Structural characterization of the mycobacterium tuberculosis biotin biosynthesis enzymes 7,8-diaminopelargonic acid synthase and dethiobiotin synthetase

Article abstract of DOI:10.1021/bi902097j

Mycobacterium tuberculosis (Mtb) depends on biotin synthesis for survival during infection. In the absence of biotin, disruption of the biotin biosynthesis pathway results in cell death rather than growth arrest, an unusual phenotype for an Mtb auxotroph. Humans lack the enzymes for biotin production, making the proteins of this essential Mtb pathway promising drug targets. To this end, we have determined the crystal structures of the second and third enzymes of the Mtb biotin biosynthetic pathway, 7,8-diaminopelargonic acid synthase (DAPAS) and dethiobiotin synthetase (DTBS), at respective resolutions of 2.2 and 1.85 A. Superimposition of the DAPAS structures bound either to the SAM analogue sinefungin or to 7-keto-8-aminopelargonic acid (KAPA) allowed us to map the putative binding site for the substrates and to propose a mechanism by which the enzyme accommodates their disparate structures. Comparison of the DTBS structures bound to the substrate 7,8-diaminopelargonic acid (DAPA) or to ADP and the product dethiobiotin (DTB) permitted derivation of an enzyme mechanism. There are significant differences between the Mtb enzymes and those of other organisms; the Bacillus subtilis DAPAS, presented here at a high resolution of 2.2 A, has active site variations and the Escherichia coli and Helicobacter pylori DTBS have alterations in their overall folds. We have begun to exploit the unique characteristics of the Mtb structures to design specific inhibitors against the biotin biosynthesis pathway in Mtb.

Full text of DOI:10.1021/bi902097j

6746 Biochemistry 2010, 49, 6746–6760
DOI: 10.1021/bi902097j
Structural Characterization of the Mycobacterium tuberculosis Biotin Biosynthesis
Enzymes 7,8-Diaminopelargonic Acid Synthase and Dethiobiotin Synthetase†,‡
Sanghamitra Dey,# James M. Lane, Richard E. Lee,^ Eric J. Rubin, and James C. Sacchettini*,§,z
§Department of Biochemistry and Biophysics, Texas A&M University, College Station, Texas 77843, Department of Immunology and
Infectious Diseases, Harvard School of Public Health, Boston, Massachusetts 02115, ^Chemical Biology and Therapeutics, St. Jude
Children’s Research Hospital, Memphis, Tennessee 38105, and #Department of Biochemistry and Biophysics, Texas A&M University,
College Station, TX 77843. zCurrent address: University of California, San Diego, La Jolla, California 92093.
Received December 7, 2009; Revised Manuscript Received March 12, 2010
ABSTRACT: Mycobacterium tuberculosis (Mtb) depends on biotin synthesis for survival during infection. In the
absence of biotin, disruption of the biotin biosynthesis pathway results in cell death rather than growth arrest, an
unusual phenotype for an Mtb auxotroph. Humans lack the enzymes for biotin production, making the proteins
of this essential Mtb pathway promising drug targets. To this end, we have determined the crystal structures of the
second and third enzymes of the Mtb biotin biosynthetic pathway, 7,8-diaminopelargonic acid synthase
˚
(DAPAS) and dethiobiotin synthetase (DTBS), at respective resolutions of 2.2 and 1.85 A. Superimposition
of the DAPAS structures bound either to the SAM analogue sinefungin or to 7-keto-8-aminopelargonic acid
(KAPA) allowed us to map the putative binding site for the substrates and to propose a mechanism by which the
enzyme accommodates their disparate structures. Comparison of the DTBS structures bound to the substrate
7,8-diaminopelargonic acid (DAPA) or to ADP and the product dethiobiotin (DTB) permitted derivation of
an enzyme mechanism. There are significant differences between the Mtb enzymes and those of other organisms;
˚
the Bacillus subtilis DAPAS, presented here at a high resolution of 2.2 A, has active site variations and
the Escherichia coli and Helicobacter pylori DTBS have alterations in their overall folds. We have begun to exploit
the unique characteristics of the Mtb structures to design specific inhibitors against the biotin biosynthesis
pathway in Mtb.
Biotin is a cofactor for essential enzymes that function in
carboxylation, decarboxylation, and transcarboxylation reac-
tions found in many processes such as fatty acid biosynthesis,
gluconeogenesis, and amino acid metabolism (1). Whereas plants
and bacteria, including Mtb,1 have enzymes for production of
this cofactor, mammals do not, instead obtaining biotin through
diet and the action of gut bacteria. This makes the Mtb biotin
synthetic pathway an attractive drug target.
Mtb relies on biotin utilizing pathways, especially fatty acid
biosynthesis, during latency (2), and multiple lines of evidence
suggest that biotin production is essential for mycobacterial sur-
vival in vitro and in vivo. In the model organism Mycobacterium
smegmatis, the bioA gene, which encodes the second enzyme in
the biosynthetic pathway 7,8-diaminopelargonic acid synthase
(DAPAS), is required for optimal stationary phase growth in
rich, biotin-replete medium (3). This suggests that de novo biotin
biosynthesis is required during the stationary phase and that
mycobacteria in this phase are unable to acquire biotin from the
medium. In Mtb expression of bioD, the gene encoding dethio-
biotin synthase (DTBS), the third enzyme in the biosynthetic
pathway, is upregulated after 96 h of nutrient starvation (4), a
condition thought to mimic bacterial persistence in vivo (2).
Finally, disruption of the Mtb biotin biosynthesis machinery by
transposon mutagenesis results in rapid clearance of the mutants
in the early stages of infection (5). This is unsurprising, as biotin
auxotrophy has been linked to the attenuation of virulence in
other bacterial pathogens (6).
The Mtb H37Rv genome encodes all four enzymes needed to
convert pimeloyl-CoA to biotin (Figure 1) (7), although the source
of pimeloyl-CoA in Mtb has not yet been identified. KAPA synthase
(BioF: Rv1569), a pyridoxal 50-phosphate (PLP) dependent enzyme,
catalyzes the first step of biotin biosynthesis, the decarboxylative
†This work was supported by National Institutes of Health Grant
PO1AIO68135 and by funds from the Robert A. Welch Foundation for
J.C.S. and by the American Lebanese Syrian Associated Charities
(ALSAC) for R.E.L.
condensation of pimeloyl-CoA with L-alanine to produce 7-keto-
8-aminopelargonic acid (KAPA). In the second step of the
pathway, the PLP-dependent DAPA synthase (BioA: Rv1568)
uses S-adenosylmethionine (SAM) to transaminate KAPA into
7,8-diaminopelargonic acid (DAPA). Dethiobiotin synthetase
(BioD: Rv1570) carboxylates DAPA to form the ureido ring of
dethiobiotin (DTB) in the ATP- and magnesium-dependent third
step of the pathway. Finally, biotin synthase (BioB: Rv1589), an
iron-sulfur cluster enzyme, converts DTB into biotin, generating a
free radical via SAM in the fourth and final step of biotin production.
There have been initial biochemical characterizations of the bio-
tin biosynthetic enzymes in microbes such as Bacillus sphaericus (8)
‡The coordinates for these structures are available from the Protein
Data Bank (ID codes 3BV0, 3LV2, 3DRD, 3DOD, 3DU4, 3FGN,
3FMF, 3FMI, and 3FPA).
*To whom correspondence should be addressed: e-mail, sacchett@
tamu.edu; phone, 979-862-7636; fax, 979-862-7638.
1Abbreviations: Mtb, Mycobacterium tuberculosis; DAPAS, 7,8-
diaminopelargonic acid synthase; DTBS, dethiobiotin synthetase; PLP,
pyridoxal 50-phosphate; KAPA, 7-keto-8-aminopelargonic acid; SAM,
S-adenosylmethionine; DAPA, 7,8-diaminopelargonic acid; DTB,
dethiobiotin; PMP, pyridoxamine phosphate; AMPSO, N-(1,1-dimethyl-
2-hydroxyethyl)-3-amino-2-hydroxypropanesulfonic acid; SAH,
S-adenosylhomocysteine; ATP, adenosine triphosphate; ADP, adeno-
sine diphosphate.
r
pubs.acs.org/Biochemistry
Published on Web 06/21/2010
2010 American Chemical Society

Article
Biochemistry, Vol. 49, No. 31, 2010 6747
FIGURE 2: The chemical structures of the known inhibitors of the
biotin biosynthetic pathway: (A) acidomycin; (B) amiclenomycin.
E. coli counterpart (kcat for Mtb and E. coli DAPAS are 1 ( 0.2/min
and 0.78/min, respectively) (14, 28); however, the Km value for its
amino donor SAM is higher (750 ( 200 μM in Mtb compared to
150 μM in E. coli) (14, 28). We present evidence that differences
between the active site architecture of the Mtb and E. coli
DAPAS enzymes may account for the small but noticeable
differences in the kinetic behavior between them. Structures of
substrate-bound Mtb DAPAS and of its B. subtilis homologue
also provide insight into the mode of and variability in dual
substrate recognition at the active site of this aminotransferase.
Finally, comparison of the apo and substrate-bound forms of
Mtb DTBS delineates the binding mode of the substrate as well as
the mechanism of enzyme catalysis, while comparison of DTBS
structures from multiple species highlights the substantial natural
variability in structural folds.
MATERIALS AND METHODS
Growth of Mycobacterium tuberculosis. Mtb H37Rv
ΔbioF was maintained in Middlebrook 7H9 medium with glycerol
and Tween 80 or in Sauton’s medium supplemented with 50 μM
biotin (5).
FIGURE 1: The Mtb biotin synthetic pathway showing the four
enzymes required to convert pimeloyl-CoA and alanine to biotin.
Cloning, Protein Expression, and Purification. The bioA
(Rv1568) and bioD (Rv1570) genes from Mtb H37Rv genomic
DNA were amplified and cloned into the pET28b vector
(Novagen) with the TEV site using the NdeI and HindIII sites.
The Y25A mutant of Mtb DAPAS was constructed using the
QuikChange site-directed mutagenesis kit (Stratagene) and
cloned into the same modified pET28b vector. The NdeI site
containing the ATG start codon and the stop codon precedes the
HindIII site. Both clones were then transformed into the E. coli
overexpression host BL21(DE3). The transformed cell culture
was grown in LB medium with kanamycin (50 μg/mL) at 37 ꢀC
until the A600 reached 0.8 and then was induced with 0.5 mM
IPTG at 25 ꢀC and grown overnight. There was no overexpres-
sion seen at temperatures lower than 25 ꢀC. The cells were
harvested by centrifuging at 4000 rpm for 30 min and resus-
pended in 50 mM Tris buffer (pH 7.5), 500 mM NaCl, 1 mM
PMSF, DNase (20 μg/mL), and 100 μM PLP (for DAPAS only).
The cells were lysed using a French press at 1200 psi, and cell
debris was removed by centrifuging the cell lysate at 15000 rpm
for 1 h. The filtered supernatant was then loaded on a preequili-
brated (50 mM Tris buffer (pH 7.5), 500 mM NaCl) Hitrap nickel
column (GE Healthcare) and washed. Protein elution with an
imidazole gradient (50-400 mM) followed. The fractions con-
taining pure protein were pooled and dialyzed overnight against
and Bacillus subtilis (9), the lower eukaryote Saccharomyces
cerevisiae (10), and in the plant Arabidopsis thaliana (11), but
the pathway has been most thoroughly studied in Escherichia
coli (12-20). Indeed, the E. coli enzymes’ crystal structures
have served as the basis for several studies that attempted to
define inhibitors of the plant biotin pathway for herbicide
development (21-23).
Here we present the structural characterization of two Mtb
biotin biosynthetic enzymes, DAPAS and DTBS, and our
subsequent efforts to develop inhibitors of this Mtb pathway.
Although there are two known natural inhibitors with strong
whole cell activity, amiclenomycin (MIC = 3.1 μg/mL) (24) and
acidomycin (MIC = 0.0625-0.125 μg/mL) (25), both amicleno-
mycin (26) and acidomycin (27) were demonstrated to be inactive
in a mouse model of tuberculosis (Figure 2). Previous transposon
mutagenesis work indicates that Mtb is not able to scavenge
biotin from the host, and the failure of acidomycin to work in vivo
appears to be the result of limited bioavailability (27). We show
here that the biotin biosynthesis pathway is required for Mtb
survival in vitro, a surprising finding for an Mtb auxotroph.
The Mtb DAPAS and DTBS structures also shed light on the
enzymes’ kinetics and mechanisms of action. With a 48%
sequence identity, the efficiency of Mtb DAPAS is similar to its

6748 Biochemistry, Vol. 49, No. 31, 2010
Dey et al.
a buffer containing 25 mM Tris buffer (pH 7.5), 50 mM NaCl,
1 mM EDTA, and 1 mM DTT. The purity of the proteins was
judged from a 4-12% gradient SDS-PAGE gel.
The B. subtilis bioA gene was PCR-amplified from B. subtilis
168 genomic DNA (Bacillus Genetic Stock Centre, Ohio State
University) and further cloned and purified in the same way as
Mtb DAPAS and DTBS.
1 mM PLP and 5 mM KAPA, respectively, in hanging drop plates.
Simple soaking of DAPAS crystals in drops containing PLP or
KAPA resulted in no diffraction.
The native Mtb DTBS protein crystals were obtained in 1 M
sodium citrate and 0.1 M imidazole (pH 7.5) (Wizard I no. 36).
Molecular replacement using the E. coli DTBS structure (PDB
code: 1BYI) (31) as a template was not successful. Thus native
DTBS crystals were soaked in 50 mM samarium chloride for 1 h,
Spectroscopic Studies of Mtb DAPAS and DTBS.
Absorbance spectra for wild-type Mtb DAPAS with 1 mM
SAM, sinefungin, or SAH and for the Y25A mutant DAPAS
with SAM and DAPA were measured on the Cary 50 spectro-
photometer (Varian) in the range of 300-600 nm. Fluorescence
emission scans for Mtb DAPAS with 1 mM sinefungin and SAH
were followed in the range of 300-450 nm on the LS 55 lumine-
scence spectrometer (Perkin-Elmer Instruments) with constant
stirring. All experiments were carried out at room temperature.
The buffer used was 50 mM AMPSO (pH 8.5), 150 mM NaCl,
10% glycerol, 4 mM β-mercaptoethanol, and 100 μM PLP.
Coupled assays for wild-type and D47A mutant Mtb DTBS
with pyruvate kinase/lactate dehydrogenase (29) were carried out
to determine the kinetic parameters of ATP and DAPA by
keeping the other substrate at a saturating concentration of 3 mM
and 50 μM, respectively. The buffer used for the 500 μL assay con-
tained 50 mM Tris (pH 7.5), 150 mM NaCl, 10 mM NaHCO3,
and 5 mM MgCl2. ADP formation in the DTBS reaction was
measured by monitoring the decrease in NADH absorbance
at 340 nm. These assays were carried out using a Cary 100
spectrophotometer. DAPAS and DTBS concentrations were
calculated from their absorbance at 280 nm using extinction
coefficients (ε = 64490 mM-1 cm-1 and ε = 12840 mM-1
cm-1, respectively) (30).
˚
and a data set at 2.6 A resolution was collected using a Rigaku
˚
Raxis IVþþ detector at Cu wavelength (1.54 A). The DAPA-,
KAPA-, and DTB-ADP-complexed DTBS structures were ob-
tained by soaking the native crystals overnight in 10 mM DAPA,
10 mM KAPA, or 10 mM DTB, 10 mM ADP, and 10 mM
MgCl2, respectively. The DAPA- and KAPA-bound data sets
were collected from the crystals obtained in the same condition as
the native. The DTB-ADP complexed structure was obtained
from crystals in 0.4 M sodium phosphate/1.6 M potassium
phosphate, 0.1 M imidazole (pH 8.0), and 0.2 M NaCl.
All diffraction data were collected at 120 K using DAPAS and
DTBS crystals cryoprotected in paratone oil.
Data Processing, Structure Determination, and Refine-
ment. Mtb and B. subtilis DAPAS and Mtb DTBS data were
reduced and scaled using the HKL 2000 suite (32), d*trek (33),
mosflm (34), and SCALA (35).
The structure of PLP-bound Mtb DAPAS was determined by
molecular replacement using PHASER (36) to provide clear
solutions. The E. coli DAPAS monomer (PDB code: 1QJ5) (15)
was used as the model (LLG = 1437.551). The structure of
sinefungin-bound DAPAS was solved by molecular replacement
using the PLP-bound Mtb DAPAS as the model (LLG =
4189.75). Phasing of the B. subtilis DAPAS data set was also
achieved by the molecular replacement method using the co-
ordinates of Mtb DAPAS chain A (LLG = 279) as the search
model for PHASER. The LLG was 243 using the E. coli DAPAS
monomer (PDB code: 1QJ5).
For determination of the Mtb DTBS structure, four sites for
samarium (Sm) were found as the anomalous difference Patter-
son peaks using SHELXD (37). After refining these Sm sites and
performing initial phasing with AutoSHARP (38), we obtained a
density-modified map by solvent flattening using DM. The
phases were further extended to a high-resolution native data
Crystallization and Data Collection. The crystal trials for
Mtb and B. subtilis DAPAS and Mtb DTBS were carried out at
18 ꢀC using Crystal Screens I and II, Index, Membfac, and PEG/
Ion screens (Hampton Research) and the Wizard I, II, and III
screens (Emerald Biosystems). Initial hits were obtained using
robotic screening with a 1:1 ratio of protein to precipitant on
96-well Intelli-Plates (Hampton Research).
Yellow crystals for Mtb DAPAS appeared after heavy pre-
cipitation under different conditions with PEG 8000 as the com-
mon precipitant. These crystals were obtained from a DAPAS
protein with a random mutation at position 315 (H315R). A
high-resolution data set for the Mtb DAPAS crystal obtained in
10% PEG 8000, 0.1 M Tris buffer (pH 7.0), and 0.1 M MgCl2
(Wizard II no. 43) was collected at Cu wavelength using the home
˚
set of 1.85 A for the DTBS apo structure. DAPA-, KAPA-, and
DTB-, ADP-, and Pi-complexed structures were obtained by
refining the DTBS apo structure against these data, which have
similar unit cell dimensions.
˚
source to a resolution of 2.2 A. The data set for sinefungin-bound
Iterative rounds of model building were performed using
Xtalview (39) and Coot (40) with the Shake&wARP (41) un-
biased electron density maps. Initially, one round of simulated
annealing at 4000 K using CNS (42) was performed to improve
the geometry of the models and to remove bias from the model.
Water molecules were then added to the models after the R-factor
and Rfree were below 30% and inspected manually after refine-
ment. The quality of the models was validated by Ramachandran
plot using the program PROCHECK (43).
DAPAS was obtained by soaking the crystal in 10 mM sinefungin
for 45 min. The crystal was obtained in Hampton Research’s
Index screen no. 46 (20% PEGMME 5000, 0.1 M Bis-Tris buffer
(pH 7.5)). Soaking the crystals for a longer amount of time or
using a higher concentration of sinefungin resulted in significant
deterioration of diffraction or no diffraction at all. The data set
was collected in-house on a Rigaku R-axis IVþþ detector at a
˚
resolution of 2.2 A.
Crystals for B. subtilis DAPAS were obtained in a 4 μL
hanging drop with a 1:1 ratio of protein (10 mg/mL) to
precipitant in 20% PEG 3350 and 0.2 M sodium thiocyanate
(PEG/ion screen no. 13; Hampton Research). The quality of the
crystal greatly improved with the addition of 5% xylitol from the
additive screen (no. 56 of Hampton Research). The PLP-bound
structure and PLP- and KAPA-bound structures were obtained
by cocrystallization of the DAPAS protein with 1 mM PLP or
RESULTS AND DISCUSSION
Biotin Synthesis Is Required for Bacterial Survival in
Vitro. We previously constructed an Mtb strain with an in-frame
deletion of bioF, the gene that encodes KAPA synthase. This
mutant is unable to establish an infection in mice and is rapidly
cleared (5). We found that in vitro growth of the bioF strain is

Article
Biochemistry, Vol. 49, No. 31, 2010 6749
F
IGURE 3: Survival of the bioF Mtb mutant is dependent on biotin.
A ΔbioF Mtb strain was grown in Sauton’s medium in the presence or
absence of 50 μM biotin. Surviving colonies were quantified by
measuring growth on biotin-containing Middlebrook 7H10 medium
at the indicated time points. Data represent mean( SEM of triplicate
samples. The threshold of detection is 100 colonies.
completely dependent on the addition of exogenous biotin to the
culture medium (Figure 3). Unexpectedly, there is no measurable
survival of the mutant following 2 weeks of biotin depletion even
after replating on medium supplemented with biotin (Figure 3).
This irreversible cell death phenotype is unusual as Mtb auxo-
trophs typically do not die when starved for their given nutrient in
vitro but instead enter a nonreplicative phase that can be reversed
by addition of the nutrient (44, 45). Biotin appears to be different;
mycobacteria may be unable to sense biotin depletion and grow
in the absence of essential metabolites whose synthesis is biotin-
dependent. In this regard, the Mtb bioF mutant is reminiscent of
the M. smegmatis dapA strain, which lyses because it continues to
grow despite being unable to synthesize a normal cell wall (46).
Alternatively, biotin may aid in defense against a continual,
unknown stress.
Analysis of whole genome screening data suggests that it is
unlikely that bioF has a unique role in mycobacterial physiology
and virulence. Mutations in two (bioA and bioB) of the three
other predicted biotin synthesis genes lead to decreased in vivo
growth (5). Although the degree of attenuation is not as profound
as that produced by the bioF mutation, this type of screen is not
perfectly quantitative. Thus, biotin synthesis at any step is likely
essential for Mtb growth during infection and for survival in
culture.
F
fungin and S-adenosylhomocysteine.
IGURE 4: The chemical structures of SAM and its analogues sine-
sinefungin, resulting in an increase in absorbance at 335 nm that
is characteristic of the PMP form of the enzyme. Similar results
were observed when the PLP-complexed DAPAS was mixed with
the product DAPA. However, the intermediate quinonoid form
of the aminotransferase reaction observed at 487 nm in the E. coli
DAPAS reaction with SAM was not detected (50). This suggests
that there is no accumulation of the intermediate in the Mtb
DAPAS reaction and that the quinonoid is quickly converted
into product.
Fluorescence emission spectra of PLP-complexed Mtb DAPAS
on excitation at 280 nm result in a peak at 335 nm. The reaction
with sinefungin on excitation at 280 nm results in a decrease
in this emission peak and a subsequent increase in the peak at
€
Biochemical Characterization of Mtb DAPAS. The PLP-
dependent enzyme DAPAS transfers the amino group from SAM
to KAPA to form DAPA. Aminotransferases are generally able
to use a variety of substrates, including Asp, Glu, Met, and
aromatic amino acids Phe, Tyr, and Trp to donate amino groups
for the transamination reaction (47, 48). Besides SAM, however,
no other substrate has been found to act as an amino donor for
Mtb DAPAS, even at concentrations of 5 mM (28).
385 nm. The increase in emission at 385 nm represents a Forster
resonance energy transfer (FRET) effect in which PMP absorbs
light at 335 nm emitted by the aromatic residues and, in turn,
emits at 385 nm. This result confirms PMP formation in the
aminotransferase reaction with sinefungin and indicates that the
SAM analogue can bind at the DAPAS active site and donate its
amino group to PLP to form PMP (Figure 5). Thus, sinefungin
can act as a substrate for Mtb DAPAS.
The kinetic properties of Mtb DAPAS with the substrate SAM
have been reported (28, 49). To investigate DAPAS catalysis
further, we took advantage of the chromophore properties of the
cofactor PLP to follow the half-reaction of Mtb DAPAS, forma-
tion of the pyridoxamine phosphate (PMP) complex, with the
SAM analogues sinefungin and S-adenosylhomocysteine (SAH)
by spectroscopic methods. SAM, sinefungin, and SAH differ in
their δ carbons; SAM and SAH have a sulfonium ion at that
position (this is methylated in SAM) whereas sinefungin has an
amine (Figure 4). An absorbance scan of Mtb DAPAS showed an
absorption maximum at 420 nm, characteristic of the PLP form
of the enzyme. This absorption reduced upon reaction with 1 mM
Similar kinetic studies with the SAM analogue SAH showed
no absorption peak at 335 nm, even at concentrations up to 3 mM.
This suggests that SAH is unable to donate its amino group to
PLP and corresponds well with studies that demonstrate the same
inability of the analogue to act as the amino donor for E. coli (51)
and B. sphaericus DAPAS (8).
We also employed a single-turnover half-reaction with Mtb
DAPAS to measure the increase in absorbance at 335 nm. We
app
obtained the Km
and kmax values for sinefungin as 0.7 (
0.2 mM and 0.009/s, respectively. The equivalent values for SAM
are 0.45 ( 0.2 mM and 0.012/s. kmax/Kmapp for the Mtb DAPAS
reaction with sinefungin (0.013 mM-1 s-1) are lower than those

6750 Biochemistry, Vol. 49, No. 31, 2010
Dey et al.
Table 1: DataCollection, Refinement, andGeometryStatisticsofMtb DAPAS
Mtb-DAPAS
Mtb-DAPAS
3BV0
þ sinefungin
PDB code
3LV2
data collection
space group
P212121
P212121
unit cell dimensions
˚
a, b, c (A)
63.1, 66.5, 203.4 62.6, 83.4, 157.9
R = β = γ (deg)
90
90
molecules per ASUa
2
2
˚
wavelength (A)
1.54
1.54
˚
resolution range (A)
35-2.2
2.3-2.2
253218
40606
35-2.2
2.3-2.2
545792
43549
˚
highest resolution bin (A)
observed reflections
unique reflections
completeness (%)c
average redundancyc
I/σ(I)c
92.9 (78.4)
6.2 (3.6)
21.3 (3.5)
0.07 (0.31)
98.7 (95.7)
12.5 (12.3)
33.4 (7.0)
0.07 (0.36)
F
IGURE 5: Fluorescencespectra ofMtb DAPAS with SAM analogue
sinefungin. Fluorescence emission spectra of 12 μM DAPAS alone
(blue) and its reaction with 1 mM sinefungin (purple) were recorded
at room temperature in the 300-450 nm range at 30 s.
c
Rsym
refinement statistics (REFMAC)
R value (%)
19
18
R
free value (5%)
24
23
no. of protein residues
no. of water molecules
824
200
0.009
1.2
836
309
0.009
1.2
b
˚
rmsd bond lengths (A)
rmsdb bond angles (deg)
2
˚
average B factor (A )
Wilson plot
32.5
22.5
21.6
19.8
32.5
27.3
27.6
18
protein
water
PLP
sinefungin
31.4
Ramchandran plot (PROCHECK)
most favored region (%)
additional allowed regions (%)
generously allowed regions (%)
disallowed regions (%)
629 (90.4)
60 (8.6)
3 (0.4)
634 (90.2)
60 (8.5)
5 (0.7)
4 (0.6)
4 (0.6)
aASU, asymmetric unit. brmsd, root-mean-square deviation. cValues in
parentheses for the highest resolution bin.
catalysis, we carried out the coupled reaction of Mtb DTBS with
pyruvate kinase and lactate dehydrogenase and tracked ADP
formation by monitoring the decrease in NADH absorbance at
340 nm. We measured the Km and Vmax for ATP as 29 μM and
3.5 μM min-1 mg-1, respectively, and for DAPA as 2 μM and
6 μM min-1 mg-1, with the other substrates at saturation. The
Km for ATP in Mtb DTBS is four times higher than that of E. coli
DTBS (7 μM) (52).
F
IGURE 6: Absorption spectra of wild-type Mtb DAPAS with
DAPA. Absorption spectra of 5 μM DAPAS alone (red), immedi-
ately after addition of 0.1 mM DAPA (blue), and after 30 s of the
reaction (green) were recorded in the range of 300-600 nm at room
temperature.
Structure of Mtb DAPAS. We solved the three-dimensional
obtained with SAM (0.03 mM-1 s-1), indicating that the
analogue has a lower affinity for Mtb DAPAS than the natural
substrate and that the corresponding reaction is slower.
˚
structure of the 46 kDa Mtb DAPAS at 2.2 A resolution by
molecular replacement using E. coli DAPAS as a template. Based
on the Ramachandran plot, the structure has good geometry,
with the exception of two residues: Lys283, which is covalently
linked to PLP, and Val222. However, these residues have good
electron density in both chains. The structure was refined to an
R-factor of 19% and an Rfree of 24%. The electron densities for
the six histidine tag and the preceding six residues at the
N-terminus are not visible, perhaps the result of flexibility in a
region containing three alanine and two glycine residues. The
data collection, refinement, and geometric statistics are provided
in Table 1.
Since aminotransferases catalyze a reversible reaction, we also
studied the half-reaction of the PLP-complexed Mtb DAPAS
with the DAPA product. Between 300 and 500 nm, the absorp-
tion spectra of this reaction show a decrease in the absorption
peak at 420 nm and a subsequent increase in absorbance at
335 nm. This change in the absorption peaks represents the
formation of PMP in the single-turnover reaction with DAPA
(Figure 6). However, comparison of the absorbance spectra
reveals that this half-reaction (0.5 mM) is much slower than
the half-reaction of Mtb DAPAS with SAM (1 mM) (49).
Biochemical Characterization of Mtb DTBS. DTBS
catalyzes the carboxylation of DAPA, closing the ureido ring
to form DTB. To determine the kinetic parameters of Mtb DTBS
Like E. coli DAPAS, the Mtb DAPAS subunit structure
indicates two distinct domains, the small domain, composed of
amino acid residues 1-60 and 339-437, and the large domain,
containing residues 61-338. The N-terminus of the small domain

Article
Biochemistry, Vol. 49, No. 31, 2010 6751
0
˚
Phe93 (2.8 A) main-chain oxygens. Helix R1 of the small domain
from one subunit crosses over helix R30 of the large domain of the
other subunit, primarily involving hydrophobic interactions with
Pro, Val, and Ile residues from both. Superimposition of the two
subunits of Mtb DAPAS using 435 CR atoms results in an rmsd of
˚
0.38 A, indicating that there is no major flexibility between the
domains in contrast to what has been reported for aminotrans-
ferases of subgroup I (53, 54).
Although Mtb DAPAS, at 437 residues, is eight residues longer
than its E. coli counterpart, the two share a similar overall second-
ary structure. The orientation and length of the Mtb DAPAS
˚
loop regions (rmsd for 420 CR atoms=1.07 A) Asp51-Gly52,
Thr111-Asp116, Pro189-Asp196, and Ala307-Asn322, how-
ever, differ from those of the E. coli enzyme, as does the helix
˚
Pro197-Glu213 with a transition of 2 A. Furthermore, there are
a large number of interactions unique to Mtb DAPAS which are
not observed in E. coli DAPAS. These include a salt bridge
˚
between Arg246 and Glu204 (2.7 and 3.1 A) and hydrogen bonds
between Tyr195-OH-Asp239-OD2-Ser200-γO, Lys105-NZ-
˚
˚
Asp109 (2.9 and 2.7 A), and Trp45-NE1-Glu418-OE2 (3 A).
Analysis of DAPAS Active Site Architecture. The Mtb
˚
˚
DAPAS active site is a 24 A deep pocket with an entrance of 22 A ꢀ
˚
˚
˚
23 A that narrows to 14 A ꢀ 10 A at the bottom. The pocket
is composed of loop regions Pro24-Ser34, Arg156-Asp160,
Gln224-Gly228, Arg400-Arg403, Ala3070-Asn3220, andMet870-
His970 from the small and large domains of one subunit and the
large domain of the neighboring subunit to accommodate two
dissimilar substrates, SAM or KAPA/DAPA (the prime (0)
indicates the residues in the adjacent subunit). Residues from
both chains make up the catalytic site, and the two active sites
F
IGURE 7: Ribbon diagram representation of the PLP-bound Mtb
DAPAS crystal structure showing domain arrangement in the com-
pact dimeric form of Mtb DAPAS. The N-terminal residues of the
small domainare shown inpurpleand the C-terminalresiduesinblue.
The large domain is colored in different shades of green to differ-
entiate between the two chains. As seen, the active sites are present at
˚
are approximately 18 A apart. The entrances to the cavities are at
opposite sides of the dimer axis (Figure 7). In the Mtb structure,
the loop at the entrance of the active site (Tyr25-Val33) and the
long loop that lines the active site (Gly3070-Pro3160) are in
disorder. The latter has four alanine and three glycine residues in
the sequence at the entrance of the tunnel (AGAAGALMHGP),
rendering it flexible on solvent exposure. The contrasting lack of
flexibility at the dimer interface of the E. coli structure may
explain its higher efficiency, although there are also 13 residues
lining the Mtb DAPAS active site that are not conserved in the
E. coli enzyme.
The continuous electron density between PLP and Lys283 in
the Mtb DAPAS structure highlights an inherent feature of
aminotransferases, the covalent link between the cofactor and
the Nε group of the conserved Lys283. Like E. coli DAPAS, Mtb
DAPAS is a PLP fold type I enzyme in which all residues
interacting with PLP are conserved.
˚
the interface, 18 A apart. This dimeric interface primarily involves the
large domain interactions. PLP molecules are shown in space-filling
representation (carbon atoms are in yellow, oxygen atoms in red,
nitrogen atoms in blue, and phosphorus atoms in aquamarine color)
to indicate the position of the active site.
consists of one helix (R1) and three antiparallel β-strands (β1-
β3) connected by a long loop, and the C-terminus consists of a
helix-loop-helix containing three antiparallel β-sheets (β11-
β13) and three helices (R10-R12) which lie against the β-strands.
The large domain consists of seven β-strands (β4-β10-β9-β8-β7-
β5-β6) that are parallel with the exception of β10. The β-strands
are interconnected by helices R2-R9, which also contain the
cofactor binding site.
The asymmetric unit of the crystal consists of a homodimer
that is formed by extensive contact between the large domains of
both subunits (Figure 7). The buried surface area between these
DAPAS Substrate Binding Site. The binding site of the
amino donor in the DAPAS active site has not been reported to
date. Initial crystallization trials of the Mtb DAPAS with SAM,
both by soaking and cocrystallization, were not fruitful. Our
spectrophotometric analysis demonstrates that the SAM analo-
2
˚
subunits is approximately 4000 A , 13.6% of the total subunit
surface area. In this dimer formation, helix R2 (Pro74-Arg85) of
the large domain of subunit A is antiparallel to helix R2 of the
large domain of subunit B at a 2-fold axis of noncrystallographic
symmetry. Both ends of the helix have one hydrogen bond
contact between the main-chain oxygen of Leu84 and Asp77-
OD1 through an ordered water molecule. There are several other
hydrophobic interactions throughout this interface. Additional
subunit contact is present in the hydrogen bond between His71-
app
gue sinefungin can act as a substrate for this enzyme, with a Km
of 0.7 ( 0.2 mM for the first half-reaction (Km
app
for SAM is
0.45 ( 0.2 mM). Therefore, we determined the crystal structure of
Mtb DAPAS complexed with sinefungin at 2.2 A and refined to
˚
an R-factor of 18% and an Rfree of 23%. Sinefungin occupies the
extra electron density adjacent to the C40 atom of PLP at the
active site of DAPAS chain A. In chain B the electron density for
sinefungin is broken due to disorder. In the structure of subunit
A, PLP is covalently attached to Lys283, and there is nothing in
the PMP form that would be indicative of product formation.
˚
ND1 and Asp88-OD1 (2.7 A) at the loop region, connected with
helix R2. There are also hydrogen bonds between the N-terminal
part of the small domain and the0 large domains of the two sub-
˚
units at the His20-ND1-Asp116 (2.5 A) and the His23-ND1-

6752 Biochemistry, Vol. 49, No. 31, 2010
Dey et al.
F
IGURE 9: Substrate recognition at the Mtb DAPAS active site. (A)
Interaction of amino group donor SAM, shown in green ball and
stick form, with DAPAS. The DAPAS active site residues are shown
in yellow stick form. The amino group of SAM points toward PLP,
the carboxy group interacts with the network of Tyr157 and Asp160,
and the Tyr25 residue forms a stacking interaction with the adenine
ring of SAM. Trp64 and Trp65 have changed their orientations to
accommodate SAM at the active site. (B) Interaction of amino group
acceptor KAPA, shown in green ball and stick form, with DAPAS.
KAPA binds at a similar position as SAM at the active site. Here,
Tyr25 changes its orientation to form a hydrogen bond with the
amino group of KAPA. The carboxy group of KAPA forms a salt
bridge with Arg400.
F
IGURE 8: Precatalytic binding of sinefungin at the Mtb DAPAS
active site. This figure shows the stereoview of the precatalytic state of
sinefungin binding at the catalytic site of the Mtb DAPAS. The
electron density (2Fo - Fc) for sinefungin (light blue) is contoured at
the 1σ level. Most of the surrounding residues at the active site have
been shown. Arg315 has been truncated for the clarity of the figure.
The amino donor sinefungin is shown in ball and stick form, cofactor
PLP is shown in green, and the surrounding residues are shown in
yellow stick representation. The water molecule is shown in purple.
Oxygen atoms are shown in red, nitrogen atoms are in blue, and
phosphorus atoms are in cyan.
site for SAM in a wide active site cleft. We speculate that the
change in the unit cell dimension of the sinefungin-bound DAPAS
This may result from the short soaking period for sinefungin
(10 mM for 45 min); longer soaking or higher substrate con-
centrations resulted loss of diffraction. The sinefungin-bound
Mtb DAPAS structure thus depicts the precatalytic state of
substrate binding (Figure 8).
From chain A the amino group of sinefungin points toward the
internal aldimine in this structure, oriented to form the external
aldimine complex required for subsequent amino group dona-
tion. The carboxy oxygen atoms of sinefungin form hydrogen
bonds with the main-chain amide of Gly316 (3.4 A) and the
hydroxyl group of Tyr157 via a water molecule at the active site.
The five-membered aliphatic chain of sinefungin spans the entire
length of the active site pocket and forms van der Waals contacts
with the side chains of Trp64, Trp65, and Phe402. At the entrance
of the active site pocket, the electron density for part of the ribose
sugar and the adenine ring of sinefungin is highly disordered
from complete solvent exposure. A loop (Ser26-Ala32) near the
entrance of the pocket is also disordered, similar to that of the
native structure. However, unlike the apo structure, the loop
region in chain A (Gly3070-Pro3160) lining the active site at the
dimer interface of the sinefungin structure is ordered from the
presence of the ligand. Tyr25, which may be involved in a
stacking interaction with the adenine ring of sinefungin, is also
more ordered than it is in the PLP enzyme structure. Support for
the interaction of Tyr25 with the adenine ring is analogous to the
interaction of Tyr85 in the structure of 1-aminocyclopropane-
1-carboxylate (ACC) synthase complexed with the amine-oxy
SAM analogue [2-(aminooxy)ethyl](50-deoxyadenosin-50-yl)-
(methyl)sulfonium (AMA) (PDB code: 1M4N) (55).
˚
˚
˚
Mtb structure (a=62.6 A, b=83.4 A, c=157.9 A, R=β=γ=90ꢀ)
˚
˚
compared to the PLP-bound structure (a=63 A, b=66.5 A, c=
203.4 A, R=β=γ=90ꢀ) is the result of crystal packing.
Based on the precatalytic sinefungin-bound structure, we
˚
˚
further modeled SAM to a position 2 A inside the Mtb DAPAS
active site pocket (Figure 9A). Movement of SAM inside the
active site would allow its amino group to form a Schiff’s base
with the C40 group of PLP. The carboxy oxygen atoms of SAM
would then interact with the hydrogen bond network of Tyr157-
OH-Asp160-OD2, the aliphatic chain would span the hydro-
phobic active site pocket, and the adenine ring would form a
stacking interaction with the phenol ring of Tyr25. Most sub-
group II aminotransferases, in contrast, have hydrogen bonds
between the carboxy groups of the amino donor and arginine
residues in the active site (56, 57). The sulfonium ion of SAM may
hydrogen bond with the hydroxyl group of the conserved Tyr407,
and the oxygen atom of the ribose sugar may form similar
interactions with Arg400 and the N-atom of Trp65 imidazole ring.
We were unable to produce Mtb DAPAS crystals complexed
with KAPA, the amino group acceptor. Therefore, to locate the
substrate binding site, we superimposed the KAPA-bound E. coli
DAPAS structure (PDB code: 1QJ3) (15) on the Mtb enzyme
structure. In the E. coli DAPAS structure, KAPA is bound to the
PLP form of the enzyme, an unproductive enzyme complex (15).
Based on this superimposition, the 7-keto group of KAPA points
0
˚
˚
toward the internal aldimine and is approximately 2.6 A, and the
8-amino group forms a hydrogen bond with the hydroxyl group
0
˚
of Tyr25 (2.5 A) as well as with the main-chain oxygen of Gly316
˚
Superimposition of the sinefungin-bound Mtb DAPAS struc-
ture on the PLP-bound structure shows that the enzyme does not
undergo major conformational changes (the rmsd for 429 CR
(2.5 A) in Mtb DAPAS. Conserved residue Tyr25 is thought to
stabilize KAPA binding by forming a hydrogen bond with the
amino group. The carboxyl group of KAPA has specificity for
˚
˚
atoms is 0.3 A) except for movement of the side chains of Trp64,
the conserved Arg400, whose side chain moves 3 A toward
˚
Trp65, Met409, and Phe402 to accommodate sinefungin at the
active site pocket; this may be partially accounted for by the
flexibility of subunit B. The indole side chains of both Trp64
and Trp65 rotate 90ꢀ, causing the side chain of Met409 to move
KAPA to form a salt bridge (2.6 and 3.15 A) and break the
hydrogen bond between itself and Thr389. The movement of
the Arg400 guanidinium group in response to KAPA binding at
the active site is not seen with sinefungin binding. The rest of the
KAPA aliphatic chain is within van der Waals distance to active
˚
by 2 A. Comparison of these structures provides the putative binding

Article
Biochemistry, Vol. 49, No. 31, 2010 6753
Table 2: Data Collection, Refinement, and Geometry Statistics of Bs DAPAS
Bs-DAPAS
Bs-DAPAS þ PLP
Bs-DAPAS þ PLP þ KAPA
PDB code
3DRD
3DOD
3DU4
data collection
space group
P21
P21
P21
unit cell dimensions
˚
a, b, c (A)
58.1, 102.8, 74.5
90
58.1, 102.8, 74.5
90
57.7, 105.1, 75.1
90
R = γ (deg)
β (deg)
105.1
105.1
105.2
molecules per asymmetric unit
2
2
2
˚
wavelength (A)
1.54
1.54
1.54
˚
resolution range (A)
35-2.2
2.3-2.2
279606
41693
35-1.9
1.97-1.9
397953
65688
30-2.2
2.28-2.2
184127
43651
˚
highest resolution bin (A)
observed reflections
unique reflections
completeness (%)b
average redundancyb
I/σ(I)b
93 (63.5)
6.7 (4.4)
26.1 (3.8)
0.063 (0.32)
99.1 (98)
6.1 (6.2)
10.5 (2.8)
0.072 (0.48)
99.4 (95)
4.2 (3.5)
7.2 (2.8)
0.11 (0.37)
b
Rsym
refinement statistics (REFMAC)
R value (%)
20
20
20
R
free value (5%)
26
24
26
no. of protein residues
no. of water molecules
836
250
0.009
1.28
833
385
0.009
1.26
894
261
0.01
1.4
a
˚
rmsd bond lengths (A)
rmsda bond angles (deg)
2
˚
average B factor (A )
Wilson Plot
32.5
27.6
22.7
33.8
48.2
48.7
42.6
31.5
28.8
25.3
27.2
38.2
protein
water
PLP
KAPA
Ramachandran plot (PROCHECK)
most favored region (%)
additional allowed regions (%)
generously allowed regions (%)
disallowed regions (%)
654 (89.1)
71 (9.7)
5 (0.7)
658 (90.0)
68 (9.3)
1 (0.1)
681 (87.3)
94 (12.1)
1 (0.1)
4 (0.5)
4 (0.5)
4 (0.5)
armsd, root-mean-square deviation. bValues in parentheses for the highest resolution bin.
site residues Tyr25, Trp64, Tyr157, and Phe402 (Figure 9B). Ana-
logous to sinefungin binding, modeled KAPA binding to DA-
PAS does not result in major conformational changes (rmsd=
The reaction catalyzed by Mtb DAPAS proceeds via a ping-
pong bi-bi reaction mechanism (28). The enzyme differs from
most aminotransferases in that its amino donor, SAM, does not
resemble its amino acceptor, KAPA. To accommodate such
disparate substrates, DAPAS appears to use a local induced fit
mechanism in which certain residues at the active site change their
side chain orientations but avoid large conformational changes
in the form of domain movement. When SAM binds at the
DAPAS active site, its amino group forms a Schiff’s base with
PLP and its carboxy group hydrogen bonds to the Tyr157-OH-
Asp160-OD2 network. The conserved Tyr25 residue changes the
orientation of the phenol ring by approximately 180ꢀ to provide a
platform for stacking interactions with the adenine ring. Follow-
ing the release of SAM, KAPA binds at the same active site
region of the PMP form of DAPAS, and the Tyr25 side chain
again changes its orientation by rotating 90ꢀ clockwise to form a
hydrogen bond interaction with the 8-amino group of KAPA.
This bond positions KAPA to accept the amino group from
PMP. In response to this interaction, the guanidinium group of
˚
0.3 A for 429 CR atoms) apart from the movement of a few
residues’ side chains (Trp64, Trp65, and Met409). Higher con-
centrations of KAPA (14 ( 2 μM) inhibit DAPAS by forming a
dead end complex in the active site (28). These data suggest that
KAPA and SAM, despite differences in size and other character-
istics, bind at the same place in the active site.
Possible DAPAS Substrate Recognition. Aminotrans-
ferases use two main mechanisms, alone or in combination, to
identify dual substrates at the same active site: the “induced fit”
mechanism (58, 59), in which the conformation of the active site
changes on substrate binding, and the “lock and key” mechan-
ism (60), in which the active site and substrate are complemen-
tary to each other. Subgroup I enzymes aspartate aminotransfer-
ase (53, 54) and glutamine phenylpyruvate aminotransferase (48)
employ an induced fit mechanism, with large movement of the
small domain (13ꢀ rotation) as a rigid body for substrate recogni-
tion. Other transaminases such as lysine aminotransferase use the
glutamate switch (61), while tyrosine aminotransferase (57) and
GABA aminotransferase (56) use the arginine switch. Acetylor-
nithine aminotransferase employs a combination of a lock and key
and induced fit mechanisms for recognition of its acetylornithine
and R-ketoglutarate substrates (57).
˚
Arg400 moves 3 A toward KAPA to form a salt bridge with the
carboxy group of the substrate (Figure 9B).
Variation in the Mode of DAPAS Substrate Binding.
To confirm the role of Tyr25 in Mtb DAPAS substrate recogni-
tion suggested by our structural studies, we constructed a Y25A
mutant and spectrophotometrically followed the half-reactions

6754 Biochemistry, Vol. 49, No. 31, 2010
Dey et al.
of the mutant enzyme with SAM and DAPA. Compared to the
native protein, the Y25A mutant does not show detectable
activity at 335 nm with SAM, even up to concentrations of
3 mM, and shows approximately 70% reduced activity with high
concentrations of DAPA (0.5 mM). These results support a role
for Mtb DAPAS Tyr25 in both a stacking interaction with the
adenine ring of SAM and proper positioning of KAPA for amino
group acceptance.
In E. coli DAPAS, the Tyr to Phe mutation results in a 60-fold
reduction in enzyme activity in the second half of the reaction
with DAPA but has no effect on the first half of the reaction with
SAM (13) because Phe can still stack with SAM’s adenine ring.
Interestingly, based on sequence alignment of different DAPAS
proteins, Bacillus species B. subtilis and B. cereus as well as
Brucella melitensis and Aquifex aquenox have a Phe residue in this
position of the protein sequence, raising the question of how
substrate binding, especially KAPA binding, occurs in these
organisms.
F
IGURE 10: Stereoview of KAPA binding at the B. subtilis DAPAS
active site. Superimposition of KAPA-bound E. coli DAPAS (PDB
code: 1QJ3) in blue and B. subtilis DAPAS in orange shows the bind-
ing pattern ofKAPA inthe absence ofTyr17. KAPA isrepresentedin
stick form, dark green in E. coli and light green in Bacillus. There is a
transition of KAPA in Bacillus by approximately 1 A for hydrogen
bond formation with Tyr145.
˚
these reactions occur involves aldimine formation with PLP. This
interaction is the target for many of the most potent inhibitors
against this class of enzymes.
To investigate alternative mechanisms of KAPA binding, we
solved the crystal structure of the B. subtilis DAPAS apoprotein,
its PLP-bound complex, and its ternary complex with PLP and
˚
The natural antibiotic amiclenomycin irreversibly inhibits Mtb
DAPAS in vitro (Ki=12 ( 2 μM) (16) and is effective against Mtb
in cell culture (24). To examine the mechanism of irreversible
inhibition, we superimposed the cis-amiclenomycin-bound struc-
ture of E. coli DAPAS (PDB code: 1MLY) (16) on the Mtb
enzyme. In this model, amiclenomycin occupies the same position
as KAPA, forming an irreversible aromatic adduct with PLP by a
covalent bond with the C40 carbon atom of the cofactor and
tilting it at an angle of 17ꢀ. Hydrophobic interactions with Trp64,
Trp65, and Phe402 further stabilize the aromatic adduct. Tyr25,
however, does not seem to form hydrogen bonds with the amino
group of amiclenomycin.
KAPA at resolutions of 2.2, 1.9, and 2.2 A, respectively (Table 2).
Sequence alignment of these enzymes shows that B. subtilis
DAPAS has a 32% sequence identity with Mtb DAPAS and
33% with the E. coli enzyme. The overall structure of B. subtilis
DAPAS is similar to its Mtb and E. coli counterparts, with a
˚
pairwise rmsd of ∼1 A for 325 CR atoms. Unlike the Mtb and
E. coli DAPAS, however, a large part of the active site (Lys143-
Glu172) is completely disordered in both chains. PLP binding
does not help to order this region of the active site. However, we
were able to trace the active site of the enzyme using the KAPA-
bound structure in which electron density for these 29 residues is
clearly visible in both chains.
˚
The side chain of Arg400 translates 3 A to form a salt bridge
In the unproductive KAPA-bound B. subtilis structure, the
carboxy group of the substrate forms a salt bridge with the
˚
with the carboxy group of the antibiotic (2.8 and 2.5 A) and
˚
provides specificity to the inhibition. This may be the reason that
drugs such as cycloserine (62) and gabaculine (63, 64) do not
inhibit Mtb DAPAS despite working well against alanine race-
mase (Ki=360 μM) and GABA aminotransferase (Ki=2.86 μM),
respectively, though they employ similar mechanisms of inhibit-
ing aromatic adduct formation with PLP (28). We also find that
vigabatrin, a potent inhibitor of GABA aminotransferase and
an effective epilepsy treatment (65), is ineffective against Mtb
DAPAS, possibly because the compound is not long enough to
form bonds with both PLP and Arg400. There are no other Arg
residues lining the active site of the enzyme to provide specificity
by interacting with the carboxy groups of vigabatrin covalently
bonded to PLP.
guanidinium group of Arg403 (2.8 A). The Arg residue moves
˚
backward 2.2 A, in contrast to E. coli DAPAS. In the absence
of Tyr17, the amino group of the KAPA hydrogen binds to the
˚
OH group of Tyr146 (2.8 A) and to the main-chain oxygen of
0
˚
Gly315 (2.65 A) while maintaining the hydrogen bond network
of Tyr146-OH-Asp149-OD2. Comparing this to the KAPA-
˚
bound E. coli DAPAS, these interactions result in a 1 A shift in
the active site’s KAPA binding position where it approaches
Tyr146 for hydrogen bond formation (Figure 10). The substrate
hence stabilizes the loop region of the B. subtilis active site. In the
E. coli Y17F mutant structure (PDB code: 1S0A) (13), a water
molecule fills the role of the hydroxyl group of Tyr17 in the
active site, further indicating the importance of the Tyr157-
Asp160-Tyr25 hydrogen bond network in maintaining active
site structure. We note that two consecutive Trp residues and a
Gly are present in the Mtb and E. coli DAPAS active sites,
replacing the Val53, Trp54, and Leu820 residues found in the
B. subtilis enzyme. The void created by Val53 at the active site is
filled by Leu820 from the neighboring subunit. Thus the B. subtilis
DAPAS structures shed light on the maintenance of enzyme
activity in the absence of Tyr25. Analysis of a DAPAS phylo-
genetic tree shows that the B. subtilis enzyme evolved long before
Mtb or E. coli DAPAS, suggesting that the Tyr residue may
represent a recent adaptation toward greater efficiency.
We are currently pursuing the design of more effective
inhibitors of Mtb DAPAS using a structure-based drug design
approach including virtual screening experiments of the rigid
active site against commercially available libraries of drug-like
compounds.
Structure of Mtb DTBS. In the penultimate step of biotin
biosynthesis, DTBS carboxylates DAPA to form the ureido ring
of dethiobiotin (DTB) in an ATP- and magnesium-dependent
manner. We solved the crystal structure of the Mtb enzyme using
single anomalous dispersion (SAD) with the lanthanide metal
˚
samarium at Cu wavelength (1.54 A). The phases obtained were
˚
further extended to high-resolution native data of 1.85 A and
Search for an Mtb DAPAS Inhibitor. PLP is a cofactor for
several enzymes that catalyze diverse reactions, some of which are
implicated in human disease. Part of the mechanism by which
refined to a final R-factor of 22% and an Rfree of 26%. On
PROCHECK evaluation (43), no outliers were apparent in the

Article
Biochemistry, Vol. 49, No. 31, 2010 6755
Table 3: Data Collection, Refinement, and Geometry Statistics of Mtb DTBS
DTBS
DTBS
DTBS þ DTB
þ phosphate
DTBS-Sm
native DTBS
3FGN
þ DAPA
þ KAPA
PDB code
3FMF
3FMI
3FPA
data collection
space group
P212121
P212121
P212121
P212121
P212121
unit cell dimensions
˚
a, b, c (A)
55.9, 104.2, 152.3
90
55.4, 104.7, 151.4
90
55.3, 103.8, 151.5
90
55.3, 105.1, 151.6
90
55.1, 104.5, 151.1
90
R = β = γ (deg)
molecules per asymmetric unit
4
4
4
4
4
˚
wavelength (A)
1.54
1.54
1.54
1.54
1.54
˚
resolution range (A)
22-2.6
2.7-2.6
397900
52748
32-1.85
1.95-1.85
312728
31.5-2.05
2.12-2.05
270190
30-2.18
2.26-2.18
640379
35- 2.3
˚
highest resolution bin (A)
2.4-2.3
315813
39525
observed reflections
unique reflections
completeness (%)b
average redundancyb
I/σ(I)b
73844
50636
46066
100 (99.8)
7.5 (7.3)
9.5 (4.9)
0.15 (0.41)
99 (93.7)
4.2 (2.3)
17.2 (1.4)
0.16 (0.54)
91.2 (80.6)
5.3 (4.9)
29.5 (8.4)
0.05 (0.17)
97.6 (96.3)
13.9 (13.5)
38.0 (7.3)
0.06 (0.39)
99.9 (99.8)
8.0 (7.7)
17.0 (2.0)
0.05 (0.21)
b
Rsym
refinement statistics (REFMAC)
R value (%)
21
20
21
21
R
free value (5%)
25
24
27
26
no. of protein residues
no. of water molecules
907
320
0.01
1.3
904
298
0.01
1.3
904
195
0.016
1.8
904
180
0.012
1.6
a
˚
rmsd bond lengths (A)
rmsda bond angles (deg)
2
˚
average B factor (A )
Wilson plot
31.5
35.5
31.7
25.3
23.3
21.3
22.5
40.4
36.7
31.7
45.3
37.2
32.1
protein
water
DAPA
KAPA
34.7
DTB
44.0
33.2
PO4
Ramachandran plot (PROCHECK)
most favored region (%)
additional allowed regions (%)
generously allowed regions (%)
disallowed regions (%)
726 (95.2)
36 (4.7)
1 (0.1)
0
714 (95.1)
36 (4.8)
1 (0.1)
0
712 (94.9)
37 (4.9)
1 (0.1)
0
696 (92.4)
50 (6.6)
5 (0.7)
0
armsd, root-mean-square deviation. bValues in parentheses for the highest resolution bin.
disallowed region of the Ramachandran plot. The data collec-
tion, refinement, and geometry statistics of this enzyme are pro-
vided in Table 3.
subunit-subunit interface, the inner helices, R5 and R50, of both
subunits, and the outer helices, R6 and R30 and R3 and R60, which
are perpendicular to each other. Because the dimer forms at the
2-fold noncrystallographic axis, all residual contact occurs twice.
The interface of helices R5 and R50 involves a stacking interaction
of the imidazole rings of His148 and His1480. The interface of
helices R3 and R60 ma0kes hydrogen bond contacts between
The asymmetric unit of the Mtb DTBS structure has four
molecules, and each pair of molecules (chains A and B and chains
C and D) forms a dimer. The electron density for the entire
polypeptide chain is well-defined in all four subunits. Each
subunit of the 226 amino acid DTBS consists of one globular
domain containing seven parallel β-sheets (β3-β2-β4-β1-β5-β6-
β7) interconnected with R-helices. Like other synthetases, the
structure consists of the classical P-loop motif (also called the
Walker A motif), Gly8-X-Gly10-X-Gly12-Val13-Gly14-Lys15-
Thr16, which binds the phosphate group of the nucleotide. Here,
the motif is in a loop region between strand β1 and helix R1 at
the N-terminus of the protein. There is also a Walker B motif
at strand β4 (Leu104-Thr105-Leu106-Val107-Glu108). The
˚
Arg188-NH2 and Glu79 -OE2 (3.5 A) and hydrophobic interac-
tions with Met72, Leu177, and Val178. Other interactions at the
dimer interface include hydrogen bonding between the Thr9
˚
main-chain oxygen of the P-loop and His148-NE2 (2.7 A),
hydrogen bonding between the Asn147-OD1 of one monomer
and the main-chain amide nitrogen of Leu114 of the other
˚
(2.9 A), and a salt bridge between Arg125 and Glu154 of both
˚
chains (2.85 and 3.0 A). Superimposition of chain A on chain B of
the dimer shows little variation in the overall structure (rmsd=
˚
dimer formation occurs in a buried surface area of approximately
0.62 A), indicating a lack of flexibility.
2
˚
˚
1300 A (7.5% of total surface area of the dimer). The 37.2 A
elongated dimer interface mainly involves the R3 (Pro74-Ala81),
R6 (Gly144-Gln159), and R7 (Leu177-Ile189) helices, the loop
(Gly109-Leu124) between strand β5 and helix R5, and the
phosphate-binding loop (Thr9-Gly14) between strand β1 and
helix R1 of both the subunits (the prime (0) indicates the residues
in the adjacent subunit). There are six helices present at the
Binary Complexes of DTBS with DAPA and with Its
Analogue KAPA. To understand the substrate-binding mode at
the Mtb DTBS active site, we soaked the enzyme crystals with the
˚
DAPA substrate. The 2.2 A crystal structure of Mtb DTBS
reveals a long, distinct density at the dimer interface that fits
DAPA in all four chains of the asymmetric unit as well as
additional electron density attached to the N7 atom of DAPA

6756 Biochemistry, Vol. 49, No. 31, 2010
Dey et al.
F
IGURE 11: Mode of DAPA-carbamate binding at the Mtb DTBS active site. (A) There is clear electron density (2Fo - Fc) for DAPA-carbamate,
shown in yellow stick form, at the 3 σ level. (B) Residues interacting with the substrate DAPA-carbamate in the DTBS active site are shown in this
figure. Thisplot has beenprepared using Ligplot (72). The protein residuesare colored orange, and the ligand DAPA-carbamateiscolored purple.
The water molecules are colored cyan. Hydrogen bonds are shown in green dotted lines. Hydrophobic interactions are shown in brick red
sunbursts. Atoms are colored by type (C, black; O, red; N, blue; S, yellow).
that fits the carbon dioxide molecule (Figure 11). The carboxy
group of the carbamate occupies the position of the water molecule
(W11) at the active site of the apo structure, and the carboxy
oxygen atoms are within hydrogen bond distance to the catalytic
Similar interactions are also visible for KAPA at the DTBS
active site. As evident from the weak electron density for KAPA
and its high B factor, however, the occupancy of KAPA (0.5) is
much lower than DAPA (1.0) at the active site of all four chains
of the asymmetric unit. The low occupancy of KAPA suggests
that it may have low affinity as a substrate for DTBS.
˚
Lys37 (3 and 3.4 A) and to the Thr41 main-chain amide nitrogen
˚
(2.8 A). There are hydrogen bond links between the N7 atom of
˚
DAPA-carbamate and Thr41-OG1 (2.89 A) and between the N8
The KAPA-bound enzyme structure has a sulfate ion from
the crystallization solution bound at the P-loop (Gly8-Thr16),
mimicking the β-phosphate of the nucleotide. The 7-keto oxygen
atom of KAPA points toward the sulfate ion with an ordered
water molecule occupying a similar position to the γ-phosphate
atom and Thr11-OG1, both through water molecules. The long
aliphatic chain of this unnatural amino acid spans the active site
pocket, which is located at the dimer interface and is composed of
a stretch of residues from helix R6 (Leu1430-Asn1470) and loops
Pro71-Pro74 and Ala110-Gly112.
˚
of the nucleotide at the active site (3.2 A). The N8 atom of KAPA
˚
One of the C1-carboxy oxygen atoms of DAPA coordinates
forms a hydrogen bond with Thr41-OG1 (3 A) at the active site.
0
˚
with the main-chain amide of Leu146 (2.78 A) and an ordered
Thus, KAPA and DAPA occupy similar positions in the hydro-
phobic pocket of the enzyme, away from the P-loop.
˚
water molecule (2.6 A), while the other hydrogen bonds with the
0
˚
main-chain amide of Asn147 (2.6 A) in helix R6 of the adjacent
The substrate-bound Mtb DTBS structures illustrate that the
first substrate, DAPA, can bind to the enzyme in the absence of
the second substrate, ATP. They may also indicate Mtb DTBS
specificity for N7-DAPA-carbamates as the crystallization con-
ditions did not introduce bicarbonate.
DTBS Interactions with ATP/ADP. To further character-
ize the interaction of ATP with Mtb DTBS and the resulting
conformational changes upon binding, we attempted to deter-
mine the crystal structures of the enzyme bound to ADP or to the
ATP analogues AMP-PCP and ATP-γ-S. Only partial electron
density for these entities at the active site was visible, phosphate
subunit and with the R7 helix’s Asn1820-OD1 through a neigh-
˚
boring water molecule (3 A). This water molecule is absent from
the active site of the Mtb DTBS apo structure. The presence of
the water molecule causes a peptide flip at Gly111 of the loop
region 110-112, a phenomenon not seen in E. coli DTBS despite
identical sequence in the 106-112 region (LVEGAGG). Finally,
Tyr1870 in helix R6 of the substrate-bound E. coli structure,
replaced by Asn1820 in the corresponding Mtb structure, changes
its side chain orientation approximately 90ꢀ upon DAPA binding
and hydrogen bonds with the carboxy oxygen of DAPA.

Article
Biochemistry, Vol. 49, No. 31, 2010 6757
˚
in the orientations of the Thr9-Gly14 (2.4 A), Gly109-Leu113
0
0
˚
˚
(1 A), and Thr140 -Gly144 (0.8 A) loop regions upon substrate
binding, and these alterations appear to compress the active site
for better interactions.
Analysis of the DTBS-Product Complex at the Active
Site. To study the mechanism of Mtb DTBS catalysis, we
attempted to capture the ternary complex of the products by
soaking the enzyme crystal with DTB, ADP, and Mg2þ. The
˚
2.3 A electron density map of Mtb DTBS clearly reveals the posi-
tion of the product DTB just opposite the P-loop in the active site.
As with DAPA, the C1-carboxyl group of DTB buried inside the
pocket forms a hydrogen bond with the main-chain amide
F
IGURE 12: Surface representation of the Mtb DTBS active site. The
Mtb DTBS active site is composed of both subunit A, colored in gray,
and subunit B, colored in rose brown. DTB is positioned inside the
pocket and is anchored to the other subunit. Adjacent to DTB, Pi and
ADP are completely exposed to the solvent. DTB, Pi, and ADP
molecules are shown in stick form.
0
0
0
˚
˚
nitrogen atoms of Leu146 (2.8 A), Asn147 (2.6 A), and Asn182
via a water molecule. This water molecule also causes a peptide
flip at Gly111 such that the main-chain oxygen of the residue
forms a hydrogen bond with Thr7-OG1. The aliphatic chain of
DTB spans the two subunits, making hydrophobic contacts with
Pro74, Ala73, Gly111, and Val115. The N7 atom of the ureido
ring of DTB is hydrogen bonded to the Asp47-OD1 of the P-loop
ions were clear at the P-loop, but electron density for the rest of
the molecules was poorly defined. Hence, to pinpoint the residues
involved in the interactions of the nucleotides at the active site, we
superimposed the ADP-complexed E. coli DTBS structure (PDB
code: 1DAM) (18) on the Mtb structure. Surface representation
˚
(3.3 A), and the N8 atom is hydrogen bonded to Thr11-OG1
˚
(3.4 A). This breaks the salt bridge between Arg45 and Asp47 at
˚
of the Mtb DTBS active site shows that the 17 A long, nucleo-
the active site. Comparison of the DAPA-bound and the DTB-
and ADP-bound Mtb DTBS structures shows clearly that the
substrate and product occupy the same position in the active site.
The carboxy end of the tail of the ligand is always involved in
anchoring to the other subunit of the dimer, and the diamino end
participates in catalysis. The DTB aliphatic chain is in a different
confirmation when compared to E. coli DTBS because the N1-
ureido ring hydrogen bonds with Asp47 rather than Thr41 (18).
The Mtb DTBS active site is lined by a short loop (Gly42-
Asp47) whereas the active site of the E. coli enzyme is an exten-
sion of two antiparallel β-strands (Gly42-Ser53). There are three
continuous Asp residues in this region of the Mtb DTBS active
site: Asp47, which hydrogen bonds with Arg45 as well as the
substrate or product, Asp48, which hydrogen bonds with Arg67
tide-binding region is completely exposed to solvent, resulting
in disordered electron density for ATP/ADP (Figure 12). This
might be due to Mtb variation in the conformations of the Gln168-
Val176 loop region and of Leu196-Asp204 near the adenine ring
binding pocket. A unique feature of the loop region is the pre-
sence of multiple proline residues, Pro172, Pro174, and Pro175,
which likely decrease loop flexibility. The only apparent interac-
tion of DTBS with the nucleotide ribose sugar occurs between the
O20 atom and the carboxy group of Glu52 via a water molecule.
˚
This glutamic acid residue is present on helix R2 and is 5.3 A
closer to the active site than its E. coli counterpart, Glu211. The
E. coli DTBS Glu211 residue makes direct hydrogen bond
contact with the O20 of the nucleotide ribose sugar and is present
in the loop region (Trp209-Glu212) between strand β7 and helix
R9. This interaction orders the ADP-bound E. coli DTBS loop
region (PDB code: 1DTS), an event that has been reported for
other GTPases (66, 67). Intriguingly, the unliganded Mtb DTBS
structure also displays clear density for this loop, possibly the
contribution of the long helix R7 that is absent in the E. coli
enzyme and resides just behind the loop region. The nitrogen
atoms of the adenine ring make hydrogen bond interactions with
the peptide loop (Leu196-Gly199), the N1 atom with the main-
˚
(3.5 and 3.3 A), and Asp49-OD1, which hydrogen bonds with
Lys37-NZ (2.9 A) or a Mg2þ ion.
˚
It appears that the phosphate ion present in the crystallization
condition of the DTB- and ADP-bound Mtb DTBS structure
outcompetes ADP for the nucleotide-binding site. However,
there is one phosphate ion bound at the P-loop that is similar
to the β-phosphate of ADP and a second phosphate ion present
between the P-loop-bound phosphate and DTB that occupies the
γ-phosphate position. This mimics the position of the intermedi-
ate DTB-Pi-ADP complex at the active site just prior to
product release. The phosphate bound at the P-loop causes all
of the main-chain amide nitrogen atoms in this region, Thr11-
Lys15, to orient inward and create a positively charged environ-
ment for tight binding of the nucleotide phosphate group. The
Thr11 side chain also changes its rotamer 90ꢀ in response to
hydrogen bond formation with the γ-phosphate-O4, with a
˚
chain amide of Gly199 (2.7 A), and the N7 atom with the main-
chain oxygen of Gly169 (2.7 A). Superimposition of the ADP-
˚
complexed E. coli DTBS structure on the Mtb enzyme structure
thus shows that the solvent-exposed nucleotide-binding site is
located adjacent to the hydrophobic DAPA binding pocket in
Mtb DTBS.
DAPA and ATP have distinct binding sites in both E. coli and
Mtb DTBS. When the γ-phosphate of ATP is bound to one mono-
˚
transition of 1 A toward the active site. These phosphate ions
˚
mer, the distance (23.8 A) to the carboxy group of the DAPA-
are kept in position by conserved lysine residues (15 and 37) and
the magnesium ion. Analysis of the DTB- and phosphate-bound
Mtb DTBS structure thus demonstrates that the P-loop is
involved in binding both substrates at the enzyme active site,
providing a rationale for the unusually long peptide sequence at
this region.
DTBS, like other nucleotide-binding proteins, uses divalent
metal ions such as Mg2þ for stabilizing the phosphate groups,
mostly β and γ, of the nucleotides. The DTB- and ADP-bound
Mtb DTBS structure displays octahedral coordination of the
carbamate bound to the same monomer is too far to perform
the catalytic reaction. We therefore hypothesize that the two
DTBS active sites are composed of two subunits each such that
the enzyme reaction combines the ATP from one monomer with
the DAPA substrate from the other. The active sites of the enzyme
are approximately 13 A apart at the dimer interface and have
entrances facing the same side (Figure 12).
Substrate binding does not produce drastic changes in DTBS
conformation. However, there are local conformational changes
˚

6758 Biochemistry, Vol. 49, No. 31, 2010
Dey et al.
Asn, respectively. In E. coli DTBS, Lys148 forms a salt bridge
with Glu12. Binding of the nucleotide to the Walker motif breaks
this bond, and the Glu12 side chain turns toward the β-phosphate
of ATP and DAPA-carbamate at the active site. Indeed, Lys148
stabilizes the P-loop by supporting the pocket formed at the back
of the loop through its interaction with Glu12. The Mtb DTBS
has glycine and aspartate residues in place of Glu12 and Lys148
and therefore lacks the ionic interaction found in the active site of
the E. coli enzyme.
Possible Mechanism of DTBS Reaction. Based on the Mtb
DTBS structures complexed with substrate and products, we can
outline a reaction mechanism for this amidoligase. Like E. coli
DTBS catalysis (68), the Mtb enzyme reaction involves three
steps: carbamylation of the N7 position of DAPA, formation of a
carbamate-phosphoric anhydride intermediate, and, ultimately,
closure of the ureido ring to form DTB and release of the
inorganic phosphate.
F
IGURE 13: Superimposition of CR atoms of E. coli and Mtb DTBS
monomers. Ribbon diagram of an overlay of E. coli DTBS (PDB
code: 1BYI) structure (cyan) on Mtb DTBS (blue). All core β-strands
(gold) from both of the structures overlay very well. The arrows point
toward the major differences in the two structures (red). The rmsd for
Carbamylation is possible at both the N7 and N8 positions of
DAPA in solution (69). The DAPA-bound Mtb DTBS structure
suggests that this enzyme has specificity for N7-carbamate, but it
is not clear whether the carbamylation is catalyzed or whether it
happens freely in solution. In the next step of the reaction, one of
the carbamate oxygen atoms attacks the γ-phosphate of ATP,
evident by the positions of the carbamate-carboxyl group and
Pi in the DAPA- and DTB-complexed structures. Protonation of
the linking oxygen atom facilitates cleavage of the bond between
the β- and γ-phosphates of ATP. A water molecule (W221 in
chain A) in the active site abstracts one of the protons from the
adjacent N8 amino group. The Lys37 remains in close enough
proximity to the phosphate-carbamate anhydride intermediate
to stabilize the complex. Closure of the DTB ureido ring requires
a second deprotonation of the N8 atom of the carbamate
phosphoric anhydride for attack on the carboxy carbon. Pre-
viously, it was thought that Asp47 acted as a base for this proton
abstraction. Asp47 is present in the loop region of the active site,
a unique feature of this DTBS enzyme. However, kinetic studies
show that a D47A mutation does not abolish enzyme activity,
suggesting that the residue may not be involved in proton
abstraction (data not shown). In the Mtb enzyme the lack of a
carbamate phosphate anhydride-complexed DTBS prevents un-
ambiguous identification of the residue responsible for proton
abstraction. However, based on a structural comparison of
carbamate phosphate anhydride complex and native E. coli
DTBS (19) as well as analogy to the p21ras protein mechan-
ism (70, 71), the oxygen atom on the phosphate of the inter-
mediate complex may act as a base. A second oxygen atom in the
same phosphate group might extract another proton from the N8
atom of the complex after ureido ring closure, thereby neutraliz-
ing one of its negative charges, releasing the inorganic phosphate,
and forming DTB.
˚
224 CR atoms for both chains is 1.79 A.
Mg2þ ion between the two oxygen atoms from each of the β- and
γ-phosphate groups and the side chains of residues Thr16-OG1,
Asp49-OD1, and Glu108-OE1. These residues are conserved in
E. coli DTBS and nearly conserved in Helicobacter pylori DTBS,
with Asp49 replaced by Asn46.
Comparison with Other DTBS Structures. Superimposi-
tion of the Mtb and E. coli DTBS structures (PDB code: 1BYI,
sequence identity 28%) results in an rmsd of approximately
˚
1.79 A and reveals significant structural differences, justifying the
failure of molecular replacement using the E. coli DTBS structure
as a search model (Figure 13). Interestingly, when the Mtb DTBS
structure was overlaid with a recently solved DTBS structure
from H. pylori (PDB code: 2QMO, sequence identity 18%, rmsd=
˚
1.05 A for 111 CR atoms), similar differences in the overall
structure were found. First, there are two extended antiparallel
β-strands of unknown function between residues Gly42-Ser53 at
the E. coli DTBS active site in place of the short loop (Gly42-
Asp47) present in Mtb. The active site of the H. pylori enzyme
contains a long loop (Lys35-Ser49) in this region. Second,
whereas the E. coli and H. pylori enzymes have a long loop
followed by a short, one-turn helix in residues Arg62-Pro72
(E. coli) and Leu61-Ile70 (H. pylori), Mtb DTBS has a single
β-strand (β3) in the corresponding Gln61-Leu65 region. Third,
there is a stretch of three short one-turn helices connected by
loops in the E. coli DTBS C-terminal region (Pro209-Asn224),
with some disordered parts in both the apo and substrate
complexed enzymes (PDB codes: 1DTS, 1DAI, 1DAE), while
the Mtb DTBS has a long helix (R9, Ala205-Ala215) followed by
a short one-turn helix in the same region. The short stretches of
helix at the C-terminus of E. coli DTBS might be responsible for
its flexibility and subsequent disordered electron density upon
solvent exposure. The H. pylori DTBS has an entirely different
structure at its C-terminus, a short loop with one-turn helix in lieu
of the β7 strand and R-helices 8 and 9 of the Mtb enzyme. Since
we have seen that the adenine ring of the nucleotide binds to this
region in Mtb and E. coli DTBS, the exact mode of nucleotide
binding in the H. pylori DTBS is unclear and awaits elucidation
by a nucleotide-bound enzyme structure.
Implications for DTBS Inhibitors. The natural antibiotic
acidomycin has a chemical structure similar to DTB and
biotin and has good whole cell activity against Mtb (MIC=
0.0625-0.125 μg/mL) (25). However, our studies show that
Mtb DTBS is not the target of acidomycin as the drug fails to
inhibit the DTBS reaction at concentrations up to 200 μM.
Low yield of the enzyme precluded further kinetic analysis
with different compounds. Future candidates for Mtb DTBS
inhibitors include DAPA carbamate analogues and vari-
able length analogues of DAPA and KAPA. Mimics of the
carbamate-phosphoric anhydride complex may also be good
inhibitors.
The residues in the P-loop motif are different in all three DTBS
proteins: the Mtb enzyme has glycine residues and its E. coli and
H. pylori counterparts have charged residues, Glu and Asp, and

Article
Biochemistry, Vol. 49, No. 31, 2010 6759
CONCLUSION
Badcock, K., Basham, D., Brown, D., Chillingworth, T., Connor, R.,
Davies, R., Devlin, K., Feltwell, T., Gentles, S., Hamlin, N., Holroyd,
S., Hornsby, T., Jagels, K., Krogh, A., McLean, J., Moule, S.,
Murphy, L., Oliver, K., Osborne, J., Quail, M. A., Rajandream,
M. A., Rogers, J., Rutter, S., Seeger, K., Skelton, J., Squares, R.,
Squares, S., Sulston, J. E., Taylor, K., Whitehead, S., and Barrell,
B. G. (1998) Deciphering the biology of Mycobacterium tuberculosis
from the complete genome sequence. Nature 393, 537–544.
8. Izumi, Y., Kano, Y., Inagaski, K., and Yamada, H. (1981) Char-
acterisation of biotin biosynthetic enzymes of Bacillus sphaericus: a
desthiobiotin producing bacterium. Agric. Biol. Chem 45, 1983–1989.
9. Van Arsdell, S. W., Perkins, J. B., Yocum, R. R., Luan, L., Howitt,
C. L., Chatterjee, N. P., and Pero, J. G. (2005) Removing a bottleneck
in the Bacillus subtilis biotin pathway: bioA utilizes lysine rather than
S-adenosylmethionine as the amino donor in the KAPA-to-DAPA
reaction. Biotechnol. Bioeng. 91, 75–83.
10. Phalip, V., Kuhn, I., Lemoine, Y., and Jeltsch, J. M. (1999) Char-
acterization of the biotin biosynthesis pathway in Saccharomyces
cerevisiae and evidence for a cluster containing BIO5, a novel gene
involved in vitamer uptake. Gene 232, 43–51.
11. Pinon, V., Ravanel, S., Douce, R., and Alban, C. (2005) Biotin
synthesis in plants. The first committed step of the pathway is
catalyzed by a cytosolic 7-keto-8-aminopelargonic acid synthase.
Plant Physiol.y 139, 1666–1676.
12. Stoner, G. L., and Eisenberg, M. A. (1975) Purification and properties
of 7,8-diaminopelargonic acid aminotransferase. J. Biol. Chem. 250,
4029–4036.
13. Sandmark, J., Eliot, A. C., Famm, K., Schneider, G., and Kirsch, J. F.
(2004) Conserved and nonconserved residues in the substrate binding
site of 7,8-diaminopelargonic acid synthase from Escherichia coli are
essential for catalysis. Biochemistry 43, 1213–1222.
14. Eliot, A. C., Sandmark, J., Schneider, G., and Kirsch, J. F. (2002) The
dual-specific active site of 7,8-diaminopelargonic acid synthase and
the effect of the R391A mutation. Biochemistry 41, 12582–12589.
15. Kack, H., Sandmark, J., Gibson, K., Schneider, G., and Lindqvist, Y.
(1999) Crystal structure of diaminopelargonic acid synthase: evolu-
tionary relationships between pyridoxal-50-phosphate-dependent
enzymes. J. Mol. Biol. 291, 857–876.
We have determined the crystal structures of two Mtb biotin bio-
synthetic enzymes, 7,8-diaminopelargonic acid synthase (DAPAS)
and dethiobiotin synthetase (DTBS). These enzymes are promis-
ing drug targets as the biotin biosynthesis pathway is required for
Mtb survival during infection but absent in humans. Moreover,
abolition of biotin production leads to irreversible cell death
when Mtb is starved for the compound, suggesting that inhibitors
of the pathway could be bactericidal.
Beyond their utility as potential drug targets, the Mtb DAPAS
and DTBS crystal structures yield basic biological insights into
their mechanisms of catalysis. The mode by which the DAPAS
recognizes dual substrates, the amino donor SAM and acceptor
KAPA, was not known despite the availability of the E. coli crystal
structure. The Mtb and E. coli DAPAS structures have similar
folds and conserved active site residues, making comparison
studies feasible. Superimposition of the Mtb DAPAS structure
bound to the SAM analogue sinefungin on the KAPA-complexed
E. coli enzyme indicates that the mechanism of substrate recogni-
tion for this enzyme is different than other aminotransferases.
Both Mtb and E. coli DAPAS have an active site tyrosine that is
critical for substrate interaction. Interestingly, the DAPAS of
certain Bacillus species lack this residue. We solved the apo and
KAPA-bound B. subtilis DAPAS structures to show that there is
an alternate KAPA binding pattern in the absence of the tyrosine.
The B. subtilis, E. coli, and Mtb DAPAS structures also reveal
the importance of the hydrogen bond network Tyr25-Tyr157-
Asp160 for maintenance of the active site structure.
The overall fold of Mtb DTBS differs from other organisms.
On the basis of our substrate- and product-bound DTBS struc-
tures, we identified key components required for catalysis, some
of which are not conserved in E. coli, and from these derived an
enzyme mechanism.
Finally, we have found that the active site of Mtb DTBS is
more solvent exposed than Mtb DAPAS, which has a tunneled
active site. Our characterization of the Mtb DAPAS and DTBS
active sites and of the mechanism by which they bind substrates
and products is a starting point in the design of new inhibitors
against these enzymes and, by extension, a metabolic pathway
that is critical for Mtb survival.
16. Sandmark, J., Mann, S., Marquet, A., and Schneider, G. (2002)
Structural basis for the inhibition of the biosynthesis of biotin by
the antibiotic amiclenomycin. J. Biol. Chem. 277, 43352–43358.
17. Huang, W., Lindqvist, Y., Schneider, G., Gibson, K. J., Flint, D., and
Lorimer, G. (1994) Crystal structure of an ATP-dependent carboxy-
˚
lase, dethiobiotin synthetase, at 1.65 A resolution. Structure 2,
407–414.
18. Kack, H., Sandmark, J., Gibson, K. J., Schneider, G., and Lindqvist,
Y. (1998) Crystal structure of two quaternary complexes of dethio-
biotin synthetase, enzyme-MgADP-AlF3-diaminopelargonic acid
and enzyme-MgADP-dethiobiotin-phosphate; implications for cata-
lysis. Protein Sci. 7, 2560–2566.
19. Kack, H., Gibson, K. J., Lindqvist, Y., and Schneider, G. (1998)
Snapshot of a phosphorylated substrate intermediate by kinetic
crystallography. Proc. Natl. Acad. Sci. U.S.A. 95, 5495–5500.
20. Alexeev, D., Baxter, R. L., Smekal, O., and Sawyer, L. (1995)
Substrate binding and carboxylation by dethiobiotin synthetase;a
kinetic and X-ray study. Structure 3, 1207–1215.
21. Ashkenazi, T., Widberg, A., Nudelman, A., Wittenbach, V., and
Flint, D. (2005) Inhibitors of biotin biosynthesis as potential herbi-
cides: part 2. Pest Manag. Sci. 61, 1024–1033.
22. Ashkenazi, T., Pinkert, D., Nudelman, A., Widberg, A., Wexler, B.,
Wittenbach, V., Flint, D., and Nudelman, A. (2007) Aryl chain
analogues of the biotin vitamers as potential herbicides. Part 3. Pest
Manag. Sci. 63, 974–1001.
23. Rendina, A. R., Taylor, W. S., Gibson, K., Lorimer, G., Rayner,
D., Lockett, B., Kranis, K., Wexler, B., Marcovici-Mizrahi, D.,
Nudelman, A., Nudelman, A., Marsilii, E., Chi, H., Wawrzak, Z.,
Calabrese, J., Huang, W., Jia, J., Schneider, G., Lindqvist, Y., and
Yang, G. (1999) The design and synthesis of inhibitors of dethiobiotin
synthetase as potential herbicides. Pestic. Sci. 55, 236–247.
24. Okami, Y., Kitahara, T., Hamada, M., Naganawa, H., and Kondo,
S. (1974) Studies on a new amino acid antibiotic, amiclenomycin.
J. Antibiot. (Tokyo) 27, 656–664.
25. Grundy, W. E., Whitman, A. O., Rdzok, E. G., Hanes, M. E., and
Sylvester, J. C. (1952) Actithiazic acid. I. Microbiological studies.
Antibiot. Chemother. II, 399–408.
26. Kitahara, T., Hotta, K., Yoshida, M., and Okami, Y. (1975) Bio-
logical studies of amiclenomycin. J. Antibiot. (Tokyo) 28, 215–221.
27. Hwang, K. (1952) Actithiazic acid: pharmacological Studies. Antibiot.
Chemother. II, 453–459.
ACKNOWLEDGMENT
The authors thank Tracey Musa and Siaska Castro for assi-
stance on the manuscript and Dr. Kimberly D. Grimes for the
preparation of acidomycin.
REFERENCES
1. Moss, J., and Lane, M. D. (1971) The biotin-dependent enzymes. Adv.
Enzymol. Relat. Areas Mol. Biol. 35, 321–442.
2. Gomez, J. E., and McKinney, J. D. (2004) M. tuberculosis persistence,
latency, and drug tolerance. Tuberculosis (Edinburgh) 84, 29–44.
3. Keer, J., Smeulders, M. J., Gray, K. M., and Williams, H. D. (2000)
Mutants of Mycobacterium smegmatis impaired in stationary-phase
survival. Microbiology 146 (Part 9), 2209–2217.
4. Betts, J. C., Lukey, P. T., Robb, L. C., McAdam, R. A., and Duncan,
K. (2002) Evaluation of a nutrient starvation model of Mycobacter-
ium tuberculosis persistence by gene and protein expression profiling.
Mol. Microbiol. 43, 717–731.
5. Sassetti, C. M., and Rubin, E. J. (2003) Genetic requirements for
mycobacterial survival during infection. Proc. Natl. Acad. Sci. U.S.A.
100, 12989–12994.
6. Chiang, S. L., and Mekalanos, J. J. (1998) Use of signature-tagged
transposon mutagenesis to identify Vibrio cholerae genes critical for
colonization. Mol. Microbiol. 27, 797–805.
7. Cole, S. T., Brosch, R., Parkhill, J., Garnier, T., Churcher, C., Harris,
D., Gordon, S. V., Eiglmeier, K., Gas, S., Barry, C. E., 3rd, Tekaia, F.,

6760 Biochemistry, Vol. 49, No. 31, 2010
Dey et al.
28. Mann, S., and Ploux, O. (2006) 7,8-Diaminoperlargonic acid amino-
transferase from Mycobacterium tuberculosis, a potential therapeutic
target. Characterization and inhibition studies. FEBS J. 273, 4778–
4789.
29. Cleland, W. W. (1979) Optimizing coupled enzyme assays. Anal.
Biochem. 99, 142–145.
52. Eisenberg, M. A., and Krell, K. (1979) Dethiobiotin synthetase.
Methods Enzymol. 62, 348–352.
53. Jager, J., Moser, M., Sauder, U., and Jansonius, J. N. (1994) Crystal
structures of Escherichia coli aspartate aminotransferase in two
conformations. Comparison of an unliganded open and two liganded
closed forms. J. Mol. Biol. 239, 285–305.
30. Mach, H., Middaugh, C. R., and Lewis, R. V. (1992) Statistical
determination of the average values of the extinction coefficients of
tryptophan and tyrosine in native proteins. Anal. Biochem. 200, 74–80.
31. Sandalova, T., Schneider, G., Kack, H., and Lindqvist, Y. (1999)
54. Hohenester, E., and Jansonius, J. N. (1994) Crystalline mitochondrial
aspartate aminotransferase exists in only two conformations. J. Mol.
Biol. 236, 963–968.
55. Capitani, G., Eliot, A. C., Gut, H., Khomutov, R. M., Kirsch, J. F.,
and Grutter, M. G. (2003) Structure of 1-aminocyclopropane-
1-carboxylate synthase in complex with an amino-oxy analogue of
the substrate: implications for substrate binding. Biochim. Biophys.
Acta 1647, 55–60.
˚
Structure of dethiobiotin synthetase at 0.97 A resolution. Acta
Crystallogr., Sect. D: Biol. Crystallogr. 55, 610–624.
32. Otwinowski, Z., and Minor, W. (1997) Methods Enzymol. 276,
307–326.
33. Pflugrath, J. W. (1999) The finer things in X-ray diffraction data
collection. Acta Crystallogr., Sect. D: Biol. Crystallogr. 55, 1718–
1725.
34. Leslie, A. G. W. (1992) in Jt. CCP4/ESF-EAMCB Newslett. Protein
Crystallogr.
35. Evans, P. R. (1993) Data Collection and Processing, Daresbury
Laboratory, Warrington, U.K.
36. Storoni, L. C., McCoy, A. J., and Read, R. J. (2004) Likelihood-
enhanced fast rotation functions. Acta Crystallogr. 60, 432–438.
37. Schneider, T.R., and Sheldrick, G.M. (2002) Substructure solution
with SHELXD. Acta Crystallogr. D 58, 1772–1779.
56. Toney, M. D., Pascarella, S., and De Biase, D. (1995) Active site
model for gamma-aminobutyrate aminotransferase explains sub-
strate specificity and inhibitor reactivities. Protein Sci. 4, 2366–2374.
57. Hirotsu, K., Goto, M., Okamoto, A., and Miyahara, I. (2005) Dual
substrate recognition of aminotransferases. Chem. Rec. 5, 160–172.
58. Koshland, D. E. (1958) Application of a theory of enzyme specificity
to protein synthesis. Proc. Natl. Acad. Sci. U.S.A. 44, 98–104.
59. Koshland, D. E. (1994) The key-lock theory and the induced fit
theory. Angew. Chem., Int. Ed. Engl. 33, 2375–2378.
60. Fischer, E. (1894) Einfluss der configuration auf die wirkung der
enzyme. Ber. Deut. Chem. Ges. 27, 2985–2993.
38. Cowtan, K. D., and Main, P. (1996) Phase combination and cross
validation in iterated density-modification calculations. Acta Crystal-
logr. 52, 43–48.
39. McRee, D. E. (1999) XtalView/Xfit;a versatile program for mani-
pulating atomic coordinates and electron density. J. Struct. Biol. 125,
156–165.
40. Emsley, P., and Cowtan, K. (2004) Coot: model-building tools for
molecular graphics. Acta Crystallogr., 60, 2126–2132.
41. Reddy, V., Swanson, S. M., Segelke, B., Kantardjieff, K. A.,
Sacchettini, J. C., and Rupp, B. (2003) Effective electron-density
map improvement and structure validation on a Linux multi-CPU
web cluster: The TB Structural Genomics Consortium Bias Removal
Web Service. Acta Crystallogr. D Biol. Crystallogr. 59, 2200–2210.
42. Brunger, A. T., Adams, P. D., Clore, G. M., DeLano, W. L., Gros, P.,
Grosse-Kunstleve, R. W., Jiang, J. S., Kuszewski, J., Nilges, M., Pannu,
N. S., Read, R. J., Rice, L. M., Simonson, T., and Warren, G. L. (1998)
Crystallography & NMR system: a new software suite for macro-
molecular structure determination. Acta Crystallogr. 54, 905–921.
43. Laskowski, R. A., MacArthur, M. W., Moss, D. S., and Thornton,
J. M. (1993) J. Appl. Crystallogr. 26, 283–291.
44. Parish, T. (2003) Starvation survival response of Mycobacterium
tuberculosis. J. Bacteriol. 185, 6702–6706.
45. Hondalus, M. K., Bardarov, S., Russell, R., Chan, J., Jacobs, W. R.,
Jr., and Bloom, B. R. (2000) Attenuation of and protection induced by
a leucine auxotroph of Mycobacterium tuberculosis. Infect. Immun. 68,
2888–2898.
46. Pavelka, M. S., Jr., and Jacobs, W. R., Jr. (1996) Biosynthesis of
diaminopimelate, the precursor of lysine and a component of pepti-
doglycan, is an essential function of Mycobacterium smegmatis.
J. Bacteriol. 178, 6496–6507.
47. Han, Q., and Li, J. (2004) pH dependence, substrate specificity and
inhibition of human kynurenine aminotransferase I. Eur. J. Biochem.
271, 4804–4814.
61. Mani Tripathi, S., and Ramachandran, R. (2006) Direct evidence for
a glutamate switch necessary for substrate recognition: crystal struc-
tures of lysine epsilon-aminotransferase (Rv3290c) from Mycobacter-
ium tuberculosis H37Rv. J. Mol. Biol. 362, 877–886.
62. Fenn, T. D., Stamper, G. F., Morollo, A. A., and Ringe, D. (2003) A
side reaction of alanine racemase: transamination of cycloserine.
Biochemistry 42, 5775–5783.
63. Shah, S. A., Shen, B. W., and Brunger, A. T. (1997) Human ornithine
aminotransferase complexed with L-canaline and gabaculine: struc-
tural basis for substrate recognition. Structure 5, 1067–1075.
64. Hennig, M., Grimm, B., Contestabile, R., John, R. A., and Jansonius,
J. N. (1997) Crystal structure of glutamate-1-semialdehyde amino-
mutase: an alpha2-dimeric vitamin B6-dependent enzyme with asym-
metry in structure and active site reactivity. Proc. Natl. Acad. Sci. U.S.
A. 94, 4866–4871.
65. Storici, P., De Biase, D., Bossa, F., Bruno, S., Mozzarelli, A., Peneff,
C., Silverman, R. B., and Schirmer, T. (2004) Structures of gamma-
aminobutyric acid (GABA) aminotransferase, a pyridoxal 50-phos-
phate, and [2Fe-2S] cluster-containing enzyme, complexed with gamma-
ethynyl-GABA and with the antiepilepsy drug vigabatrin. J. Biol.
Chem. 279, 363–373.
66. Pai, E. F., Krengel, U., Petsko, G. A., Goody, R. S., Kabsch, W., and
Wittinghofer, A. (1990) Refined crystal structure of the triphosphate
˚
conformation of H-ras p21 at 1.35 A resolution: implications for the
mechanism of GTP hydrolysis. EMBO J. 9, 2351–2359.
67. Silva, M. M., Poland, B. W., Hoffman, C. R., Fromm, H. J., and
Honzatko, R. B. (1995) Refined crystal structures of unligated
adenylosuccinate synthetase from Escherichia coli. J. Mol. Biol. 254,
431–446.
68. Yang, G., Sandalova, T., Lohman, K., Lindqvist, Y., and Rendina,
A. R. (1997) Active site mutants of Escherichia coli dethiobiotin
synthetase: effects of mutations on enzyme catalytic and structural
properties. Biochemistry 36, 4751–4760.
48. Goto, M., Omi, R., Miyahara, I., Hosono, A., Mizuguchi, H.,
Hayashi, H., Kagamiyama, H., and Hirotsu, K. (2004) Crystal
structures of glutamine:phenylpyruvate aminotransferase from Ther-
mus thermophilus HB8: induced fit and substrate recognition. J. Biol.
Chem. 279, 16518–16525.
69. Gibson, K. J., Lorimer, G. H., Rendina, A. R., Taylor, W. S., Cohen,
G., Gatenby, A. A., Payne, W. G., Roe, D. C., Lockett, B. A., and
Nudelman, A.; et al. (1995) Dethiobiotin synthetase: the carbonyla-
tion of 7,8-diaminonanoic acid proceeds regiospecifically via the
N7-carbamate. Biochemistry 34, 10976–10984.
49. Bhor, V. M., Dev, S., Vasanthakumar, G. R., and Surolia, A. (2006)
Spectral and kinetic characterization of 7,8-diaminopelargonic acid
synthase from Mycobacterium tuberculosis. IUBMB Life 58, 225–233.
50. Breen, R. S., Campopiano, D. J., Webster, S., Brunton, M., Watt, R.,
and Baxter, R. L. (2003) The mechanism of 7,8-diaminopelargonate
synthase; the role of S-adenosylmethionine as the amino donor. Org.
Biomol. Chem. 1, 3498–3499.
51. Stoner, G. L., and Eisenberg, M. A. (1975) Biosynthesis of 7,8-
diaminopelargonic acid from 7-keto-8-aminopelargonic acid and
S-adenosyl-L-methionine. The kinetics of the reaction. J. Biol. Chem.
70. Schweins, T., Geyer, M., Scheffzek, K., Warshel, A., Kalbitzer, H. R.,
and Wittinghofer, A. (1995) Substrate-assisted catalysis as a mechan-
ism for GTP hydrolysis of p21ras and other GTP-binding proteins.
Nat. Struct. Biol. 2, 36–44.
71. Schweins, T., Geyer, M., Kalbitzer, H. R., Wittinghofer, A., and
Warshel, A. (1996) Linear free energy relationships in the intrinsic
and GTPase activating protein-stimulated guanosine 50-triphosphate
hydrolysis of p21ras. Biochemistry 35, 14225–14231.
72. Wallace, A. C., Laskowski, R. A., and Thornton, J. M. (1995)
LIGPLOT: a program to generate schematic diagrams of protein-
ligand interactions. Protein Eng. 8, 127–134.
250, 4037–4043.