent inhibitors that are acceptable drug leads can take years
and varies in difficulty from protein to protein, being nearly
impossible for some protein targets, leading to the notion of
“druggable” protein targets.8,9 There is a vital need to identify
“druggable targets” (those for which potent and selective
inhibitors can be designed) as early in the drug discovery process
as possible. To address this second concern, compounds can
be designed based on “privileged scaffolds”,10–13 which are
druglike14,15 molecular structures that provide baseline affinity
for a whole protein family. These scaffolds then serve as starting
points for optimization against specific protein targets of interest
in the family, usually by building a focused combinatorial library
off of the scaffold. To this end, privileged scaffolds have been
reported for kinases,16 proteases,17,18 and GPCRs.19,20 We have
recently reported the first privileged scaffold for NAD(P)(H)-
binding proteins,21 based on a catechol-rhodanine ring system.
Proteins in this family include the oxidoreductases (aka dehy-
drogenases), with drug targets such as HMG-CoA reductase
(statin drugs), steroid-5R-reductase (finasteride), aldose reductase
(diabetes), and a large number of infectious disease targets,22,23
including enoyl CoA reductase, deoxyxylulose-5-phosphate
reductoisomerase (DOXPR), and dihydrodipicolinate reductase
(DHPRa); this family even includes enzymes other than oxi-
doreductases, such as sirtuins, ADP-ribosylating enzymes, and
ligases. The catechol-rhodanine privileged scaffold has served
as a template for building biligand libraries, where the ligand
attached to the scaffold is situated in the substrate pocket,
thereby giving specificity to a particular enzyme in the family
(Figure 1). It has been used to generate multiple potent (Kd e
200 nM) and selective inhibitors for dehydrogenases, including
DHPR and DOXPR,21 with affinity and selectivity readily tuned
by varying the fragment attached to the scaffold.
Despite the power of this scaffold, it has never been properly
verified as being specific for NAD(P)(H)-binding enzymes in a
proteome-wide manner. This is because a strategy was not
available to assess cross-reactivity (off-target binding) with other
family members, in the context of a whole proteome, whether
for the catechol-rhodanine scaffold itself or for biligand drug
leads constructed from it, for specific targets. This gets to the
a Abbreviations: ADME, absorption, distribution, metabolism, and excre-
tion; CRAA, catechol-rhodanine acetic acid; DCC, N,N′-dicyclohexylcar-
bodiimide; DCU, dicyclohexylurea; DHPR, dihydrodipicolinate reductase;
DMAP, 4-(dimethylamino)pyridine; ESI, electrospray ionization; HPLC,
high performance liquid chromatography; IPTG, Iisopropyl ꢀ-D-1-thioga-
lactopyranoside; LC-MS, Liquid Chromatography-Mass Spectrometry;
LTQ, Linear Trap Quadrupole; NHS, N-hydroxysuccinimide ester; NMR-
SOLVE, structurally oriented library valency engineering; PAGE, poly-
acrylamide gel electrophoresis; PBS, phosphate buffered saline; SDS,
sodium dodecyl sulfate; TB, tuberculosis; Tris, tris(hydroxymethyl)ami-
nomethane; TEMED, tetramethylethylenediamine.
* To whom correspondence should be addressed. Phone: 414-288-7859.
Fax: 414-288-7066. E-mail: Daniel.Sem@marquette.edu.
†
Marquette University.
Medical College of Wisconsin.
‡
10.1021/jm8002284 CCC: $40.75
2008 American Chemical Society
Published on Web 07/11/2008

4572 Journal of Medicinal Chemistry, 2008, Vol. 51, No. 15
Ge et al.
herein, with the caveat that the o-catechol may need to be
replaced if there is any toxicity.
The chemical proteomic strategy proposed herein also relies
on attaching a dehydrogenase-specific ligand to a resin, using
that affinity column with subsequent digestion of the eluted
proteins, and subjecting the tryptic peptides to electrospray LC/
MS followed by searching the MS/MS data against an appropri-
ate subset of the Uniprot database to identify all (reasonably
abundant) proteins in a proteome that bind the ligand. While
affinity purification using native cofactor has been applied to
dehydrogenases for over 30 years,31–33 it has never been coupled
to tandem MS to probe binding profiles for a dehydrogenase-
targeted privileged scaffold. And more broadly, although there
is emerging interest in using affinity chromatography coupled
to mass spectrometry to probe protein-ligand interactions across
a proteome,34 there is a need for more efficient coupling of this
assay methodology earlier in the drug design process, using the
chemical leverage provided by privileged scaffolds to create
an integrated drug discovery approach that blends (a) a broad
assessment of target/antitarget binding profiles, (b) a pragmatic
selection of druggable targets, and (c) an efficient chemical
strategy for tuning target/antitarget affinity. This paper presents
a foundation for such a strategy, applied using the first such
privileged scaffold for NAD(P)(H) binding proteins.
Figure 1. Catechol-rhodanine privileged scaffold (CRAA, 1) and its
use in creating biligand inhibitors with high affinity and specificity for
specific dehydrogenase targets.
first concern mentioned above. Recent advances in chemical
proteomics6,24,25 now permit proteome-wide binding studies of
the scaffold (and later, of biligands) by covalently attaching
scaffold to a resin, binding all protein family members in a
proteome sample, then eluting with free scaffold (or biligand)
and identifying proteins with tandem MS. Such a strategy was
recently used to assess binding profiles for currently prescribed
drugs, such as imatinib4–6 and isoniazid.7 Both of these drugs
were thought to bind tightly to a single target, and it was later
discovered that their biological efficacy might actually be due
to binding to multiple targets. The strategy and tools presented
in this paper would now permit the assessment and optimization
of cross-target binding profiles (target/antitarget) across a
proteome as an integral part of the drug design process; in this
manner, binding profiles could be correlated with biological
efficacy up front in a rational manner rather than relying on
serendipitous and unknown off-target effects.
The strategy proposed herein depends crucially on the
availability of a privileged scaffold that binds to a protein family
(dehydrogenases, in this case) and that has been designed in
such a way that it can be quickly modified to produce potent
inhibitors for a given family member (building biligands, in this
case). The latter has already been verified21 for the privileged
scaffold that is the topic of this paper, which is based on the
catechol-rhodanine acetic acid 1 (CRAA) shown in Figure 1.
But is this scaffold a viable starting point for drug discovery?
While the thiazolidine ring (rhodanine is a type of thiazolidine)
has been reported by Poupaert et al.26 as a frequently occurring
heterocyclic motif in anti-inflammatory, antipsychotic, and
anticonvulsant drugs, the rhodanine ring is less common. But
it does occur in drugs such as epalrestat ((2-[(5Z)-5-[(E)-3-
cyclohexyl-2-methylprop-2-enylidene]-4-oxo-2-thioxo-3-thiazo-
lidinyl]acetic acid)), a potent inhibitor of aldose reductase (AR),
and has been shown to have no significant toxicity in recent
clinical trials.27,28 The catechol group, though present in a
number of plant-derived natural products, can have toxicity
in some cases when it is oxidized to an o-quinone, which can
then alkylate cellular macromolecules or generate reactive
oxygen species.29,30 As such, 1 does seem to be a viable scaffold
upon which to build drug leads, using the strategies described
Results
2 (NHS-CRAA) Uptake into E. coli Cells and Labeling
of DHPR. To assess whether 1 (CRAA) can make it across
bacterial cell walls and therefore whether 1 is a viable scaffold
for anti-infective drug design efforts, experiments were per-
formed to determine if its N-hydroxysuccinimide ester, 2 (NHS-
CRAA, Figure 2), could enter E. coli and label intracellular
DHPR (dihydrodipicolinate reductase). DHPR is an anti-
infective drug target and is known to bind the CRAA scaffold
(1).21 The NHS-CRAA active ester (2)35 was synthesized as in
Figure 2. The NHS (N-hydroxysuccinimide) group reacts with
amines, and since it is attached to the linker position of 1 (where
the acetic acid chain is attached), it should reside at the interface
of the NADH and substrate binding sites21 (Figure 1), near lysine
16336 (see Supporting Information Figure S1). Indeed, DHPR
is labeled with the NHS-CRAA (2) active ester, based on
imaging of an SDS-PAGE gel of labeled protein (Supporting
Information Figure S1). Labeling is partially blocked by NADH
(Figure S2), indicating that the NHS-CRAA (2) probe is in fact
binding and labeling (at least partially) in the active site of
DHPR. Next, to explore whether cell wall penetration occurs,
uptake of 2 was measured into E. coli that was expressing E.
coli DHPR. Since we have recently shown that the CRAA
scaffold (1) is bifunctional, in that it is itself weakly fluores-
cent,33 protein labeling could be monitored by fluorescence
imaging of SDS-PAGE gels of crude cell extracts. In-cell
studies were performed by overexpressing DHPR in E. coli,
then exposing intact cells to the NHS-CRAA probe (2), washing
and lysing cells, and then running an SDS-PAGE gel to see if
the probe was able to covalently label the DHPR (Figure 3a).
Any NHS-CRAA probe (2) that was nonspecifically associated
with the cells was quenched by treatment with 100 mM Tris
before lysis. Gels show that significant labeling of the DHPR
did occur within the intact E. coli cells (Figure 3b), indicating
that scaffold can cross the cell wall in order to gain access to
DHPR. That there is more labeling in cells expressing DHPR
is evident based on the more intense color for cells expressing
DHPR (Figure 3c) and on fluorescence images of the cells
(Figure 4a).

Chemical Proteomics-Based Drug Design
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 15 4573
Figure 2. Synthesis of the NHS-CRAA active ester (2) and CRAA aminohexylagarose matrix.
Figure 3. (a) E. coli uptake study. Cartoon representation of the uptake study demonstrating that NHS-CRAA ester (2) can cross the E. coli cell
wall to react with overexpressed DHPR and other intracellular proteins. (b) SDS-PAGE analysis of the crude cell lysate from the experiment in
part a. Lane 1: protein marker. Lanes 2 and 4: lysate of cells with DHPR present (+IPTG). Lanes 3 and 5: lysate of cells without DHPR present
(-IPTG). Lanes 2 and 3 were fluorescently scanned using a Kodak Image Station. Lanes 4 and 5 were scanned with a CanonScan D1250U2F
document scanner after Coomassie blue staining, used with the same Gel. (c) Vials of E. coli cells just prior to lysis, showing the CRAA-associated
color change in the cells containing overexpressed DHPR (left) relative to those without DHPR (right).
CRAA (1) Affinity Chromatography and Nanospray-
LC/MS/MS. Our proteome fishing studies (Figure 5) require a
resin with a privileged scaffold, in this case 1, covalently
attached. The NHS-CRAA (2) active ester (Figure 2) was used
to prepare this affinity resin to permit purification of dehydro-
genase (NAD(P)(H)-binding protein) subproteomes from protein
mixtures. Crude cell lysates from E. coli and M. tuberculosis
were both loaded onto the affinity column, and proteins were
eluted using free CRAA (1) probe, as shown in Figure 5.
SDS-PAGE analysis of both microbial samples showed very
similar patterns (Figure 6a), although analysis of proteins that
were identified from M. tuberculosis (vide infra, Table 2)
indicates that some proteins have no E. coli homologues, and
for those that do, masses differ. So the apparent similarity in
gel patterns is more likely due to the prevalence of proteins in
these molecular weight ranges in both microbes. As with the
microbial samples, human liver proteins were loaded onto the
affinity resin and eluted with 1 (Supporting Information and
Figure 6b), to determine the CRAA-binding profile for the liver
proteome. It is noteworthy in this gel, which shows wash and
elution fractions, that proteins are specifically eluted by free 1.
For both human liver and M. tuberculosis eluents, nanospray-
LC/MS/MS analysis was performed followed by searching the
MS/MS data against an appropriate subset of the Uniprot
database to determine which CRAA-binding proteins were
present in reasonably high abundance. To complement this

4574 Journal of Medicinal Chemistry, 2008, Vol. 51, No. 15
Ge et al.
Figure 4. E. coli fluorescence labeling: (a) fluorescence and (b) bright field images of E. coli cells containing overexpressed DHPR, after incubation
with NHS-CRAA (2) and subsequent washing with PBS. A 100× objective and 495 nm/520 nm excitation/emission filters were used.
Figure 5. Proteome fishing with a privileged scaffold. Cartoon representation of how the CRAA (1) affinity column is used for target fishing in
a proteome pool either to initially identify potential targets and antitargets (top branch) or to later characterize the binding profile for a particular
biligand drug lead candidate (bottom branch). The top branch also demonstrates how one assesses whether a privileged scaffold really is targeting
a gene family (as intended), such as NAD(P)(H) binding proteins.
Figure 6. SDS-PAGE analysis of the proteome fishing experiments described in Figure 5. (a) SDS-PAGE analysis of CRAA-captured proteins
from E. coli and M. tuberculosis H37Rv proteomes. Lane 1: protein marker. Lane 2: E. coli crude cell lysate. Lane 3: E. coli column wash. Lanes
4-6: E. coli fractions after elution with free CRAA (1). Lane 7: M. tuberculosis crude cell lysate. Lane 8: M. tuberculosis column wash. Lanes
9-11: M. tuberculosis fractions after elution with free CRAA (1). (b) SDS-PAGE analysis of CRAA-captured proteins from human liver proteome.
Lane 1: Protein marker. Lane 2, crude cell lysate. Lanes 3-5: column wash. Lanes 6-12: fractions after elution with free CRAA (1). (c) SDS-PAGE
analysis of human liver and M. tuberculosis H37Rv protein fractions after CRAA (1) affinity column chromatography, showing the protein bands
that were cut and extracted for nanospray-LC/MS/MS proteomic analysis. Samples identical to those loaded on the gel were polymerized in a gel
piece as described in Methods and subjected to whole proteomic analyses. Labeled bands A-D are referred to in Table 3 (database search results).
Gels were silver-stained.
whole proteome (actually subproteome) analysis, CRAA-eluted
fractions were also separated using SDS-PAGE, and protein
bands at ∼35 and ∼55 kDa (Figure 6c) were in-gel-digested.
Then peptides were extracted from the gel and analyzed as
above. In both cases, proteins were first digested with trypsin,
then zip-tip-cleaned and injected into an LC-MS system (LTQ
with a linear ion trap from Thermo-Fisher). Whole subproteome
analyses are in Table 1 (human liver) and Table 2 (M.
tuberculosis), while analysis of extracted bands is in Table 3.
Complete data sets, even for very low scoring hits, are given in
Supporting Information (Figures S8 and S9). Generally, LC/
MS data indicate that >50% of these proteins are dehydroge-
nases or other NAD(P)(H) binding proteins, as expected. Better
identifications were obtained from the human liver sample,
perhaps because the M. tuberculosis sample workup involves
irradiation, which may cause some protein damage. In any case,

Chemical Proteomics-Based Drug Design
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 15 4575
Table 1. Mass Spectrometry-Based Identification of Proteins Identified in the Target Fishing Study Using the CRAA Affinity Column: Human Liver
Proteomea
accession number (gi)
6648067
annotated function (human)
molecular weight (Da)
35 531
coverage (%)
52
score
1557
malate dehydrogenase,
mitochondrial precursor
1346343
113611
keratin, type II cytoskeletal 1
66 018
39 473
23
24
852
643
fructose-bisphosphate aldolase B
(liver-type aldolase)
118504
118541
118495
aldehyde dehydrogenase,
mitochondrial precursor
56 381
61 398
54 862
28
29
22
526
523
377
glutamate dehydrogenase 1,
mitochondrial precursor
retinal dehydrogenase 1 (aldehyde
dehydrogenase family 1 member
A1)
81175178
59802911
keratin, type I cytoskeletal 9
(cytokeratin-9) (CK-9) (keratin-9)
(K9)
62 129
98 829
30
15
346
252
10-formyltetrahydrofolate
dehydrogenase (aldehyde
dehydrogenase family 1 member
L1)
a Pyridine nucleotide (NAD(P)(H)) binding proteins are indicated in bold. Annotation is directly from the database.
Table 2. Mass Spectrometry-Based Identification of Proteins Identified in the Target Fishing Study Using the CRAA Affinity Column: M. tuberculosis
Proteomea
accession number (gi)
81671721
annotated function (Mycobacterium tuberculosis)
possible oxidoreductaseb
molecular weight (Da)
33 220
coverage (%)
30
score
80
54036852
2829534
chaperone protein clpB
92 568
35 366
15
8
31
26
riboflavin biosynthesis protein ribD [includes
diaminohydroxyphosphoribosylaminopyrimidine deaminase]
1706274
bifunctional enzyme cysN/cysC [includes sulfate adenylyltransferase
subunit 1]
67 839
3
20
81671959
81340808
putative uncharacterized proteinc
putative uncharacterized proteind
30 296
38 520
30
16
17
13
a Pyridine nucleotide (NAD(P)(H)) binding proteins are indicated in bold. Annotation is directly from the database. Of the M. tuberculosis nucleotide
(NAD(P)(H)) binding proteins indicated, two have homologues in E. coli: gi81671721 is similar to the E. coli protein gi75240619 (Mr ) 42 233 g/mol), and
gi2829534 is similar to the E. coli protein gi75230139 (Mr ) 39 456 g/mol). b A subsequent NCBI search indicates this protein has homology to coenzyme
F420-dependent N5,N10-methylene tetrahydromethanopterin reductase or other flavin-dependent oxidoreductases. c A subsequent NCBI search indicates
this protein has homology to 17-ꢀ-hydroxysteriod dehydrogenase and to hydratase-dehydrogenase-epimerase. It contains the R-hydratase-like hot dog fold.
Other proteins with this fold include fatty acid synthase beta subunit and MaoC dehydratase. d Protein of closest homology with annotated function, based
on a BLAST search (E ) 10-24), is the nitroreductase from Burkholderia dolosa (gi:124901246). Enzymes in this family catalyze the NAD(P)H dependent
reduction of flavin or nitro compounds using FMN or FAD as cofactor.
several M. tuberculosis proteins could be identified with high
certainty. Analysis of extracted bands was intended as a check
on the whole subproteome analyses, although in general there
was lower signal-to-noise (and, as a consequence, scores) for
these samples. Still, there is generally good agreement between
extracted band data and whole subproteome analysis, especially
when scores are higher (>10) and percent coverage of the
protein sequence is more complete (g7%).
Of the highest scoring human liver proteins (Table 1), 5 out
of 6 (excluding keratin, a very abundant protein) were dehy-
drogenases. The top hit, malate dehydrogenase, has more than
50% peptide coverage and a very high score, while glutamate,
aldehyde, and retinal dehydrogenases also had high percent
coverage (>20%). Binding of 1 to two of these dehydrogenases
(glutamate and malate) was subsequently verified experimentally
in NMR STD (saturation transfer difference) binding assays
(Supporting Information Figure S7). Other dehydrogenases that
appear to bind 1, based on lower but still statistically significant
scores and percent coverage, include (Supporting Information
Figure S8) alcohol dehydrogenases, isocitrate dehydrogenase,
R-aminoadipic semialdehyde dehydrogenase (gi116241244), and
NADP-dependent leukotriene B4 12-hydroxydehydrogenase
(gi23503081). It is noted that for tandem MS analysis of bands
at around 55 and 35 kDa, isocitrate/aldehyde and malate
dehydrogenases, respectively, are again identified with high
certainty, confirming that they do indeed bind to 1.
As with the liver proteins, M. tuberculosis proteins were
bound to the CRAA-affinity resin, then eluted with free CRAA
(1), and fractions were analyzed using electrospray LC/MS/
MS. Of the highest scoring (score >13) M. tuberculosis proteins
(Table 2), there were four possible pyridine nucleotide-binding
proteins out of six proteins identified. The other proteins bind
ATP, so the CRAA scaffold (1) may have some modest affinity
for ATP binding sites as well. Interestingly, three of the proteins

4576 Journal of Medicinal Chemistry, 2008, Vol. 51, No. 15
Ge et al.
Table 3. Mass Spectrometry-Based Identification of Proteins Captured in the Target Fishing Study Using the CRAA Affinity Column: Analysis of
Proteins Extracted from Bandsa
accession number (gi)
annotated function (human)
molecular weight (Da)
coverage (%)
score
(A) Human Liver Proteins in Band A (45-60 kDa)
74762137
21903432
tubulin ꢀ-2A chainb
49 906
46 659
26
24
148
84
isocitrate dehydrogenase [NADP] cytoplasmic (cytosolic
NADP-isocitrate dehydrogenase)b
118504
aldehyde dehydrogenase, mitochondrial precursor (ALDHclass 2)b
56 381
50 135
16
15
36
19
55977864
tubulin R-1A chain (tubulin B-R-1) (tubulin R-3 chain) (R-tubulin 3)b
(B) Human Liver Proteins in Band B (30-35 kDa)
malate dehydrogenase, mitochondrial precursorb
6648067
35 531
7
3
accession number (gi)
annotated function (Mycobacterium tuberculosis)
molecular weight (Da)
coverage (%)
score
(C) Mycobacterium tuberculosis Proteins in Band C (50-65 kDa)
2829534
1706274
2497387
riboflavin biosynthesis protein ribD
35 366
67 839
45 494
8
3
2
4
3
2
bifunctional enzyme cysN/cysCb
putative transposase Rv3428c
(D) Mycobacterium tuberculosis Proteins in Band D (30-35 kDa)
81671721
81668779
1706274
2829534
possible oxidoreductaseb
33 220
60 268
67 839
35 366
27
2
78
7
exopolysaccharide phosphotransferase cpsY
bifunctional enzyme cysN/cysC
3
3
riboflavin biosynthesis protein ribDb
8
3
a Pyridine nucleotide (NAD(P)(H)) binding proteins are indicated in bold. b Only these proteins have the correct mass for the extracted band(s).
had no annotated function, but a subsequent NCBI search (i.e.,
updated annotation) and BLAST alignments identified the
closest homologues to in fact be NAD(P)(H) binding proteins.
This highlights the potential value of CRAA (1) target fishing
in functional proteomic efforts, by even capturing uncharacter-
ized proteins and providing suggestive data on their cofactor
binding preferences, as well as the start of a chemical genetic
probe.
such a probe for NAD(P)(H)-binding proteins,21 which is shown
in Figure 1. The study presented herein extends this work by
(a) showing that this scaffold is able to cross bacterial cell walls
(Figures 3 and 4), (b) demonstrating that it truly is a privileged
scaffold for NAD(P)(H) binding proteins (Tables 1–3) based
on proteome-wide profiling (Figures 5 and 6), and (c) identifying
potential drug targets (Table 2) and antitargets (Table 1) to be
pursued in future drug design efforts, coupled with proteome-
wide profiling studies (Figure 5).
Discussion
With regard to penetrating bacterial cell walls, uptake studies
were performed by monitored labeling of intracellular proteins
using the CRAA (1) privileged scaffold tethered to an amine-
reactive reagent. This is effectively an activity-based probe,
analogous to those described for other protein families in the
field of chemical proteomics38,39 but not yet reported for
NAD(P)(H) binding proteins. The attachment point for the NHS
group was chosen to be at the end of the CRAA (1) linker, in
the position that is normally proximal to or in the substrate site
(Figure 1), so it will only label proteins that have an amine in
that position. Fortunately (and by design), DHPR has an amine
in this position, and so it could be labeled. Incubation of purified
DHPR with NHS-CRAA (2) does in fact lead to covalent
labeling (Supporting Information Figures S1 and S2). Labeling
is also observed if intact E. coli cells that are overexpressing
DHPR are exposed to probe (Figures 3 and 4). This could only
happen if the probe can cross the cell wall, providing unambigu-
ous evidence that there is nothing about the CRAA scaffold
(1) that would inherently preclude transport across cell walls.
Of course, cell wall penetration will vary significantly depending
on the bacteria in question and is based on what is attached to
the CRAA (1) linker (as in Figure 1), but these results are
encouraging in that at least in some cases cell wall penetration
will be possible.
The methods presented herein were developed to address two
of the major roadblocks in drug design projects and in the
development of chemical genetic probes (i.e., functional ge-
nomics): (a) there is a need to know the binding profile (target,
antitarget binding) for a molecule as broadly as possible, whether
it is a privileged scaffold that targets many proteins in a gene
family or a highly specific drug lead intended for one protein,
and (b) there is a desperate need to speed up chemistry by
including, integral to this process, a strategy for rapid tuning
of a binding profile; this is accomplished by using a privileged
scaffold that can be rationally modified to target a protein of
interest (Figure 1). A central element of this pragmatic approach
to drug discovery is to make sure that protein targets are only
pursued if a druglike inhibitor is already in hand, which can be
rationally modified for higher affinity with minimal effort. This
addresses up front the common concern over whether a protein
target is “druggable”. One additional drug discovery challenge,
in the case of anti-infectives, is the formidable barrier of needing
to cross the microbial cell wall. The studies presented herein
present a strategy to drug discovery that attempts to address all
of these problems, with a focus on NAD(P)(H)-binding proteins.
The approach relies on the availability of a privileged scaffold
that targets a gene family and that is easily modified to achieve
higher affinity for a given target. We have previously described

Chemical Proteomics-Based Drug Design
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 15 4577
Next, to assess whether the CRAA scaffold (1) is a privileged
scaffold for NAD(P)(H)-binding proteins and to identify po-
tential target and antitarget proteins, crude cell lysates from E.
coli and M. tuberculosis were both loaded onto the affinity
column and proteins eluted using free CRAA (1) probe (Figure
5). SDS-PAGE analysis of both microbial samples showed very
similar patterns (Figure 6a), suggesting some overlap in their
dehydrogenase subproteomes and corresponding binding profiles
for the CRAA privileged scaffold (1), although some of this
apparent overlap may also be due to the prevalence of proteins
in this molecular weight range. Because Mycobacterium tuber-
culosis is of greater interest as a drug target,40,41 proteomics
studies were pursued to identify potential targets in its proteome.
Interestingly, 4 out of the top 6 scoring proteins from the
Mycobacterium tuberculosis proteome were NAD(P)(H)-binding
proteins, although this was not obvious based on the initial
annotation of the database (3 out of the 4 hits were for proteins
of undefined function). An analogous study with a human liver
proteome sample also resulted in the identification of proteins
that were mostly (>50%) NAD(P)(H)-binding proteins. Given
that most proteomes comprise <5% dehydrogenases,23 the
CRAA scaffold (1) appears to have good selectivity for this
gene family. Now, any proteins that were identified in either
the human or Mycobacterium tuberculosis proteomes have a
baseline affinity for 1, so more potent inhibitors could easily
be made for a target of interest using the biligand design strategy
outlined in Figure 1 and previously validated.21 Only pursuing
protein targets for which the start of a potent inhibitor/drug lead
is available is highly pragmatic because it identifies “druggable”
targets at the start of the drug discovery process. But are any
of the identified proteins in Tables 1 and 2 drug targets, and/or
are they worth pursuing as targets of chemical genetic probes
for basic research objectives (i.e., functional genomics)?
prevented using antiepileptic drugs, they can be avoided by
treatment with pyridoxine.46 So it appears that AASD is an
antitarget to be avoided. But any potential problems from
transient inhibition of AASD are likely to be less severe than
the genetic knockout just described and in any case could be
alleviated by treatment with pyridoxine. Conversely, the CRAA
scaffold (1) could be used as a starting point for designing a
more potent inhibitor of AASD (Figure 1) for chemical genetic
studies in model organisms that contain close homologues of
human AASD, such as zebrafish (gi27882244), rat (gi149064286),
and Xenopus (gi51703516).
If any human proteins are to be pursued as drug targets,
specificity should be checked against the other metabolically
important dehydrogenases listed in Table 1 to avoid toxicity
and to achieve an acceptable therapeutic index. So important
outcomes of the human proteome data are (a) a list of targets
that could be pursued in subsequent drug discovery efforts,
especially for diabetes (aldose reducatse) and inflammation
(NADP-dependent leukotriene B4 12-hydroxydehydrogenase),
(b) a list of human antitargets for these drug discovery efforts,
and (c) a proteomics-based strategy for assessing binding profiles
(described in Figure 5) to assess off-target effects. Finally, these
data confirm that 1 is behaving as a privileged scaffold for
dehydrogenases, in the context of the human liver proteome.
Toward the goal of using the CRAA privileged scaffold (1)
in anti-infective drug discovery efforts, analogous proteome
fishing studies were performed using crude cell lysates from
Mycobacterium tuberculosis. As with the human liver proteins,
M. tuberculosis proteins were bound to the CRAA-affinity resin,
then eluted with free 1, and fractions were analyzed using
tandem MS. Of the highest scoring M. tuberculosis proteins
(score >13, Table 2), there were four possible pyridine
nucleotide-binding proteins out of six proteins identified.
Interestingly, three of the captured proteins had no annotated
function, but subsequent NCBI searches and BLAST alignments
identified the closest homologues to be NAD(P)(H)-binding
proteins; this highlights the value of CRAA-based proteome
fishing in functional proteomic efforts, even providing the start
of a chemical genetic probe to later explore function. There is
also some likelihood that one or more of these proteins could
be drug targets. For example, the top scoring protein in Table
2 has high homology to a coenzyme F420-dependent N5,N10-
methylene tetrahydromethanopterin reductase. Coenzyme F420
was first discovered in methanogenic archaea47,48 and is now
known to be present in mycobacteria. Indeed, Daniels has
suggested that targeting of F420-dependent enzymes might be
pursued as a new strategy for killing mycobacteria.49 RibD,
another Mycobacterium tuberculosis hit (Table 2), is essential
for synthesis of riboflavin. While this may not be a viable drug
target, a potent inhibitor of RibD would provide a chemical
knockout to complement genetic knockouts of RibD (such
mutants are riboflavin auxotrophs50), to explore function. One
potential application might be to create transient vitamin B2
auxotrophy if one wanted to incorporate isotopically labeled
riboflavin into a microbially expressed protein. Finally, the two
“putative uncharacterized proteins” in Table 2 are also of interest
not just as potential drug targets but because chemical genetic
probes might help to better define their function. One of these
proteins has the highest homology to 17-ꢀ-hydroxysteroid
dehydrogenase/hydratase-dehydrogenase-epimerase, but very
little is known about the role of 17-ꢀ-ketodehydrogenases in
microbes. The human homologue (17-ꢀ-hydroxysteroid dehy-
drogenase) is involved in the synthesis of estradiol from estrone
and so is a target for breast cancer and endometriosis.51 What
Any drug designed to be an anti-infective would need to be
optimized to not disrupt function of vital proteins in the human
proteome. And since the liver is the body’s first line of defense
(after passage through the intestinal mucosa) before drugs go
into the general circulation, proteome profiling was done against
the human liver proteome. Of the human liver proteins identified
(Table 1), 5 out of 6 (excluding keratin) were dehydrogenases.
In terms of antitargets of concern, any drug leads designed using
the CRAA (1) privileged scaffold (Figure 1) should certainly
be tested against malate, glutamate, isocitrate, and the various
aldehyde dehydrogenases listed in Table 1 and in Supporting
Information. It is also noted that some of these proteins may
prove to be useful targets for human disease in their own right.
In this regard, it is interesting that CRAA (1) has affinity for
various aldehyde dehydrogenases. This is perhaps not surprising
because the drug epalrestat, also known as ONO-2235,27,28 also
contains a rhodanine core and is an aldose reductase inhibitor
used to treat diabetes. Indeed, this suggests that our CRAA core
(1) might be used as a starting point for building other aldose
reductase inhibitors, with different and tunable off-target binding
profiles. Another human enzyme that may bind 1 is NADP-
dependent leukotriene B4 12-hydroxydehydrogenase, which is
involved in eicosanoid inactivation and is a target of indometha-
cin42 as well as other nonsteroidal anti-inflammatory drugs
(NSAIDs).43 Our proteome fishing data suggest that 1 might
also be pursued as a starting point for inhibitors of this enzyme
by properly tuning affinity based on what fragments are added
to the scaffold (Figure 1). Another enzyme that may bind 1 is
R-aminoadipic semialdehyde dehydrogenase (AASD). Genetic
deficiency in AASD is known to cause pyridoxine-dependent
epilepsy.44,45 While seizures in such individuals cannot be

4578 Journal of Medicinal Chemistry, 2008, Vol. 51, No. 15
Ge et al.
cinimide Ester).55 Under a N2 atmosphere, a mixture of 6.22 g of
1, 5.75 g of N-hydroxysuccinimide, 20.6 g of DCC, 50 mL of
DMSO, and a small amount of DMAP catalyst was reacted at room
temperature overnight. The next day the reaction was monitored
by NMR (Supporting Information Figure S6) and with TLC (using
EMD silica gel 60 F254 developed with chloroform/methanol/acetic
acid, 12:3:1 v/v/v), visualized using a 254 nm UV light (the Rf of
1 is 0.39 and the new spot’s Rf is 0.68). The DCU (dicyclohexy-
lureo) was vacuum-filtered off, and the NHS-CRAA (2, Figure 2)
DMSO solution was used in the next step without further
purification.
Synthesis of CRAA (1) Agarose Matrix (5-[(3,4-Dihydroxy-
phenyl)methylene]-4-oxo-2-thioxo-3-thiazolidineacetic ω-Ami-
nohexylagarose Amide).56 NHS-CRAA ester (2) DMSO solution
was added dropwise into 100 mL of ω-aminohexylagarose sus-
pended in 600 mL of 100 mM phosphate buffer, pH 10.0. During
this process, the pH was maintained at 10.0, and then the reaction
was run at 7 °C in a refrigerator overnight. The next day, 60 mL
of 1 M Tris-HCl buffer, pH 6.5, was added to the reaction mixture
to stop the reaction. Then 47.7 g of sodium chloride was added to
form a final 0.5 M saline solution. The liquid layer was decanted,
and the labeled matrix (Figure 2) was washed with a large amount
of deionized water. Then ∼10 mL of matrix was packed into a 1
cm × 20 cm column for column chromatography.
metabolic role the microbial enzyme plays, and whether it is a
viable drug target, is not known52 but could certainly be probed
with chemical genetic probes based on the CRAA scaffold (1).
The other uncharacterized protein identified in Table 2 is in
the nitroreductase family. Purkayastha et al.53 have noted that
nitroreductases may play a role in helping mycobacteria respond
to different host conditions; for example, a nitroreductase is up-
regulated when mycobacteria are inside the macrophage.
Because mycobacteria survive and multiply inside macroph-
ages,54 it is important to better understand the enzymes that are
up-regulated and perhaps facilitate their survival in this environ-
ment. Dissecting this regulatory regulatory cascade might
uncover new drug targets and could possibly provide a better
basic understanding of how the bacteria can hide within the
host’s own defense system. Higher affinity ligands constructed
off the CRAA scaffold (1) would minimally serve as chemical
genetic probes and perhaps even as drug leads.
Of the scaffold-binding M. tuberculosis proteins identified,
it is certainly not yet known which (if any) will be useful drug
targets because of a lack of proper annotation. But the above
discussion points out an especially useful feature of the CRAA
probe (1): it can be used both as a platform for drug design and
for development of chemical genetic probes. That is, biligands
designed with specificity for these proteins of unknown function
could then be used to explore phenotypic effects of a chemical
knockout. If the phenotypic effect suggests a mechanism to kill
the microbe, then at least there is the start a drug lead in hand.
The intention of this study, then, is to prepare a foundation for
future drug design and chemical genetic initiatives by providing
a chemical scaffold for optimization (CRAA, 1), a strategy for
using it to generate new and potent biligand inhibitors (Figure
1), a proteome-wide method for assessing target and antitarget
binding broadly (Figures 5 and 6) to correlate with phenotypic
effects, and a list of targets and antitargets in both human and
Mycobacterium tuberculosis to begin pursuing (Tables 1–3).
NHS-CRAA Ester (1) in In-Cell Uptake and Labeling
Study.57 E. coli containing the pET11a DHPR expression construct
was inoculated into 30 mL of LB culture medium, growing
overnight at 37 °C, 225 rpm. The next day, 10 mL of this culture
was added to two flasks (flasks A and B, each containing 800 mL
of LB medium with 50 µg/mL carbenicillin). The OD600 was
monitored until it reached ∼1.0. To flask A was added 0.8 mL of
a 0.4 M IPTG stock to start induction.58 Flask B was used as a
control without induction. Five hours later, cells were collected
(centrifuged 10 min at 4000 rpm) and washed once with 100 mM
PBS buffer, pH 7.4. The cells were then suspended in 100 mL of
pH 7.4 PBS buffer (8 g of NaCl, 0.2 g of KCl, 1.44 g of Na2HPO4,
0.24 g of KH2PO4 in 1.0 L) and incubated for 5 min. Then 1 mL
of NHS-CRAA ester (2) was added to each flask and incubated at
room temperature for about 30 min (Figure 3a). An amount of 10
mL of 1 M Tris-HCl, pH 6.5, was added to each flask and shaken
for another 5 min to quench the reaction. The cells were collected
again by centrifuging and washed twice with PBS buffer. The cells
were lysed with SDS loading buffer (50 mM Tris-HCl (pH 6.8),
100 mM dithiothreitol, 2% SDS, 0.1% bromophenol blue, 10%
glycerol) and were run on a 4-12% Bis-Tris SDS-PAGE gel
(Figure 3b). The gel was fluorescently imaged on a Kodak Image
Station to selectively detect 1 which has λmax for absorbance and
emission at 465 and 535 nm.37 Cells were fluorescently imaged in
PBS buffer after the wash (before lysis) using 495 nm/520 nm
excitation/emission filters on the fluorescence microscope. Comple-
mentary fluorescence and bright field images may be shifted slightly
relative to each other because of bacterial motion between image
captures. Exposure time was 1/3.5 s, and 1000× magnification was
used (Figure 4). Fluorescence images of control cells (no IPTG
treatment, no DHPR) indicate much less labeling of cells, with the
majority of the fluorescence coming from cellular debris (Supporting
Information Figure S3).
General Procedure for CRAA (1) Affinity Column Chroma-
tography and Target Fishing.32 The CRAA (1) affinity column
was equilibrated with buffer A, which contains 25 mM Tris-HCl,
50 mM NaCl, and 0.1% NaN3, pH 7.8. Washing was done until
the eluent was nearly colorless (1 is intensely colored). Then the
protein sample (E. coli, M. tuberculosis or human liver) was loaded
onto the affinity column and washed with a large amount of buffer
A until no protein sample was detected using a Bradford assay
(BioRad). The buffer volume used was usually 10-fold of the
packing volume of the column. Then the affinity column was eluted
with buffer B, which is the same as buffer A except containing 4
mM 1. Fractions were collected, then separated on an SDS-PAGE
gel and stained using a SilverQuest kit. Fractions from E. coli and
M. tuberculosis were compared and showed very similar banding
Methods
Equipment and Materials. Nano-HPLC-mass spectrometry was
performed using an LTQ mass spectrometer (Thermo-Fisher)
coupled to a Surveyor HPLC system (Thermo Fisher) equipped
with a Finnigan Micro AS autosampler. The instrument was
interfaced with an Aquasil C18 PicoFrit capillary column (75 µm
× 10 cm) from New Objective. A Kodak Image Station 2000MM
System was used for gel fluorescence scanning (Figure 3b), and
an Olympus BX60 microscope was used for fluorescene imaging
of cells (Figure 4). All Novex gel products for the SDS-PAGE
experiments were from Invitrogen, as was the SilverQuest staining
kit. All salts, buffers, enzymes, and other chemical reagents are
from Sigma-Aldrich and are of biochemical reagent grade, unless
specified otherwise. The ω-aminohexylagarose and the human liver
proteins (cytoplasmic) are also from Sigma. The M. tuberculosis
H37Rv whole cell lysate was from Colorado State University. These
proteins are from cells that were grown in glycerol-alanine stocks
for 14 days, then washed with PBS. After γ-irradiation (to
inactivate), cells were disrupted (French Press) and the lysate was
centrifuged to remove cell debris. Lysis buffer was PBS with 8
mM EDTA and protease inhibitors. Further details are available at
www.cvmbs.colostate.edu/microbiology/tb/cellysate.htm.
Synthesis of 1 (CRAA): 5-[(3,4-Dihydroxyphenyl)methylene]-
4-oxo-2-thioxo-3-thiazolidineacetic Acid. Synthesis was largely
as described before.37 Briefly, 3-rhodanine acetic acid was reacted
with 3,4-dihydroxybenzaldehyde in acetic acid/acetate at 90 °C for
6 h. After the mixture was cooled, yellow crystals were poured
into cold water, filtered, washed, and then crystallized from acetic
acid.
Synthesis of 2 (NHS-CRAA): (5-[(3,4-Dihydroxyphenyl)-
methylene]-4-oxo-2-thioxo-3-thiazolidineacetic N-Hydroxysuc-

Chemical Proteomics-Based Drug Design
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 15 4579
(6) Peters, E. C.; Gray, N. S. Chemical proteomics identifies unanticipated
targets of clinical kinase inhibitors. ACS Chem. Biol. 2007, 2, 661–
664.
(7) Argyrou, A.; Jin, L.; Siconilfi.Baez, L.; Angeletti, R. H.; Blanchard,
J. S. Proteome-wide profiling of isoniazid targets in Mycobacterium
tuberculosis. Biochemistry 2006, 45, 13947–13953.
profiles based on SDS-PAGE gel analysis (Figure 6a). Human
liver and M. tuberculosis fractions either were used directly for
mass spectral analysis (next section) or were separated using
SDS-PAGE, with protein extracted from the bands indicated in
Figure 6c.
Sample Preparation for Mass Spectrometry. Pooled fractions,
after elution from the CRAA (1) affinity column, were concentrated
using a Centricon filter with 10 kDa cutoff (Millipore). Then 100
µL of affinity purified protein mixtures were polymerized in the
presence of 100 µL of acrylamide/bis (30% T/2.67% C), 2 µL of
10% ammonium persulfate, and 2 µL of TEMED. With this mixture
a 15% gel piece was formed. Polymerization was performed in the
cap of an Eppendorf tube. The polymerized gel pieces were then
transferred to the corresponding Eppendorf tube in 1 mL of 40%
methanol and 7% acetic acid and incubated for 30 min. The gel
pieces were washed twice in water for 30 min each time while
sonicating. Gel pieces were then washed twice in 50% acetonitrile
for 30 min each time while sonicating. The gel pieces were then
washed twice again, this time in 50% acetonitrile in 50 mM
ammonium bicarbonate, pH 8.0. The gel pieces were then dried
using a speed vac from Savant. To each gel piece was added 200
µL of 20 mM ammonium bicarbonate, pH 8.0, containing 1 µg of
trypsin (Promega); this was incubated overnight at 37 °C. Each
gel piece with the digested proteins was then extracted twice with
70% acetonitrile in 0.1% formic acid. From this step onward, all
water used was MS quality water. Corresponding extracts of each
gel were pooled together and dried. To each dried sample was added
6 M guanidine ·HCl in 5 mM potassium phosphate and 1 mM DTT,
pH 6.5. This was sonicated, and peptides were extracted using a
(8) Schalon, C.; Surgand, J. S.; Kellenberger, E.; Rognan, D. A simple
and fuzzy method to align and compare druggable ligand-binding sites.
Proteins, in press.
(9) Hajduk, P. J.; Huth, J. R.; Tse, C. Predicting protein druggability.
Drug DiscoVery Today 2005, 10, 1675–1682.
(10) DeSimone, R. W.; Currie, K. S.; Mitchell, S. A.; Darrow, J. W.; Pippin,
D. A. Privileged structures: applications in drug discovery. Comb.
Chem. High Throughput Screening 2004, 7, 473–494.
(11) Duarte, C. D.; Barreiro, E. J.; Fraga, C. A. Privileged structures: a
useful concept for the rational design of new lead drug candidates.
Mini-ReV. Med. Chem. 2007, 7, 1108–1119.
(12) Costantino, L.; Barlocco, D. Privileged structures as leads in medicinal
chemistry. Curr. Med. Chem. 2006, 13, 65–85.
(13) Muller, G. Medicinal chemistry of target family-directed masterkeys.
Drug DiscoVery Today 2003, 8, 681–691.
(14) Bemis, G. W.; Murcko, M. A. The properties of known drugs. 1.
Molecular frameworks. J. Med. Chem. 1996, 39, 2887–2893.
(15) Brustle, M.; Beck, B.; Schindler, T.; King, W.; Mitchell, T.; Clark,
T. Descriptors, physical properties, and drug-likeness. J. Med. Chem.
2002, 45, 3345–3355.
(16) Peters, U.; Cherian, J.; Kim, J. H.; Kwok, B. H.; Kapoor, T. M. Probing
cell-division phenotype space and Polo-like kinase function using small
molecules. Nat. Chem. Biol. 2006, 2, 618–626.
(17) Smith, A. B., 3rd; Favor, D. A.; Sprengeler, P. A.; Guzman, M. C.;
Carroll, P. J.; Furst, G. T.; Hirschmann, R. Molecular modeling,
synthesis, and structures of N-methylated 3,5-linked pyrrolin-4-ones
toward the creation of a privileged nonpeptide scaffold. Bioorg. Med.
Chem. 1999, 7, 9–22.
(18) Amano, M.; Koh, Y.; Das, D.; Li, J.; Leschenko, S.; Wang, Y. F.;
Boross, P. I.; Weber, I. T.; Ghosh, A. K.; Mitsuya, H. A novel bis-
tetrahydrofuranylurethane-containing nonpeptidic protease inhibitor
(PI), GRL-98065, is potent against multiple-PI-resistant human im-
munodeficiency virus in vitro. Antimicrob. Agents Chemother. 2007,
51, 2143–2155.
(19) Guo, T.; Hobbs, D. W. Privileged structure-based combinatorial
libraries targeting G protein-coupled receptors. Assay Drug DeV.
Technol. 2003, 1, 579–592.
C18 ZipTip from Millipore. Extracted peptides were then collected
into an insert in a vial to be used for mass spectrometry and dried
in the inserts. To each dried sample was added 5 µL of 0.1% formic
acid in MS water containing 5% acetonitrile. Samples were then
ready for mass spectrometry and were injected into the LTQ LC/
MS. The MS/MS data were collected and searched against the
appropriate subset of the Uniprot database.
Acknowledgment. We thank Dr. John Blanchard (Albert
Einstein College of Medicine, NY) for the DHPR (dapB) expres-
sion constructs, and Dr. M. Behnam Ghasemzadeh for use of the
KODAK Image Station at the College of Health Sciences of
Marquette University. M. tuberculosis H37Rv whole cell lysate
was received as part of the National Institutes of Health and
National Institute of Allergy and Infectious Disease Contract
HHSN266200400091C, entitled Tuberculosis Vaccine Testing and
Research Materials, which was awarded to the Colorado State
University. Fluorescence microscopy was performed at the Chil-
dren’s Environmental Health Sciences Core (EHSC) Center, with
expert assistance of Dr. Henry Tomasiewicz. The EHSC is
supported in part by NIH Grant ES-04184.
(20) Sasse, B. C.; Mach, U. R.; Leppaenen, J.; Calmels, T.; Stark, H. Hybrid
approach for the design of highly affine and selective dopamine D(3)
receptor ligands using privileged scaffolds of biogenic amine GPCR
ligands. Bioorg. Med. Chem. 2007, 15, 7258–7273.
(21) Sem, D. S.; Bertolaet, B.; Baker, B.; Chang, E.; Costache, A. D.;
Coutts, S.; Dong, Q.; Hansen, M.; Hong, V.; Huang, X.; Jack, R. M.;
Kho, R.; Lang, H.; Ma, C. T.; Meininger, D.; Pellecchia, M.; Pierre,
F.; Villar, H.; Yu, L. Systems-based design of bi-ligand inhibitors of
oxidoreductases: filling the chemical proteomic toolbox. Chem. Biol.
2004, 11, 185–194.
(22) Kho, R.; Newman, J. V.; Jack, R. M.; Villar, H. O.; Hansen, M. R.
Genome-wide profile of oxidoreductases in viruses, prokaryotes, and
eukaryotes. J. Proteome Res. 2003, 2, 626–632.
(23) Kho, R.; Baker, B. L.; Newman, J. V.; Jack, R. M.; Sem, D. S.; Villar,
H. O.; Hansen, M. R. A path from primary protein sequence to ligand
recognition. Proteins 2003, 50, 589–599.
(24) Godl, K.; Wissing, J.; Kurtenbach, A.; Habenberger, P.; Blencke, S.;
Gutbrod, H.; Salassidis, K.; Stein-Gerlach, M.; Missio, A.; Cotten,
M.; Daub, H. An efficient proteomics method to identify the cellular
targets of protein kinase inhibitors. Proc. Natl. Acad. Sci. U.S.A. 2003,
100, 15434–15439.
Supporting Information Available: Additional gels, NMR
characterization, and tables of mass spectral analyses. This material
is available free of charge via the Internet at http://pubs.acs.org.
(25) Bantscheff, M.; Eberhard, D.; Abraham, Y.; Bastuck, S.; Boesche,
M.; Hobson, S.; Mathieson, T.; Perrin, J.; Raida, M.; Rau, C.; Reader,
V.; Sweetman, G.; Bauer, A.; Bouwmeester, T.; Hopf, C.; Kruse, U.;
Neubauer, G.; Ramsden, N.; Rick, J.; Kuster, B.; Drewes, G.
Quantitative chemical proteomics reveals mechanisms of action of
clinical ABL kinase inhibitors. Nat. Biotechnol. 2007, 25, 1035–1044.
(26) Poupaert, J.; Carato, P.; Colacino, E.; Yous, S. 2(3H)-Benzoxazolone
and bioisosters as “privileged scaffold” in the design of pharmacologi-
cal probes. Curr. Med. Chem. 2005, 12, 877–885.
(27) Matsuoka, K.; Sakamoto, N.; Akanuma, Y.; Hotta, N.; Shichiri, M.;
Toyota, T.; Oka, Y.; Kawamori, R.; Shigeta, Y. ADCT Study Group.
A long-term effect of epalrestat on motor conduction velocity of
diabetic patients: ARI-Diabetes Complication Trial. Diabetes Res. Clin.
Pract. 2007, 77, S263-268.
References
(1) Fitzgerald, G. A. Opinion: anticipating change in drug development:
the emerging era of translational medicine and therapeutics. Nat. ReV.
Drug DiscoVery 2005, 4, 815–818.
(2) Tolliday, N.; Clemons, P. A.; Ferraiolo, P.; Koehler, A. N.; Lewis,
T. A.; Li, X.; Schreiber, S. L.; Gerhard, D. S.; Eliasof, S. Small
molecules, big players: the National Cancer Institute’s Initiative for
Chemical Genetics. Cancer Res. 2006, 66, 8935–8942.
(3) Strausberg, R. L.; Schreiber, S. L. From knowing to controlling: a
path from genomics to drugs using small molecule probes. Science
2003, 300, 294–295.
(4) Krishnamurty, R.; Maly, D. J. Chemical genomic and proteomic
methods for determining kinase inhibitor selectivity. Comb. Chem.
High Throughput Screening 2007, 10, 652–666.
(5) Daub, H.; Godl, K.; Brehmer, D.; Klebl, B.; Muller, G. Evaluation of
kinase inhibitor selectivity by chemical proteomics. Assay Drug DeV.
Technol. 2004, 2, 215–224.
(28) Hotta, N.; Akanuma, Y.; Kawamori, R.; Matsuoka, K.; Oka, Y.;
Shichiri, M.; Toyota, T.; Nakashima, M.; Yoshimura, I.; Sakamoto,
N.; Shigeta, Y. ADCT Study Group. Long-term clinical effects of
epalrestat, an aldose reductase inhibitor, on diabetic peripheral
neuropathy. Diabetes Care 2006, 29, 1538–1544.

4580 Journal of Medicinal Chemistry, 2008, Vol. 51, No. 15
Ge et al.
(29) Bolton, J. L.; Pisha, E.; Shen, L.; Krol, E. S.; Iverson, S. L.; Huang,
Z.; van.Breemen, R. B.; Pezzuto, J. M. The reactivity of o-quinones
which do not isomerize to quinone methides correlates with alkylcat-
echol-induced toxicity in human melanoma cells. Chem. Biol. Interact.
1997, 106, 133–148.
(30) Hutzler, J. M.; Melton, R. J.; Rumsey, J. M.; Thompson, D. C.; Rock,
D. A.; Wienkers, L. C. Chem. Res. Toxicol., in press.
(31) Barry, S.; O’Carra, P. Affinity chromatography of nicotinamide-
adenine dinucleotide-linked dehydrogenases on immobilized deriva-
tives of the dinucleotide. Biochem. J. 1973, 135, 595–607.
(32) Morelli, A.; Benatti, U. Simple chemical synthesis of a specific effector
for the affinity chromatography of nicotinamide adenine dinucleotide
phosphate-dependent dehydrogenases. Ital. J. Biochem. 1974, 23, 279–
291.
(33) Trayer, I. P.; Trayer, H. R. Affinity chromatography of nicotinamide
nucleotide-dependent dehydrogenases on immobilized nucleotide
derivatives. Biochem. J. 1974, 141, 775–787.
(34) Ng, E.; Schriemer, D. C. Emerging challenges in ligand discovery:
new opportunities for chromatographic assay. Expert ReV. Proteomics
2005, 2, 891–900.
(35) Bertrand, R.; Derancourt, J.; Kassab, R. Fluorescence characteriza-
tion of structural transitions at the strong actin binding motif in
skeletal myosin affinity labeled at cysteine 540 with novel
spectroscopic cysteaminyl mixed disulfides. Biochemistry 2000, 39,
14626–14637.
(36) Scapin, G.; Reddy, S. G.; Zheng, R.; Blanchard, J. S. Three-
dimensional structure of Escherichia coli dihydrodipicolinate reductase
in complex with NADH and the inhibitor 2,6-pyridinedicarboxylate.
Biochemistry 1997, 36, 15081–15088.
(37) Ge, X.; Sem, D. S. Affinity-based chemical proteomic probe for
dehydrogenases: fluorescence and visible binding assays in gels. Anal.
Biochem. 2007, 370, 171–179.
(44) Mills, P. B.; Struys, E.; Jakobs, C.; Plecko, B.; Baxter, P.; Baumgartner,
M.; Willemsen, M. A.; Omran, H.; Tacke, U.; Uhlenberg, B.; Weschke,
B.; Clayton, P. T. Mutations in antiquitin in individuals with
pyridoxine-dependent seizures. Nat. Med. 2006, 12, 307–309.
(45) Bok, L. A.; Struys, E.; Willemsen, M. A.; Been, J. V.; Jakobs, C.
Pyridoxine-dependent seizures in Dutch patients: diagnosis by elevated
urinary alpha-aminoadipic semialdehyde levels. Arch. Dis. Child. 2007,
92, 687–689.
(46) Salomons, G. S.; Bok, L. A.; Struys, E. A.; Pope, L. L.; Darmin, P. S.;
Mills, P. B.; Clayton, P. T.; Willemsen, M. A.; Jakobs, C. An intriguing
”silent” mutation and a founder effect in antiquitin (ALDH7A1). Ann.
Neurol. 2007, 62, 414–418.
(47) Eirich, L. D.; Vogels, G. D.; Wolfe, R. S. Distribution of coenzyme
F420 and properties of its hydrolytic fragments. J. Bacteriol. 1979,
140, 20–27.
(48) Eirich, L. D.; Vogels, G. D.; Wolfe, R. S. Proposed structure for
coenzyme F420 from Methanobacterium. Biochemistry 1978, 17,
4583–4593.
(49) Purwantini, E.; Daniels, L. Molecular analysis of the gene encoding
F420-dependent glucose-6-phosphate dehydrogenase from Mycobac-
terium smegmatis. J. Bacteriol. 1998, 180, 2212–2219.
(50) Salvetti, S.; Celandroni, F.; Ghelardi, E.; Baggiani, A.; Senesi, S. Rapid
determination of vitamin B2 secretion by bacteria growing on solid
media. J. Appl. Microbiol. 2003, 95, 1255–1260.
(51) Messinger, J.; Hirvela, L.; Husen, B.; Kangas, L.; Koskimies, P.;
Pentikainen, O.; Saarenketo, P.; Thole, H. New inhibitors of 17beta-
hydroxysteroid dehydrogenase type 1. Mol. Cell. Endocrinol. 2006,
248, 192–198.
(52) Rizner, T. L.; Zakelj-Mavric, M. Characterization of fungal 17beta-
hydroxysteroid dehydrogenases. Comp. Biochem. Physiol., Part B:
Biochem. Mol. Biol. 2000, 127, 53–63.
(53) Purkayastha, A.; McCue, L. A.; McDonough, K. A. Identification of
a Mycobacterium tuberculosis putative classical nitroreductase gene
whose expression is coregulated with that of the acr gene within
macrophages, in standing versus shaking cultures, and under low
oxygen conditions. Infect. Immun. 2002, 70, 1518–1529.
(54) Srivastava, V.; Rouanet, C.; Srivastava, R.; Ramalingam, B.; Locht,
C.; Srivastava, B. S. Macrophage-specific Mycobacterium tuberculosis
genes: identification by green fluorescent protein and kanamycin
resistance selection. Microbiology 2007, 153, 659–666.
(55) Vanin, E. F.; Ji, T. H. Synthesis and application of cleavable
photoactivable heterobifunctional reagents. Biochemistry 1981, 20,
6754–6760.
(56) Witzemann, V.; Muchmore, D.; Raftery, M. A. Affinity-directed cross-
linking of membrane-bound acetylcholine receptor polypeptides with
photolabile alpha-bungarotoxin derivatives. Biochemistry 1979, 18,
5511–5518.
(57) Waugh, S. M.; DiBella, E. E.; Pilch, P. F. Isolation of a proteolytically
derived domain of the insulin receptor containing the major site of
cross-linking/binding. Biochemistry 1989, 28, 3448–3455.
(58) Reddy, S. G.; Sacchettini, J. C.; Blanchard, J. S. Expression,
purification, and characterization of Escherichia coli dihydrodipicoli-
nate reductase. Biochemistry 1995, 34, 3492–3501.
(38) Barglow, K. T.; Cravatt, B. F. Activity-based protein profiling for the
functional annotation of enzymes. Nat. Methods 2007, 4, 822–827.
(39) Evans, M. J.; Saghatelian, A.; Sorensen, E. J.; Cravatt, B. F. Target
discovery in small-molecule cell-based screens by in situ proteome
reactivity profiling. Nat. Biotechnol. 2005, 23, 1303–1307.
(40) Dye, C.; Scheele, S.; Dolin, P.; Pathania, V.; Raviglione, M. C.
Consensus statement. Global burden of tuberculosis: estimated inci-
dence, prevalence, and mortality by country. WHO Global Surveillance
and Monitoring Project. JAMA, J. Am. Med. Assoc. 1999, 282, 677–
686.
(41) Okada, M.; Kobayashi, K. Recent progress in mycobacteriology.
Kekkaku 2007, 82, 783–799.
(42) Hori, T.; Ishijima, J.; Yokomizo, T.; Ago, H.; Shimizu, T.; Miyano,
M. Crystal structure of anti-configuration of indomethacin and
leukotriene B4 12-hydroxydehydrogenase/15-oxo-prostaglandin 13-
reductase complex reveals the structural basis of broad spectrum
indomethacin efficacy. J. Biochem. 2006, 140, 457–466.
(43) Clish, C. B.; Sun, Y. P.; Serhan, C. N. Identification of dual
cyclooxygenase-eicosanoid oxidoreductase inhibitors: NSAIDs that
inhibit PG-LX reductase/LTB(4) dehydrogenase. Biochem. Biophys.
Res. Commun. 2001, 288, 868–874.
JM8002284