Crystal structure of the Eg5 - K858 complex and implications for structure-based design of thiadiazole-containing inhibitors

Article abstract of DOI:10.1016/j.ejmech.2018.07.006

The thiadiazole scaffold is an important core moiety in a variety of clinical drug candidates targeting a range of diseases. For example, the 2,4,5-substituted 1,3,4-thiadiazole scaffold is present in a lead compound and at least two clinical candidates t

Full text of DOI:10.1016/j.ejmech.2018.07.006

Accepted Manuscript
Crystal structure of the Eg5 - K858 complex and implications for structure-based
design of thiadiazole-containing inhibitors
Sandeep K. Talapatra, Chuin Lean Tham, Paolo Guglielmi, Roberto Cirilli, Balakumar
Chandrasekaran, Rajshekhar Karpoormath, Simone Carradori, Frank Kozielski
PII:
S0223-5234(18)30562-2
10.1016/j.ejmech.2018.07.006
EJMECH 10541
DOI:
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 10 April 2018
Revised Date: 20 June 2018
Accepted Date: 3 July 2018
Please cite this article as: S.K. Talapatra, C.L. Tham, P. Guglielmi, R. Cirilli, B. Chandrasekaran, R.
Karpoormath, S. Carradori, F. Kozielski, Crystal structure of the Eg5 - K858 complex and implications
for structure-based design of thiadiazole-containing inhibitors, European Journal of Medicinal Chemistry
(2018), doi: 10.1016/j.ejmech.2018.07.006.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT
Crystal structure of the Eg5 - K858 complex and implications for structure-
based design of thiadiazole-containing inhibitors
Sandeep K. Talapatraa,1, Chuin Lean Thama,1, Paolo Guglielmib, Roberto Cirillic,
Balakumar Chandrasekarand, Rajshekhar Karpoormathd, Simone Carradorie
and Frank Kozielskia*
aSchool of Pharmacy, University College London, 29-39 Brunswick Square, London,
WC1N 1AX, United Kingdom,
bDepartment of Drug Chemistry and Technologies, Sapienza University of Rome,
P.le A. Moro 5, 00185 Rome, Italy,
cNational Institute of Health, Centre for the Control and Evaluation of Medicines,
Viale Regina Elena 299, 00161 Rome, Italy.
dDiscipline of Pharmaceutcial Sciences, College of Health Sciences, University of
KwaZulu-Natal (UKZN), Westville, Durban, 4001, South Africa,
eDepartment of Pharmacy, G. D'Annunzio University of Chieti/Pescara, Via dei
Vestini 31, 66100 Chieti, Italy,
Corresponding author: f.kozielski@ucl.ac.uk
1Both authors contributed equally
1

ACCEPTED MANUSCRIPT
Highlights
• Eg5 is a potential target for drug development in cancer chemotherapy
• The thiadiazole scaffold containing inhibitors target the Eg5 motor domain
• (S)-ARRY-520 has advanced to clinical phase III trials against multiple myeloma
• The Eg5-K858 complex provides crucial insights into this important chemical
scaffold
• Using the crystal structure of the complex we rationalised our structure-activity
relationship study
Keywords
1,3,4-thiadiazole, K858 enantiomers, Eg5, antimitotic drugs, Eg5-K858 complex,
rational drug design
Abbreviations
KSP, Kinesin spindle protein; MT, Microtubule; PDB, Protein Data Bank; PK,
Pyruvate kinase; LDH, Lactate dehydrogenase; ITC, Isothermal titration calorimetry;
PIPES, Piperazine-N,N′-bis(2-ethanesulfonic acid); βMe, β-Mercaptoethanol; DMSO,
Dimethylsulfoxide; rpm, rounds per minute; HEPES, 4-(2-Hydroxyethyl)-1-
piperazineethanesulfonic
acid;
PEG,
Polyethyleneglycol;
MES,
2-(N-
morpholino)ethanesulfonic acid; AU, Asymmetric unit; SAR, Structure-activity
relationship.
Abstract
The thiadiazole scaffold is an important core moiety in a variety of clinical drug
candidates targeting a range of diseases. For example, the 2,4,5-substituted 1,3,4-
thiadiazole scaffold is present in a lead compound and at least two clinical
candidates targeting the human motor protein Eg5, against neoplastic diseases. An
inhibitor named K858 has in vivo activity in various mouse xenografts whereas the
clinical candidates (S)-ARRY-520 and (R)-Litronesib have entered clinical trials with
2

ACCEPTED MANUSCRIPT
the former one in phase III clinical trials either alone or in combination with a
proteasome inhibitor against relapsed/refractory multiple myeloma. Astonishingly,
structural data are lacking for all thiadiazole-containing Eg5 inhibitors. Here we
report the structure determination of two crystal forms of the ternary Eg5-ADP-K858
complex, locking the motor in the so-called final inhibitor bound state, thus blocking
ADP release, a crucial stage for Eg5 activity. K858 acts at the established allosteric
inhibitor-binding pocket formed of helix α2, loop L5 and helix α3. The structure of the
complex has far reaching consequences for thiadiazole containing Eg5 inhibitors.
For example, we could rationalise the structure-activity relationship in the crucial 5-
position of the thiadiazole scaffold and the complex will serve in the future as a basis
for strucutre-based drug design.
1. Introduction
Established tubulin-targeting chemotherapeutic drugs such as taxanes and vinca
alkaloids are clinically proven anti-cancer agents. They target the mitotic spindle, a
molecular machinery made of microtubules (MTs), which is responsible for
chromosome segregation during cell division [1-3]. However, MTs are also key
components in neuronal transport, and their disruption can lead to serious toxic side
effects such as neurotoxicity [4]. Therefore, certain mitotic kinesins are investigated
for their potential as novel drug targets with the hope to develop complementary
treatment options and to reduce side effects.
The mitotic kinesin Eg5 (also called Kif11, Kinesin Spindle Protein, KSP), a member
of the kinesin-5 subfamily [5], is essential for establishing the bipolar mitotic spindle
[6]. Structurally, Eg5 exists as a homotetramer via oligomerisation of the central stalk
domain [7, 8]. As there is a second MT-binding site in the C-terminal tail domain, the
quaternary structure of homotetrameric Eg5 imparts the functionality of crosslinking
and sliding anti-parallel MTs apart, which is essential for centrosome separation
during early stages of cell division [9].
3

ACCEPTED MANUSCRIPT
Figure 1: Chemical structures of related Eg5 inhibitors. Racemic K858, (S)-
ARRY-520 (also known as Filanesib) and (R)-Litronesib (LY2523355) contain the
1,3,4-thiadiazole scaffold, which is coloured in blue. MK-0731 is a dihydropyrrole
containing Eg5 inhibitor developed by Merck whereas compound 9h contains a
dihydropyrazole scaffold. Please also note that the 5-phenyl group is present in the
chemical structures shown here and all inhibitors display a very similar 2,4,5
substitution pattern, sometimes even sharing identical substituents.
Eg5 has become an attractive target for drug development in cancer chemotherapy
since the discovery of monastrol, the first selective inhibitor of Eg5 [10]. The
inhibition of Eg5 results in the formation of monopolar spindles, which leads to
mitotic arrest and may lead to subsequent apoptotic cell death. As Eg5 is crucial and
specific for mitosis, its inhibition would achieve anti-mitotic effects with fewer and
less undesirable effects than have been observed for tubulin targeting agents [11-14].
Today, several inhibitors targeting Eg5 have entered phase I, II and III clinical trials
either as monotherapies or in combination with other drugs [15].
4

ACCEPTED MANUSCRIPT
Clinical candidates targeting Eg5 show a variety of chemically distinct scaffolds, with
thiadiazole-based (S)-ARRY-520 (Figure 1) being the most encouraging inhibitor to
enter phase III clinical trials as a combination therapy with a proteasome inhibitor for
the treatment of multiple myeloma [16, 17]. Despite of the promising clinical results, it
remains to be established where exactly (S)-ARRY-520 binds to the motor domain of
Eg5 to exert its highly specific inhibitory effects. The structural information is urgently
needed to understand why (S)-ARRY-520 is so successful in advancing into phase
III clinical trials.
Another clinically relevant Eg5-targeting inhibitor containing the thiadiazole scaffold
is (R)-Litronesib (also named LY2523355). (R)-Litronesib (Figure 1) causes mitotic
arrest and apoptotic cell death in tumour cells and has been shown to be active in a
variety of mouse xenograft tumour models [18]. Dose finding and dose-escalating
studies were conducted in a range of phase I clinical trials in advanced solid tumours
[19, 20] in preparation for phase II trials. Later, it was shown that (R)-Litronesib
displayed chiral instability leading to racemisation of the active enantiomer [21].
Eventually, the drug candidate was discontinued for strategic reasons [22].
K858, the third example of an Eg5 inhibitor containing the thiadiazole as a core
scaffold and which is structurally related to (S)-ARRY-520 and (R)-Litronesib (Figure
1), was first discovered in a phenotype-based screen as a novel Eg5-specific
inhibitor [23]. Racemic K858 (rK858) inhibited the MT-stimulated Eg5 ATPase
activity (IC50 = 1.3 µM) in an ATP-uncompetitive manner. It induced monoastral
spindles, mitotic arrest and subsequent cells death in HTC116 tumour cells. It did not
inhibit other human kinesins tested and did not influence MT dynamics in vitro nor
had a detectable effect on the MT network in cells, indicating specificity for Eg5.
Even though it was not a very potent inhibitor against the MT-stimulated ATPase
activity, it inhibited tumour growth in an A2780 ovarian cancer xenograft model and
did not exhibit any neurotoxicity in this animal model. Importantly, it was also able to
induce cell death in paclitaxel-resistant cancer cells, suggesting that rK858 may
serve as a starting point for the development of more potent anti-cancer agents,
based on the thiadiazole scaffold.
Recently, it has been shown that rK858 induces apoptosis and reverses the
malignant invasive phenotype in human glioblastoma cells [24] and also influences
5

ACCEPTED MANUSCRIPT
survivin-related chemoresistance in breast cancer cells [25]. While the anti-Eg5 and
anti-tumour activities of rK858 have been characterised in cell-based assays and in
vivo, the biochemical and structural basis for inhibiting Eg5 not only by rK858, but
also for the other two thiadiazole containing derivatives (S)-ARRY-520 and (R)-
Litronesib, are still unknown. Here, we further describe the inhibition effects of rK858
using biochemical and biophysical assays. We also determined two distinct crystal
structures of the ternary complex of the human Eg5 motor domain in complex with
Mg2+ADP and K858. Our crystal structures reveal important structural insights into
how thiadiazole-containing compound binds to the allosteric drug-binding pocket of
Eg5. This provides a first glimpse of the important thiadiazole scaffold as a potent
inhibitor of Eg5 and has implications for further structure-based design. Finally, we
conducted a structure-activity relationship (SAR) study by determining the IC50
values for 46 analogues of K858 and rationally interpreted the results using the
structure of the complex as a basis.
2. Results and Discussion
For biochemical experiments and co-crystallisation, rK858 was synthesised as
previously described [26, 27] and its synthetic route is presented in Scheme 1 as a
representative example for the thiadiazole derivatives (1-46) measured in this study.
Acetophenone (a) in methanol was reacted with thiosemicarbazide (b) under
vigorous stirring to afford (E)-2-(1-phenylethylidene)hydrazine-1-carbothioamide (c)
1
quantitatively. The H NMR data of c revealed that a singlet peak resonating at δ
10.27 ppm was attributed to the –NH proton. One of the thioamidic protons (-
C=(S)NH2) appeared as a singlet at δ 8.26 ppm, while the other proton had merged
with the aromatic protons in the aromatic region between δ 7.91 - 7.93 ppm.
Moreover, the methyl (CH3) protons resonated at δ 2.29 ppm. In 13C NMR, we
observed a characteristic peak at δ 178.93 ppm which indicated the presence of the
C=S functionality. The compound c undergoes a cyclisation and a subsequent
acetylation reaction in acetic anhydride to give N-(4-acetyl-5-methyl-5-phenyl-4,5-
dihydro-1,3,4-thiadiazol-2-yl)acetamide (1) as a racemic mixture. The formation of 1
was confirmed by the presence of two additional methyl group protons appearing as
6

ACCEPTED MANUSCRIPT
1
distinctive singlets at δ 2.01 ppm and δ 2.19 ppm in the shielded region of H NMR.
13
This has been further established in C NMR where two new methyl group (CH3)
carbons appeared at δ 23.6 ppm and δ 26.5 ppm. In addition, two new carbonyl
carbons (C=O) were also recorded at δ 167 and 169 ppm, respectively. Thus the
structure of compound 1 (rK858), was confirmed based on 1H and 13C NMR spectral
data, which corroborated with previous literature [23]. The syntheses and
characterisation of the 46 selected K858 analogues investigated in our SAR study
were recently described [28].
Scheme 1: Synthetic route for rK858 [26, 27].
2.1 Biochemical and biophysical characterisation of inhibition of Eg5 by rK858
Although the inhibitory effect of rK858 on the MT-stimulated Eg5 ATPase activity has
been reported [23], we were curious whether rK858 was also able to directly bind to
and inhibit the basal ATPase activity of Eg5. rK858 was capable of inhibiting the
basal ATPase activity of Eg5 with an IC50 value of 740 ± 0.4 nM (Figure 2A). To
verify the inhibition of the MT-stimulated ATPase activity, the assays were repeated
in the presence of MTs and we obtained an IC50 value of 738 ± 0.3 nM, similar to the
value obtained by Nakai and close to the IC50 value for the inhibition of the basal
ATPase activity (Figure 2B) [23, 29]. As a complementary assay, we also employed
isothermal titration calorimetry (ITC) to determine the binding affinity of rK858 for
Eg5. ITC analysis revealed that rK858 is a nanomolar inhibitor that tightly binds to
Eg5 with a Kd value of 664.5 ± 0.7 nM, with one inhibitor molecule bound per
catalytic domain of Eg5 (Figures 2C).
7

ACCEPTED MANUSCRIPT
Figure 2: Biochemical and biophysical characterisation of the Eg5-K858
complex. A) Concentration-response plots of K858 for the inhibition of the basal
Eg5 ATPase activity. Inhibition of the basal ATPase activity by rK858 (●), the first
eluted K858 enantiomer on the Chiralpak IA column (K858-1) (■), and the second
eluted enantiomer K858-2 (□). B) Concentration-response plots of K858 for the
inhibition of the MT-stimulated Eg5 ATPase activity. Inhibition of the MT-stimulated
ATPase activity by rK858 (●), the first eluted K858 enantiomer (K858-1) (■), and
the second eluted enantiomer K858-2 (□). C) ITC analysis of the Eg5-K858
interaction. The raw ITC data of the binding reaction between Eg5 and rK858 is
shown in the upper panel. The lower panel shows normalised ITC data plotted
against the molar ratio of inhibitor to protein. The data fits very well to the single-
site binding model. D) Thermodynamic parameters, Kd of binding and
stoichiometry for rK858 and the two enantiomers K858-1 and K858-2 obtained
from the ITC experiments.
8

ACCEPTED MANUSCRIPT
2.2 Overall structure of the ADP-Eg5-K858 ternary complex
Since rK858 was capable to bind to and inhibit the basal ATPase activity of Eg5, this
allowed us to set up crystallisation trials of the Eg5-K858 complex, in order to
determine the structure of the ADP-Eg5-K858 ternary complex. We were able to
crystallise the Eg5-rK858 complex in two distinct space groups and to determine the
structures for both complexes, albeit at distinct resolutions (Table 1). In space group
C2221, crystals diffracted to 1.8 Å resolution and contained one Eg5-rK858 complex
per asymmetric unit (AU). In contrast, in the second condition, crystals formed in
space group P6522 with one molecule in the AU, but diffracted to only 2.8 Å
resolution. In both crystal forms the complex structures represent the final inhibitor
bound state. In this conformation, the switch II cluster is in the permissive position,
allowing the neck-linker to dock to the motor domain (Figure 3A and B). In the two
molecules of the two distinct crystal forms, rK858 binds to the same pocket in an
identical orientation. Due to the significantly higher resolution of the first crystal form,
we will concentrate our subsequent structure interpretation using the high-resolution
complex present in the C2221 complex.
9

ACCEPTED MANUSCRIPT
Figure 3: Overall structure of the Eg5-ADP-(R)-K858 complex. A) Front view of
the overall structure of the Eg5 motor domain with bound Mg2+ADP in the nucleotide-
binding pocket. (R)-K858 (magenta) is located in the allosteric helix α2/loop L5/helix
α3 inhibitor-binding pocket. Both, Mg2+ADP and (R)-K858 are shown as ball-and-
stick models. B) Back view showing that the switch II region is in the so-called
permissive orientation (coloured in orange) allowing the neck linker (shaded in purple)
to dock to the motor domain. C) (R)-K858 and the fo-fc map contoured at 3σ, showing
the quality of the electron density map. D) Stereoplot of the magnification of the
inhibitor-binding pocket. Residues, which are part of the pocket (green) and (R)-
K858 (magenta), are shown as ball-and-stick models. Water molecules are shown as
red cross and hydrogen bond interactions between the inhibitor and residues of Eg5
are indicated as dotted lines. Leu214, which has been shown to be crucial for
conferring resistance to K858 when mutated to alanine, is coloured in black.
10

ACCEPTED MANUSCRIPT
2.3. Molecular insights into the K858 binding pocket
The magnification of the inhibitor binding pocket with bound K858 is shown in Figure
3D. K858 binds to the allosteric inhibitor-binding pocket formed by helix α2, loop L5
and helix α3. Although additional inhibitor binding pockets have previously been
identified in Eg5 such as the α4/α6 pocket, an additional site in the MT binding
region and a fourth site in or close to the nucleotide binding pocket, K858 binds to
the well-established allosteric pocket, which accommodates several other Eg5
inhibitor scaffolds. Although we used the racemic mixture of K858 for crystallisation
of the complex we exclusively observed electron density for the (R)-enantiomer in
the loop L5 pocket, indicating enantio-selectivity for K858 (Figure 3D). Enantio-
selectivity has also been observed for ARRY-520 and Litronesib with the S and the
R-enantiomers being the active compounds, respectively, although no structural data
are available for these two thiadiazole containing clinical inhibitors yet.
Although (R)-K858 is a relatively non-sophisticated Eg5 inhibitor, compared to (S)-
ARRY-520 (inhibition of the MT-stimulated Eg5 ATPase activity: IC50 = 6 nM, [30])
and (R)-Litronesib (inhibition of the MT-stimulated Eg5 ATPase activity: IC50 = 26 nM
[31]), it nevertheless displays a complex network of interactions with residues in the
inhibitor-binding pocket (Figure 3D). There are a variety of structural water molecules
establishing interactions with (R)-K858, in particular the nitrogen atoms of the 1,3,4-
thiadiazole moiety and the N-acetamido group. The 2-N-acetamido group is shielded
from the solvent and points towards the interior of the inhibitor-binding pocket
displaying hydrophobic interactions with Ile136, Leu160 and Phe239. The 4-acetyl
group is on the opposite side of the thiadiazole ring pointing towards the solvent
showing a hydrogen bond interaction with Tyr211 and some more distant water
molecules. The asymmetric C5 atom of the thiadiazole ring houses the 5-methyl and
5-phenyl groups and is the reason for K858’s enantio-selectivity resulting in the
active R-enantiomer. Whereas the 5-methyl group is oriented towards the solvent
region, the 5-phenyl group is pointing towards a more hydrophobic portion of the
binding pocket establishing hydrophobic interactions with the side chains of residues
Arg119, Trp127 and Pro137. By switching the positions of two substituents, thus
generating a virtual S-enantiomer, the phenyl group with a significantly larger
hydrophobic character than the methyl would be placed towards the solvent region
resulting in larger non-favourable interactions rationally explaining why the R-
11

ACCEPTED MANUSCRIPT
enantiomer is favoured over the S-enantiomer. Indeed, we do not observe any
electron density indicative for mixed binding of both enantiomers, further supported
by the high resolution of our structure of 1.8 Å. Finally, the sulphur atom of the ring
system interacts with the main chain carbonyl oxygen of Glu116, acting as the
electron partner through a hydrogen bond interaction.
To further confirm enantioselectivity of Eg5 for the (R)-enantiomer, we resolved the
racemic mixture of K858 into its enantiomers by HPLC on the polysaccharide-based
IA chiral stationary phase and used the collected enantiomeric pure forms of K858 in
comparative biological experiments. The first eluting enantiomer on the Chiralpak IA
column has been named K858-1 and the more retained enantiomer K858-2. The
enantiomer K858-1 displayed inhibitory activity at an IC50 of 580 ± 0.015 nM while
K858-2 does not inhibit Eg5 at all, when measuring the inhibition of the basal Eg5
ATPase activity (Figure 2A). Similarly, when measuring the inhibition if the MT-
stimulated ATPase acitvity, we obtained an IC50 of 493 ± 0.037 nM for K858-1
whereas K858-2 was inactive (Figure 2B). From ITC experiments a Kd of 220.0 ± 4.2
nM was recorded for K858-1, whereas K858-2 was again inactive (Figures 2C & D),
a comparable result as obtained in ATPase activity experiments. Therefore from the
biochemical assays, one could easily deduce that K858-1 is the active (R)-
enantiomer, whereas K858-2 represents the inactive (S)-enantiomer, further
displaying enantioselectivity.
K858 is essentially a “stripped-down” version of the two inhibitors, which reached
clinical development, (S)-ARRY-520 and (R)-Litronesib, albeit already a very potent
nanomolar inhibitor (Figure 1). None of the three inhibitors with a thiadiazole core
moiety has been solved in complex with Eg5 and our Eg5-(R)-K858 structure allows
a first glimpse into this important scaffold and how it inhibits Eg5 with exceptional
potency. The thiadiazole moiety and the phenyl group are present in all three
chemical structures. The substitution pattern is identical with substituents at the 2, 4
and 5 positions on the thiadiazole ring system. However, the three inhibitors vary in
the nature of their substituents, which account for their different IC50 values,
significantly better for the clinical candidates. Whereas K858 has an acetamido
group in the 2-position the two clinical candidates possess either a 2,5-difluorophenyl
group or a 2-pivalamide substituent. The 4-acetyl group in K858 can still be
recognised in the same position on the two other clinical candidates, but as more
12

ACCEPTED MANUSCRIPT
elaborated substituents such as the N-methoxy-N-methyl or the 4-pivaloyl
substituent. This trend is even more pronounced in the 5th position, which displays a
single methyl group in K858, whereas this small substituent is replaced by a 3-
aminopropyl group in (S)-ARRY-520 and a sulfonamido containing substituent in
Litronesib. In the case of (S)-ARRY-520 we predict that the 3-aminopropyl group is
destined to interact with Glu166 in Eg5, similar to the other inhibitors containing the
aminopropyl group such as STLC, ispinesib and its second generation analogue SB-
743921. Although the primary amine is known to be prone to hERG liabilities, this
has not been reported in any of the clinical trials for (S)-ARRY-520.
2.4. Structure-based, rational interpretation of binding of a range of K858 analogues
to the Eg5 inhibitor-binding pocket
Recently, some of us [28] synthesised about one hundred K858 analogues and
tested their anti-proliferative potential against the human prostate cancer cell line
(PC3) and two melanoma cell lines (SK-MEL-5 and SK-MEL-28). We were capable
of identifying ten compounds with anti-proliferative activity and compared the results
to the control compound rK858. These results were challenging to fully interpret in
terms of the SAR, because drug-like properties of these compounds such as
solubility, lipophilicity and cell permeability may have significant effects in cell-based
assays as opposed to cell membrane-free biochemical assays. We therefore
embarked on determining the IC50 values of about 50 selected rK858 analogues all
carrying modifications in the 5-position by quantifying the inhibition of the basal Eg5
ATPase activity. The results are summarised in Tables 1 to 4.
We first investigated the influence of the 5-methyl group on the Eg5 inhibitory activity
(Table 1). Whereas analogue 2 with a 5-ethyl group displayed only a modest drop in
activity (1.46 ± 0.3 ꢀM), analogue 3 lacking the 5-methyl group displayed a
significantly reduced IC50 value of 20.0 ± 7.3 ꢀM, compared to 0.74 ± 0.04 ꢀM for the
parent compound 1. A similar strong and systematic trend was observed for all other
analogues that either contained or were lacking the 5-methyl group (cmpd pairs 1/3,
26/27, 28/29, 33/34, 35/36, 41/42, 44/45) and will therefore not be discussed further.
Exchanging the planar 5-phenyl group with a staggered cyclohexyl substituent (4)
13

ACCEPTED MANUSCRIPT
led to a 10-fold drop in activity (7.5 ± 2.3 ꢀM), indicating that the planar phenyl ring
establishes better hydrophobic interactions in this part of the inhibitor-binding pocket.
Table 1: Inhibition of the basal Eg5 ATPase activity by a variety of rK858 analogues
with alkyl substituents in the 5-position. MIA: Maximum inhibition attained.
Compd
No
IC50 [ꢀM]
(MIA)
R1
R2
1
0.74 ± 0.04
(96)
Control
rK858
Me
Ph
1.46 ± 0.3
(95)
20.0 ± 7.3
(75)
7.5 ± 2.3
(85)
n.i.
2
3
Et
H
Ph
Ph
4
Me
Me
Me
Me
Et
Cyclohexyl
5
Me
(30)
n.i.
(5)
n.i.
(30)
n.i.
(35)
n.i.
(45)
n.i.
(30)
n.i.
(20)
n.i.
(30)
n.i.
(45)
p.i.
(50)
p.i.
6
Et
7
Pr
8
Et
9
Me
Me
Et
i-Bu
10
11
12
13
14
15
16
Bu
Pr
2-Methylprop-1-enyl
But-3-enyl
Pentyl
Me
Me
Me
Et
Bu
(60)
p.i.
(60)
Me
n-Hexyl
14

ACCEPTED MANUSCRIPT
We then investigated the inhibitor effects of a range of analogues in which the 5-
phenyl group was substituted with a variety of alkyl groups of varying length (cmpds
5 to 16). Although we expected at least some inhibitory activity because placing a
hydrophobic substituent into an essentially hydrophobic pocket should lead to at
least some favourable interactions, to our surprise none of the tested analogues fully
inhibited Eg5 activity. However, there is a clear trend between the length of the alkyl
chain and the extend of inhibition with the most extended alkyl chains in this position
(cmpds 15 and 16) indicating partial inhibition of Eg5 activity of up to 60% (Table 1).
From the crystal structure of Eg5-(R)-K858 complex, an interesting edge-to-face (T-
shaped) aromatic stacking interaction was observed between the phenyl ring (edge)
of Trp211 and phenyl group (face) of thiadiazole scaffold, which is essential for the
effective binding and subsequent inhibition of Eg5 besides having good hydrophobic
interactions with key residues Arg119 and Pro137. This specific edge-to-face
stacking interaction was not exhibited by the analogues containing hydrophobic alkyl
chains. Although being hydrophobic in nature, they can only establish weaker long
distance hydrophobic interactions disfavored by a higher conformational freedom,
which are, however, not sufficient for effective inhibition. In our previous published
work, we had observed similar crucial edge-to-face contacts involving phenyl groups
between STLC (Eg5 inhibitor) and Eg5 [32]. These data clearly identify the phenyl
group as a crucial substituent of the thiadiazole scaffold for effectively inhibiting Eg5.
We then examined rK858 analogues, in which the methyl and phenyl substituents at
the 5-position are combined into alicyclic substituents of increasing ring sizes and
additional substituents at the cyclohexyl ring system (Table 2). Neither 5-, 6-,
substituted 6-, 7- or 8-membered ring systems displayed any inhibitory activity
against Eg5 thus suggesting that a certain degree of conformational freedom is
requested in this position as they can be considered as cyclic analogues of
derivatives 7-16. In summary synthesising analogues, which combine the two
substituents at the asymmetric carbon atom into non-aromatic, cyclic substituents
(17-25) did also not lead to more potent K858 analogues.
15

ACCEPTED MANUSCRIPT
Table 2: Inhibition of the basal Eg5 ATPase activity by a variety of rK858 analogues
substituted with varying alicyclic ring systems in the 5-position. MIA: Maximum
inhibition attained.
Compd
No
IC50 [ꢀM]
(MIA)
n.i.
Cycle
17
18
19
20
21
22
23
24
25
cyclopentyl
(40)
p.i.
(60)
p.i.
(60)
n.i.
(40)
n.i.
(40)
n.i.
(10)
n.i.
(30)
n.i.
(30)
n.i.
cyclohexyl
2-methylcyclohexyl
3-methylcyclohexyl
4-methylcyclohexyl
3,3-dimethylcyclohexyl
2-tert-butylcyclohexyl
cycloheptyl
cyclooctyl
(15)
A group of analogues incorporating various heterocyclic rings in the 5-position
instead of the phenyl group resulted in a variety of active analogues (cmpds 26 to 37,
Table 3). Whereas cmpd 26, in which the phenyl group is substituted by a furan ring
system has a slightly lower IC50 value of 4.3 ± 0.9 ꢀM, cmpd 28 with an unsaturated
thiophen-2-yl substituent (bioisosteric replacement) displayed an IC50 value of 1.7 ±
0.1 ꢀM. When the thiophene system is coordinated to the thiadiazole ring system
through its 3-position, the IC50 is still 2.2 ± 0.3 ꢀM. For the analogues containing an
unsaturated heterocyclic ring system, the most potent analogue was cmpd 30 with a
halogen substituent at the 5-position of the thiophene group. We also investigated 6-
membered unsaturated pyridine substituents linked to the 5-position via the 2, 3 or 4
position of the heterocyclic pyridine ring. Whereas cmpd 33 still displays reasonable
inhibitory activity against Eg5 with an IC50 value of 8.5 ± 0.8 ꢀM, cmpds 32 and 35, in
16

ACCEPTED MANUSCRIPT
which the pyrimidine rings are linked to the 5-position of the thiadiazole scaffold via
their 2- and 4-positions are significantly less active with IC50 values of 37.2 ± 8.8 ꢀM
and 59.9 ± 9.2 ꢀM, respectively. Not surprisingly, cmpd 37, containing the 1,4-
diazine scaffold is a weak inhibitor of Eg5 (IC50 value of 46.9 ± 6.9 ꢀM).
Table 3: Inhibition of the basal Eg5 ATPase activity by a variety of rK858 analogues
with heterocyclic ring systems in the 5-position. MIA: Maximum inhibition attained.
Compd
No
IC50 [ꢀM]
(MIA)
4.3 ± 0.9
(95)
n.i.
(20)
1.7 ± 0.1
(95)
n.i.
(30)
0.95 ± 0.15
(95)
2.2 ± 0.3
(90)
37.2 ± 8.8
(65)
8.5 ± 0.8
(90)
n.i.
(30)
59.9 ± 9.2
(60)
n.i
(5)
46.9 ± 6.9
(70)
R1
Me
H
R2
Furan-2-yl
26
27
28
29
30
31
32
33
34
35
36
37
Furan-2-yl
Me
H
Thiophen-2-yl
Thiophen-2-yl
5-Chloro-thiophen-2-yl
Thiophen-3-yl
Pyridin-2-yl
Me
Me
Me
Me
H
Pyridin-3-yl
Pyridin-3-yl
Me
H
Pyridin-4-yl
Pyridin-4-yl
Me
Pyrazin-2-yl
Table 4 contains K858 analogues with ortho-, meta- and para-substitutions at the
structural phenyl ring system (38-42) as well as fused ring substituents (43-47).
17

ACCEPTED MANUSCRIPT
Compounds 41 and 42 are characterised by the same 2,5-difluoroaryl substituent
present in (R)-Litronesib; the former was endowed with an inhibitory potency
comparable to rK858. Cmpd 44 containing a 1-naphthyl group in the 5-position still
shows good activity (IC50 = 1.9 ꢀM), proving that this part of the inhibitor-binding
pocket can accommodate larger aromatic substituents as long as these are planar
and hydrophobic. Also in this case, the concurrent presence of a 5-methyl group was
important for the inhibitory activity (compare 44-46).
Table 4: Inhibition of the basal Eg5 ATPase activity by a variety of rK858 analogues
with substitutions in the phenyl ring, fused aromatic and heterocyclic systems in the
5-position. MIA: Maximum inhibition attained.
Compd
No
IC50 [ꢀM]
(MIA)
n.i.
R1
Me
Me
Me
Me
H
R2
38
39
40
41
42
43
44
45
46
47
4-Nitrophenyl
(25)
n.i.
(20)
4-Acetylphenyl
4-Acetyloxyphenyl
2,5-Difluorophenyl
2,5-Difluorophenyl
Benzo[d][1,3]dioxol-5-yl
1-Naphthyl
5.1 ± 0.2
(95)
0.84 ± 0.3
(97)
41.1 ± 8.1
(60)
n.i.
(10)
1.9 ± 0.1
(95)
53.9 ± 7.1
(60)
H
Me
H
1-Naphthyl
15.6 ± 2.6
(80)
Me
Me
2-Naphthyl
n.i.
(20)
2-Oxo-2H-chromen-3-yl
Table 5 lists selected rK858 analogues with a notable inhibitory activity against Eg5
and a range of their calculated drug-like properties. Interestingly, all analogues
18

ACCEPTED MANUSCRIPT
display favourable lipophilicity (clogP < 3.5), PSA values (<75) and good calculated
solubility with the exception of cmpd 44 (calc. LogS = -4.486), which contains the
much more lipophilic 1-naphthyl substituent. Remarkably, the phenyl group in rK858
is the most crucial structural requirement, responsible for imparting optimal steric
and electronic properties for all analogues tested. While rK858 remains the most
active Eg5 inhibitor tested in this study, compounds 28, 30, 31 and 41, display
inhibitory activity and ligand efficiency values close to rK858.
It is interesting to note that all three thiadiazole-based Eg5 inhibitors (Figure 1) retain
the 5-phenyl group as a substituent and we can safely assume that the clinically
relevant compounds such as (S)-ARRY-520 and (R)-Litronesib have certainly gone
through extensive chemical optimisation of this substituent (Figure 1 & 3D), although
a SAR study has not been made available for any of these clinically important drugs.
This phenyl group seems to fit snuggly into this pocket and represent an ideal
substituent with optimal binding features in this predominantly hydrophobic part of
the pocket (Figure 4C). To gain more insights about potential structural relationships
we compared all available Eg5 inhibitor scaffolds for which structural information is
available in the PDB with the binding mode of K858. Interestingly, two types of
scaffolds, dihydropyrrole and dihydropyrazole containing Eg5 inhibitors, which are
structurally related to the thiadiazole scaffold (Figure 1, lower panel) can be fully
Table 5: Summary of inhibitory activity of rK858 and analogues against Eg5 (IC50
values < 4.3 ꢀM) and some calculated drug-like properties such as ligand efficiency,
clogP, solubility LogS, and polar surface area (PSA). MIA: Maximum inhibition
attained.
Solubility
(calc.
LogS)
Compd
No
IC50 [ꢀM]
(MIA)
Analogue Structure
LE
clogP
PSA
1
0.74 ± 0.04
(96)
0.324
2.324
-3.108
61.77
Control
19

ACCEPTED MANUSCRIPT
1.46 ± 0.3
0.292
2
2.853
-3.376
61.77
(95)
4.3 ± 0.9
0.298
(95)
26
28
1.5
-2.085
-2.553
74.91
61.77
1.7 ± 0.1
0.321
(95)
1.97
0.95 ± 0.15
0.317
30
2.683
-2.972
61.77
(95)
2.2 ± 0.3
0.314
(90)
31
41
1.97
-2.768
-3.643
61.77
61.77
0.84 ± 0.3
0.289
0.308
(97)
1.9 ± 0.1
0.249
(95)
51
3.498
-4.486
61.77
overlaid with (R)-K858, displaying the same binding mode (Figures 4A and B): all
scaffolds show the same substitution pattern with the thiadiazole, dihydropyrrole and
dihydropyrazole ring systems at the same position, as in the case of the
unsubstituted phenyl group. The 2,5-difluorophenyl substituent can be found in two
20

ACCEPTED MANUSCRIPT
clinical candidate molecules such as MK-0731 and (S)-ARRY-520 (Figure 1). The
combination of medicinal chemistry efforts from Merck by combining the
dihydropyrrole and dihydropyrazole scaffolds resulted in the MK-0731 [33]. Although
the chemical and the structural resemblance as well as their mode of action on Eg5
of both candidates is fully obvious, it clearly demonstrates how important small
differences in promising candidate compounds can be: whereas MK-0731 is mostly
insoluble and did not proceed past phase I clinical trials, its more successful “brother”
entered phase III clinical trails against multiple myeloma.
21

ACCEPTED MANUSCRIPT
Figure 4: Comparision of the interactions of different chemical scaffolds with
key residues of the allosteric Eg5 inhibitor-binding pocket. A) Stereoplot of the
magnification of the inhibitor-binding pocket showing the overlay of (R)-K858
(thiadiazole core, magenta) and MK-0731 (dihydropyrrole core, blue, PDB ID: 3CJO,
[33]). B) Stereoplot of the overlay between (R)-K858 (magenta) and a
dihydropyrazole containing scaffold (yellow, PDB ID: 2G1Q, [34]). All three core
structures possess the same substitution pattern and the core as well as the phenyl
groups are located in the same part of the inhibitor - binding pocket.
22

ACCEPTED MANUSCRIPT
3. Conclusions
In this study we determined the crystal structure of an Eg5-inhibitor complex
containing the 1,3,4-thiadiazole core scaffold, which is present in clinically relevant
Eg5 targeting inhibitors such as (S)-ARRY-520 and (R)-Litronesib. We were also
able to determine the IC50 values of 46 K858 analogues by measuring the inhibition
of the basal Eg5 ATPase activity and based on the structural insights gained from
the complex we could use this structure for rationalising our SAR study. This will
serve as a basis for future structure-based drug design of this very important
chemical scaffold.
4. Materials and methods
4.1. Chemistry
All chemicals and reagents (AR grade) were purchased from Sigma–Aldrich and
Merck Millipore, South Africa. Solvents except laboratory reagent grade were dried
and purified as per the reported literature methods when necessary. The progress of
reactions and purity of compounds were monitored by thin-layer chromatography
(TLC) on pre-coated aluminum silica gel plates procured from E. Merck and Co.
(Darmstadt, Germany) using methanol (10%) in chloroform as mobile phase and
1
13
iodine vapor as a visualising agent. The H NMR and C NMR were recorded on a
Bruker AVANCE 400 and 100 MHz (Bruker, Rheinstetten/Karlsruhe, Germany) using
DMSO-d6 or CDCl3. Chemical shifts are reported in δ ppm units with respect to TMS
as an internal standard.
4.1 Synthesis of racemic K858
Racemic K858 was synthesised as previously described [27, 28].
(E)-2-(1-phenylethylidene)hydrazine-1-carbothioamide (c)
To a solution of acetophenone a (33.3 mmol, 4 g) and thiosemicarbazide (b) (34.6
mmol, 3.15 g) in 90 ml of methanol, 0.3 ml of concentrated hydrochloric acid was
slowly added dropwise and stirred for 15 h. After completion of the reaction (as
monitored by TLC), 30 ml of water was added to the mixture to obtain product c. The
obtained solids were filtered and washed with diisopropyl ether to yield white crystals
23

ACCEPTED MANUSCRIPT
1
of b in pure form (4.7 g, 81% yield). H NMR (400 MHz, DMSO-d6) δ: 2.30 (s, 3H,
CH3), 7.37 – 7.39 (m, 3H, ArH), 7.91 – 7.93 (m, 2H, ArH, 1H, NH2), 8.27 (s, 1H, NH2),
13
10.2 (s, 1H, NH) ppm; C NMR (100 MHz, DMSO-d6) δ: 14.0 (CH3), 126.6, 128.2,
129.1, 137.6, 147.8, 178.9 ppm.
N-(4-acetyl-5-methyl-5-phenyl-4,5-dihydro-1,3,4-thiadiazol-2-yl)acetamide (1) or
K858
Compound c (0.889 mmol; 0.3 g) in acetic anhydride (11 mmol; 1 ml) was refluxed
until a homogenous mixture appeared. Then, the solution was cooled to room
temperature to obtain compound 1. Diethyl ether (120 ml) was added to remove non-
polar impurities to yield white crystals which were dried under vacuum (0.413g,
86 %). 1H NMR (400 MHz, DMSO-d6) δ: 2.01 (s, 3H, CH3), 2.19 (s, 3H, CH3), 2.28 (s,
3H, CH3), 7.24 – 7.25 (s, 1H, ArH), 7.25 – 7.35 (m, 4H, ArH), 11.63 (s, 1H, NH) ppm;
13C NMR (100 MHz, DMSO-d6) δ: 22.4, 23.7, 26.6 (3CH3), 78.6, 124.4, 127.3, 128.4,
142.2, 143.8, 167.7 and 169.3 (2C=O) ppm.
The syntheses and characterisation of the other 46 rK858 analogues (2-46) have
recently been described in detail [26].
4.2 Enantioselective HPLC.
Enantioselective HPLC analyses were performed by using stainless-steel Chiralpak
IA (250 mm × 4.6 mm i.d. and 250 mm × 10 mm i.d.) (Daicel, Chemical Industries,
Tokyo, Japan) columns. HPLC-grade solvents were used as supplied by Aldrich
(Milan, Italy). The HPLC apparatus consisted of a PerkinElmer (Norwalk, CT, USA)
200 LC pump equipped with a 542 Rheodyne (Cotati, CA, USA) injector, a 1000 ꢀL
sample loop, a HPLC Perkin-Elmer oven and a Perkin-Elmer 290 detector. The
signal was acquired and processed by Clarity software (DataApex, Prague, The
Czech Republic).
4.3 Subcloning, expression and purification of human Eg5
The human Eg51-368 motor domain was expressed and purified as previously
described [32].
24

ACCEPTED MANUSCRIPT
4.4 Steady-state ATPase activity assays
Steady-state basal and MT-stimulated Eg5 ATPase activities were measured using
the pyruvate kinase (PK)-lactate dehydrogenase (LDH) coupled assay as described
[35]. The quantities of Eg5 used in the basal and MT-stimulated conditions were 150
nM and 10 nM, respectively. The Mg2+ATP concentration was fixed at 1 mM and the
MT-stimulated ATPase activity was carried out in the presence of 1 µM MTs. For the
inhibition of the basal Eg5 ATPase activity, assays were carried out in the presence
of 150 mM KCl, whereas in the presence of MTs, no salt was added to the buffer
conditions. ATPase measurements were performed at 25°C using a 96-well Sunrise
photometer (Tecan, Mannesdorf, Switzerland). The IC50 values were calculated
using Kaleidagraph 4.0 (Synergy Software). MTs were polymerised from lyophilised
tubulin (Tebu-Bio catalogue No. 027T240-B) as described in [36].
4.5 Isothermal titration calorimetry (ITC)
Interactions between Eg5 and K858 were studied by ITC using a MicroCal PEAQ-
ITC (Malvern) as previously described, but with some modifications [37, 38]. Prior to
ITC analysis, Eg5 was dialysed in dialysis buffer (20 mM PIPES pH 6.9, 250 mM
NaCl, 2 mM β-mercaptoethanol, 0.5 mM ADP) overnight at 4°C. For the ITC analysis,
Eg5 was used at 10 µM and rK858, K858-1 and K858-2 at 100 µM. To compensate
for additional DMSO in the inhibitor solution, 1% DMSO was added to Eg5 before the
experiment. rK858 and its enantiomers were titrated into the Eg5 solution in 30
injections with spacing of 120 sec between each injection and a constant stirring
speed of 750 rpm. To calculate heats of dilution, rK858 and its enantiomers were
injected into buffer with 1% DMSO that was subtracted prior to data analysis using
the MicroCal PEAQ-ITC Analysis Software (Malvern). All ITC experiments were
performed at 25°C. The measured thermodynamic param eters n (stoichiometry), Ka
(association constant) and ∆H (enthalpy change) were acquired through fitting of the
experimental data using the single-site binding model. The binding free energy (∆G)
and entropy change (∆S) were then calculated from the fitted values. The ITC
analysis was performed in duplicate and the averaged values are presented.
25

ACCEPTED MANUSCRIPT
4.6 Crystallisation of the Eg5-rK858 complex
Purified Eg5 (10 mg/mL) was incubated with 1 mM rK858 and 1 mM Mg2+ATP
overnight at 4°C. Following incubation, the sample was centrifuged at 13,000 rpm for
10 min at 4°C to remove any protein aggregate or un dissolved inhibitor. The complex
was then used to set up crystallisation trials.
Crystal obtained in space group C2221: Needle-like crystals of the complex
appeared overnight at 18°C in hanging drops consist ing of 1 µl Eg5-rK858 complex
and 1 µl reservoir solution (0.1 M HEPES pH 8.0, 12.5% PEG3350, and 0.1 M
MgCl2). The crystals initially obtained did not diffract to a minimum threshold
resolution required to study protein-inhibitor interactions. Therefore, we employed
crystal dehydration techniques. Crystals were transferred to different drops to allow
further vapour diffusion with increasing amounts of PEG3350 with increments of 0.5%
to 10% compared to the amount originally present in the reservoir solution. The time
for dehydration varied between three days to a week. Eventually, plate-like crystals
appeared and these crystals diffracted to a significantly higher resolution compared
to the original needle-like crystals.
Crystals obtained in space group P6522: Rhombohedral shaped crystals appeared
after a fortnight in a mixture of 1 µl of the Eg5-rK858 complex and 1 µl of the
reservoir solution containing 0.1 M MES, pH 5.5, 24.5% PEG3350, 0.3 M (NH4)2SO4
and 0.1 M Trimethylamine at 4°C.
Crystals from both conditions were cryoprotected in a solution containing 1.2 fold of
the crystallisation solution present in the reservoir supplemented with ~15% glycerol
and flash-frozen in liquid nitrogen for subsequent data collection.
4.7 Data collection, structure determination, refinement and model quality
Diffraction data for the two distinct crystal forms were collected at the European
Synchrotron Radiation Facility (ESRF), beamline ID30A-1. Data were processed
using either XDS [39] or iMosflm [40] and scaled to resolutions as mentioned in Table
6 [41]. The structures of the Eg5-rK858 complex were solved in two space groups by
molecular replacement (PHASER MR in CCP4 suite) using a previously published
Eg5 structure (PDB ID: 1X88,[42]) as a search model. All structures were initially
26

ACCEPTED MANUSCRIPT
refined with REFMAC5 [43]. Electron density and difference density maps, all σA-
weighted, were inspected, and the models were improved using Coot [44]. Further
refinement of the structures was performed using PHENIX [45]. The calculation of
Rfree used 5% of data. Crystallographic and refinement statistics are given in Table 6.
Both space groups contain one Eg5-rK858 complex per AU. The complexes cover
residues Lys17 to Glu364 and Asn18 to Glu367 in space groups C2221 and P6522,
respectively. In the former space group, loops containing residues 56-60, 176-179,
249-243 and 272-286 are missing in the structure, whereas in the latter space group
the regions 34-35 and 272-287 are absent.
Table 6: Data collection, structure determination and refinement statistics for
the ternary ADP-Eg5–(R)-K858 complex.a, b
Statistics
Eg5-(R)-K858
complex in
space group
C2221
Eg5-(R)-K858
complex in
space group
P6522
PDB ID
6G6Y
ID30A-1
1
6G6Z
Beamline
ID30A-1
1
Molecules per asymmetric unit
Resolution range [Å]
Unit cell parameters [Å, °]
50 - 1.8
50 - 2.8
a = 62.5, b =
108.4, c =
111.4, α = β =
γ = 90
a = b = 100.3, c =
185.4, α = β = 90,
γ = 120
Completeness [%]
Rmerge
96.0 (93.5)
2.9 (51.6)
1.9 (1.8)
99.3 (96.5)
2.1 (41.8)
2.0 (3.3)
Multiplicity
27

ACCEPTED MANUSCRIPT
Mean I/σ(I)
13.9 (2.1)
22.8 (1.8)
CC1/2
99.9 (77.4)
99.9 (79.9)
Total reflections
Unique reflections
Wilson B-factor (Å2)
Refinement statistics
Rwork / Rfree [%]
Average B-factor (Å2):
- Overall
64132 (5878) 28050 (2454)
33988 (3262) 14182 (1326)
26.5
75.7
21.2 / 25.6
23.9 / 26.1
37.2
77.3
- Eg5 motor domain
- Solvent
37.1
77.6
41.2
70.2
- Mg2+ADP
25.8
70.9
- (R)-K858
25.7
76.3
No. of Mg2ADP / Inhibitor /
water
1 / 1 / 229
1 / 1 / 41
r.m.s.d. bond lengths [Å]
r.m.s.d. bond angles [˚]
Ramachandran plot statistics (%):
- Favoured
0.012
1.48
0.013
1.56
98.4
1.6
96.6
2.2
- Allowed
- Outliers
0.0
1.2
aThe racemic mixture of K858 was used for crystallisation. bValues in parentheses
pertain to the highest-resolution shell.
28