Báo cáo khoa học: The bI/bIII-tubulin isoforms and their complexes with antimitotic agents pptx

The bI/bIII-tubulin isoforms and their complexes with
antimitotic agents
Docking and molecular dynamics studies
Matteo Magnani
1
, Francesco Ortuso
2
, Simonetta Soro
3
, Stefano Alcaro
2
, Anna Tramontano
3
and Maurizio Botta
1
1 Dipartimento Farmaco Chimico Tecnologico, Universita
`
degli Studi di Siena, Italy
2 Dipartimento di Scienze Farmacobiologiche ‘Complesso Nin’ Barbieri’ Universita
`
degli Studi di Catanzaro ‘Magna Graecia’, Roccelletta di
Borgia (CZ), Italy
3 Dipartimento di Scienze Biochimiche ‘A. Rossi Fanelli’, Universita
`
degli Studi ‘La Sapienza’, Rome, Italy
Microtubules are filamentous dynamic polymers com-
posed of a ⁄ b-tubulin heterodimers involved in a
diverse range of cellular functions including motility,
morphogenesis, intracellular trafficking of macromole-
cules and organelles, and mitosis and meiosis [1,2].
The role played by microtubules in the cell division

process makes them attractive targets for anticancer
therapy [3], a perspective that has been explored by
using tubulin-binding agents [4]. These compounds are
able to disrupt microtubule dynamics and can act
either as microtubule destabilizers (such as vinca alka-
loids) or as microtubule stabilizers (such as taxanes).
Among the latter, paclitaxel (Fig. 1 left) has been dem-
onstrated to be effective for the treatment of ovarian,
breast, and nonsmall cell lung carcinomas [5]. These
molecules have the drawback of being scarcely select-
ive, but an even more significant problem that limits
their usage in the treatment of malignancies is the
emergence of resistance. There are essentially two
routes to resistance [4]: (i) expression of the P-glyco-
protein [6], which is able to pump the antitumoral
compounds out of the tumor cell; and (ii) emergence
of structural modification of the microtubules them-
selves, both via mutations and modifications of their
isotype composition, in particular that of their b-sub-
unit [7–9]. In humans, seven isoforms of b-tubulin,
displaying different patterns of tissue expression,
have been identified [10,11]. In particular, bIis
Keywords
docking; epothilone A; IDN5390; paclitaxel;
tubulin
Correspondence
M. Botta, Dipartimento Farmaco Chimico
Tecnologico, Universita
`
degli Studi di Siena,

Via Alcide de Gasperi, 2, I-53100 Siena, Italy
Fax: +39 577 234333
Tel: +39 577 234306
E-mail:
(Received 7 April 2006, revised 16 May
2006, accepted 23 May 2006)
doi:10.1111/j.1742-4658.2006.05340.x
Both microtubule destabilizer and stabilizer agents are important molecules
in anticancer therapy. In particular, paclitaxel has been demonstrated to be
effective for the treatment of ovarian, breast, and nonsmall cell lung carci-
nomas. It has been shown that emergence of resistance against this agent
correlates with an increase in the relative abundance of tubulin isoform
bIII and that the more recently discovered IDN5390 can be effectively used
once resistance has emerged. In this paper, we analyze the binding modes
of these antimitotic agents to type I and III isoforms of b-tubulin by com-
putational methods. Our results are able to provide a molecular explan-
ation of the experimental data. Using the same protocol, we could also
show that no preference for any of the two isoforms can be detected for
epothilone A, a potentially very interesting drug for which no data about
the emergence of resistance is currently available. Our analysis provides
structural insights about the recognition mode and the stabilization mech-
anism of these antimitotic agents and provides useful suggestions for the
design of more potent and selective antimitotic agents.
Abbreviation
PDB, Protein Data Bank.
FEBS Journal 273 (2006) 3301–3310 ª 2006 The Authors Journal compilation ª 2006 FEBS 3301
constitutively expressed and represents, in general, the
most abundant isotype, whereas bIII expression is
restricted to neuronal tissues and testis. There are sev-
eral studies reporting that an increase in the relative

abundance of isoform III destabilizes the microtubules
[12,13] and clear indications that it correlates with
paclitaxel resistance, both in vitro and in vivo [14–18].
Recent studies have shown that a seco-taxan (IDN5390,
Fig. 1 middle) [19], although less potent than paclitaxel,
is active on tumor cells overexpressing isoform III and
therefore could be used in cases where resistance to
paclitaxel has emerged [20]. In this study, we use a com-
bination of molecular modeling, docking and molecular
dynamics techniques to investigate the molecular basis
of paclitaxel resistance and IDN5390 sensitivity,
through the analysis of the complexes between these two
ligands and the isotypes I and III of the human b-tubu-
lin. We also investigated the complexes involving the bI
and bIII isoforms with epothilone A (Fig. 1 right). Epo-
thilones are microtubule stabilizing agents, sharing a
common mechanism of action with taxanes [21]. To the
best of our knowledge, no data is available about the
activity of this class of compounds on different isoforms
of tubulin, even though these molecules are gaining
more and more attention in antitumoral therapy [22].
Results
Analysis of tubulin crystallographic models and
docking with the three ligands
The structures of ligands used in this study are repor-
ted in Fig. 1; with respect to paclitaxel and IDN5390,
epothilone A is characterized by a less complex
molecular structure. There is some confusion in the lit-
erature and in databases about the nomenclature of
the various tubulin isoforms. The bI and bIII genes

have been recently re-sequenced [17], and we noticed
that the protein annotated as tubulin bII chain (Code:
TBB2_HUMAN, P07437) in SwissProt corresponds to
the sequence of the tubulin bI chain. We also checked
that the confusion did not reflect population polymor-
phisms: no single nucleotide polymorphism is reported
in the human genome in positions that are different
between bI and bIII. The problem has now been
brought to the attention of the database curators.
There are several structural determinations of the
tubulin dimer from different sources, some of which
have been obtained by binding tubulin to a zinc sheet
in order to obtain a bidimensional crystal, some others
by fitting the zinc sheet structures in electron microscopy
data, some by X-ray diffraction of crystals containing
a tubulin dimer in complex with small ligands and ⁄ or
other proteins, and some by modeling. No structure
determination is available for the human proteins and
therefore we needed to build comparative models for
the human proteins. In particular, we concentrated on
isoforms I and III of the human tubulin b subunit, as
this subunit hosts the common binding site for our
molecules of interest [23,24].
The sequence identity between tubulin from different
sources and their human counterpart is very high
(between 89 and 94%), nevertheless it is very difficult
to assess which of the available structural determina-
tions better reflects the conformation of the protein in
physiological conditions and is therefore better suitable
to be used as template in the model-building proce-

dure. In order to select the appropriate template, we
analyzed all tubulin-related entries in the Protein Data
Bank (PDB) and performed an all-against-all compar-
ison of the complete structures and of the most rele-
vant parts of the structures (dimer interface and
binding site). The differences in the overall structure
are rather high in terms of rmsd (up to 4 A
˚
and more,
see supplementary material, Table S1). More limited is
the structural variation of the set of residues involved
in the interaction with paclitaxel and epothilone A,
which are relevant for our purposes (see supplementary
material, Table S2). The variability seems to be mostly
correlated with the nature of the ligand and, conse-
Fig. 1. Chemical structures of paclitaxel (1), IDN5390 (2) and epothilone A (3).
Docking of antimitotic agents to tubulin isoforms M. Magnani et al.
3302 FEBS Journal 273 (2006) 3301–3310 ª 2006 The Authors Journal compilation ª 2006 FEBS
quently, the best choice seemed to be to select as tem-
plates the structures bound to the ligands that we
planned to study and we used 1JFF as template to
build our comparative model, named hTUB. We used
a standard comparative modeling procedure, replacing
the residues of the template with those of the target
according to the sequence alignment obtained using
clustalw [25] with standard parameters. The replaced
sidechains were positioned in their most commonly
observed conformation [26] and the model was opti-
mized using 100 cycles of steepest descent energy mini-
mization using discover [27] and the CVFF force

field. For the initial positioning of the paclitaxel moi-
ety in the complex, we used the orientation found in
the 1JFF structure, which contains a dimer of tubulin
complexed with paclitaxel [23], to position the ligand
in hTUB. In the case of IDN5390, the crystallographic
structure of its complex with b-tubulin is not available.
Consequently, the ligand was docked into the same
binding site of paclitaxel (and epothilone A), assuming
that these two very similar molecules act in a similar
fashion. As a result of docking (for details, see Experi-
mental procedures), IDN5390 was located within the
binding site in a conformation which closely resembles
that of paclitaxel, with the macrocyclic moiety and the
lateral chains occupying the same regions of the pocket
(Fig. 2A,B). For simulations involving epothilone, we
took advantage of the availability of the 1TVK struc-
ture, which contains the structure of a tubulin dimer
complexed with epothilone A [24]. 1TVK was superim-
posed to hTUB and the ligand positioned in the con-
text of the model structure in the same relative
orientation as observed in 1TVK (Fig. 2C). Thus, a
common tubulin structure (hTUB) was used for all
three ligands under analysis (Fig. 3), in an attempt to
limit the biases that could derive from using different
starting protein structures. The procedures described
above were followed to obtain the starting structures
of the six complexes of paclitaxel, IDN5390 and epo-
thilone A with both bI and bIII isotypes of tubulin.
Molecular dynamics and thermodynamics
calculations

The starting complexes, built as described above, were
analyzed by means of molecular dynamics. Such analy-
sis involved, at first, paclitaxel and IDN5390, for
which data about the activity towards microtubules
with different composition in terms of b-tubulin iso-
types are reported, and was subsequently extended to
epothilone A, for which to date no data is available.
Human tubulin bI and bIII isoforms complexed with
paclitaxel (referred to as P1 and P3, respectively), and
A
B
C
Fig. 2. Location of ligands within the hTUB binding site (solvent-
accessible surface representation) in the starting complexes (A)
paclitaxel, (B) IDN5390, and (C) epothilone A. Nonpolar hydrogen
atoms of the ligands are not shown.
M. Magnani et al. Docking of antimitotic agents to tubulin isoforms
FEBS Journal 273 (2006) 3301–3310 ª 2006 The Authors Journal compilation ª 2006 FEBS 3303
with IDN5390 (referred to as I1 and I3, respectively),
were first energy minimized and then subjected to
backbone-constrained molecular dynamics. In the
course of each simulation, 100 ligand–protein com-
plexes were sampled at regular time intervals and were
fully minimized. A conformational clustering analysis
was applied to the resulting structures. Representative
structures for the whole set of collected and optimized
frames were selected by performing a Boltzmann ana-
lysis, in order to take into account both the energy
and the number of structures within each cluster. The
reduced number of selected structures was used to

investigate the molecular basis of the difference in the
calculated binding energies of both ligands for the bI
to bIII-tubulin isoform. Finally, in all simulations, the
average drug–protein binding energies (DG-, DH- and
DS-values) were computed according to the MOLINE
methodology reported by some of us [28]. The results
are reported in Table 1. The data predict a higher
affinity of paclitaxel for the bI isoform than for the
bIII isoform, and an opposite behavior of IDN5390.
This is in good agreement with experimental data, as
we are able to correctly reproduce the differences in
sensitivity to paclitaxel and IDN5390 observed for
microtubules with different isotype composition.
Figure 4 shows the region around the ligand for both
the P1 and P3 complexes, in one of the representative
sampled structures (for other representative structures,
quite similar considerations can be made). The binding
of paclitaxel to bI-tubulin (Fig. 4A) involves both
hydrogen bonds and multiple hydrophobic contacts;
most interactions are in agreement with the crystallo-
graphic structure of the complex (taken as starting
structure) and have already been described [22,29,30]
(supplementary material, Table S3). In particular, the
C2 phenyl ring is involved in hydrophobic interactions
with Leu217, His229 and Leu230, while the C4 acetate
makes hydrophobic contacts with Phe272, Pro274 and
Leu371. Two hydrogen bonds are established between
the oxygen of the oxetane ring and the NH backbone
of Thr276, and between the C2¢ hydroxyl group and
the NH backbone of Gly370. However, as expected,

Fig. 3. Ribbon representation of the main secondary structure ele-
ments characterizing the b-tubulin binding site for the ligands under
analysis: paclitaxel (green), IDN5390 (magenta), epothilone A
(orange).
Table 1. Free energy, enthalpy and entropy for the drug-protein
complexes computed at 300 K. P1 and P3 refer to paclitaxel bound
to human b-tubulin isoforms I and III, respectively. I1 and I3 to
IDN5390 bound to human b-tubulin isoforms I and III.
Complex
DG
(kcalÆmol
)1
)
DH
(kcalÆmol
)1
)
DS
(calÆmol
)1
)
P1 )64.47 )64.16 1.01
P3 )54.67 )54.64 0.11
I1 )55.89 )55.49 1.32
I3 )67.52 )67.12 1.33
A
B
Fig. 4. Paclitaxel in complex with (A) bI-tubulin (P1), and (B) bIII-
tubulin (P3). For sake of clarity, nonpolar hydrogen atoms are omit-
ted and hydrogen bond interactions are represented by black

dashed lines.
Docking of antimitotic agents to tubulin isoforms M. Magnani et al.
3304 FEBS Journal 273 (2006) 3301–3310 ª 2006 The Authors Journal compilation ª 2006 FEBS
the molecular dynamics simulation is also accompan-
ied by structural variations of the initial complex.
Especially the flexible M-loop experiences a slight rear-
rangement that results in additional interactions with
the ligand. As shown in Fig. 4A, in the P1 complex
the C7 hydroxyl group of paclitaxel forms two hydro-
gen bonds with the main chain carbonyl group and
amino group of Ser277 and Gln282, respectively. A
hydrogen bond interaction is also established between
the C10 acetate and the lateral chain of Arg284. Such
hydrogen bonds are stable during the course of the
whole simulation, being present in >95% of the
sampled structures. Finally, in P1, paclitaxel is also
engaged in hydrogen bond interactions with Glu27.
Differently from the interactions described above, the
latter hydrogen bonds are less stable: in fact, due to
the high mobility of the lateral chain of Glu27 they
are observed in most of the sampled structures only
after minimization. In the complex with bIII-tubulin
(P3, Fig. 4B), the ligand is characterized by a very sim-
ilar binding conformation compared with P1, while the
structure of the M-loop is somewhat different in the
two complexes, owing to the replacement of Ser277 in
isoform bI with Ala277 in isoform bIII. This results in
a different and less effective interaction with the C7–
C10 moiety of the ligand. In comparison with P1,
Arg278 is directed towards the inside of the binding

pocket, establishing two hydrogen bonds with carbonyl
at C9, only one of which is consistently observed in
the course of the molecular dynamics simulations.
Interestingly, in P1 Ser277 forms a hydrogen bond
with Ser280, thus directing its carbonyl group towards
the C7 position of the paclitaxel ring and forming a
hydrogen bond with its OH. The substitution Ser277-
Ala, present in the bIII isoform, does not allow this
interaction to take place. Similarly, the different rear-
rangement of the M-loop in bIII prevents the C10
acetate from interacting with Arg284. In P1, Glu27 is
hydrogen bonded to the ligand, while in P3 this inter-
action is absent. In this isoform, Glu27 interacts with
Arg320, which, in turn, forms a hydrogen bond with
Ser241. Such an interaction network cannot take place
in P1, where Ser241 is replaced by a cysteine, which
does not interact with Arg320.
The models of the complexes of IDN5390 with the
bI and bIII-tubulin isoforms are shown in Fig. 5. The
bound conformation of the ligand is quite similar in
the I1 and I3 complexes, in part resembling that of
paclitaxel. In fact, similarly to paclitaxel, hydrophobic
interactions are established between the C2 phenyl ring
of IDN5390 and Leu217, His229 and Leu230, as well
as between the C4 acetate and Phe272, Pro274 and
Leu371. IDN5390 too is engaged in two hydrogen
bonds involving the oxetane ring and the C2¢ hydroxyl
group with Thr276 and Gly370, respectively. Further-
more, the hydroxyl group in the C2¢ position of paclit-
axel forms a second hydrogen bond with Glu27 of bI

isotype, while the equivalent atom of IDN5390 is
involved in a second hydrogen bond with the sidechain
of Asp26 in both isoforms. However, a pattern of
interactions different from those observed for paclit-
axel is established with the M-loop, whose structural
rearrangement is, also in this case, different in the bI
and bIII isotypes. In both complexes, the C1 and the
C9 hydroxyl groups interact with His229 and Gln282,
respectively, through hydrogen bond interactions. Nev-
ertheless, the different rearrangement of the M-loop in
the two isoforms (due to the replacement of Ser277 in
I1 with Ala277 in I3) results in some important differ-
ences in the binding of IDN5390. Remarkably, only in
the bIII isoform does the conformation of the M-loop
allow the lateral chain of Arg278 to move towards the
ligand and to favorably interact with it through hydro-
gen bonds with the C1 hydroxyl group and the C2
A
B
Fig. 5. IDN5390 in complex with (A) bI-tubulin (I1), and (B) bIII-
tubulin (I3). For sake of clarity, nonpolar hydrogen atoms are omit-
ted and hydrogen bond interactions are represented by black
dashed lines.
M. Magnani et al. Docking of antimitotic agents to tubulin isoforms
FEBS Journal 273 (2006) 3301–3310 ª 2006 The Authors Journal compilation ª 2006 FEBS 3305
benzoyl chain (Fig. 5B). Moreover, such location of
the sidechain of Arg278 also results in a better ‘entrap-
ment’ of the ligand within the binding site compared
to isoform bI, as shown in Fig. 6. As mentioned
above, in the P3 complex, Arg278 also points towards

the inside of the pocket, but it does so to a much
lower extent than in the I3 complex. In fact, the differ-
ent structure of the macrocycle in paclitaxel with
respect to IDN5390 induces a different rearrangement
of the M-loop, which does not allow Arg278 to inter-
act with the ligand as closely as in the IDN5390 bound
structure and therefore the paclitaxel in bIII isotype is
less well packed within the protein structure.
Taken together, our analysis of the P and I com-
plexes reveals that for both paclitaxel or IDN5390 the
different calculated binding energies can be mainly
ascribed to the replacement of Ser277 of the bI iso-
form with Ala277 in bIII. Such a residue is located
within the M-loop (which constitutes an important
part of the binding pocket, as shown in Fig. 3). The
importance of the role of this residue in ligand binding
and in the tubulin structure has been pointed out
recently [11]. According to our findings, the role
played by residue 277 is crucial not only because
Ser277 is directly involved in the binding of paclitaxel
with the bI isotype, but also because its replacement
with Ala277 in bIII induces a conformational rear-
rangement of the M-loop which, in turn, results in dif-
ferent interactions of the ligands with other residues in
the M-loop. In particular, paclitaxel interacts through
hydrogen bonds with Ser277, Gln282 and Arg284 in
the case of the bI isoform and only with Arg278 in the
case of bIII. Similarly, even though IDN5390 interacts
with Gln282 in both complexes, it is hydrogen bonded
to Arg278 only in the I3 complex. As mentioned

above, our results are in accordance with the known
pharmacological effects of paclitaxel and of IDN5390
and are able to provide a rational structural basis for
them. This prompted us to investigate the mode of
binding of the much less well characterized epothilone
A. No data about the effect of different isotype com-
position of tubulin on the activity of this molecule
have been reported so far, therefore we used our pro-
cedure to investigate the interactions of epithilone A
with the bI- and bIII-tubulin isoforms (indicated as E1
and E3, respectively). The average calculated binding
energies of complexes sampled during molecular
dynamic simulations and subsequently optimized are
shown in Table 2. Our data suggest that epothilone A
does not preferentially bind to one of the two iso-
forms. The molecular details of the predicted interac-
tions are shown in Fig. 7. The position of the ligand in
the E1 and E3 structures is not as similar as in the
case of paclitaxel and IDN5390 and, especially in E3,
substantially differs from the starting complex. In E1,
the epothilone is located between the M-loop and helix
H7, interacting with them essentially through: (i)
hydrogen bonds involving C3 and C7 hydroxyl groups
and lateral chains of Arg278 and Gln282, respectively;
and (ii) p–p interactions between the thiazole ring and
His229. In E3, the ligand is farther away from the
M-loop and shifted towards helix H1 and the S9–S10
loop with respect to the E1 complex, and therefore is
Table 2. Free energy, enthalpy and entropy of the drug-protein
complexes computed at 300 K between epothilone A and isoforms

I (E1) and III (E3) of human b-tubulin.
Complex
DG
(kcalÆmol
)1
)
DH
(kcalÆmol
)1
)
DS
(calÆmol
)1
)
E1 )28.01 )27.99 0.08
E3 )31.00 )30.36 2.13
A
B
Fig. 6. Solvent accessible surfaces of the bI (A) and bIII (B) iso-
forms in complex with IDN5390.
Docking of antimitotic agents to tubulin isoforms M. Magnani et al.
3306 FEBS Journal 273 (2006) 3301–3310 ª 2006 The Authors Journal compilation ª 2006 FEBS
able to interact with Gly370 (even if this hydrogen
bond is not consistently observed during throughout
the whole simulation). However, epothilone is still
within the van der Waals distance with His229 and the
carbonyl in C5 is engaged in a hydrogen bond with
Arg278. Also in this case, the structure of the M-loop
differs in the two complexes and is stabilized by a dif-
ferent pattern of hydrogen bonds involving its residues

(in particular, residues 277–280). Similarly to the case
of the P3 and I3 complexes, in E3 the rearrangement
of the loop directs the sidechain of Arg278 towards
the binding pocket, allowing Arg278 to maintain
hydrogen bond interactions with the ligand, notwith-
standing the shift in the ligand position with respect to
the E1 complex. Due to the described differences
between the two binding modes, the interactions invol-
ving the ligand in E1 and E3 are quite difficult to com-
pare. Nevertheless, the analysis of the two complexes
suggests that there should be no significant difference
in binding energies of epothilone for the two isoforms.
These observations, together with the data reported in
Table 2, suggest that epothilone A is able to interact
with similar affinities with both the bI and bIII iso-
forms of tubulin. As a consequence, according to our
analysis, it should be useful in cases where resistance
mediated by overexpression of the bIII-tubulin isotype
arises.
Discussion
A combination of molecular modeling and molecular
dynamics techniques has been applied to investigate
the binding modes of three microtubule stabilizing
agents, namely paclitaxel, IDN5390 and epothilone A,
with isotypes I and III of human b-tubulin. Increased
expression of bIII isoform in cancer cells has been cor-
related with paclitaxel resistance in several studies,
whereas recent findings revealed that the activity of
IDN5390 is not affected by bIII-tubulin levels. To our
knowledge, no data about the activity of epothilones

on tumors characterized by different b-tubulin isotype
composition have been reported so far. Six complexes
of the three ligands under analysis with the human bI
and bIII-tubulin were first built and subjected to
molecular dynamics. The average binding energies for
structures sampled during the simulations were calcula-
ted after energy optimization. Our data rationalize the
experimental observations, suggesting a higher affinity
of paclitaxel for the bI than for the bIII isoform and
an opposite behavior for IDN5390. Interestingly, the
calculated binding energies of complexes involving
epothilone A are very similar for the bI and bIII iso-
forms. Although docking simulation results have to be
taken with caution, especially when based on modeled
structures, our results suggest an equally effective
interaction of this molecule with microtubules with dif-
ferent isoform composition. Representative structures
of complexes sampled during the course of molecular
dynamic simulations were subsequently analyzed, with
the aim of detecting specific interactions responsible
for the differences in the calculated binding energies of
paclitaxel and IDN5390 with bIorbIII isotypes. Such
analysis highlighted the crucial role played by the dif-
ferent residue present in the 277 position in the two
isoforms (serine in bI and alanine in bIII) in determin-
ing the different binding affinities of paclitaxel and
IDN5390 to the two distinct isoforms. In short, such
substitution is responsible for a different rearrange-
ment of the M-loop, whose final outcome is a more
favorable interaction of paclitaxel and IDN5390 with

the bI and bIII isotypes, respectively. Our study sup-
ports the hypothesis that the molecular basis of the
different activities of paclitaxel and IDN5390 against
microtubules expressing variable levels of bIII isoform
A
B
Fig. 7. Epothilone A in complex with (A) bI-tubulin (E1), and (B) bIII-
tubulin (E3). For sake of clarity, nonpolar hydrogen atoms are omit-
ted and hydrogen bond interactions are represented by dashed
lines.
M. Magnani et al. Docking of antimitotic agents to tubulin isoforms
FEBS Journal 273 (2006) 3301–3310 ª 2006 The Authors Journal compilation ª 2006 FEBS 3307
lie in the different ligand binding mode of the two
molecules to the bI and bIII isotypes. The same analy-
sis, when applied to epothilone A, predicts that its
binding should not be affected by the isoptype compo-
sition of tubulin, suggesting that this molecule can
have a broader efficacy than paclitaxel and IDN5390
and perhaps be less prone to inducing resistance in
tumor cells.
Experimental procedures
Comparative modeling
Human tubulin isoform sequences were downloaded from
the Swiss-Prot database (). The
identifiers of human bI- and bIII-tubulins are P07437
(TBB2_human) and Q13509 (TBB3-human), respectively.
The multiple sequence alignments were obtained using
clustalw [25].
The comparative modeling protocol consisted of import-
ing the main-chain coordinates of the conserved regions

from the template and positioning the replaced sidechains
in their most commonly observed conformation [26]. The
model was optimized using 100 cycles of steepest descent
energy minimization using discover [27].
The PDB identifiers of the three-dimensional structures
used in this work are as follows: 1FFX, 1IA0, 1JFF, 1SA0,
1TUB and 1TVK.
Molecular dynamics
Each complex was subjected to 2000 ps of molecular
dynamic simulations with a time step of 1.5 fs. The calcu-
lations were performed using macromodel version 7.2
[31] with the AMBER* united atom force field [32]. Sol-
vent effects were taken into account by means of the
implicit GB ⁄ SA water model [33]. A force constant of
23.9 kcalÆmolÆA
˚
)1
was applied to the protein backbone,
while sidechains and ligands were left free. One hundred
frames were sampled at regular time intervals for each
drug–protein complex and subjected to 5000 steps of the
Polak-Ribiere Conjugate Gradient energy minimization
algorithm with the same force field and parameters as
above. During these optimizations all constrains were
removed allowing full relaxation of the system internal
degrees of freedom. In order to select the most representa-
tive binding modes, a clustering analysis of the optimized
conformational ensemble was performed. In details, con-
formations with an internal energy difference lower than
1 kcalÆmol

)1
were duplicated if their RMS deviation, after
superposition of the whole coordinate set, was lower than
0.25 A
˚
. Binding energies and Boltzmann analysis were car-
ried out using the thermodynamic module of the moline
program [28].
Docking experiments
Docking simulations of IDN5390 in the paclitaxel (and
epothilone A) binding site of b-tubulin were performed
using autodock 3.0.5 software [34]. Both the modeled
protein (hTUB) and the ligand (IDN5390, after building
and minimization with macromodel version 7.2 [31]) were
imported in autodock. Kollman’s united-atoms partial
charges and solvent parameters were added to the protein,
while Gasteiger atomic charges were calculated for the lig-
and. Next, a grid including the binding site of interest was
defined and several atom probes (corresponding to the
atom types of the ligand) were placed at the grid nodes, in
order to calculate the interaction energies between the
probe and the protein. Grid maps were generated for each
atom probe. The Lamarckian genetic algorithm [34] was
employed to explore the orientation ⁄ conformation space of
the ligand within the binding pocket. In Lamarckian genetic
algorithm docking, the number of individuals within the
population and the number of runs were both set to 200. A
maximum number of 2 · 10
6
energy evaluations and 50 000

generations was allowed, while all the remaining parameters
were kept to their default values. Finally, the conformation
with the lowest estimated free energy of binding (and
belonging to a well populated cluster) was selected. The
reliability of the docking protocol was first assessed by
simulating the known binding of paclitaxel. The protocol
described above was able to correctly reproduce the X-ray
coordinates of paclitaxel binding conformation (the super-
position between the modeled and the experimental struc-
ture of the ligand had an RMSD of 0.91) and all the
known interactions between the protein and the ligand were
reproduced [23].
References
1 Valiron O, Caudron N & Job D (2001) Microtubule
dynamics. Cell Mol Life Sci 58, 2069–2084.
2 Desai A & Mitchison TJ (1997) Microtubule poly-
merization dynamics. Annu Rev Cell Dev Biol 13,
83–117.
3 Jordan MA & Wilson L (2004) Microtubules as a target
for anticancer drugs. Nat Rev Cancer 4, 253–265.
4 Drukman S & Kavallaris M (2002) Microtubule altera-
tions and resistance to tubulin-binding agents. Int J
Oncol 21, 621–628.
5 Slichenmyer WJ & Von Hoff DD (1991) Taxol: a new
and effective anti-cancer drug. Anticancer Drugs 2,
519–530.
6 Horwitz SB, Cohen D, Rao S, Ringel I, Shen HJ &
Yang CP (1993) Taxol: mechanisms of action and resis-
tance. J Natl Cancer Inst Monogr 15, 55–61.
7 Giannakakou P, Sackett DL, Kang YK, Zhan Z, Buters

JT, Fojo T & Poruchynsky MS (1997) Paclitaxel-resis-
tant human ovarian cancer cells have mutant b-tubulins
Docking of antimitotic agents to tubulin isoforms M. Magnani et al.
3308 FEBS Journal 273 (2006) 3301–3310 ª 2006 The Authors Journal compilation ª 2006 FEBS
that exhibit impaired paclitaxel-driven polymerization.
J Biol Chem 27, 17118–17125.
8 Gonzalez-Garay ML, Chang L, Blade K, Menick DR &
Cabral F (1999) A b-tubulin leucine cluster involved in
microtubule assembly and paclitaxel resistance. J Biol
Chem 34, 23875–23882.
9 Burkhart CA, Kavallaris M & Horwitz SB (2001) The
role of b-tubulin isotypes in resistance to antimitotic
drugs. Biochim Biophys Acta 1471 , O1–O9.
10 Luduen
˜
a RF (1998) Multiple forms of tubulin: different
gene products and covalent modifications. Int Rev Cytol
178, 207–275.
11 Verdier-Pinard P, Shahabi S, Wang F, Burd B, Xiao H,
Goldberg GL, Orr GA & Horwitz SB (2005) Detection
of human bV-tubulin expression in epithelial cancer cell
lines by tubulin proteomics. Biochemistry 44, 15858–
15870.
12 Panda D, Miller HP, Banerjee A, Luduen
˜
a RF & Wil-
son L (1994) Microtubule dynamics in vitro are regu-
lated by the tubulin isotype composition. Proc Natl
Acad Sci USA 91, 11358–11362.
13 Kavallaris M, Kuo DY, Burkhart CA, Regl DL, Norris

MD, Haber M & Horwitz SB (1997) Taxol-resistant
epithelial ovarian tumors are associated with altered
expression of specific b-tubulin isotypes. J Clin Invest
100, 1282–1293.
14 Derry WB, Wilson L, Khan IA, Luduen
˜
a RF & Jordan
MA (1997) Taxol differentially modulates the dynamics
of microtubules assembled from unfractionated and pur-
ified b-tubulin isotypes. Biochemistry 36, 3554–3562.
15 Verdier-Pinard P, Wang F, Martello L, Burd B, Orr
GA & Horwitz SB (2003) Analysis of tubulin isotypes
and mutations from taxol-resistant cells by combined
isoelectrofocusing and mass spectrometry. Biochemistry
42, 5349–5357.
16 Kamath K, Wilson L, Cabral F & Jordan MA (2005)
bIII-tubulin induces paclitaxel resistance in association
with reduced effects on microtubule dynamic instability.
J Biol Chem 280, 12902–11297.
17 Mozzetti S, Ferlini C, Concolino P, Filippetti F, Raspa-
glio G, Prislei S, Gallo D, Martinelli E, Ranelletti FO,
Ferrandina G et al. (2005) Class III b-tubulin overex-
pression is a prominent mechanism of paclitaxel resist-
ance in ovarian cancer patients. Clin Cancer Res 11,
298–305.
18 Dumontet C, Isaac S, Souquet PJ, Bejui-Thivolet F,
Pacheco Y, Peloux N, Frankfurter A, Luduen
˜
aR&
Perol M (2005) Expression of class III b-tubulin in

non-small cell lung cancer is correlated with
resistance to taxane chemotherapy. Bull Cancer 92,
E25–E30.
19 Appendino G, Danieli B, Jakupovic J, Belloro E,
Scambia G & Bombardelli E (1997) Synthesis and
evaluation of C-seco paclitaxel analogues. Tetrahedron
Lett 38, 4273–4276.
20 Ferlini C, Raspaglio G, Mozzetti S, Cicchillitti L,
Filippetti F, Gallo D, Fattorusso C, Campiani G &
Scambia G (2005) The seco-taxane IDN5390 is able to
target class III b-tubulin and to overcome paclitaxel
resistance. Cancer Res 65, 2397–2405.
21 Bollag DM, McQueney PA, Zhu J, Hensens O, Koupal
L, Liesch J, Goetz M, Lazarides E & Woods CM (1995)
Epothilones, a new class of microtubule-stabilizing
agents with a taxol-like mechanism of action. Cancer
Res 55, 2325–2333.
22 Altmann KH (2005) Recent developments in the chemical
biology of epothilones. Curr Pharm Des 11, 1595–1613.
23 Lo
¨
we J, Li H, Downing KH & Nogales E (2001)
Refined structure of ab-tubulin at 3.5 A
˚
resolution.
J Mol Biol 313, 1045–1057.
24 Nettles JH, Li H, Cornett B, Krahn JM, Snyder JP &
Downing KH (2004) The binding mode of epothilone A
on a,b-tubulin by electron crystallography. Science 305,
866–869.

25 Higgins DG, Thompson JD & Gibson TJ (1996) Using
CLUSTAL for multiple sequence alignments. Methods
Enzymol 266, 383–402.
26 Dunbrack RL (1999) Comparative modeling of CASP3
targets using PSI-BLAST and SCWRL. Proteins Suppl
3, 81–87.
27 Discover. Accelrys Software Inc, ª.
28 Alcaro S, Gasparrini F, Incani O, Mecucci S, Misiti D,
Pierini M & Villani C (2000) A quasi-flexible automatic
docking processing for studying stereoselective recogni-
tion mechanisms. Part I. Protocol validation. J Comput
Chem 21, 515–530.
29 Snyder JP, Nettles JH, Cornett B, Downing KH &
Nogales E (2001) The binding conformation of Taxol in
b-tubulin: a model based on electron crystallographic
density. Proc Natl Acad Sci USA 98, 5312–5316.
30 Johnson SA, Alcaraz AA & Snyder JP (2005) T-Taxol
and the electron crystallographic density in b-tubulin.
Org Lett 7, 5549–5552.
31 Mohamadi F, Richards NGJ, Guida WC, Liskamp R,
Lipton M, Caufield C, Chang G, Hendrickson T & Still
WC (1990) MacroModel – an integrated software system
for modeling organic and bioorganic molecules using
molecular mechanics. J Comput Chem 11, 440–467.
32 Weiner SJ, Kollman PA, Case DA, Singh UC, Chio C,
Alagona G, Profeta S & Weiner P (1984) A new force
field for molecular mechanical simulation of nucleic
acids and proteins. J Am Chem Soc 106, 765–784.
33 Hasel W, Hendrickson TF & Still WC (1988) A rapid
approximation to the solvent accessible surface areas of

atoms. Tetrahedron Comput Methodol 1, 103–116.
34 Morris GM, Goodsell DS, Halliday RS, Huey R, Hart
WE, Belew RK & Olson AJ (1998) Automated docking
using a lamarckian genetic algorithm and an empirical
binding free energy function. J Comput Chem 19, 1639–
1662.
M. Magnani et al. Docking of antimitotic agents to tubulin isoforms
FEBS Journal 273 (2006) 3301–3310 ª 2006 The Authors Journal compilation ª 2006 FEBS 3309
Supplementary material
The following supplementary material is available
online:
Table S1. RMSd values (A
˚
) after optimal superposition
of the backbone atoms of known tubulin structures.
Table S2. RMSd values (A
˚
) after optimal superposition
of the backbone atoms of ligand binding residues.
These are defined as the residues within 4 A
˚
of any
atom of either paclitaxel in the 1JFF structure or epo-
thilone in 1TVK. They are: Glu2 2, Val23, Asp26,
Glu27, Leu217, Gly225, Asp226, His229, Leu230,
Ala233, Ser236, Phe272, Pro274-Arg278, Arg284,
Pro360, Arg369-Leu371 (1JFF numbering). The num-
ber of superimposed atoms is 176 per pair with the
exception of the superpositions involving 1SA0 due to
missing residues in this latter structure.

Table S3. Interactions between paclitaxel and tubulin
observed in the 1JFF entry as reported by Ligplot
( />pdbsum/).
This material is avalilable as part of the online arti-
cle from .
Docking of antimitotic agents to tubulin isoforms M. Magnani et al.
3310 FEBS Journal 273 (2006) 3301–3310 ª 2006 The Authors Journal compilation ª 2006 FEBS