The structure of PknB in complex with mitoxantrone, an ATP-competitive
inhibitor, suggests a mode of protein kinase regulation in mycobacteria
Annemarie Wehenkela,1, Pablo Fernandeza,1, Marco Bellinzonia, Vincent Catherinotb,
Nathalie Barilonec, Gilles Labesseb, Mary Jacksonc, Pedro M. Alzaria,*
aUnite ´ de Biochimie Structurale and CNRS URA2185, Institut Pasteur, 25 rue du Docteur Roux, 75724 Paris, France
bCentre de Biochimie Structurale, INSERM U414, CNRS UMR5048, Universite ´ Montpellier 1, 15 Avenue Charles Flahault,
34060 Montpellier, France
cUnite ´ de Ge ´ne ´tique Mycobacte ´rienne, Institut Pasteur, 25 rue du Docteur Roux, 75724 Paris, France
Received 2 March 2006; revised 15 April 2006; accepted 18 April 2006
Available online 27 April 2006
Edited by Hans Eklund
receptor-like protein kinase involved in cell growth control.
Here, we demonstrate that mitoxantrone, an anthraquinone
derivative used in cancer therapy, is a PknB inhibitor capable
of preventing mycobacterial growth. The structure of the com-
plex reveals that mitoxantrone partially occupies the adenine-
binding pocket in PknB, providing a framework for the design
of compounds with potential therapeutic applications. PknB
crystallizes as a ‘back-to-back’ homodimer identical to those ob-
served in other structures of PknB in complex with ATP analogs.
This organization resembles that of the RNA-dependent protein
kinase PKR, suggesting a mechanism for kinase activation in
? ? 2006 Federation of European Biochemical Societies. Published
by Elsevier B.V. All rights reserved.
Mycobacterium tuberculosis PknB is an essential
Keywords: Drug design; Back-to-back dimerization; Crystal
structure; Ser/Thr protein kinase-inhibitor complex;
The ability of Mycobacterium tuberculosis, the pathogen
responsible for tuberculosis (TB), to adapt to changing envi-
ronmental conditions requires an efficient way of sensing and
transducing extracellular signals. One of the mechanisms used
in mycobacteria to assure a tight regulation of cell growth and
division involves the reversible phosphorylation on serine/thre-
onine residues, a well-established process for eukaryotic signal-
ing networks .
M. tuberculosis PknB is a trans-membrane Ser/Thr protein
kinase (STPK) highly conserved in Gram-positive bacteria
and apparently essential for mycobacterial viability . The
crystal structure of the kinase domain of PknB in complex with
an ATP analogue [3,4] showed a striking conservation of both
protein fold and catalytic mechanism between eukaryotic and
prokaryotic STPKs. We have previously shown that PknB is
regulated by autophosphorylation and dephosphorylation by
the Ser/Thr protein phosphatase PstP [5,6] and recent work
showed that PknB is predominantly expressed during exponen-
tial growth, where its overexpression causes morphological
changes linked to defects in cell wall synthesis and cell division
Aberrant kinase activity is implicated in numerous human
diseases and, not surprisingly, protein kinases represent today
one of the most important groups of drug targets [8,9]. Here
we report that mitoxantrone, a compound used in cancer treat-
ment, is a PknB inhibitor capable of preventing mycobacterial
cell growth, suggesting that bacterial kinases may also repre-
sent a potential target for drug design. The crystal structure
of the complex demonstrates that mitoxantrone is an ATP-
competitive inhibitor of PknB and suggests a mode of regula-
tion of protein kinases in mycobacteria.
2. Materials and methods
2.1. In silico screening
Over 40000 compounds from different chemical libraries, including
the Comprehensive Medicinal Chemistry database, were docked into
the nucleotide-binding pocket of the M. tuberculosis PknB structure
(pdb ID 1O6Y ) using the program FlexX .
2.2. Kinase assays
The kinase assays were carried out in 15 ll kinase buffer (50 mM
HEPES, pH 7.0, 1 mM DTT, 0.01% Brij35, 5% glycerol, and 2 mM
MnCl2) using GarA as a substrate  (kinase:substrate molar ratio
1:2000). The reactions were started with the addition of 2.25 lM final
ATP (containing 1 lCi of [c-33P]ATP), and carried out for 20 min at
30 ?C. For the inhibition experiments, each compound was pre-incu-
bated for 30 min at 4 ?C with the reaction mixture (without ATP).
The reactions were stopped by heat inactivation and the mixture trans-
ferred onto P81 paper (phosphocellulose, Whatman). The paper was
washed with 1% phosphoric acid, rinsed with acetone and allowed to
dry. Radiolabelled spots were analyzed with a PhosphorImager
(Storm, Molecular Dynamics). IC50 values were determined using
KaleidaGraph (Synergy Software).
2.3. Determination of MIC values
Minimal inhibitory concentration (MIC) values for mitoxantrone
against different mycobacteria (Fig. 1c) were determined using the col-
orimetric resazurin microtiter assay in 7H9-OADC broth (Difco) at
37 ?C, as described .
*Corresponding author. Fax: +33 145688604.
E-mail address: firstname.lastname@example.org (P.M. Alzari).
1These authors contributed equally to this work.
0014-5793/$32.00 ? 2006 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved.
FEBS Letters 580 (2006) 3018–3022
The cytoplasmic domain of PknB (residues 1–279) was produced
and purified as described . Crystals of PknB (11 mg/ml) in complex
with mitoxantrone (0.2 mM) (Sigma–Aldrich) appeared after 2 or 3
days in 1.2 M sodium acetate, 50 mM sodium cacodylate, pH 5.6, at
18 ?C. After testing several crystals grown in slightly different condi-
tions, we retained the best dataset from a single flash-frozen crystal col-
lected on the ID14.3 beamline (ESRF, Grenoble) for further analysis
(Table 1). The data are highly anisotropic (2.9 A˚resolution limit along
the c*axis, but only 3.5 A˚along the a*and b*axes), which accounts
for the poor data completeness (data are 99% complete at 3.49 A˚,
45% in the 3.49–3.21 A˚shell, but only 16% in the 3.21–2.9 A˚shell)
and implies that the effective resolution may be around 3.2 A˚. All crys-
tallographic calculations were carried out using programs from the
CCP4 software package . Four independent monomers of PknB
were positioned in the asymmetric unit by molecular replacement
methods using the program PHASER  and the previously deter-
mined PknB structure (1O6Y) as a search probe. The bound ligand
was clearly visible in the initial Fourier difference map for three out
of the four independent monomers. After a few rounds of simulated
annealing refinement using the program CNS  and manual model
building with the program O , the crystallographic refinement
was continued with the program REFMAC from CCP4 using tight
non-crystallographic symmetry restraints and a translation/libration/
screw model with eight different groups (corresponding to the
N- and C-terminal lobes of each kinase domain in the asymmetric
unit). The final parameters for the refined model are given in Table
1. Structure factors and atomic coordinates have been deposited with
the Protein Data Bank (Accession Code 2FUM).
2.5. Sequence and structural comparisons
A homology search using the PknB sequence against all finished bac-
terial genomes (NCBI, as of December 2005) identified 39 trans-mem-
brane kinases with predicted extracellular PASTA domains. The
sequences were aligned and the residue conservation pattern was
mapped onto the structure (Fig. 3b) using the program ConSurf .
3. Results and discussion
3.1. Mitoxantrone inhibits mycobacterial growth in culture
Wecarried out an in silico screening to search for PknB inhib-
itors (see Section 2), and 20 commercially available compounds
from the first 60 hits were then tested for their inhibitory
properties. Kinase assays revealed that mitoxantrone (1,4-
anthracenedione) was able to inhibit PknB with an IC50in the
micromolar range (IC50= 0.8 ± 0.05 lM), comparable to that
observed for the cytotoxic large-spectrum kinase inhibitor
staurosporine (IC50= 0.6 ± 0.05 lM), see Fig. 1a/b. Mitoxan-
trone is a DNA-reactive agent that has been used for several
years in cancertreatment . Besides its DNA-binding proper-
may involve free radical production  and inhibition of Ser/
Thr protein kinases [19,20].
Mitoxantrone showed an inhibitory effect on cell growth
(Fig. 1c) when tested on cultures of M. tuberculosis
(MIC = 400 lM), M. smegmatis mc2155 (MIC = 100 lM),
and M. aurum A+ (MIC = 25 lM) using the resazurin micro-
titer assay. Differences in the permeability of the cell envelope
and/or in the structure of the targets of mitoxantrone in the
Fig. 1. Mitoxantrone inhibits PknB and prevents mycobacterial
growth. (a) Radiolabelled spots of twofold serial dilutions of stauro-
sporine (S; first spot: 10 lM) and mitoxantrone (M; first spot: 20 lM).
A control without inhibitor (+) is included. (b) IC50 values for
staurosporine and mitoxantrone. (c) Minimal inhibitory concentra-
tions (MIC) values of mitoxantrone for different mycobacteria. (M.tb:
M. tuberculosis H37Rv; M.au: M. aurum A+; M.sg: M. smegmatis
mc2155; M.sg+pknb: M. smegmatis mc2155 overexpressing pknB; and
M.sg+pOMK: M. smegmatis mc2155 transformed with the control
vector). Histogram values are representative of 2 or 3 independent
experiments in each case. The chemical formula of mitoxantrone is
shown in the inset.
Data collection and refinement statistics
Cell dimensions [a = b,c] (A˚)
Number of refined atoms
Bond lengths (A˚)
Bond angles (?)
Ramachandran outliers (%)
aNumbers in parentheses correspond to the highest resolution shell.
cThe number of reflections used for free R-factor calculation is shown
jIh;jÞ=nhand nhis the multiplicity of reflection h.
hkljjFoj ? kjFcjj=P
A. Wehenkel et al. / FEBS Letters 580 (2006) 3018–3022
three mycobacterial species may account for the different
MICs observed. The effect of mitoxantrone on M. smegmatis
was partially reversed when the wild-type strain was trans-
formed with a multicopy replicative plasmid (pOMK)  car-
rying a wild-type copy of the M. tuberculosis pknB gene
expressed from its own promoter. The MIC of the pknB over-
expressor was twofold those of the wild-type strain (200 lM)
or the strain transformed with the control vector alone
(Fig. 1c), suggesting that PknB is at least one of the lethal tar-
gets of mitoxantrone in this species.
3.2. Structure of the PknB-inhibitor complex
To investigate the mode of action of mitoxantrone on PknB,
we crystallized the complex and determined its 3D structure by
X-ray diffraction methods (Table 1). The overall structure of
the enzyme is similar to those previously described for PknB
in complex with ATP analogs [3,4], with the kinase domain
in an overall closed conformation and a disordered activation
loop. The most noticeable structural change involves the gly-
cine-rich loop, which in the absence of ATP moves further
towards the C-terminal lobe (Fig. 2a).
Clear electron density is observed for mitoxantrone in the
nucleotide-binding cleft of PknB (Fig. 2b). The planar dihy-
droxy anthraquinone moiety of the inhibitor occupies the
hydrophobic cage that binds the adenosine moiety of ATP.
An important number of residues of both the N- and C-termi-
nal lobes makes van der Waals contacts with the inhibitor,
including Leu17, Gly18, Val25, Ala38, Met 92, Glu93, Tyr94
and Val95 in the N-terminal lobe, Met145 and Met155 in the
C-terminal lobe. The main-chain amide group of Val95, which
in the PknB-AMPPCP complex is hydrogen bonded to the N1
atom of adenosine , now forms a hydrogen bonding interac-
tion with one hydroxyl group of the inhibitor (Fig. 2c). This
interaction may account for the observed lateral positioning
of the inhibitor within the wide hydrophobic binding pocket
(Fig. 2b). The partial occupancy of the cleft leaves space to
accommodate bulkier substituents at the three-ring moiety,
which might be exploited to improve the inhibitory properties
of the compound.
The limited resolution of this study precludes a detailed
analysis of the interactions made by the flexible hydroxyethyla-
mino moieties of the ligand, although in at least two of the four
independent PknB molecules the side-chain of Asn143 makes
additional hydrogen bonding interactions with the nitrogen
atom of one hydroxyethylamino moiety. As observed in other
protein kinase-inhibitor complexes , it is possible that these
flexible extended moieties, which protrude away from the
ATP-binding pocket, could interfere with the active conforma-
tion of the kinase and account in part for the inhibitory prop-
erties of mitoxantrone.
3.3. The conserved dimeric arrangement of PknB is similar to
that of PKR
When expressed as a recombinant protein, the catalytic do-
main of PknB was observed to behave as a mixture of mono-
mers and dimers in solution (Ref.  and data not shown).
Interestingly, in the PknB-mitoxantrone complex the kinase
domain crystallized as a ‘back-to-back’ homodimer (Fig. 3a).
The two crystallographically independent dimers in the com-
plex are very similar to each other and to those observed in
two other structures of PknB crystallized in different space
Fig. 2. Structure of the PknB-mitoxantrone complex. (a) Super-
position of the PknB-mitoxantrone complex (in yellow) and the
PknB-AMPPCP complex (1O6Y, in cyan). Note the movement of the
Gly-rich loop (black arrow). (b) Observed (yellow) and predicted (thin
lines) orientations of mitoxantrone within the adenosine-binding cavity
(represented as a molecular surface). The electron density map for the
inhibitor is contoured at 1r. (c) Schematic view (represented as in Ref.
) of the PknB ATP-binding site showing hydrogen bonding
interactions with both the inhibitor (in blue) and AMP-PCP (PDB
A. Wehenkel et al. / FEBS Letters 580 (2006) 3018–3022
groups [3,4], lending strong support to the hypothesis that this
homodimeric arrangement is physiologically relevant .
The dimer interface, composed almost exclusively of residues
from the N-lobe, is largely conserved in PknB-like bacterial
protein kinases (Fig. 3b). When the amino acid sequences of
39 pknB putative ortholog genes are mapped onto the PknB
structure, 8 positions at the dimer interface are almost invari-
ably conserved (i.e. conserved in at least 35 of the 39 protein
sequences): Arg10, Leu33, Arg35, Ala60, Pro69, Asp76,
Gly78 and Glu93. The strictly invariant residues Arg10 and
Asp76 form a double intermolecular salt bridge in the homodi-
mer. Two other highly conserved residues are also engaged in
intermolecular hydrogen-bonding interactions: the guanidium
group of Arg35 forms two hydrogen bonds with the main-
chain oxygen atoms of residues Ala64 and Val74, and the
side-chain of Asn67 forms a hydrogen bond with the carboxyl-
ate group of Glu93. Interestingly, the main-chain atoms of
Glu93 are hydrogen-bonded to the ATP analog in the binary
complex (and are also in contact with mitoxantrone in the
PknB-inhibitor complex), thus establishing a direct link be-
tween the dimer interface and the nucleotide-binding site.
Remarkably, the PknB homodimer strongly resembles that
recently observed for the RNA-dependent antiviral protein ki-
nase PKR  (Fig. 3c). In both cases the same equivalent
positions are involved in the interface and a similar surface
area is occluded upon dimer formation: the PknB interface is
made up of 23 residues (80 atoms) that contribute 800 A˚2to
the contact surface area, while the PKR dimer interface in-
cludes 26 residues (78 atoms) and has a buried surface area
of 730 A˚2per monomer. Whereas a precise understanding of
how PknB dimerization can directly influence the catalytic
activity must await the structural study of the enzyme in a re-
pressed monomeric state, the striking similarity between the
PknB and PKR homodimers allows us to speculate on a pos-
sible role of dimerization on PknB activity regulation, based
on the analogy with PKR [22,23]. Thus, in its monomeric state,
PknB would be inactive due, for instance, to a misplacement of
helix aC (whose C-terminal end is within the dimer interface),
in much the same way as proposed for other eukaryotic pro-
tein kinases. Upon ligand binding, the extracellular region
would then promote ‘back-to-back’ dimerization of the cata-
lytic domain, as observed in the crystal structure. Indeed, the
influence of the extracellular domain on dimerization was dem-
onstrated for the PknB-like protein kinase PrkC from B. sub-
tilis . In turn, PknB homodimer formation would then
promote autophosphorylation at Ser/Thr residues in the acti-
vation loop [5,25] and subsequent substrate recruitment [6,7].
While many aspects of the above model remain necessarily
speculative at this stage, it suggests a possible strategy for drug
design based on ATP-competitive lead compounds such as
mitoxantrone derivatives. Given the direct structural links be-
tween PknB dimerization, catalytic activity and nucleotide
binding, it should be possible to obtain specific ATP-competi-
tive inhibitors that either block the catalytic domain in an inac-
tive state or just preclude ‘back-to-back’ dimerization.
Acknowledgements: The authors thank Martin Gran ˜a for helpful dis-
cussions. This work has been supported by grants from the Institut
Pasteur (GPH-Tuberculose) and the European Commission (X-TB,
 Hunter, T. (2000) Signaling-2000 and beyond. Cell 100, 113–
 Sassetti, C.M., Boyd, D.H. and Rubin, E.J. (2003) Genes required
for mycobacterial growth defined by high density mutagenesis.
Mol. Microbiol. 48, 77–84.
 Ortiz-Lombardia, M., Pompeo, F., Boitel, B. and Alzari, P.M.
(2003) Crystal structure of the catalytic domain of the PknB
serine/threonine kinase from Mycobacterium tuberculosis. J. Biol.
Chem. 278, 13094–13100.
 Young, T.A., Delagoutte, B., Endrizzi, J.A., Falick, A.M. and
Alber, T. (2003) Structure of Mycobaterium tuberculosis PknB
Fig. 3. The conserved PknB homodimer. (a) Superposition of the two
mitoxantrone complex (in red and green) with those observed in the
PknB-nucleotide complexes 1O6Y  (in blue) and 1MRU  (in
yellow). (b) Overall view of the PknB monomer (rotated 90? along the
vertical axis with respect to the right monomer in Fig. 3a), color-coded
according to amino acid conservation (red: highly conserved) in 39
PknB-like protein sequences from 35 different bacterial species
(Bacillus anthracis, B. cereus, B. clausii, B. licheniformis, B. subtilis,
Bifidobacterium longum, Clostridium acetobutylicum, C. perfringens, C.
tetani, Corynebacterium diphtheriae, C. efficiens, C. glutamicum,
Enterococcus faecalis, Geobacillus kaustophilus, Lactobacillus acidoph-
ilus, L. johnsonii, Listeria monocytogenes, Mycobacterium avium, M.
bovis, M. leprae, M. tuberculosis, Nocardia farcinica, Nocardioides,
Leifsonia xyli, Oceanobacillus iheyensis, Propionibacterium acnes,
Staphylococcus haemolyticus, S. saprophyticus, Streptococcus agalac-
tiae, S. mutans, S. pyogenes, Streptomyces coelicolor, Symbiobacterium
thermophilum, Thermoanaerobacter tengcongensis, Thermobifida fusca).
(c) Comparison of the PknB and RNA-dependent PKR dimer
interfaces. The side-chain residues belonging to the interfaces are
shown (PknB color-coded as in (b)).
A. Wehenkel et al. / FEBS Letters 580 (2006) 3018–3022
supports a universal activation mechanism for Ser/Thr protein
kinases. Nat. Struct. Biol. 10, 168–174.
 Boitel, B., Ortiz-Lombardia, M., Duran, R., Pompeo, F., Cole,
S.T., Cerven ˜ansky, C. and Alzari, P.M. (2003) PknB kinase
activity is regulated by phosphorylation in two Thr residues and
dephosphorylation by PstP, the cognate phospho-Ser/Thr phos-
phatase, in Mycobacterium tuberculosis. Mol. Microbiol. 49,
 Villarino, A., Duran, R., Wehenkel, A., Fernandez, P., England,
P., Brodin, P., Cole, S.T., Zimny-Arndt, U., Jungblut, P.R.,
Cerven ˜ansky, C. and Alzari, P.M. (2005) Proteomic identification
of M. tuberculosis protein kinase substrates: PknB recruits GarA,
a FHA-containing protein, through activation loop-mediated
interactions. J. Mol. Biol. 350, 953–963.
 Kang, C.M., Abbott, D.W., Park, S.T., Dascher, C.C., Cantley,
L.C. and Husson, R.N. (2005) The Mycobacterium tuberculosis
serine/threonine kinases PknA and PknB: substrate identification
and regulation of cell shape. Genes Dev. 19, 1692–1704.
 Cohen, P. (2002) Protein kinases, the major drug targets of the
21st century? Nat. Rev. Drug Discov. 1, 309–316.
 Noble, M.E.M., Endicott, J.A. and Johnson, L.N. (2004) Protein
kinase inhibitors: insights into drug design from structure. Science
 Rarey, M., Wefing, S. and Lengauer, T. (1996) Placement of
medium-sized molecular fragments into active sites of proteins. J.
Comp. Aided Mol Des. 10, 41–54.
 Martin, A., Camacho, M., Portaels, F. and Palomino, J.-C. (2003)
Resazurin microtiter assay plate testing of Mycobacterium tuber-
culosis susceptibilities to second-line drugs: rapid, simple, and
inexpensive method. Antimicrob. Agents Chemotherap. 47, 3616–
 CCP4, Collaborative Computational Project Number (1994) Acta
Cryst. D 50, 760–763.
 McCoy, A.J., Grosse-Kunstleve, R.W., Storoni, L.C. and Read,
R.J. (2005) Likelihood-enhanced fast translation functions. Acta
Cryst. D 61, 458–464.
 Brunger, A.T., Adams, P.D., Clore, G.M., 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 (CNS): a new software
system for macromolecular structure determination. Acta Cryst.
D 54, 905–921.
 Jones, T.A., Zhou, J.Y., Cowan, S.W. and Kjeldgaard, M. (1991)
Improved methods for building protein models in electron density
maps and the location of errors in these models. Acta Cryst. A 47,
 Glaser, F., Pupko, T., Paz, I., Bell, R.E., Bechor-Shental, D.,
Martz, E. and Ben-Tal, N. (2003) ConSurf: identification of
functional regions in proteins by surface-mapping of phylogenetic
information. Bioinformatics 19, 163–164.
 Shenkenberg, T.D and Von Hoff, D.D. (1986) Mitoxantrone: a
new anticancer drug with significant clinical activity. Ann. Int.
Med. 105, 67–81.
 Basra, J., Wolf, C.R., Brown, J.R. and Patterson, L.H. (1985)
Evidence for human liver mediated free-radical formation by
doxorubicin and mitoxantrone. Anticancer Drug Des. 1, 45.
 Jinsart, W., Ternai, B. and Polya, G.M. (1992) Inhibition of
myosin light chain kinase, cAMP-dependent protein kinase,
protein kinase C and of plant Ca(2+)-dependent protein kinase
by anthraquinones. Biol. Chem. Hoppe-Seyler 373, 903–910.
 Takeuchi, N., Nakamura, T., Takeuchi, F., Hashimoto, E. and
Yamamura, H. (1992) Inhibitory effect of mitoxantrone on
activity of protein kinase C and growth of HL60 cells. J. Biochem.
(Tokyo) 112, 762–767.
 Jackson, M., Berthet, F.-X., Otal, I., Rauzier, J., Martin, C.,
Gicquel, B. and Guilhot, C. (1996) The Mycobacterium tubercu-
losis purine biosynthetic pathway: isolation and characterization
of the purC and purL genes. Microbiology 142, 2439–2447.
 Dar, A.C., Dever, T.E. and Sicheri, F. (2005) Higher-order
substrate recognition of eIF2a by the RNA-dependent protein
kinase PKR. Cell 122, 887–900.
 Dey, M., Cao, C., Dar, A.C., Tamura, T., Ozato, K., Sicheri, F.
and Dever, T.E. (2005) Mechanistic link between PKR dimeriza-
tion, autophosphorylation and eIF2a substrate recognition. Cell
 Madec, E., Laszkiewicz, A., Iwanicki, A., Obuchowski, M. and
Seror, S. (2002) Characterization of a membrane-linked Ser/Thr
protein kinase in Bacillus subtilis, implicated in developmental
processes. Mol. Microbiol. 2, 571–586.
 Duran, R., Villarino, A., Bellinzoni, M., Wehenkel, A., Fernan-
dez, P., Boitel, B., Cole, S.T., Alzari, P.M. and Cervenansky, C.
(2005) Conserved autophosphorylation pattern in activation
loops and juxtamembrane regions of M. tuberculosis Ser/Thr
 Traxler, P. and Furet, P. (1999) Strategies toward the design of
novel and selective protein tyrosine kinase inhibitors. Pharmacol.
Ther. 82, 195–206.
A. Wehenkel et al. / FEBS Letters 580 (2006) 3018–3022